Health Informatics and Medical Systems [1 ed.] 9781683921936

Citation preview

PROCEEDINGS OF THE 2017 INTERNATIONAL CONFERENCE ON HEALTH INFORMATICS AND MEDICAL SYSTEMS

Editors Hamid R. Arabnia Leonidas Deligiannidis, Michael B. O'Hara Associate Editors Ashu M. G. Solo, Fernando G. Tinetti

CSCE’17 July 17-20, 2017 Las Vegas Nevada, USA americancse.org ©

CSREA Press

This volume contains papers presented at The 2017 International Conference on Health Informatics and Medical Systems (HIMS'17). Their inclusion in this publication does not necessarily constitute endorsements by editors or by the publisher.

Copyright and Reprint Permission Copying without a fee is permitted provided that the copies are not made or distributed for direct commercial advantage, and credit to source is given. Abstracting is permitted with credit to the source. Please contact the publisher for other copying, reprint, or republication permission.

© Copyright 2017 CSREA Press ISBN: 1-60132-459-6 Printed in the United States of America

Foreword It gives us great pleasure to introduce this collection of papers to be presented at the 2017 International Conference on Health Informatics and Medical Systems (HIMS’17), July 17-20, 2017, at Monte Carlo Resort, Las Vegas, USA. An important mission of the World Congress in Computer Science, Computer Engineering, and Applied Computing, CSCE (a federated congress to which this conference is affiliated with) includes "Providing a unique platform for a diverse community of constituents composed of scholars, researchers, developers, educators, and practitioners. The Congress makes concerted effort to reach out to participants affiliated with diverse entities (such as: universities, institutions, corporations, government agencies, and research centers/labs) from all over the world. The congress also attempts to connect participants from institutions that have teaching as their main mission with those who are affiliated with institutions that have research as their main mission. The congress uses a quota system to achieve its institution and geography diversity objectives." By any definition of diversity, this congress is among the most diverse scientific meeting in USA. We are proud to report that this federated congress has authors and participants from 64 different nations representing variety of personal and scientific experiences that arise from differences in culture and values. As can be seen (see below), the program committee of this conference as well as the program committee of all other tracks of the federated congress are as diverse as its authors and participants. The program committee would like to thank all those who submitted papers for consideration. About 65% of the submissions were from outside the United States. Each submitted paper was peer-reviewed by two experts in the field for originality, significance, clarity, impact, and soundness. In cases of contradictory recommendations, a member of the conference program committee was charged to make the final decision; often, this involved seeking help from additional referees. In addition, papers whose authors included a member of the conference program committee were evaluated using the double-blinded review process. One exception to the above evaluation process was for papers that were submitted directly to chairs/organizers of pre-approved sessions/workshops; in these cases, the chairs/organizers were responsible for the evaluation of such submissions. The overall paper acceptance rate for regular papers was 24%; 15% of the remaining papers were accepted as poster papers (at the time of this writing, we had not yet received the acceptance rate for a couple of individual tracks.) We are very grateful to the many colleagues who offered their services in organizing the conference. In particular, we would like to thank the members of Program Committee of HIMS’17, members of the congress Steering Committee, and members of the committees of federated congress tracks that have topics within the scope of HIMS. Many individuals listed below, will be requested after the conference to provide their expertise and services for selecting papers for publication (extended versions) in journal special issues as well as for publication in a set of research books (to be prepared for publishers including: Springer, Elsevier, BMC journals, and others). • • • • • •

Prof. Abbas M. Al-Bakry (Congress Steering Committee); University President, University of IT and Communications, Baghdad, Iraq Prof. Nizar Al-Holou (Congress Steering Committee); Professor and Chair, Electrical and Computer Engineering Department; Vice Chair, IEEE/SEM-Computer Chapter; University of Detroit Mercy, Detroit, Michigan, USA Prof. Hamid R. Arabnia (Congress Steering Committee); Graduate Program Director (PhD, MS, MAMS); The University of Georgia, USA; Editor-in-Chief, Journal of Supercomputing (Springer); Fellow, Center of Excellence in Terrorism, Resilience, Intelligence & Organized Crime Research (CENTRIC). Prof. Dr. Juan-Vicente Capella-Hernandez; Universitat Politecnica de Valencia (UPV), Department of Computer Engineering (DISCA), Valencia, Spain Prof. Juan Jose Martinez Castillo; Director, The Acantelys Alan Turing Nikola Tesla Research Group and GIPEB, Universidad Nacional Abierta, Venezuela Prof. Kevin Daimi (Congress Steering Committee); Director, Computer Science and Software Engineering Programs, Department of Mathematics, Computer Science and Software Engineering, University of Detroit Mercy, Detroit, Michigan, USA

• • • • • • •

• • • • • • • • • • • •

•

• • • •

Prof. Zhangisina Gulnur Davletzhanovna; Vice-rector of the Science, Central-Asian University, Kazakhstan, Almaty, Republic of Kazakhstan; Vice President of International Academy of Informatization, Kazskhstan, Almaty, Republic of Kazakhstan Prof. Leonidas Deligiannidis (Congress Steering Committee); Department of Computer Information Systems, Wentworth Institute of Technology, Boston, Massachusetts, USA; Visiting Professor, MIT, USA Prof. Mary Mehrnoosh Eshaghian-Wilner (Congress Steering Committee); Professor of Engineering Practice, University of Southern California, California, USA; Adjunct Professor, Electrical Engineering, University of California Los Angeles, Los Angeles (UCLA), California, USA Hindenburgo Elvas Goncalves de Sa; Robertshaw Controls (Multi-National Company), System Analyst, Brazil; Information Technology Coordinator and Manager, Brazil Prof. Byung-Gyu Kim (Congress Steering Committee); Multimedia Processing Communications Lab.(MPCL), Department of Computer Science and Engineering, College of Engineering, SunMoon University, South Korea Prof. Tai-hoon Kim; School of Information and Computing Science, University of Tasmania, Australia Prof. Louie Lolong Lacatan; Chairperson, Computer Engineerig Department, College of Engineering, Adamson University, Manila, Philippines; Senior Member, International Association of Computer Science and Information Technology (IACSIT), Singapore; Member, International Association of Online Engineering (IAOE), Austria Prof. Dr. Guoming Lai; Computer Science and Technology, Sun Yat-Sen University, Guangzhou, P. R. China Prof. Hyo Jong Lee; Director, Center for Advanced Image and Information Technology, Division of Computer Science and Engineering, Chonbuk National University, South Korea Dr. Muhammad Naufal Bin Mansor; Faculty of Engineering Technology, Department of Electrical, Universiti Malaysia Perlis (UniMAP), Perlis, Malaysia Dr. Andrew Marsh (Congress Steering Committee); CEO, HoIP Telecom Ltd (Healthcare over Internet Protocol), UK; Secretary General of World Academy of BioMedical Sciences and Technologies (WABT) a UNESCO NGO, The United Nations Michael B. O'Hara (Vice Chair and Editor, HIMS); CEO, KB Computing, LLC, USA; Certified Information System Security Professional (CISSP); Certified Cybersecurity Architect (CCSA); Certified HIPAA Professional (CHP); Certified Security Compliance Specialist (CSCS) Prof. Dr., Eng. Robert Ehimen Okonigene (Congress Steering Committee); Department of Electrical & Electronics Engineering, Faculty of Engineering and Technology, Ambrose Alli University, Nigeria Prof. James J. (Jong Hyuk) Park (Congress Steering Committee); Department of Computer Science and Engineering (DCSE), SeoulTech, Korea; President, FTRA, EiC, HCIS Springer, JoC, IJITCC; Head of DCSE, SeoulTech, Korea Dr. Akash Singh (Congress Steering Committee); IBM Corporation, Sacramento, California, USA; Chartered Scientist, Science Council, UK; Fellow, British Computer Society; Member, Senior IEEE, AACR, AAAS, and AAAI; IBM Corporation, USA Ashu M. G. Solo (Publicity), Fellow of British Computer Society, Principal/R&D Engineer, Maverick Technologies America Inc. Prof. Dr. Ir. Sim Kok Swee; Fellow, IEM; Senior Member, IEEE; Faculty of Engineering and Technology, Multimedia University, Melaka, Malaysia Prof. Fernando G. Tinetti (Congress Steering Committee); School of CS, Universidad Nacional de La Plata, La Plata, Argentina; Co-editor, Journal of Computer Science and Technology (JCS&T). Prof. Hahanov Vladimir (Congress Steering Committee); Vice Rector, and Dean of the Computer Engineering Faculty, Kharkov National University of Radio Electronics, Ukraine and Professor of Design Automation Department, Computer Engineering Faculty, Kharkov; IEEE Computer Society Golden Core Member; National University of Radio Electronics, Ukraine Prof. Shiuh-Jeng Wang (Congress Steering Committee); Director of Information Cryptology and Construction Laboratory (ICCL) and Director of Chinese Cryptology and Information Security Association (CCISA); Department of Information Management, Central Police University, Taoyuan, Taiwan; Guest Ed., IEEE Journal on Selected Areas in Communications. Dr. Yunlong Wang; Advanced Analytics at QuintilesIMS, Pennsylvania, USA Prof. Layne T. Watson (Congress Steering Committee); Fellow of IEEE; Fellow of The National Institute of Aerospace; Professor of Computer Science, Mathematics, and Aerospace and Ocean Engineering, Virginia Polytechnic Institute & State University, Blacksburg, Virginia, USA Prof. Jane You (Congress Steering Committee); Associate Head, Department of Computing, The Hong Kong Polytechnic University, Kowloon, Hong Kong Dr. Farhana H. Zulkernine; Coordinator of the Cognitive Science Program, School of Computing, Queen's University, Kingston, ON, Canada

We would like to extend our appreciation to the referees, the members of the program committees of individual sessions, tracks, and workshops; their names do not appear in this document; they are listed on the web sites of individual tracks. As Sponsors-at-large, partners, and/or organizers each of the followings (separated by semicolons) provided help for at least one track of the Congress: Computer Science Research, Education, and Applications Press (CSREA); US Chapter of World Academy of Science; American Council on Science & Education & Federated Research Council (http://www.americancse.org/); HoIP, Health Without Boundaries, Healthcare over Internet Protocol, UK (http://www.hoip.eu); HoIP Telecom, UK (http://www.hoip-telecom.co.uk); and WABT, Human Health Medicine, UNESCO NGOs, Paris, France (http://www.thewabt.com/ ). In addition, a number of university faculty members and their staff (names appear on the cover of the set of proceedings), several publishers of computer science and computer engineering books and journals, chapters and/or task forces of computer science associations/organizations from 3 regions, and developers of high-performance machines and systems provided significant help in organizing the conference as well as providing some resources. We are grateful to them all. We express our gratitude to keynote, invited, and individual conference/tracks and tutorial speakers - the list of speakers appears on the conference web site. We would also like to thank the followings: UCMSS (Universal Conference Management Systems & Support, California, USA) for managing all aspects of the conference; Dr. Tim Field of APC for coordinating and managing the printing of the proceedings; and the staff of Monte Carlo Resort (Convention department) at Las Vegas for the professional service they provided. Last but not least, we would like to thank the Co-Editors of HIMS’17: Hamid R. Arabnia, Leonidas Deligiannidis, and Michael B. O'Hara. We present the proceedings of HIMS’17.

Steering Committee, 2017 http://americancse.org/

Contents SESSION: MEDICAL SYSTEMS, DEVICES AND SERVICES + MONITORING SYSTEMS + TOOLS FOR REHABILITATION Integration of Haptic Rendering with Eye-Gaze Tracking using a Multi-Platform GUI Manuel Segura, Norali Pernalete, Amar Raheja

3

A Treatment Adherence Mobile Application Designed Specifically for Patients after Open-Heart Surgery Arshia Khan, Nam Phung, Janna Madden

10

HFinder: Anti-Emergency Medical Service Late Response Time System Thusoyaone Joseph Moemi, Bassey Isong, Boloka Jonathan

15

Counting Multiple People on a Floor Based Array Sensor System Fadi Muheidat, Harry Tyrer

21

Human Movements Monitoring Using Smartphone Sensors Hala Bin Saidan, Hmood Al-Dossari

28

Open System for Monitoring Vital Signs of Babies to Help in the Prevention and Diagnosis of Sudden Death Kleisson Tedesco, Robison Cris Brito, Eduardo Todt, Fabio Luiz Bertotti

32

Smartphones Application in Ophthalmology - Keratoconus Detection Rafael Seiji Ishibe, Maximiliam Luppe, Jean Jacques Groote

38

IoT Support for Dementia Patients Bernd Muller

41

Undergraduate Experience Developing a Medication Reminder App Suhair Amer, Steven Sebastian

44

Digital Filter Array Optimization for Directivity Pattern Soon Jarng, You Kwon, Dong Jarng

50

SESSION: HEALTH INFORMATICS, HEALTHCARE AND PUBLIC HEALTH RELATED SYSTEMS An HL7 v2 Platform for Standards Development and Testing Sandra Martinez Robles, Robert Snelick

57

Infrastructure for Health Care Simulation: Recommendations from the Model for Telecare Alarm Services Project Berglind Fjola Smaradottir, Rune Werner Fensli, Elin Sundby Boysen, Santiago Gil Martinez

64

Examining Locking Down of Electronic Medical Records Ryan Carr, Sudip Chakraborty, Leah Johnson, Jasmin Miravete, Joshua Vega

70

Targeted Data Swapping and K-Means Clustering for Healthcare Data Privacy and Usability Kato Mivule

77

A Study on the Regulatory Oversight of Direct-to-consumer Genetic Testing in USA Jiantao Ding, Eunjoo Huisung Pacifici

81

Mobile Rescue Management for Medical Emergencies: Introducing a Novel Application Framework Martin Zsifkovits, Stefania Madalina Toader, Stefan Wolfgang Pickl

85

What do College Undergraduates Know about Zika and What Precautions Are They Willing to 92 Take to Prevent its Spread? Michele Miller, William Romine, Megan Rua TB Portals Program Image Analysis: Can chest X-ray similarity identify drug resistant Tuberculosis patients? Octavio Juarez-Espinosa, Andrei Gabrieli , Eric Engle, Alex Rosenthal

99

SESSION: TOOLS FOR DECISION MAKING AND DIAGNOSTICS Prediction with Multiple, Multi-class Models and Dempster-Shafer Theory Michael Bauer

107

E-HandicapScale : An Open and Secure Way to Promote and Improve Diagnostics of Disabled Patients Laurent Bobelin, Andres Felipe Gil Salcedo, Christian Toinard, Marie-Elisabeth Labat

114

Machine Learning for Autism Diagnostics: Applying Support Vector Classification Florian Hauck, Natalia Kliewer

120

Predicting Cardiovascular Events: Sensitivity and Specificity Nishigandha Kale, Joshua Thomas, Rupesh Agrawal, Bruce Benjamin, Johnson P Thomas

124

Prediction of Dengue Cases in Paraguay using Artificial Neural Networks Victor Ughelli, Yohanna Lisnichuk, Julio Paciello, Juan Pane

130

SESSION: POSTER PAPERS Visualizing Healthcare Fraud Detection - Doctor Patient Network - Demo Javed Iqbal, Saba Riaz, Franklin Din

139

Visualizing the Effectiveness of Behavior Change in Combatting Fraudulent Healthcare Claims - Demo Javed Iqbal, Saba Riaz, Franklin Din

141

Intra-body Communication Using Ultrasonic Wave Propagation Meina Li, Youn Tae Kim

143

Detection of Epileptic Seizure Channel from EEG Signal Jinhyuck Kim, Jeongung Kim, Seungbum Shim, Mi-Sun Yum, Sunwoong Choi

145

Physical Fitness Assessment by Using Heart Rate and Physical Activity Sensor Meina Li, Youn Tae Kim

147

SESSION: LATE PAPERS - HEALTH INFORMATICS AND MEDICAL SYSTEMS 3D Heart Reconstruction using Thoracic Computer Tomography for Computational 151 Holography Applications Arthur Bucioli, B. Gerson Lima, Edgard Lamounier, Isabela Peres PERES, Alexandre Cardoso, Roberto Vieira Botelho Creating an Ontology for Family Diseases Prognosis Zahra Rabinia, Amin Rabinia

155

The Data Donation Pass: Enabling Sovereign Control of Personal Healthcare Data Matthieu-P. Schapranow, Janos Brauer, Hasso Plattner

159

The Use and Adoption of HIS and EMR in Botswana Lilybert Machacha, Moses Kanjadza, Babli Kumari

164

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

SESSION MEDICAL SYSTEMS, DEVICES AND SERVICES + MONITORING SYSTEMS + TOOLS FOR REHABILITATION Chair(s) TBA

ISBN: 1-60132-459-6, CSREA Press ©

1

2

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

3

Integration of Haptic Rendering with Eye-Gaze Tracking Using a Multi-Platform GUI M. Segura1 , N. Pernalete1 ,A. Raheja1, , 1 California State Polytechnic University, Pomona, California. U.S.A.

Abstract – This paper describes the extension of a haptic feedback software program for eye-hand coordination therapy described in [1]. The new system is capable of running on multiple computing platforms (Windows, Linux, Mac), and capable of working with multiple haptic devices. This involves creating a new and intuitive user interface using a GUI API library capable of running on multiple computing platforms, and using a haptic API that can also be run on multiple computing platforms. The haptic API is also able to support and control several different types of haptic devices. The haptic rendering takes place with a bounding box, while restraining the movements of the user in the real world. The force effects rendered by the inside of the bounding box include stiffness of the box, the damping of movement along the surface of the box, and dynamic friction while moving on the surface of the box. To guide the user along a chosen pattern a haptic tunnel calculated by using B-spline is used. Two types of haptic tunnels are presented and evaluated: one that follows the mid path of the pattern, and one that takes the smoothest path through the pattern. Finally, the Pearson coefficient was chosen as a metric to correlate the haptic device and the eye-gaze coordinates recorded simultaneously while the user traces a path. Keywords: Multi-Platform Haptics, Eye-Hand Coordination, Eye-gaze tracking.

1

Introduction

There are various methods for implementing the control of a haptic device for haptic rendering. Common comparable properties which can be found in technical specifications of haptic devices include: workspace specifying a maximal reach of a touch tool and maximal rotation abilities if appropriate, position resolution of a touch tool measured in dots per inch, maximal force specified in newton unit or as a force capability in kilograms or pounds and stiffness of a haptic device along a degree of freedom measured in Newtons per meter [2]. These methods can be categorized into layers ordered from the lowest level of control to the highest. The lowest layer in which a program can communicate with a haptic device is a device driver of the operating system [2]. At this layer the driver receives data through a serial bus (such as USB and IEEE 1394) from the device’s encoders. Kinematic algorithms are then used to transform the encoder readings to Cartesian coordinates. This layer also controls the initialization of the

device, opening and closing communications with the device, and performing inverse kinematic algorithms that convert a commanded force into the appropriate haptic angle joint movement [2]. Guidance and navigation control of mobile robotic agents has been an important research topic for several decades. A better solution to the guidance and navigation control problem takes into account dynamic constraints resulting in a smooth, flyable path that passes through the way-points [3]. The way-points must be connected such that the generated smooth paths preserve the continuity of the curvature between line and arc segments, while minimizing the maximum curvature [4]. One way of generating smooth paths through waypoints is with the usage of B-splines. The advantage of employing B-splines in generating a smooth path stems from the fact that the path can be represented using a smaller number of parameters than using a complete geometric description of the path. Accordingly, it is straightforward to deal with a relatively small number of parameters both for path optimization and for on-line implementation. The path planning problem using Bsplines involves finding the solution of a constrained optimization problem not only to avoid forbidden regions, but also to generate flyable trajectories [5]. A B-spline curve that remains within the channel is found by quadratic programming.

2

Multi-platform Software Libraries

The haptic feedback software program for hand-eye coordination therapy was written in C++ using the Qt 5 library. The Qt library is a multi-platform application framework that is widely used for developing application software that can be run on various software and hardware platforms with little or no change in the underlying codebase, while still functioning like a native application (The Qt Company 2016). The Qt library’s vast GUI creation and event handling capabilities allows for the creation of a GUI-based program that will be portable to multiple platforms, including Windows, Linux, and Mac based Operating Systems (OS), as shown in Figure 1. The Qt library is available with both commercial and open source GPL v3, LGPL v3 and LGPL v2 licenses (The Qt Company 2016). The application will also use SensseGraphics HAPI library as a low-level haptic API to control haptic devices, as graphic rendering will performed by Qt. To facilitate the use of multiple types of haptic devices, the application will use the

ISBN: 1-60132-459-6, CSREA Press ©

4

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

HAPI library’s AnyHapticDevice class, which will try to find a haptic device to use by trying to initialize all the devices that are supported; the haptic device chosen will be the first haptic device initialized successfully. The HAPI library is also a multi-platform application library that functions on multiple platforms (SenseGraphics AB. 2012). The application also performs some image processing in order to extract the boundaries of the path region of the pattern and other pattern details pertaining to the calculation of the haptic tunnel used to guide the haptic device’s avatar. Image processing is also required for scoring of a user’s recorded performance using the application. The OpenCV library is an open source library created and maintained by the Itseez computer vision research consultant company, and is used to provide the necessary image processing algorithm implementations for the application [6].

transform corresponding to the Viewport. The transform also accounts for the change in aspect ratio from normalized coordinates to the Viewport’s aspect ratio. See Figure 3 for an illustration of the Viewport transform mapping to the Viewport. Aspect ratio = 1:1

y

Aspect ratio = width:height

(1, 1)

y

(0, 0) x

(width-1, height-1)

Viewport transform

x

(0, 0) (-1, 1)

Default Clipping Area

Viewport

Figure 3: OpenGL Coordinates and Viewport Transform Both the trace pattern and the haptic device avatar’s tooltip are graphically rendered to the Viewport portion of the visible space, to produce a two-dimensional virtual environment. Since OpenGL coordinates are in a normal coordinate system, the coordinates of the haptic device avatar’s tooltip need to be normalized as well before being graphically rendered. We bound the avatar’s tooltip physical movement using a haptically rendered bounding box, which we can define in terms of its maximum corner and minimum corner in the haptic device’s coordinate system (see Figure 4). (1,1,1)

h max

Affine Transform

n

h

Figure 1. Haptic Feedback Application

3

Graphics and Haptic Rendering of the Virtual Environment

The patterns for tracing and the cursor used to display the user’s position on the trace are rendered using OpenGL 2.0, specifically OpenGL function wrappers provided by the Qt 5 library’s QOpenGLFunctions class and QOpenGLWidget class. To render to the screen using OpenGL, one must establish a Viewport, which is the start of the visible region of the virtual environment and represents the computer screen in the virtual environment (Figure 2). Camera/ Frustum Location

h min Haptic Rendered Bounding Box (in haptic device coordinates)

(-1,-1,-1) Haptic Rendered Bounding Box (normalized coordinates)

Figure 4. Haptic Bounding Box and Affine Transformation from Haptic to Normalized Coordinates Haptic rendering is done with the god object/proxy force response algorithm using HAPI library’s RuspiniRenderer class, which uses a sphere proxy. The haptically rendered bounding box is implemented using the HAPI library’s AABox class, and the surface in blue corresponds to the Viewport of the graphics rendering. Mapping a coordinate from this bounding box to a normalized coordinate is done with an affine transform

height

. near

far

Visible space

Figure 2: OpenGL Viewport in 3D Virtual Environment The visible region expands the further from the frustum where the camera is located. When drawing to this visible region, OpenGL uses normalized coordinates. OpenGL automatically transforms these normalized coordinates to the appropriate coordinates in the visible space by using a

See Figure 4 for the transformation. The affine transform is composed of scale transform and a translation transform as follows: . The scale transform can be defined as

ISBN: 1-60132-459-6, CSREA Press ©

,

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

where

5

abrupt turns in the movement through the pattern. The user is able to choose between these two B-spline approaches.

,

4.1

, . The translation transform is applied after the scale transform and can be defined as

B-Spline of Center Path

The general procedure for creating a B-spline from the white center line of the path region of the pattern is outlined as follows:

, where , , . Using the normalized coordinate , we use the derived to draw the avatar’s tooltip to the coordinate Viewport, where is the distance from the frustum to the Viewport. The normalized coordinate is also recorded when recording a user’s trace. To simulate inertial forces during haptic rendering, the bounding box is given a friction rendering surface, using the HAPI library’s FrictionSurface class. The FrictionSurface class allows the program to set stiffness, damping, static friction and dynamic friction parameters for the surface (SenseGraphics AB. 2012). The force produced is a linear spring-damper force between the proxy position and the avatar position, , where is the proxy position, is the avatar position, and is the avatar velocity. The friction parameters control how the proxy moves over the surface during contact. The static friction parameter controls how hard it is to start moving across the surface from a resting contact, while the dynamic friction parameter controls how hard it is to move across the surface when the proxy has started moving. The stiffness, damping, and static friction settings can be set through the GUI with values ranging from 0% - 100% of the maximum force output of the haptic device being used.

Figure 5. Multi-Platform Haptic System GUI 1. Extract white center path line points from pattern using color filtering implemented with the OpenCV library (Itseez 2016). [6] 2. Choose every 20th point of the center path line, save to center path polygonal chain points . 3. Fit B-spline control points to using least squares fitting. The data points to be fitted can be written as , where are uniform knots. Hence the least squares error function (based on the control points to be determined) between the prospective B-spline curve and the data points is the function

We solve for that minimizes . This is a nonlinear optimization problem. The application uses the open source nonlinear optimization solver library Nlopt [11] to solve this problem with the sequential quadratic programming (SQP) algorithm [7].

4.2

4

Haptic Tunnel Guidance

The application allows the user the option of using a haptically rendered tunnel to guide and restrain the haptic device’s avatar’s tooltip movement as the user moves from the start position of the pattern to the end position. To shape this haptic tunnel, a B-spline is calculated for the pattern and represents the ideal path from the start position of the pattern to the end position. Two approaches are taken to calculate a B-spline representing an ideal path. The first is a B-spline that is fitted to the white center line of the path region of the pattern. This path tends to be smooth but can have abruptly sharp turns if the pattern has abrupt turns. The second is a Bspline that is second derivative continuous, thereby smoothing

B-Spline of Smooth Movement Path

To create a B-spline with smooth curvature that is second derivative continuous, we define the MVB problem for the pattern as explained and applied by Berglund et al [8,9]. However, unlike Berglund et al ,we use two-dimensional control points rather than one-dimensional control points so that the prospective B-spline can yield a two dimensional point per knot (Jung et al [5]) . Then we solve the MVB problem to find the B-spline by choosing control points whose control polygons fit in a tight piecewise linear envelope within the path region of the pattern like in Jung et al [5]. The Bspline path would fit in the trace pattern as shown by the magenta line in Figure 6. Note that this B-spline path is actually graphically rendered to the Viewport.

ISBN: 1-60132-459-6, CSREA Press ©

6

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

which are the Greville abscissae. Then the piecewise linear functions that interpolate their values at each Greville abscissae are defined as follows

Figure 6. B-Spline Path for Trace Pattern (Magenta Line) 4.2.1 Tight Piecewise Linear Envelopes of B-spline curves The procedure followed is described in Jung et.al. [5]. To define a tight piecewise linear envelope for a B-spline, we must define the following concepts. The weighted second differences of control points are defined in terms of the Greville abscissae : , where is and If the dimension of we can express in terms of its components , where , where

,

and . Over the interval between two consecutive Greville abscissae (where ) the contribution of the th B-spline basis function to the distance between the B-spline and its control polygon is captured by the non-negative and convex functions

where and denote the indices of the first and last B-spline basis functions that are nonzero over the corresponding interval. Thus, the distance between the B-spline function and its control polygon is calculated as

If the dimension of

is

we have

If we define and we can bound the B-spline in each dimension as follows

,

where is the th component of , and is the th component of . Since the values are nonnegative and convex functions over the interval , the maximum function values occur at each end of the interval,

where denotes a linear interpolation between two values. The functions and then form the following tight piecewise linear bound around the th component of : If we have , then this bound would suffice as the tight piecewise linear envelope when . This is the form of tight piecewise envelope used by Berglund et al. (2003, 2010) as they used one-dimensional control points paired with corresponding knots to define their B-splines. However, if we , then this bound would form a two-dimensional have axis-aligned bounding box at each possible knot (including the Greville abscissae). Let denote the axis aligned bounding box at the Greville abscissae described by . Then the B-spline curve segment for lies in the convex combination of and the consecutive box because of the linearity of and . We can denote this convex combination as . Note that the convex hull over the knot consists of parts of the edges of and and exactly two extra line segments and chosen from the outermost possible line segments that would connect the corners of . The line segments and are separated by the and line that connects the control points and . Thus, denotes the denotes the left envelop line segment and right envelope line segment with respect to the control polygon for . These line segments and , where are joined together to form piecewise linear envelopes of the B-spline curves and , respectively. If for some reason these line segments don’t intersect, the line segments are extend to get an intersection point. For example in Figure 7, the line segments and are extended to get the intersection point , and is taken where and intersect. Both and can be characterized by these intersection points and , respectively, for ; we say that these are the feature points of the envelope. 4.2.2 Channel Constraints for Envelopes As presented in Jung et.al. [5], our goal is to find a Bspline that will fit between and that fit into the path region of the selected pattern, from the start position to the end position. To do this we need to constrain and to fit in the path region. We also assume that the path region can be described in terms of two non-intersecting

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

7

2. Define a cost function for curvature as defined in Jung et al (2008) [5]. In order to keep the the B-spline as close as possible to a straight line, for which the weighted second differences we use the cost function

polygonal chains and . To this end, we first let be the non-path region of the pattern (we will use discrete coordinates later). We define a distance-map function with respect to a polygonal line in order to provide the relative distance metric for formulating these geometric constraints, as follows: , where is the distance from a point to a corner point of the polygonal line , is the perpendicular distance from to the line segment that connects two consecutive corner points and , and is a sign value which dictates the location of the point with respect to as follows:

Using distance-map function choose envelope feature points and the following constraints:

This cost function implicitly minimizes the curvature variation of the B-spline, thus resulting in a smooth B-spline curve.

, we need to that will satisfy

. This is to ensure that each axis-aligned where bounding boxes at the Greville abscissae of the B-spline curve will be contained in the path region as well. In addition to this constraint we need to ensure that concave corner points of and of of the non-path region are placed outside the envelopes:

where is the number of concave corner points of , and is the number of concave corner points of . For example of these geometric constraints for the envelopes, see Figure 8.

Figure 7: Constructing the Bounding Envelopes from Neighboring Bounding Boxes. Source: [5] Jung and Tsiotras 4.2.3 General Procedure for Calculating B-spline for Smooth Movement The general procedure for finding a quartic (degree ) B-spline that fits between some envelopes and in the pattern path region is to manipulate control points that would fit within and that are constrained as outlined in Section 4.2.2. We also want to control the curvature of the resulting B-spline. This procedure is outlined as follows: 1. Set .

Figure 8: Geometric Constraints Formulation Source: Jung and Tsiotras 2008 [5]. 3. Extract exterior contour points from pattern color filtering implemented with the OpenCV library (Itseez 2016) [6]. 4. From contour points create 2 monotone polygonal chains and to define the MVB problem for the pattern Choose first and last points of contour points, for each chain. Choose every 5th point of contour points, for each chain. Choose every concave corner points of and of . 5. Choose uniform knot parameters for , where the first knot parameters are zero and the last knot parameters are one. 6. Set equal to the center of the starting point of the path region. 7. Set equal to the center of the ending point of the path region. 8. Choose that minimizes the cost function for curvature , and is subject to the constraints listed in Section 4.2.2. This is a nonlinear optimization problem with constraints. The application uses the open source nonlinear optimization library Nlopt (Johnson 2014) to solve this problem with the SQP algorithm (Kraft 1988) [7]. 9. The resulting control points can be used to derive the B-spline along with the chosen knots.

4.3

Rendering the Haptic Tunnel

In order to render a haptic tunnel, we create a vector field of forces pointing towards the rendered B-spline path. In turn, in order to render this vector field in a computationally efficient

ISBN: 1-60132-459-6, CSREA Press ©

8

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

manner, we discretize the B-spline path into small line segments by sampling 100 points on the B-spline path. These 100 B-spline path points need to be transformed from the Viewport coordinates to haptic coordinates using an affine transform. Then we use a custom subclass of HAPI library’s HAPIFunctionObject to loop through the line segments to find the point Q on each line segment that is closest the haptic avatar’s tooltip location , and the distance from to . We choose the line segment with the shortest distance , and its point to form the vector , as shown in Figure 9.

. As the eye gaze recording coordinates approach the haptic pattern trace coordinates, will approach the value 1; alternatively, if the value goes to zero and approaches -1, then the eye gaze recording coordinates are negatively correlated with the haptic pattern trace coordinates. Figure 10 shows the interface with the eye-gaze monitor.

P D

Q B

B A

D

Q

D

Q B A

A

Figure 9. Distance from Point to Line Segment of Discretized Curve The closest point on the line segment can be calculated as

where,

The , , and components of the vector are then output by the evaluate function and then used to define a vector field haptic force effect to draw the avatar’s tooltip towards the closest point on the B-spline path.

4.4

Figure 10. Multi-Platform Haptic System and MyGaze GUI

5

Analysis of Results

Ten users with not known eye-hand coordination problems performed the tests while the haptic effects were introduced and performance evaluation criteria previously used [1] was calculated as percentage of valid points within the path region and within the ideal path. The ideal path is predetermined as a center line with only 1-2 pixels wide (Figure 11) Two different ideal paths have been chosen: the one corresponding to the Bspline of the center path or the ideal path described by the Bspline of the smooth movement path.

Eye-Hand Number Metric for Haptic-Gaze Correlation

We record the eye gaze of the user using the MyGaze eye tracker device. We can establish a measure comparing the haptic pattern trace coordinates to the user’s eye gaze by measuring the correlation between the two. The Pearson correlation coefficient for the x-coordinates and the yare computed separately as follows: coordinates

where are the x-coordinates of the haptic pattern trace, are the y-coordinates of the of the haptic pattern trace, are the x-coordinates of the eye gaze recording, are the ycoordinates of the of the eye gaze recording, and , , , are the standard deviations of the given coordinate set ( . The Pearson correlation coefficient was chosen because it has been used widely in the sciences and is a measure of the strength of a linear association between two variables. Using and , we define a novel measurement of the hand-eye coordination of the user’s recorded trace as follows:

Pixel tolera nce radius

Ideal path point

Figure 11. Pixel Tolerance Added to Ideal Path Point. The sequence of increasing the haptic effect follows: 1. Friction 10%, center path tunnel 0% 2. Friction 30%, center path tunnel 0% 3. Friction 60%, center path tunnel 0% 4. Friction 10%, center path tunnel 30% 5. Friction 10%, center path tunnel 60% 6. Friction 10%, smooth movement tunnel 30% 7. Friction 10%, smooth movement tunnel 60% The final metric for performance evaluation is a weighted average of these two metrics

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

The value is shown by the application after the user finishes the trace and recording stops. Finally, users were chosen to perform a haptic pattern trace while using the MyGaze eye tracker. The hand-eye number metric for haptic-gaze correlation will be taken and it is expected that if is high, then will approach 1. Figure 12 shows the results for haptic-only pattern tracing. The results show that for the majority of users, higher friction force equated to higher scores. Also, the haptic tunnels equated to higher scores than raised friction force for the majority of users. However, the difference between the two types of haptic tunnels, center-path and smooth-movement-path, was minimal. The center-path haptic tunnel produced slightly higher scores on average than the smooth-path haptic tunnel, 95.961 to 93.028. It is possible that this distinction might change with a larger sample size, as it could be the users’ preference to follow lines rigidly versus smoothly for the users tested. Separately, five users performed haptic pattern traces while recording their eye-gaze data using the MyGaze eye tracker device. As seen in Table I, users’ haptic scores were above 90%, meaning they performed well during the pattern trace. Also, the Pearson coefficients of the X haptic-gaze coordinates and Y haptic-gaze coordinates of the user’s pattern traces are in the .8-.9 range, meaning that the correlation between the haptic pattern traces’ X and Y coordinates and the eye-gaze recordings’ X and Y coordinates is high. The haptic number we defined also expresses this conclusion; the higher the number the higher the correlation as it approaches 1.

6

Conclusions

This paper presents a preliminary design and the results of studies performed using a haptic device and eye-gaze data. A group of individuals tested the multi-platform haptic system with smooth tunnel assistance. The B-spline rendered haptic tunnels proved to be successful in improving accuracy of pattern traces. While the difference in performance between the center-path haptic tunnel and the smooth-movement-path haptic tunnel was not significant, it begins to establish the need for more testing and development using the haptic device for patients with eye-hand coordination difficulties. Finally a set of tests were performed while gathering data using an eye-gaze tracking device, and they also showed that the user’s eye-gaze was highly correlated to their performance on the haptic pattern trace, which helped define a novel haptic number to reflect this correlation.

9

Figure 12. Haptic Only Pattern Trace Scores for Ten Users TABLE I. HAPTIC-GAZE DATA RECORDING

7

References

[1] Pernalete, Tang, Chang, Cheng, Vetter, Stegemann, Grantner. "Development of an Evaluation Function for EyeHand Coordination Robotic Therapy". IEEE International Conference on Rehabilitation Robotics. Zurich, 2011. [2] Kadlecek, Petr. "Overview of current developments in haptic APIs ." Proceedings of CESCG, 2011. [3] Scheuer, A., and Ch. Laugier. "Planning sub-optimal and continuous-curvature paths for car-like robots." Proceedings of IEEE International Conference on Intelligent Robots and Systems, 1998: 25-31. [4] Kanayama, Y., and B.I. Hartman. "Smooth local path planning for autonomous vehicles." Proceedings of IEEE International Conference on Robotics and Automation 3 (1989): 1265-1270. [5] Jung, Dongwon, and Panagiotis Tsiotras. "On-line Generation for Small Unmanned Aerial Vehicles Using BSpline Path Templates." AIAA Guidance, Navigation and Control Conference, AIAA 7135 (2008). [6] Itseez. OpenCV Documentation. 2016. http://opencv.org/documentation.html. [7] Kraft, Dieter. "A software package for sequential quadratic programming." Technical Report DFVLR-FB 88-28 (Institut für Dynamik der Flugsysteme, Oberpfaffenhofen), July 1988 [8] Berglund, T., H. Jonsson, and I. Soderkvist. "An obstacleavoiding minimum variation B-spline problem." Proceedings of IEEE International Conference on Geometric Modeling and Graphics, 2003: 156-161. [9] Berglund, T., A. Brodnik, H. Jonsson, M. Staffanson, and I. Soderkvist. "Planning Smooth and Obstacle-Avoiding BSpline Paths for Autonomous Mining Vehicles." IEEE Transactions on Automation Science and Engineering 7, no. 1 (2010): 167-172.

ISBN: 1-60132-459-6, CSREA Press ©

10

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

A Treatment Adherence Mobile Application Designed Specifically for Patients after Open-Heart Surgery Arshia Khan PhD, Nam Phung, Janna Madden Department of Computer Science University of Minnesota, Duluth Duluth, MN, USA Abstract—Patients recovering from open heart surgery face many challenges upon returning home, primarily centered around the sudden increase in personal responsibility. Not only is this challenging because they are fatigued from the operation and decreased inpatient recovery time, but also the condition and treatments are all new to the patient and as a result, easily overwhelming. Medication adherence is a major concern and has shown to improve patient outcomes and reduction in costs. Nonadherence to medication regimen has been shown to cause an increase in cost of care, hospitalizations, and complications postsurgery. We are proposing a mobile application to address the challenges of adherence to a new treatment regimen, added responsibility and unfamiliarity of treatments, particularly in the area of medication management, cardiac rehabilitation monitoring and diet management. This application specifically designed for patients after open heart surgery also helps the patient adhere to the treatment regimen prescribed by the clinicians.

management. By providing a resource to patients upon their return home, it is also expected to decrease the feeling of abandonment common in recently discharged open heart surgery patients. Finally, this solution empowers patients to learn about their treatments but does not rush this learning. Numerous studies have shown that education is key to achieving medication regimen adherence. Our solution builds off of this research by providing patients access to information about their medications while still managing adherence to ease the transfer of responsibility to the patient. This mobile solution with features of medication management, physical therapy monitor and diet manager is expected to reduce complications after open heart surgery.

Keywords—open heart surgery recovery, myocardinal infraction, complication rates, medication adherence

Recovery time after an open heart surgery is about six to twelve weeks, of which only the first four to seven days of that recovery takes place in the hospital [2, 13]. Most patients are home less than a week after their surgery [2]. There are many factors that have been influential in shaping the current situation. The minimal length of hospital stay being seen in open heart surgery has been largely triggered due to high rates of hospital-acquired infections. Approximately 20% of open heart surgery patient contract an infection and for the most common hospital-acquired infection, pneumonia, hospital stay is typically increased by an average of 10.2 days and total costs by 29,692 dollars [8]. Because of this risk of hospital-acquired infections, patients are often discharged at the earliest possible time. In addition, as a hospital stay extends, costs quickly accumulate on an already extremely costly procedure. With an incident rate of approximately 2.75 incidents per 10,000 in the population each year, open heart surgery is both highly expensive as well as highly prevalent [8, 18]. These factors both impact the trend to-wards shorter inpatient recovery times for open heart surgery patients.

I.

INTRODUCTION

Patients who have had Open Heart Surgery following myocardial infarction face many challenges throughout their recovery. In the United States, approximately 735,000 people suffer a myocardial infarction (heart attack) every year; that’s equivalent to one heart attack every 43 seconds [1]. Of these, approximately 397,000 cases result in open heart surgery [19]. Additionally, it is typically recommended that patients maintain low levels of activity for four to six weeks and self monitor symptoms in the days and weeks that follow surgery [2]. These two recommendations turn out to be much harder than one would anticipate and puts excessive stress on the patients who are recovering from a major surgery. In addition to high patient expectations, six to twelve percent of patients encounter complications during their treatment or recovery [4, 5]. All of these factors play a role in the design of myocardial infarction treatments and interventions. The application developed records medication information, maintains record of all doses administered and a schedule of all upcoming medications, remind and walk through cardiac physical therapy, and provide recommendations and support in diet

II. BACKGROUND

However, this shift towards shorter inpatient recovery time is often challenging for patients. The first concern patients have is the lack of connection and perceived abandonment from the healthcare system upon their return home [10, 11, 12]. While in intensive care, patients’ medications are being administered

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | and managed by medical professionals. However upon leaving the hospital, this all changes; patients are responsible for the daunting task of managing all their medications. Adhering to this new regimen of prescribed medications becomes one of the greatest challenges for recovering patients [10, 14, 17]. In an analysis of elderly patients enrolled in government health programs, only 10% of patients filled prescriptions enough to have their medication daily and even among patients enrolled in no cost sharing prescription insurance plans, non-adherence rates was approximately 40 percent [14, 15]. In addition, patients recovering at home were less likely to comply with medication regimens than those in hospital or nursing facilities. It was also seen that changes in medication dosage after leaving the hospital or nursing facility were more commonly administered incorrectly compared to prescriptions that started during inpatient care [14, 15, 16]. After a major procedure, such as open heart surgery, many assume patients meticulously comply with medications. However, irrespective of diagnosis and prognosis, upwards of one third of all patients struggle to adhere to their pharmaceutical regimen [14]. It’s often not for a lack of trying; rather the expectation that patients manage their medications in light of the newness and complexity of the condition, is asking a lot of recovering patients. However, this statistic is unnerving as medication adherence is crucial for maintenance of physical functioning, and avoidance of future events; failure to correctly administer medications leads to preventable and often, costly, complications in patients of open heart surgery. In addition, the patient is prescribed to a set of physical therapy to help improve their physical activity and a strict diet management regimen. This combination of factors surrounding treatment management results in a less than ideal situation for patient recovery. Medication adherence alone has shown to reduce costs from 10.1% to 17.8%[20]. Coronary artery disease being the most common cause of death accrues healthcare costs around $475 billion. Treatment adherence can be complicated in terms of schedule, timing, dosage amount, and other recommendations. In a study only 50% of the patients adhered to prescribed medication even when their risk of acquiring coronary disease grew many fold [23]. It is critical to device mechanisms that would not only reduce healthcare costs but also help reduce deaths. The application developed records medication information, maintains record of all doses administered and a schedule of all upcoming medications, remind and walk through cardiac physical therapy, and provide recommendations and support in diet management. By providing a resource to patients upon their return home, it is also expected to decrease the feeling of abandonment common in recently discharged open heart surgery patients. Finally, this solution empowers patients to learn about their treatments but does not rush this learning. Numerous studies have shown that education is key to achieving medication regimen adherence. Our solution builds off of this research by providing patients access to information about their medications while still managing adherence to ease the transfer of responsibility to the patient. This mobile solution with features of medication management, physical therapy monitor and diet manager is expected to reduce complications after open heart surgery.

11

Fig. 2. Setting menu

III. EXISTING TECHNOLOGIES There are several medication management applications available on the market to address the problem of medication management to improve the wellbeing and health of individuals with chronic illnesses. Our mobile app is designed for the iOS and the android platforms. This application not only addresses the medication adherence but also helps follow the treatment regime that include physical therapy, social interaction and diet management. Most applications that address medication management are a general medication support application. Although there are several mobile applications for medication management none are designed specifically for use after open heart surgery [21, 22, 23]. A study that examined some of the popular medication management applications observed that the subjects were mostly content with their existing medication management system and did not feel the need for a mobile application. These subjects were frustrated with their initial interactions with the applications until they got comfortable using it. Noncompliance to medication can be attributed to social and economic factors, health system and health care team factors, therapy related factors, condition related factors and patient related factors [23]. Some of the current existing technologies are electronic pillboxes, mobile management systems, prescription reminder apps and medicine trackers. None of these offered the complete treatment adherence solution that not only addressed medication adherence but also other recommendations for physical activity, diet management, and vital monitoring. The feature of diet management will also support the adherence of dietary restrictions imposed after open heart surgery by providing diet advice and making recommendations for heart healthy meals.

ISBN: 1-60132-459-6, CSREA Press ©

12

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | 11. Recipe preparation based on the diet restrictions provided by the clinicians IV. PROPOSED SOLUTION

To address the challenge of treatment adherence after open heart surgery, we suggest the use of virtual mobile support during the post-hospital recovery phase of open heart surgery to address the pharmacological, emotional and social challenges being faced by patients during recovery. This tool will guide patients’ through medication administration, provide education on medications, such as interactions or side effects, and maintain records of medication history to achieve strong medication compliance in addition to reminding patient about physical activity, diet watch and physiological vital monitoring. The features of this application are as follows: 1.

Reminding patient of medication

2.

Displaying a picture of the medication

3.

Providing information on dosage

4.

Reminding of any dietary restrictions with the specific medication

5.

Recording of the medications taken along with the dosage and schedule

6.

Educational information on medication adherence

7.

Providing any information on side effects of the medication

8.

Physical activity tracker and monitor

9.

Demonstration of how the physical activity/therapy has to be performed

10. Diet management

Fig. 1. Medication information

These features can be categorized as follows: 1.

Medication management

2.

Physical therapy monitor

3.

Diet manager

Medication Management: Patients would ideally input their medication into the system before leaving the hospital, as seen in fig. 1. Each medication includes an image of the medication for verification, the dosage schedule and medication instructions which includes instructions on administering the medication, any recommendations made by the doctor or pharmacist, food restrictions related to the drug and any possible side effects caused by the drug. In addition, start and end dates, if applicable, are established for each drug. All imputed drugs form the “medication list”, the first option in the settings menu in fig. 2. Below that is medication history, which serves as a record of all doses administered. Finally, the patient profile serves as a means to save the patient information. Once the set up has been complete, the application opens to the default home page which shows the list of scheduled drugs for the day and the times each needs to be administered, seen in fig. 3. By selecting a particular dose, the patient can see the medication information, picture of drug, scheduled time as well as being able to mark a dose as complete, thus adding it to the medication history. This software also re-minds the patient of dosages as the scheduled time approaches if the dosage has not been marked complete. This reminder includes not only the medication name but also the image to improve compliance and insure that the correct medication is administered.

Fig. 2. Today’s medication schedule (home page)

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | Physical therapy monitor/tracker: This feature is designed to assist the physical therapist in help the patient adhere the physical activty regimen prescribed by the cardiac physical therapist. The patient has certain physical activity restrictions after an open heart surgery while at teh same time has to strengthen her/his physical condition to be able to get back to their daily routine life. The physical therapist will be able to prescribe in this mobile app a set of exercies to be performed and the frequency of these exercises. The app will also have a demonstration of these activities. Diet Manager This feature offers the patient recommendations to manage their diet by avoiding foods that are restricted by the clinicians and encourages the patient to consume foods that are beneficial to a patient with coronary artery disease. This app feature also makes recommendations and recipes for the patient to follow a healthy diet regimen. Because of the linkage between pharmaceutical effectiveness and food consumption [17], this feature will be valuable in insuring that the effectiveness of pharmaceuticals is maximized. This feature will also support the adherence of dietary restrictions imposed after open heart surgery by providing diet advice and making recommendations for heart healthy meals. Platform The app is built on the iOS platform. The apps unique features that are designed sepcifically to follow and adhere to a prescribed treatment regimen for patients after open heart surgery. V. DISCUSSION An element of effective cardiac rehabilitation interventions that stands out is the use of education in improving treatment adherence. A meta-analysis of cardiac patient interventions evaluating the role of education found that incorporating education into recovery empathized the importance of and in turn, increased adherence to treatments [16]. Another benefit of education that was described was empowerment. As previously mentioned briefly, patients recovering from open heart surgery often feel neglected upon returning home. The support and understanding gained through this application is believed to lessen the feelings of distress and abandonment and empower patients to confidently manage their recovery. The proposed intervention shows promise in these areas by balancing the need of patient education while removing the burdensome task of managing medications, physical therapy, and diet with the belief that this combination of accessibility and understanding will increase pharmaceutical adherence. This idea of accessibility is built on the idea that while there is a need for education, the amount of changes in the patients’ life means that imposing too much responsibility could easily become overwhelming and result in non-adherence simply due to the complexity and newness of the scenario. Many solutions have been proposed individually to address issues of medication adherence, diet management and physical therapy monitoring. These solutions are mostly very generic and are not specific to a disease or if they are specific they only offer one of the medication adherence, diet management and

13

physical therapy monitoring. An all encompassing solution that addresses three main elements such as medication adherence, diet management and physical therapy monitoring is critical to help support patients after open heart surgery. Our proposal addresses all of these features. VI. CONCLUSION Recovery from open heart surgery poses many challenges for patients. One of the overarching challenges is the shortness of the inpatient recovery time and lack of assistance upon returning home. Especially in the area of pharmaceutical management, patients’ struggles can lead to preventable complications. This solution manages medications to ensure that they are taken correctly and to avoid putting additional responsibility on patients, helps remind and walk through the cardiac physical rehabilitation activities and diet restrictions. In doing so, patients’ are able to learn how to manage their condition without all the responsibility being instantly put on them when they return home. Future work, currently in progress, aims to extend the services provided to include additional features that would encourage recovery after open heart surgery. Features that encourage and support emotional wellbeing are important for the complete recovery after open heart surgery. In conclusion by providing support and education to patients as they adapt to lifestyle changes preventable complications can be avoided. In light of shortened hospital stays alter-nates to inpatient education need to be explored. Not only do such alternatives limit the amount of education needed to be provided during the patients’ inpatient stay, it also provides a more gradual transition to independent life. Providing such a resource for patients upon their return home is expected to decrease feelings of abandonment and most importantly, gives patients the resources they need. A multi feature mobile application that helps recovery in the areas of medication management, cardiac rehabilitation and diet restriction management will improve recovery time, reduce complication and help reduce healthcare costs.

REFERENCES [1] [2] [3] [4]

[5]

[6]

"Heart Attack Facts & Statistics | Cdc.Gov". Cdc.gov. N. p., 2016. Web. 25 May 2016. "Overview - Coronary Bypass Surgery - Mayo Clinic". Mayoclinic.org. N. p., 2016. Web. 25 May 2016. Bradley, Claire. "Top 10 Most Expensive Medical Procedures | Investopedia". Investopedia. N. p., 2010. Web. 25 May 2016. Brown, Phillip P., et al. "The frequency and cost of complications associated with coronary artery bypass grafting surgery: results from the United States Medicare program." The Annals of thoracic surgery 85.6 (2008): 1980-1986. Dimick, Justin B., et al. "Hospital costs associated with surgical complications: a report from the private-sector National Surgical Quality Improvement Program." Journal of the American College of Surgeons 199.4 (2004): 531-537. Mocanu, Valentin, et al. "The importance of continued quality improvement efforts in monitoring hospital-acquired infection rates: A cardiac surgery experience." The Annals of thoracic surgery 99.6 (2015): 2061-2069.

ISBN: 1-60132-459-6, CSREA Press ©

14 [7]

[8]

[9] [10]

[11]

[12]

[13]

[14]

[15]

[16]

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | Jeon, Christie Y., et al. "On the role of length of stay in healthcareassociated bloodstream infection." Infection Control & Hospital Epidemiology 33.12 (2012): 1213-1218. Iribarne, Alexander, et al. "Quantifying the incremental cost of complications associated with mitral valve surgery in the United States." The Journal of thoracic and cardiovascular surgery 143.4 (2012): 864872. Civetta, Joseph M. "The inverse relationship between cost and survival."Journal of Surgical Research 14.3 (1973): 265-269. Karlsson, Ann-Kristin, Evy Lidell, and Mats Johansson. "Depressed mood over time after open heart surgery impacts patient well-being: a combined study."European Journal of Cardiovascular Nursing 7.4 (2008): 277-283. Mallik S, Krumholz HM, Zhen QL, Kasl SV, Mattera JA, Roumains SA, et al. Patients with depressive symptoms have lower health status benefits after coronary artery bypass surgery. Circulation 2005;111: 271–7. [7] Doering LV, Moser DK, Lemankiewicz W, Luper C, Khan S. Depression, healing and recovery from coronary artery bypass surgery. Am J Crit Care 2005;14(4):316–24 Nilsson, Johan, et al. "EuroSCORE predicts intensive care unit stay and costs of open heart surgery." The Annals of thoracic surgery 78.5 (2004): 1528-1534. Urquhart, J. "Patient non-compliance with drug regimens: measurement, clinical correlates, economic impact." European heart journal 17.suppl A (1996): 8-15. Cutler, David M., and Wendy Everett. "Thinking outside the pillbox— medication adherence as a priority for health care reform." New England Journal of Medicine 362.17 (2010): 1553-1555. Mullen, Patricia Dolan, Douglas A. Mains, and Ramon Velez. "A metaanalysis of controlled trials of cardiac patient education." Patient education and counseling 19.2 (1992): 143-162.

[17] Colonna, Paolo, et al. "Nonpharmacologic care of heart failure: counseling, dietary restriction, rehabilitation, treatment of sleep apnea, and ultrafiltration."The American journal of cardiology 91.9 (2003): 4150. [18] Etzioni, David A., and Vaughn A. Starnes. "The epidemiology and economics of cardiothoracic surgery in the elderly." Cardiothoracic surgery in the elderly. Springer New York, 2011. 5-24. [19] Mozaffarian. "Heart Disease and Stroke Statistics-2015 Update: A Report From the American Heart Association (vol 131, pg e29, 2015)." Circulation131.24 (2015): E535-E535. [20] Bitton, Asaf, et al. "The impact of medication adherence on coronary artery disease costs and outcomes: a systematic review." The American journal of medicine 126.4 (2013): 357-e7 [21] Chomutare T, Fernandez-Luque L, Årsand E, Hartvigsen G; Features of Mobile Diabetes Applications: Review of the Literature and Analysis of Current Applications Compared Against Evidence-Based Guidelines J Med Internet Res 2011;13(3):e65; DOI: 10.2196/jmir.1874; PMID: 21979293; PMCID: PMC3222161 [22] Grindrod KA, Li M, Gates A; Evaluating User Perceptions of Mobile Medication Management Applications With Older Adults: A Usability Study; JMIR Mhealth Uhealth 2014;2(1):e11 DOI: 10.2196/mhealth.3048 PMID: 25099993 PMCID: 4114457 [23] Silva, Juan M., Alain Mouttham, and Abdulmotaleb El Saddik. "UbiMeds: a mobile application to improve accessibility and support medication adherence." Proceedings of the 1st ACM SIGMM international workshop on Media studies and implementations that help improving access to disabled users. ACM, 2009.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

15

HFinder: Anti-Emergency Medical Service Late Response Time System Thusoyaone Joseph Moemi, Bassey Isong and Boloka Jonathan Computer Science Department, North-West University Mafikeng, South Africa {joseph.moemi, bassey.isong}@nwu.ac.za, [email protected] Abstract— In a life-threatening situation, the prompt access to Emergency Medical Service (EMS) and the ambulance response time are critical factors to the patient’s survival. However, accessing EMS constitute a great challenge during emergencies despite several solution methods that have been developed and proposed. In this paper, we developed a cost-effective approach to saving lives instead of calling and waiting endlessly for EMS or an ambulance arrival. We proposed a smart-hospital where information about hospital services, doctors on duty, availability of ambulances, and location are collected in real-time and stored in a cloud-based database. These information can then be accessed by people during emergency to assist them take important decisions that can save lives. The system utilized the technologies of Radio Frequency Identification (RFID), cloud-based databases and mobile application. Additionally, the paper developed a mobile application prototype called Hospital Finder (HFinder) and the operations/functionalities offered presented. Given its operation, we believe that if this application is adopted for use, it could go a long way to save many lives during life-threatening situation instead of searching, calling and waiting for an EMS for a long period of time. Keywords— Ambulance, Patients, Hospitals, EMS, RFID, Realtime.

I. INTRODUCTION Information and communication technologies (ICTs) have created tremendous impacts in education, finance, communication, transportation, heath and so on. In particular, ICTs in the health sector have been commendable and have played a critical role in managing and coordinating healthcare services in and out of hospitals [1]. In spite of these benefits, some challenges still exist which include the challenge of establishing affordable, accessible and high quality Emergency Medical Service (EMS) which are essential in life-threatening situations and rapid decision making [4]. EMS can be considered as a dedicated and exceptional type of smart transportation system otherwise called ambulance service. They basically depend on key critical and limited medical resources, assets, and existing transportation infrastructure [2][3][5]. EMS plays a significant role in people’s lives as well as mortality and morbidity rate reduction [2][3][5][6]. It is used for on-site medical treatment, patients’ transportation and initial safety measure during emergency situation [5][7].

Several EMS exist today and their effectiveness is critical to patients’ survival in a life-threatening condition [5]. During emergency situation EMS providers are called to request an ambulance to provide services to patients. In this case, the ambulance response time constitute a critical factor to the patient’s survival [5]. In particular, response time is decisive of mortality for patients with urgent needs such as fatal auto crash in the middle of the road. Nevertheless, EMS providers are faced with a huge challenge of deciding on which ambulance to dispatch or send to incoming incidents [12]. The problem is centered on how to minimize late arrival time than a target time. This issue has generated enormous dissatisfaction in the society and in some cease, the ambulances never arrived. EMS is very important and constitute a pre-hospital service in both the rural and urban areas, and thus, their performance has to be enhanced. This is because, in severe life-threatening situation where every second is precious, the timely arrival of emergency service can distinguish between life and death [5]. Several factors have been identified as the causes of ambulance response time such in increased human populations, long distance travel and high traffic densities [7][13]. These have all contributed to emergency vehicles finding it cumbersome to reach accident scenes at the quickest time. Today, many solutions on EMS have been offered dominated by operation research studies [6]. They all aimed at designing comprehensive EMS systems that can provide early or timely response to emergency calls [6][12] – a situation that calls for EMS locational decision making. Moreover, given the existing solutions, the link between them and the optimum solution is still much unknown, and this paper attempts to bridge this gap. The question addressed in this paper is: during emergency situations for example, a fatal auto crash in the “middle of nowhere” and ambulance services are not accessible or the response time is taking forever against optimal dispatch policies. Suppose we know in advance and in real time, the location of hospitals, availability of their services, doctors and ambulances, how much better would the performance be in saving life? Therefore, in this paper, we proposed an approach that is independent on any model. It is basically an anti-ambulance late response time which enables people involved in emergency to take early decisive action towards saving lives instead of calling

ISBN: 1-60132-459-6, CSREA Press ©

16

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

and waiting endlessly for an EMS to arrive. The objective of this paper is thus, to improve the effectiveness of pre-hospital services through quick and early planning and decision making. The proposed approach involved developing smart-hospitals which are connected and provide important information about each hospital in real-time. The information would in turn assist people to take appropriate decisions during life-threatening situation without having to wait for a long time, the arrival of medical or ambulance services. With the availability of alternative transportation, hospital of choice with required services can be located and patients taken for treatment. Additionally, our approach uses technologies of radio frequency identification (RFID), cloud computing and implemented a mobile application prototype called Hospital Finder (HFinder). RFID is a technology that is used to wirelessly and automatically identify and track objects via radio frequency electromagnetic field [14]. Its components are the RFID tag which stores the object information, the reader which captured tag information and the computer system used for information processing. RFID is a short range communication technology which is effective in the real time identification, tracking and monitoring of objects. It has widespread application especially in hospitals to perform different tasks such as the tracking of assets, drugs, location and so on [15][16]. In this paper, we utilized RFID to track the availability of ambulances and doctors on duty in the hospital in real-time which is stored in the cloud. These essential information are then accessed by the general public during emergencies using the mobile application on their phones or devices. We developed a system prototype called Hospital Finder (HFinder) that will be used by peopleWe believe that having this information on the cloud for each hospital could go a long way to save lives, costs and time. The rest of the paper is organized as follows: Section II is the related works, Section III presents the proposed system, Section IV discusses the HFinder system prototype and Section V is the paper conclusion. II. RELATED WORKS Several works have been carried out published in many articles aimed at improving the effectiveness of EMS in terms of ambulance location and allocation. These are geared towards ensuring prompt response to emergency calls, and so on [5][12]. Some of the works are discussed. Recently, Aringhieri et al. [6] performed a comprehensive literature review on present health care systems with a focus on emergency pathway. The study identified several challenges (e.g. location decisions) which affects EMS, variety of proposed models, solution methods, real-world case studies and possible research directions. Leknes et al. [10] developed a mixed integer programming model for high quality planning in EMS specifically for regions which are considered to have heterogeneous demand and multiple performance measures. The approach specifically decides on the location, the stations allocation, ambulances and the computation of the service and arrival rates for each station. In addition, it computes the

probability that a call is attended to by a certain station. The model was tested in Norway rural and urban areas. In another study, Saydam and Aytug [8] developed an approach that integrated hypercube queue system into a genetic algorithm for resolving the Maximum Expected Covering Location Problem (MEXCLP). MEXCLP was ideal for an outcome that is expected in nature other than deterministic. The solution computes the probability for ambulance availability in each station in each iteration and for finding new solutions. This solution is also used by Kepatsoglou and Karlaftis [9]. Jagtenberg at al. [12] also proposed a benchmark model which is also an offline model to compute the optimal dispatch decisions of ambulance given that all incidents are known beforehand. The model used three different techniques to calculate the optimal dispatch policy for problems involving a limited number of events. Moreover, Isong et al. [17] proposed a mobile-based approach to deal with issues of EMS. They developed a system prototype using mobile application technologies to provide emergency services to patients with the aim of lessening long queue at the hospitals and the waiting periods for an ambulance arrival which is sometimes considered endless. Utkua et al. [7] developed a solution which maintain a centralized databased where dentification and tracking technologies were used. This involve using RFID and Bluetooth low energy to identify and track each object in an ambulance. The objectives were to decrease the manual documentation and to enable medical personnel to concentrate on medical activities especially during emergencies. Pham et al. [11] proposed a cloud-based smart-parking system to assist users automatically find free parking space at a least cost. The approach applied Internet of Things (IoT) technology and use an algorithm to compute users’ parking cost via the distance and the total number of free spaces in each connected car park. The approach was found to minimize users waiting time. In the above discussed related works, the authors proposed different ways of minimizing EMS response time during emergencies. Despite these solutions, accessing and the prompt response of EMS still constitute a serious challenge today. Consequently, our approach is geared to solve this impeding issue via a cost-effective alternative means. While [11] is not related to ambulance issues, this paper adopt its idea to provide emergency solution during life-threatening situation. III. THE PROPOSED SYSTEM The proposed system is based on the idea of IoT using RFID [14] in particular, cloud computing technologies and the work by [11]. In this proposed system, RFID is used in the collection of data from the hospital about the availability of doctors, ambulances, and while the technologies of the cloud will used for storage. The proposed system is geared towards overcoming emergency situation challenges such as endless waiting for an ambulance, finding available nearest hospital, searching for doctors and services offered by certain hospital. It also assist in finding optimal direction or route to the hospitals. The proposed system is a free and on-line application

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | requiring simple registration before access is allowed. That is, users have to login for authentication basis. Due to the emergency-orientation, users or clients can only perform the tasks of view services, obtaining optimal direction, ambulances and doctor’s availability, and details of the hospital. These tasks are managed by an administrator in addition to adding hospital information to the database or perform system and database updates. Our proposed system will assist users during emergencies by accessing important information about hospitals, their proximity, services and so on which could save lives.

17 •

Local computer: This will be used to store, display and update hospital information and services in the cloud server databases. 2) Cloud-based database: A server hosted in the cloud which will be used to store and manage data collected in realtime from each participating hospital. The information will be collected by the RFID reader and transmitted in a real-time basis through the network to the cloud-based server. 3) Application software: This will be installed on the user’s computers either on desktop or smartphones which are androidoriented. On desktop, the application will be used in the hospital to either add, delete or update existing hospital information in the database. On clients’ smartphones or tablets, the application will be used to access hospitals information whenever or wherever needed. 4) Network: The prefer network is the wireless network with either 3G/4G/5G capability or Ethernet connection for the administrator and user site. Hospital info Total number of doctors/ambulance on duty Total number of Doctors/Ambul available Services available Location

Arduino Module

Display Screen RFID READER Availability = +1

RFID READER Availability = -1

Fig. 1. System architecture Exit

Entrance

A. System Architecture The architecture of the system shows the structure and the interacting components. These are the server or cloud-based database, the RFID system, the network and client application. The system architecture is shown in Fig. 1. 1) System Unit: This system unit is located in each participating hospitals. It collects and stores information of each doctor and ambulance on duty in the hospital in real-time. This is shown in Fig. 2. •

•

RFID tag or RFID-based card: The RFID tag is embedded on each ambulance with a unique code, and the RFIDbased card is given to each doctor with a tag storing a unique identity code. The information stored in each tag is used to uniquely identify each object in real-time whenever they enter and exit the hospital. Controller: This is basically an Arduino module which is directly connected to the RFID reader, the computer and the cloud server. The RFID reader at the entrance/exit scan entities information and transmit to the Arduino module which is displayed on the computer screen as well as stored in the cloud database. In this case, if there is an entrance or exit as read from the RFID tag, the module will collect and transfer data from each hospital to the cloud database.

Doctor/Nurse/Ambulance (RFID Tag-based Card)

Doctor/Nurse/Ambulance (RFID Tag-based Card)

Fig. 2. RFID-based hospital entrance/exit

B. Algorithm and System Operations In order to ensure the effective operation of the proposed system, we recommend a smart hospital complex which is RFID-based for all the participating hospitals as shown in Fig. 2. However, individual hospital is responsible for updating its information in the cloud server database. Our proposed system involve two main processes: information capturing and information accessing. 1) Information Capturing Process: As shown in Fig. 2, the proposed system collects data in real-time mode using the RFID system. This is achieved by having sophisticated RFID readers mounted at the entrances and exits of the hospital. They are placed to monitor the entering of the entities into the hospital complex while the other will monitor their exit. These RFID readers are connected to an Arduino board in the hospital and are responsible for reading information from the RFID tag and thereby updating ambulance and doctor information in the hospital in real-time in the cloud-based database through the wireless access or Ethernet connection.

ISBN: 1-60132-459-6, CSREA Press ©

18 •

•

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | Operation at the Entrance: When an ambulance or a doctor on duties enters the hospital, the RFID reader will capture the tag information automatically and transmit via the controller to the cloud-based server for update purposes. In this case, it updates the availability of an entity by +1. That is, Availability +=1. This is captured in Fig. 3. Operation at the Exit: Here, when an ambulance or a doctor on duties leaves the hospital, the RFID reader will capture the tag information automatically and transmit via the controller to the cloud-based server for update purposes. In this case, it updates the availability of an entity by -1. That is, Availability -=1.

are nearby, having required services, showing availability of ambulances, doctors and the distance from the point of incident to the hospital. The significance of this application is to ensure that life is saved during life-threatening situation. Instead of waiting for EMS for a long period of time, or driving long distances to get to the hospital, early decision can be made to use alternative means to transport the patient to a nearby hospital. The process involve is simple and straight forward which requires the user to first turn-on the GPS on his/her phone followed by opening the application. The process of information access is captured by the possible interaction sequence shown in Fig. 5. Once the application is opened, the user will logon to the system where the home page is displayed and automatically show the list of available nearby hospital names and their distance from the current location. At this point, the user can click on any hospital of choice to display their information, given the services and availability of doctors for emergency and ambulance as well as the direction to the hospital using Google map.

Fig. 3. Algorithm for data capture and update process

The algorithmic design is shown in Fig. 3. For instance, if the number of doctors and ambulances on duties are 10 and 6 respectively. Accordingly, if two doctors and three ambulances exit the hospital, the database will be updated as shown in Fig. 4 Fig. 5. Sequence of interaction with the system

IV. HOSPITAL FINDER APPLICATION PROTOTYPE

Fig. 4. Hospital information on cloud server

We assume both ambulance and paramedics are available together. Moreover, to avoid conflicts such as malfunctioned ambulances and doctors not on duty, the administrator will manually disable them. 2) Information access process: This is the core aspect of this paper. It involves using the system installed on smartphones during emergency situation to find hospitals that

In this paper, the main goal is proposed a cloud-based medical emergency assist system will assist people during emergency to make better decisions on which hospital and what mode of transport. Though the project is still on-going, we have developed an application prototype called Hospital Finder (HFinder) installable on smartphones and tablets. However, due to time and financial constraints, we were unable to integrate the hardware parts to the software to automatically collect data in real-time manner. We only provided a theoretical design of the system. The system was developed with Android and used Google cloud-based database to store hospitals and user information in the cloud severs. In this case, the user has to download the

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

19

application and install on the phone which have both GPS Internet access. Once installed, HFinder icon will appear on the phone as shown in Fig.6.

Fig. 9. Doctors/ambulance availability Fig. 6. HFinder icon

To use HFinder application during emergency, the user will have to click on the icon to perform authentication. After successful authentication, the page shown in Fig. 7 will be displayed. It will automatically display the list of all connected hospitals and their location/distance from the current location of incident.

Fig. 7. HFinder home

On the other hand, Fig. 9 shows the available doctors and total number of ambulance available. It also show service offered by the doctor (e.g. DR Test RT offers Test occupation Health). For doctor’s status, the cross sign indicates not in hospital premises. Once a choice is made for a given hospital, the system will display the address of the hospital, the route, the route options as well as the estimated time to get to selected hospital. This is shown in Fig. 10.

Fig. 10. Route options

Once on this page, the user can click on any hospital of choice to view the services offered and availability of doctors for emergency as shown in Fig. 8 and 9 respectively. For example, Fig. 8 shows the services offered by Steve Biko Hospital in Gauteng, South Africa.

In addition, by using the Google map, direction on how to get to the hospital is offered. For instance, a user at Soweto will get to the hospital at Kempton Park using R553 route.

Fig. 11. Direction to hospital Fig. 8. Hospital services

ISBN: 1-60132-459-6, CSREA Press ©

20

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | [8]

V. CONCLUSIONS In this paper, we have presented a proposed system which is a cloud-based emergency medical information assistance. The motivation was to assist people in need of EMS during lifethreatening situations by providing them with real-time hospital information such as doctor availability, ambulance availability, direction and services to get help as soon as possible. We achieved this by: using RFID system to collect data from participating hospitals and stored on the cloud-based server in real time which is accessible using smartphones by user. This paper presented the system architecture and their operations to achieve the overall system goal. Moreover, we developed a system prototype called HFinder to demonstrate the idea discussed. By observation, found that the proposed system is more efficient when compared to the current EMS system where patients have to call and wait for long time for the arrival of EMS or hospital ambulance. The proposed system provides real time information, user friendly, do not consume lot of resources. Based on its mode of operation, we believe that if HFinder system is adopted and implemented, it could go a long way to difference between life and death. Nonetheless, as limitations, this paper was unable to fully implement the components of the system due to lack of resources. Additionally, lack of Internet access will render the system inoperable. As part of the future works, we are going to fully implement the system and test it on a real-world application setting.

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

Saydam, C., & Aytu ˘g, H. (2003). Accurate estimation of expected coverage: Revisited. Socio-Economic Planning Sciences, 37 (1), 69–80. Geroliminis, N. , Kepaptsoglou, K. , & Karlaftis, M. G. (2011). A hybrid hypercube—genetic algorithm approach for deploying many emergency response mobile units in an urban network. European Journal of Operational Research, 210 (2), 287–300. Leknes, H., Aartun, E.S, Andersson, H., Christiansen, M. and Granberg, T.A. Strategic ambulance location for heterogeneous regions European Journal of Operational Research 260 (2017) 122–133 T. N. Pham, M. F. Tsai, D. B. Nguyen, C. R. Dow and D. J. Deng, "A Cloud-Based Smart-Parking System Based on Internet-of-Things Technologies," in IEEE Access, vol. 3, no. , pp. 1581-1591, 2015. Jagtenberg et al. Benchmarking online dispatch algorithms for Emergency Medical Services. European Journal of Operational Research. Volume 258, Issue 2, 16 April 2017, Pages 715–725 Health24.com,' Left for dead: The shocking state of ambulances ',2015.[online].Available: http://www.health24.com/News/PublicHealth/Left-for-dead-The-shocking-state-of-ambulances 20150331. Date Accessed: 09/02/17 Andrew Whitmore, Anurag Agarwal, Li Xu. The Internet of Things—A survey of topics and trends Information Systems Frontiers, 2015, Volume 17, Number 2, Page 261 T.R. Peabody, T. Freed, RFID technology selection andeconomic justification for healthcare asset tracking, in:Proceedings of IIE Annual Conference, 2011,pp. 1–6. W. Yao, C.H. Chu, Z. Li, The adoption and implementation ofRFID technologies in healthcare: a literature review, J. Med.Syst. 36 (6) (2012) 3507–3525. Bassey Isong, Nosipho Dladlu, Tsholofelo Magogodi. “Mobile-Based Medical Emergency Ambulance Scheduling System” International Journal of Computer Network and Information Security (IJCNIS), ISSN: 2074-9090, 2016.

ACKNOWLEDGMENT This work was supported by FRC and MaSIM in the NWUMafikeng. We express our sincere appreciation and thanks to them as well as our colleagues in the Computer Science Department. REFERENCES [1]

[2] [3]

[4]

[5]

[6]

[7]

Paul, Sharoda A. et al. “The Usefulness of Information and Communication Technologies in Crisis Response.” AMIA Annual Symposium Proceedings 2008 (2008): 561–565. Levick, N. , 2008. Emergency medical services: unique transportation safety challenge. Transportation Research Board 87th Annual Meeting. Pradhan, A.R., Laefer, D.F., Rasdorf, W.J., 2007. Infrastructure management information system framework requirements for disasters. J. Comput. Civ. Eng. 21, 90–101. doi: 10.1061/(ASCE)08873801(2007)21:2(90) V. Sriram a,*, G. Gururaj b,e, J.A. Razzak c,f, R. Naseer d,g, A.A. Hyder a,h Comparative analysis of three prehospital emergency medical services organizations in India and Pakistan. Public Health. Issue 137, pp.169-175, 2016. Chen, A.Y. and Yu, T. Network based temporary facility location for the Emergency Medical Services considering the disaster induced demand and the transportation infrastructure in disaster response. Transportation Research Part B 91 (2016) 408–423 Aringhieri, R. , Bruni, M. , Khodaparasti, S. , & van Essen, J. (2017). Emergency medi- cal services and beyond: Addressing new challenges through a wide literature review. Computers & Operations Research, 78 , 34 9–368. Utkua, S., Özcanhana, M.H., Unluturk, M.S. Automated personnelassets-consumables-drug tracking in ambulance services for more effective and efficient medical emergency interventions. Computer methods and programming in Biomedicine, Issue 127, pp.216-23, 2016

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

21

Counting Multiple People on a Floor Based Array Sensor System Fadi Muheidat2 and Harry W. Tyrer1, 2 1 Professor Emeritus 2 Electrical Engineering and Computer Science department University of Missouri, Columbia, Missouri, USA

Abstract – We have developed a context-aware system that uses the functionality associated with the Internet of Things (IOT). We have a floor based array sensor system, which we call the Smart Carpet, which recognizes a person walking or falling, reports the fall and stores the data for regular evaluation of the gait parameters. These all have medical benefits. Here we report the improvement, which counts the number of individuals walking on the carpet. We used two methods to perform the count; the number of active sensors at given time, and the number of unique subgroups formed by the activated sensors using the connected component labeling algorithm for varying number of frames in the sliding window mainly 3, 5 and 9 frames. Our results showed that we could count and monitor individual and multiple people walking on the carpet with an average accuracy of 100%. We use the carpet as a component of an automated health monitoring system, which helps enable independent living for elderly people and provide a practical smart home environment that improves quality of life, reduces healthcare costs and promotes independence. Keywords: Sensing Floor, Patient Monitoring, Detecting People, Data Acquisition, Data Mining, Internet of Things (IOT), Smart home.

1

Introduction

The number of people aged 65 and older is growing worldwide. The US population will have one in five people 65 or older by 2030 [1]. The research community is innovating new technologies to help assistive living. However, the challenge is to have unobtrusive and userfriendly, unobserved, hand-free and affordable system that supports assisted living of elderly in their homes or in nursing houses. Systems have been proposed and developed which can be categorized into wearable (accelerometer and gyroscope) [2], and non-wearable (context-aware) [3,4,5,6,7,8]. These systems can be used to detect falls, estimate gait, monitor elderly activities. In addition, they can recognize, count people, and monitor their activities. Wearables sensors systems are effective devices to detect and recognize the location and activity of people. However,

they are obtrusive, must be worn at all times and need continuous power (batteries). In addition, it is not possible to anonymously count and detect people without previous knowledge of the person’s wearable device. Alternatively, context aware systems overcome some of these issues. In video monitoring systems, the vision techniques filter the images at the device level due to privacy concerns. However, the users still have the feeling that of being watched. Kinect, a video-based system uses skeleton tracking to detect people with good resolution has value [9, 10]. However, it suffers degraded performance with occlusion, and limited depth range. Multifunction radar systems [11] proved to be promising solution in detecting humans and their activities even behind walls or foliage, yet they suffer classifications accuracy for other barriers and movement gestures. Microphone array sensors [7, 12] suffer from noise and multiple interference. Our lab uses context-aware, non-computer-vision based human recognition and fall detection system. It is a floor based array sensors system, i.e. smart carpet [13], which is completely private. One installs it in the home or apartment and additionally has usefulness in places where traditional sensing system might suffer complications like occlusion. The smart carpet system includes the sensor data acquisition, data manipulating, data reading, storage, display, and communication. The system operates by detecting the person’s movement and storing the floor sensor data. The motion on the carpet activates a set of sensors that outputs a voltage signal. The system amplifies the signal, digitizes it, and then translates it into a frame for further processing. We ran computational intelligence algorithms to measure and estimate people’s gait, and detect falls. Our goal is to accurately recognize, count and monitor the movements of the individuals walking on the smart carpet system. We organized this paper as follows. First, the methodology, which includes an overview of the system we developed and installed in lab settings. We describe algorithms used to count the number of people walking on the carpet. Third, we show experimental results for different walking scenarios performed by volunteers. Finally, we discuss the achieved results, limitations and future work.

2

Methodology

ISBN: 1-60132-459-6, CSREA Press ©

22

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

2.1

System overview

The smart carpet system, as shown in Figure 1 consists of the Smart Carpet sensors laid under the mat, data acquisition system, and processor. The signal scavenging sensors connect to the data acquisition system that scans at configurable speeds depending on the size and number of sensors. Signals convert to digital values using 10-bit Analog to Digital convertor, and microprocessor further thresholds providing a 1 for activated and 0 for not activated sensor, as well as formatting the frame with an S for start and E for end. A computer then reads this scan as an ASCII frame. The software components process the data frames, and use different computational intelligence methods to perform the required operations like fall detection, gait estimation, data

32 sensors; segment D was turned off for this project, and the sensors in each segment connects to the data acquisition system. Figure 2 shows the layout of the carpet segments. Walking across the carpet from A to C or C to A (Longitudinal Direction) will require longer time, longer travelled distance, and more activated sensors count, compared to waking from segment D direction bottom up (Transverse Direction). We made use of a binary display of the activated and non-activated sensors on the carpet to see the traversal of the individuals.

2.2

Experiments and counting algorithms

We collected data from four different people. As listed in Table I, each person, individually, performed 10 walk trials in traverse direction from bottom of segment A then to TABLE I SUBJECT’S AGE, WEIGHT, AND HEIGHT

Subject

Weight

Height

Age

Male – Adult FemaleAdult FemaleChild

200 lb, 90.72 Kg

5’9”, 174 cm

40

150 lb, 68 Kg

5’3”, 160 cm

31

98.4 lb, 44.63 Kg

4’8”, 142 cm

12

Male - Child

49.7 lb, 22.5 Kg

3’11”, cm

8

119

segment B and back to the beginning. Then, multiple persons participated in 2 people, 3 people and four people walk trials for 10 times each.

Figure 1 System overview

visualization, and notification. Additionally, the system can show the signal data scavenged by the sensor for fine-tuning of the system parameters. The system consists of sensor array

Figure 2 Carpet layout: Active segments A, B and C. Sensors in D are turned off made into four segments A, B, C, and D. Each segment has

The smart carpet data acquisition system scans the carpets at 9 frames per second. Each frame consists of 128 sensors, where all segment D sensors turned off. However, we used them to build 12x12 binary image. Where ‘1’ means, the sensor is activated, and ‘0’meas it is not activated. This image becomes the base data structure to perform computation to recognize people on the carpet. We used Connected Component Labeling (CCL) algorithm as described in [14], we applied the same procedure for both single frames and window size of frames encompassing variable number of frames: 3, 5, and 9. Each window corresponds to time (WS = 3 frames correspond to 0.2 seconds, WS = 5 frames corresponds to 0.5 seconds, and WS = 9 frames corresponds to 1 seconds) WS = total number of frames corresponds to total travel / ambulation time). We used to 8-connect neighborhood [15] for our experiments to ensure we do not ignore the effect of interference among the sensors, and so we have biased results. We used a hybrid model of both the number of subgroups formed by the neighboring activated sensors, and the count of the subgroups formed by individual activated sensor that are not direct neighbors (outside the 8 neighborhood). Another method we used is the count of the total sensors activated for full walk, and then divided by the average

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

number of sensors activated per individual(s) who performed the walk. For example, the “Two people walking in opposite directions”, we took the average count of active sensor performed by the two people when they walk alone on the carpet, and then divided the total active sensors of the full walk by this average number to count number of people.

3

23

Experimental Results

We performed 10 experiments for each scenario: individual, two people (same and opposite directions), Three people (same direction), and four people (same and opposite detections). Figure 3 shows the binary image of the carpet layout for “Two People Walking in Opposite Directions” scenario. It took 4 seconds to perform the walk. The left image shows the start of the walk (t = 1 second). Two

Segment D Sensors turned OFF

Figure 3 Active sensors map: Two people walk in Opposite directions: Longitudinal Direction (A C), frames are grouped in a window of size 9 frames/sec (i,e 1 second ambulation time). All sensors in Segment D are turned OFF. This applies for all experiments and results.

Figure 4 Active sensors map: Two people walk in the Same directions: Transverse Direction window of size 9 frames/sec (i.e. 1 second ambulation time). Segment D sensors turned OFF

(A é B), frames are grouped in a

Figure 5 Active sensors map: Four people walk in Opposite directions: Longitudinal Direction (A C), frames are grouped in a window of size 9 frames/sec (i.e. 1 second ambulation time). Segment D sensors turned OFF

ISBN: 1-60132-459-6, CSREA Press ©

24

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 6 Active sensors map: FULL Walk Sensors distributions for: (Left) Three people walk in Same directions: Transverse Direction (A é Bé C), (Middle) Four people walk in Same directions: Transverse Direction (A é Bé C), (Right) Four people walk in Opposite directions: Longitudinal Direction (A C) frames are grouped in a window of size 9 frames/sec (i.e. 1 second ambulation time). Segment D sensors turned OFF subgroups of size 4 each (four connected neighbors) are shown. The middle image shows the subgroups at the end of the walk (t = 4 seconds). This image shows three subgroups of size one, and two subgroups of size greater than one. The computational algorithm ignores these individual sensors. Hence, we got two subgroups. The Right image represents the full walk and shows the path that each person followed. Figure 4 shows the binary image for the orthogonal “Two People Walking in the Same Direction” scenario. It took 4 seconds to perform the walk. The left image shows the start of the walk (t = 1 second). Two subgroups of size 3 each (three connected neighbors) are shown. The middle image shows the subgroups at the end of the walk (t = 4 seconds), with one subgroups of size one, and one subgroup of size three. If the computational algorithm ignores the individual sensors, then the middle image would show only one person. However, the speed of walking of the two persons is different and hence one finishes before the other. So, if we change the time frame we would see the second person. The Right image represents the full walk and shows the path that each person was in a contiguous segment meaning the distance between the two segments is of one foot, and hence some sensors got activated due to interference. If Person 2 walks on Segment C, which is greater than five foot apart, such behavior did not exist. For example, the opposite scenario shown in Figure 3 does not have this problem. In Figure 5 four people walk in opposite directions. At the start of the walk left the persons were at separable distance. They were recognized and by their own subgroup. However, as shown in the middle, they became closer and were not clearly separated. Then when they reached the end of the walk, right, they were again separable. Figure 6 shows full walk image, all activated sensors during the walk, for more walk scenarios (three and four people in the same directions, and four people in the Opposite directions). We further studied one scenario for two, three and four people walking in the same direction (transverse direction). We ran the hybrid algorithm for different window sizes of frames. We applied the algorithm for 10 walk trials. Figure 7 (a,b,c) shows the count of people for “two people walking in same direction”, “three people

walking in same direction”, and “four people walking in same direction” scenarios for different sliding windows of frames. Results showed that we could reliably count the number of people for the “two” and “three” people scenarios. However, when number of people increased for the same size of the carpet used, it became difficult to count the people reliably (accuracy of 20%). Accuracy is proportional the ratio of the number of people walking on the carpet and the carpet size. We could not determine the optimal window size of frames that fits all scenarios, especially when the ratio of the number of people to the carpet size is big. Figure 8 shows that the count of people for two, three, and four people walking in the same direction at window size of nine (WS = 9 frames, i.e. the algorithm determines the count of people at time intervals of one second). As Figure 8 shows that at WS= 9 frames, the accuracy of counting people is 100% for two people, 90% for three, and 30% for four people. We evaluated the binary image by the count of the ‘1’ pixel value, which corresponds to an active sensor. Figure 9 shows the path we used to identify the average number for activated sensors for the four persons who performed the experiments. TABLE II shows the activated sensors count for each scenario, and the average time it took to perform the scenario. It is clearly evident that the bigger the area is the more people walking on the carpet. The count of the active sensors is 72 rounded to whole number. Comparing this to the 58 for the same number of people but in the same direction. The carpet layout and the time spent for the opposite directions (6 seconds) activate more sensors than and the same directions (4 seconds) of walking on the carpet. We calculated the average number of sensors activated by each walk. We then divided by the mean of active sensors produced by individual walks for the persons who performed the walk. We rounded the result to obtain the count of people. This allows us to determine the count of unknown people walking on the carpet. TABLE III shows the count of people using the average active sensors count, per individual(s) performing the walk trials, as the denominator for the total active sensors in the full walk trial.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 7-a People count for “two people walking in same direction” for different sliding windows of frames.

Figure 7-b People count for “three people walking in same direction” for different sliding windows of frames.

Figure 7-c People count for “Four people walking in same direction” for different sliding windows of frames.

4

Discussion and Conclusion

In this paper, we extended the functionality of the smart carpet to count the number of people in addition to fall detection and gait estimation. Falls for an individual are a rare but high impact event, and effect a large fraction of the population. So, it is important to monitor, but not act until necessary. Further, the system generates data 24/7 and so stored for later analysis to detect changes in gait. Since a plurality of people walking on the floor can mimic a fall

event, it is important that the system distinguish between a plurality of people and a fall event. We monitored the activity of volunteers walking on the carpet. Our algorithms were able to count the number of people at any given time with an average accuracy of 100%. This result affected by the number of people walking, and the spatial distance that separates them, and by the ratio of the people walking to the actual carpet size. The bigger the ratio the lower the counting accuracy. We used the total count of activated sensors compared to the average of individual walks sensors count,

ISBN: 1-60132-459-6, CSREA Press ©

25

26

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 8 People count for 10 walk trials average for Windows size of 9 frames for the Two, Three, and Four people walking on the carpet

Figure 9 Individual walk trials in transverse direction. Segment D sensors turned OFF and were able to count the number of people with close 100% accuracy. TABLE II AVERAGE ACTIVE SENSORS COUNT, AND TRAVEL TIME

Average active sensors count – rounded

Travel time seconds

14

6.5

Two personsSame direction

27

4

Two personsOpposite direction

25

4

Three persons Same directions

49

4

Four persons Same direction

58

4

Four persons Opposite direction

72

6

Walk Scenario One Person

Our results show high performance count and detection accuracy. Since our algorithms do not depend on the spatial location or dimensions of the sensors on the floor, we can count and track people on any sensors distributions. Future work will involve using the centroid in two settings; the directions of the centroid of the connected component, and, spatially, by computing the centroid of the actual dimensions minimum distance by which we recognize different people. We believe more information can be deduced using this technique like gait parameters (walking speed, stride length, and step length).

5

References

[1] U.S. Department of Health and Human Services: Administration on Aging, “A Profile of Older Americans: 2011,” 2011. [2] D. Giansanti, G. Maccioni and V. Macellari, The development and test of a device for the reconstruction of 3D position and orientation by means of a kinematic sensor assembly with rate gyroscopes and accelerometers, IEEE Trans. Biomed. Eng

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

TABLE III AVERAGE ACTIVE SENSORS COUNT, AND COUNT OF PEOPLE WALKING ON THE CARPET

[8] M. Popescu, Y. Li, M. Skubic and M. Rantz, An acoustic fall detector system that uses sound height information to reduce the false alarm rate, in: 30th Int. IEEE EMBS Conf, Vancouver, BC, Aug. 20–24, 2008, pp. 4628– 4631

Walk Scenario

Avera ge active sensor s count

Average active sensors count for persons who performed the walk

People Count

[9] E. E. Stone and M. Skubic, "Fall Detection in Homes of Older Adults Using the Microsoft Kinect," in IEEE Journal of Biomedical and Health Informatics, vol. 19, no. 1, pp. 290301, Jan. 2015.doi: 10.1109/JBHI.2014.2312180

One Person

13.62

13.62

1

27

14.20

2

[10] H. Y. Lin, Y. L. Hsueh and W. N. Lie, "Abnormal Event Detection Using Microsoft Kinect in a Smart Home," 2016 International Computer Symposium (ICS), Chiayi, 2016, pp. 285-289.doi: 10.1109/ICS.2016.0064

25

14.20

2

49

14.10

3

58

13.62

4

48

13.62

4

Two personsSame direction Two personsOpposite direction Three persons Same directions Four persons Same direction Four persons Opposite direction

[3] M. Addlesee, A. Jones, F. Livesey and F. Samaria, The ORL active floor, IEEE Personal Communications 4 (1997), 35–41. doi:10.1109/98.626980. [4] M. Alwan, P.J. Rajendran, S. Kell, D. Mack, S. Dalal, M. Wolfe and R. Felder, A smart and passive floor-vibration based fall detector for elderly, in: 2nd IEEE Int. Conf. on Inf. & Comm. Tech., Damascus, Syria, Apr. 24–28, Vol. 1, 2006, pp. 1003– 1007.

[11] Ram M. Narayanan, Sonny Smith, and Kyle A. Gallagher, “A Multifrequency Radar System for Detecting Humans and Characterizing Human Activities for ShortRange Through-Wall and Long-Range Foliage Penetration Applications,” International Journal of Microwave Science and Technology, vol. 2014, Article ID 958905, 21 pages, 2014. doi:10.1155/2014/958905 [12] Y. Li, Z. Zeng, M. Popescu and K. C. Ho, "Acoustic fall detection using a circular microphone array," 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology, Buenos Aires, 2010, pp. 22422245.doi: 10.1109/IEMBS.2010.5627368 [13] Neelgund R., “Floor sensor development using signal scavenging for personnel detection system”, University of Missouri, Columbia 2010. Master’s thesis for fulfillment of MS Degree. [14] https://en.wikipedia.org/wiki/Connectedcomponent_labeling. Accessed 4/4/2017 [15] https://www.mathworks.com/help/images/ref/bwlabel.ht ml. Accessed 4/4/2017

[5] D. Anderson, R.H. Luke, J. Keller, M. Skubic, M. Rantz and M. Aud, Linguistic summarization of activities from video for fall detection using voxel person and fuzzy logic, Computer Vision and Image Understanding 113(1) (Jan. 2009), 80–89. doi: 10.1016/j.cviu.2008.07.006. [6] A. Sixsmith, N. Johnson and R. Whatmore, Pyrolytic IR sensor arrays for fall detection in the older population, J. Phys. IV France 128 (2005), 153–160. doi:10.1051/jp4: 2005128024. [7] Y. Li, K. C. Ho and M. Popescu, "A Microphone Array System for Automatic Fall Detection," in IEEE Transactions on Biomedical Engineering, vol. 59, no. 5, pp. 1291-1301, May 2012.doi: 10.1109/TBME.2012.2186449

ISBN: 1-60132-459-6, CSREA Press ©

27

28

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Human Movements Monitoring Using Smartphone Sensors 1, 2

Hala Bin Saidan1, Hmood Al-Dossari2 Information System Department, College of Computer and Information Sciences, King Saud University, Riyadh, Saudi Arabia

Abstract - Nowadays, smartphones are ubiquitous; they have some powerful features such as sensors, which have recently become a topic of great interest to researchers. Smartphone sensors collect data on the user’s activities and movements, which provides meaningful knowledge if captured and analysed in an accurate way. This paper will introduce a system based on a smartphone equipped with an accelerometer sensor to monitor users’ daily activities, such as walking, jogging, stair ascents, stair descents, sitting and standing. The system will attempt to identify the maximum level of movement (MLM), meaning the time spent continuously doing an activity. To avoid any suffering, the user is notified before he or she exceeds the MLM. This system would be useful for healthy humans, and particularly elderly people or patients with medical conditions related to physical activity. Keywords: Smartphones, mobile sensor, accelerometer sensors, human activity recognition, human movement, maximum limit.

1

Introduction

The paper main goal is to improve and facilitate people’s lives, especially patients with a medical condition related to physical activity, by taking advantage of the newly available generation of smartphones equipped with a variety of sensors. Smartphones are equipped with a number of sensors, such as an accelerometer, gyroscope, magnetometer sensors and many more [1]. Accelerometers are mainly used for physical activity analysis, while a gyroscope is for capturing user motion and device orientation changes as well as the magnetometer [1]. Sensors were essentially added to smartphones to support advanced games and to enable automatic screen rotation [2]. It is possible to take advantage of sensors ability to capture meaningful information. For example, the accelerometer is the most commonly used sensor related to in human activity recognition (HAR) [9] [1], but it can also be used in conjunction with other types of sensor to gain better results, as seen in [10] [1] [13]. The motivation of this paper is to develop an application that can utilise smartphone embedded sensors to monitor users’ daily activities, to measure the maximum level of movement (MLM) that a human can physically perform on a daily basis. To produce an accurate measurement of MLM, firstly, an experimental study will be conducted on 3 categories

of people: healthy individuals, elderly people, and patients with a medical condition related to physical activity. The sensor data collection phase will be held during the experimental study. The participants of the study will install an application on their smartphones for collecting their activities and movement data while performing a regular daily activity. At this stage, they also will be asked to provide feedback on the application once they feel exhausted. Subsequently, the collected data will be analysed and classified using the classifying tools and algorithms to identify the MLM for each category. The MLM will be measured and calculated based on time spent continuously doing an activity or movement. The result will be an application capable of notifying any user before reaching their MLM to avoid any feeling of fatigue or injury. This paper seeks to increase awareness of how to control and self-monitor these daily activities to avoid any injuries. The paper is organised as follows: section 2 will review the existing literature related to the proposed approach. in section 3 the research methodology will be presented. Section 4 will present the expected outcomes and finally, the conclusion in sections 5.

2

Literature review

The validity of accelerometers in recognising the pattern of human daily activity has been approved by many studies [14] [2] [4]. One of the first studies in recognising human activity using an accelerometer was introduced in [15], in which a wearable device with three axial acceleration sensors was used to recognise different human activities. Human activity recognition sensors can be a wearable sensor as seen in [16], in which the use of wearable sensors was reviewed, and it was stated that they are experiencing rapid growth, widely used for fall detection in elderly people in the healthcare sector. Additionally, people carry their smartphone almost all the time which makes it an ideal device to monitor daily activities. Smartphones with embedded sensors have been used for human activity recognition as in [7], which presents a ready-to-use user-independent activity recognition system. In this system, the user does not need to set any parameters before starting to use the application, and it functions on several different hardware and operating systems. [17] presents a comprehensive survey of the recent advances in activity

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

recognition with smartphone sensors, along with its challenges, and some existing applications in the fields of daily movement monitoring, the care of elderly people, and localisation. [2] explores the use of the accelerometer to identify the user’s activity, to gain useful knowledge about their habits. The user only needs to carry their mobile phone, and set their accelerator data to available for public use. In [8], a scheme is invented to identify the best classifier to be used through the accelerometer. The results were that random forest (RF) has the best performance out of multilayer perceptron (MLP), J48, classification via regression (CVR), and RepTree (RT) classifiers. The following studies show the use of sensing technology in the healthcare field. For example, [5] presents a novel data analytics scheme for intelligent human activity recognition (HAR). It uses wireless body sensors and smartphone inertial sensors to assist intelligent and automatic real time human activity monitoring technology, for the development of eHealth applications. Furthermore, the study shown in [7] was more specific as it focused on the calibration of a walking speed estimation, using smartphone sensors in the field of mobile health (mHealth). In [11], a mobile phone platform is designed and developed to collect users’ psychological, physiological, and activity information. The tool collects the data using

29

this research, is that this research not only analyses user activities, but also attempts to identify the extent to which they can be active.

3

Methodology

The research will attempt to utilise the accelerometer sensors on a smartphone to monitor human activity, calculate the total time spent on activity, identify activity types and estimate the maximum level of movement that users can perform. Activity recognition is achieved by processing sensor data with appropriate data mining approaches [19]. The use of data mining techniques will first help to identify the different activity types, and secondly the system should to be able to learn from new user information as well as the sensor trained data, by applying data mining techniques. The research aims to go further than simply monitoring human activity, by also determining the maximum level of daily activity that a particular person can handle. For the sensor data collection, the plan is to review a number of off-the-shelf applications and select the most appropriate one to be used in the experimental study for the data collection that mentioned earlier. The use of publicly available datasets is also

Figure 1. System main components wireless wearable biosensors in the form of a three-axial accelerometer equipped within a wireless electrocardiogram. The purpose of the platform is to be used in mental health research. It is available as open source software. The work reported in [12] is relevant to our work in which a smartphone application was built to tracks users’ physical activities and then provide them with feedback. This application functions without user intervention during routine operation. Its purpose is to help users to be able to make decisions giving them a healthier lifestyle, based on their activities. The main difference between the work in [12] and

being considered. Therefore, the collected data will be analyzed using one of data mining tool such as WEKA tool. Generally, there are three main components in the proposed approach: the data collector, the activity extractor, and the max limit determination, as shown in Fig. 1. A. Data Collector: to collect sensor data (activity), the user profile information and user status given by users as feedback of how tired they are after being active for some time. The system will build based on the accelerometer sensor embedded into a smartphone, therefore the data is collected from the three axes acceleration of the

ISBN: 1-60132-459-6, CSREA Press ©

30

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

accelerometer: The X axis records horizontal movement; the Y axis records upward and downward motion, and the Z axis records forward movement [2]. B. Activity Extractor: Sseveral human activity recognition algorithms such as the Decision tree (J48), Naïve Bayes (NB), Neural Network (NN) (Multilayer Perceptron) and Support Vector Machine (SVM) will be reviewed, and the most appropriate algorithm will be chosen based on the test result. This algorithm will then be used to extract the activity types from the sensor data as well as the duration of each activity. C. Max limit Determination: the most important component which is responsible for identifying the maximum movement allowance for the target user. In order to do so, machine learning algorithms are used to determine the maximum limit based on the collected accelerometer data from users and their feedback, given by the participants in the data training phase during data collection

3.1 Evaluation To evaluate the system, both offline and online evaluations have been planned. Offline activity recognition means that the classification and feature selection is conducted using machine learning tools offline, and the final classified data is entered into the mobile sensor [3]. Otherwise, Online activity recognition means that the classification of activities is performed in real time on the smartphone [3]. Our solution will be evaluated as follow: A. Offline evaluation: The collected sensor data will be divided into two parts. The first part will be used to analyse the activity and determine the MLM, and the second part will be used to test the validity of this determination. B. Online evaluation: The research plans to involve the user in evaluating the accuracy of the application, by providing them with a survey to answer after they have been notified, to verify if the received notification was timely and accurate.

4

Expected outcomes

The expected result is an application functions based on accelerometer sensor to monitor user movements and activities. therefore, based on the sensor stored and trained data and the user profile information the application will be able to send alerts based on these measurements before the user reaches their own MLM to avoid any suffering or effect on their health.

5

Conclusion

To conclude, any activity, including even walking or moving, has its pros and cons. The aim of this paper is to highlight the side effects of human daily activity. In some circumstances, performing necessary activities for daily living for a long time or even in an incorrect way can cause back or joint pain [18], or possibly even fainting. The main goal of the proposed research is to develop an application that is capable of measuring the maximum level of movement and activity (MLM) that humans can perform on a daily basis. The measurements will be based on the time the user can spend active, and the type of activity. The purpose of calculating the MLM of users is to save their energy and protect their health. Consequently, a challenge for the research is the accurate identification of the maximum limit, because of the nature of human movement to change from one activity type to another. This may lead to a user not being in complete active mode for the whole calculated time. Furthermore, there are some incommensurable factors that have an effect on human activity besides age, gender and weight. For example, factors such as sleeping and eating have a huge influence on human energy and the ability to perform daily activities. However, they are difficult to measure, especially using mobile sensors.

6

References

[1] Piyare, Rajeev Kumar, and Seong Ro Lee. "Activity Recognition of Workers and Passengers onboard Ships Using Multimodal Sensors in a Smartphone."한국통신학회논문지 39.9 (2014): 811-819. [2] Kwapisz, Jennifer R., Gary M. Weiss, and Samuel A. Moore. "Activity recognition using cell phone accelerometers." ACM SigKDD Explorations Newsletter 12.2 (2011): 74-82. [3] Shoaib, Muhammad, et al. "A survey of online activity recognition using mobile phones." Sensors 15.1 (2015): 20592085. [4] Tragopoulou, Spiridoula, Iraklis Varlamis, and Magdalini Eirinaki. "Classification of movement data concerning user's activity recognition via mobile phones." Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14). ACM, 2014. [5] Chetty, Girija, Matthew White, and Farnaz Akther. "Smart phone based data mining for human activity recognition." Procedia Computer Science 46 (2015): 11811187. [6] Siirtola, Pekka, and Juha Röning. "Ready-to-use activity recognition for smartphones." Computational Intelligence and Data Mining (CIDM), 2013 IEEE Symposium on. IEEE, 2013.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[7] Altini, Marco, et al. "Self-calibration of walking speed estimations using smartphone sensors." Pervasive Computing and Communications Workshops (PERCOM Workshops), 2014 IEEE International Conference on. IEEE, 2014.

[19] Chetty, Girija, Matthew White, and Farnaz Akther. "Smart phone based data mining for human activity recognition." Procedia Computer Science 46 (2015): 11811187.

[8] Dash, Yajnaseni, Sanjay Kumar, and V. K. Patle. "A Novel Data Mining Scheme for Smartphone Activity Recognition by Accelerometer Sensor." Proceedings of the 4th International Conference on Frontiers in Intelligent Computing: Theory and Applications (FICTA) 2015. Springer India, 2016. [9] González, Silvia, et al. "Features and models for human activity recognition." Neurocomputing 167 (2015): 52-60. [10] Chetty, Girija, and Mohammad Yamin. "Intelligent Human Activity Recognition Scheme for eHealth applications." Malaysian Journal of Computer Science28.1 (2015). [11] Gaggioli, Andrea, et al. "A mobile data collection platform for mental health research." Personal and Ubiquitous Computing 17.2 (2013): 241-251. [12] Anjum, Alvina, and Muhammad U. Ilyas. "Activity recognition using smartphone sensors." 2013 IEEE 10th Consumer Communications and Networking Conference (CCNC). IEEE, 2013. [13] Kim, Tae-Seong, Jin-Ho Cho, and Jeong Tai Kim. "Mobile motion sensor-based human activity recognition and energy expenditure estimation in building environments." Sustainability in Energy and Buildings. Springer Berlin Heidelberg, 2013. 987-993. [14] Pires, Ivan Miguel, et al. "Identification of Activities of Daily Living Using Sensors Available in off-the-shelf Mobile Devices: Research and Hypothesis." Ambient IntelligenceSoftware and Applications–7th International Symposium on Ambient Intelligence (ISAmI 2016). Springer International Publishing, 2016. [15] Mantyjarvi, Jani, Johan Himberg, and Tapio Seppanen. "Recognizing human motion with multiple acceleration sensors." Systems, Man, and Cybernetics, 2001 IEEE International Conference on. Vol. 2. IEEE, 2001. [16] Mukhopadhyay, Subhas Chandra. "Wearable sensors for human activity monitoring: A review." IEEE Sensors Journal 15.3 (2015): 1321-1330. [17] Su, Xing, Hanghang Tong, and Ping Ji. "Activity recognition with smartphone sensors." Tsinghua Science and Technology 19.3 (2014): 235-249. [18] Appalachian Physical Therapy, “There’s a difference you can feel”. [Online]. Available: http://www.aptfc.com/ [Accessed 30 Dec 2016].

ISBN: 1-60132-459-6, CSREA Press ©

31

32

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Open System for Monitoring Vital Signs of Babies to Help in the prevention and Diagnosis of Sudden Death Kleisson Tedesco Federal University of Technology - Paran´a Campus Pato Branco [email protected]

Robison Cris Brito

Eduardo Todt

F´abio Luiz Bertotti

Federal University of Technology - Paran´a Campus Pato Branco [email protected]

Federal University of Paran´a UFPR [email protected]

Federal University of Technology - Paran´a UTFPR [email protected]

Abstract—This paper presents deals with the design and implementation of a system to monitor the vital signs of a baby, to support the diagnosis and pediatric use. The text presents an approach on monitoring signals, normality standards, and measurement devices. It also represents how tools used to implement the project, and how they are associated. The concept of a wearable computer is used, where computers are becoming accessories in which we wear and use on a daily basis. The system consists of a temperature sensor, a motion sensor and a heart rate sensor connected to the micro controller that sends this data over the Wi-Fi network, to a server. The data is served by an application, developed for Android devices, which performs the interface between the user and the system. At the end of the development, it was verified that the proposed objectives were achieved, the transmission of the data read in the sensors achieved the expectations. Keywords—Wearable Computer, Android, Hardware/Software, Monitoring Remote, IoT

I.

Open

I NTRODUCTION

The evolution of computing, in particular, electronic devices is increasingly fast. They became small, cheap and with more resources, it is possible to design solutions involving hardware and software to automate many tasks of everyday life. This advancement includes mobile communication techniques, embedded computing, and miniaturization of electronic devices and sensors, which are generated for the development of applications in the medical field, allowing an optimization of the of the services provided by health professionals. Wireless communication improvements have expanded the possibilities of monitoring and control in electro-medical devices remotely, increase the locomotion by the health professionals. The miniaturization of mobile electronic devices increased battery efficiency, and reduced semiconductor power consumption enhance the development of countless innovative solutions through ubiquitous computing[1]. Ubiquitous computing is the idea that computing moves out of workstations and the personal computers and becomes pervasive in people’s daily lives, wherever we are. It is a trend that has been growing exponentially in recent years [2].

As the portability and functionality of electronic equipment are increasingly improved, ”wearable” devices are revolutionizing the marketplace as a new way of taking care of health. These devices are called wearable computer [3]. The term wearable computer is defined as a wearable, portable computer that supports the human being in day to day tasks, or allows the monitoring of parameters, making this data available to the person later. Wearable computer can be classified as a computer that is always connected and always accessible [4], and can be dressed by newborn to the elderly, having functionalities such as monitoring body temperature, blood pressure, heart rate, among others. This paper present deals with the use of a wearable computer in newborn children continuous monitoring, principally during sleep. It can help parents, especially those with little experience, to deal with routine situations, such as Monitoring of fevers, movement in the cradle of newborn babies, as any anomaly detected, combined with a rapid action, may contribute to decreasing the risk of sudden death. Sudden death is still not accurately explained by science, and remains one of the leading causes of infant mortality. Sudden Infant Death Syndrome (SIDS) is defined as the sudden death of a baby less than 1 year of age, usually associated with sleep. One of the possible ways to discover the cause of sudden death is to monitor vital signs, where by crossing the data and analyzing them later, the physician can produce a more accurate and more reliable report [5]. Besides, monitoring can generate statistics that when analyzed could detect some types of anomalies, such as congenital heart diseases. Once monitored it is possible achieve more efficient prognoses, improving the cure or treatment. To perform remote monitoring, concepts about Internet of Things, known by the acronym IoT, were used to design the proposed system. This is the term used to classify common day-to-day devices that have access to the Internet. These devices are capable of measuring data, controlling systems and even serving in-house tasks. Via sensors or wearable devices, you can intermittently or continuously monitor and record relevant data on a server on the Internet, offering an important

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

solution although the limits imposed by the traditional monitoring of these signals, allowing even triggering notifications that can help prevent a worsening of that user of the system. The use of IoT concepts, together with low-cost electronic components with open standards, known as free hardware, allows a flexible development, with openness to include new features such as sensors or actuators, leaving the architecture in which are inserted open for future customizations or optimizations. Another feature of these elements is low cost, as well as easy integration with other systems, such as database servers in the cloud. This work presents a prototype of one wearable computer based on free hardware applying the principles of IoT, which allows to remotely monitor, using an Android device, the vital signs of a newborn, being that this data is stored remotely in a server in the cloud. This data can be accessed by Android Device, which also has the function of personalizing alert parameters, such as being informed when any of the vital signs are outside a previously registered range, thus considering an abnormal reading. The history of all readings performed by Wearable Computer is also available in graphics on the Android device. II.

R ESEARCH PROBLEM

According to the World Health Organization, in 2015, 6.5 million children died before reaching the age of five. Many countries still have a high infant mortality rate, particularly those in the African Region with about 81 infant deaths under five for every 1000 live births. Inequalities in infant mortality among high and low-income countries remain large. In 2015, the under-five mortality rate in low-income countries was 76 deaths per 1,000 live births, about 11 times the average rate in high-income countries (7 deaths per 1,000 live births) [6]. In addition to social inequality, several factors contribute to the composition of these numbers, among them diarrhea, pneumonia, HIV, among other diseases. Sudden infant death syndrome is classified in the other group, which is characterized by deaths during sleep or by unknown causes. As reported by the Centers for Disease Control and Preventions [7], about 3,500 babies under one year of age die every year from unknown causes in the United States. From this number 44 % of the cases were classified as sudden infant death syndrome. Another major risk for newborns is accidental asphyxia during sleep or strangulation in bed. In the United States, these numbers declined considerably following the campaigns of the American Academy of Pediatrics knowing by Safe Sleep in 1992, Back to Sleep in 1994, and the Sudden Unexplained Infant Death Investigation Reporting Form in 1996, which encourage safe sleep, counseling Parents about the best ways to put their babies to sleep [7]. In Brazil, this index is calculated by the (SUS) Sistema Unico de Saude (Unified Health System) and is included in the list of avoidable causes of death that can be reduced by adequate health promotion actions combined to health care actions. The World Health Organization classifies Sudden Infant Death Syndrome in a group called: Symptoms, signs and abnormal findings of clinical and laboratory exams, and SUS follows this classification hierarchically [8].

33

In the year 2015, 31,441 deaths of children under 4 years of age were recorded in Brazil according to the Child and Fetal Monitoring Panel. Of these deaths, 953 (2%) were classified as sudden death, represented by Chapter XVIII in the International Classification of Deseas.These cases are divided into 3 groups: Early Neonatal (full first week of life) 167 in 2015, Late Neonatal (three weeks following) 66 in 2015 and post neonatal and following 720 in 2015. The numbers total a total of 953 deaths By 2015, and are declining due to government campaigns for follow-up and counseling to prevent Sudden Infant Death Syndrome [9]. Several epidemiological studies carried out since the 1960s show some factors that contribute to an infant’s risk profile. Among them: young mother, multiparous, short interval between pregnancies, absence of prenatal care, prematurity, low birth weight, brother who died of SIDS, low socioeconomic level, and a predominance of males, ethnicities and geographic regions . Strong evidence suggests that there is a risk up to 14 % higher when infants sleep in the chest position down. From these indications, the campaign ”Reduce the risk of SIDS” was initiated in Brazil in the 1990s, and a reduction in postnatal mortality was observed up to 50 %, depending on the region of the country. A study carried out in Rio Grande do Sul, in the cities of Pelotas and Porto Alegre, showed that cases of SIDS are underdiagnosed and are not included in official statistics. This same study attributes an index of 4 % for the Pelotas region as a consequence of SIDS, and an index of 0.08 % in the region of Porto Alegre. Comparing with the numbers in this study, an estimate of approximately 6,120 cases per year can be attributed in Brazil [10]. III.

M ATERIALS AND M ETHODS

With the intention of trying to identify the cause of sudden death, as well as allowing countries to predict risks associated with newborn sleep, such as turning face down and being asphyxiated, this project aims at a flexible and low-cost solution to monitor Physiological data during sleep. For the development of the proposed system, several questions were studied, such as the correct functioning of the sensors, as well as the identification of the best place to wear a Wearable Computer in a baby, in order to obtain a more accurate measurement of the data. Another concern was the integration model of the intelligent abdominal belt with the servers, as well as its integration with the user’s Android Device. The Figure 1 presents the general scope of the system, focusing on the integration model. In the figure it is possible to observe that the abdominal belt is dressed in the baby, and has a mechanism of wireless communication to send the data read in its sensors to a remote server. The integration of this data with the Internet happens through the Wi-Fi communication (802.11), which sends the data to an Ubidots server, which stores this data. Wearable Computer communication designed with the Internet happens through a traditional Wi-Fi router, identical to what is found in most homes with Internet access.

ISBN: 1-60132-459-6, CSREA Press ©

34

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

device, such as a Smartphone or Tablet, a database service was chosen in the cloud for IoT called Ubidots. In its free version, it is used in the project, it is possible to use up to 20 devices, reading and recording. It is also possible to record the data up to 60 times per minute. There is a limitation of 500 thousand recordings, valid for 3 months, after this period this counter is reset. In Ubidots, each client has a passkey, sharing in the environment variables fed with the data read from the sensors. Communication is done through the HTTP protocol, in which the message with the information is transferred or retrieved from the server.

Fig. 1.

•

Device Android: The Android platform provides free application development tools for smartphones and tablets, and these are the focus of this work. Android is maintained by Google, and allows the application to easily integrate with native device features. The platform is developed based on the Linux operating system and is composed of a set of tools that act in all phases of the project development [13]. This platform was chosen for the facilities in programming, it contains several libraries and programmers willing to assist in problem solving. Also by the range of users, which represent a market with greater scope in all the world-wide territory.

•

Firebase: This tool provided by Google makes it possible to use the integrated notification system on the Android platform. Using the Firebase Cloud Messaging service, or FCM, it is possible to notify an application on the client device with information read from a database. Such notifications can be received by the client device from anywhere, provided that the client is connected to the Internet.

General Scope of the system.

The data stored on this server can be accessed on demand by an application developed for the Android platform, which can be run on smartphones and tablets. The android has become the most used operating system worldwide, according to the website Static Counter [11], corresponding to 37.93 % of all devices running any operating system. The Android platform was chosen for its successfully, which corresponds to 81.7 % of smartphones and 66.2 % of tablets worldwide, according to the website Statista [12]. This data can also be accessed by a Java notification service Firebase, maintained by Google, which has the function of sending messages to Android devices registered on the platform. These messages are asynchronous and do not require the user to be running the app to receive such information. For a better operation of the platform, the user can register using the Android device a Range than would be the normal data for reading the sensors, any data considered abnormal is notified to the Android device by the Firebase Notification server, as being a risk situation . An example of a parameter is the body temperature being greater than 37.5 degrees (fever) or a heart beat less than 50 (cardiac arrest).

Once the overall scope of the system was defined, the next step was to design the abdominal belt. Among the challenges is the choice of lightweight, comfortable, low-cost and low battery consumption materials. The Figure 2 show the elements of hardware present in the abdominal band.

The following are detailed about the components that are part of the system overview. •

Wearable Computer: A solution found to capture the data by sensors was the development of an abdominal belt. This range contains the temperature, beats and position sensors (accelerometer) which are coupled in strategic places to better measure such data. Also found is a micro-controller that connects Wearable Computer to the Internet via a WI-FI router. This micro-controller is powered by a battery. In the sequence more details of this belt will be presented.

•

Wi-Fi Router: A Wi-Fi router allows a connection to the wireless Internet. This type of connection is present in a large number of homes around the world. This device allows the micro-controller, like other home appliances, to connect to the Internet and transfer data in virtually real time.

•

Ubidots Server: For the exchange of information between the microcontroller and a remote monitoring

Fig. 2.

Components of the abdominal belt.

The location where the belt is dressed in the baby takes into account the objectives of the project, being defined the position of the sensors based on the information that will be measured. For the accelerometer sensor, the best position is in the middle of the baby’s chest, to be able to accurately check the movement on the three axes (x, y and z), easily identifying the most subtle movements. The heart rate sensor was on the left side of the chest, because of the proximity of the heart.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

The temperature sensor is on the side of the chest, below the armpit. The following is a quick explanation of each component that makes up the range as well as the reason for his choice: •

•

Abdominal belt: The abdominal band was developed using Neoprene material, which is a high quality, comfortable and flexible synthetic rubber. It is a highly resistant material and has great malleability. In addition, it offers elasticity in all directions and provides a perfect modeling to the body, besides not absorbing water. Slots in the material were created to improve the breathing of the baby’s skin. Temperature Sensor: As a temperature sensor, the DS18B20 sensor is used, which can measure temperatures in the range of -55o C to + 125o C. This sensor is very precise and has characteristics such as low cost and small size, which for the purpose of the project justifies its use.

•

Photoplethysmograph Sensor: This sensor is responsible for measuring heart rate. The sensor used is the pulse sensor which, when it detects a pulse in the heart, it is able to reproduce it in wave form, sending it as electrical signal signal to the microcontroller.

•

MPU 6050: The integrated module contains an accelerometer, a gyroscope and a temperature sensor. The accelerometer is used to identify when a movement occurs, and its principle of operation is based on measuring the acceleration of the object itself, through internal calculations. It also makes it possible to identify the positioning of the sensor in relation to a preset reference. The module also provides a temperature signal which is used to measure the ambient temperature, and a gyroscope signal.

•

Battery: For prototype, a rechargeable lithium ion battery was used, which is not expensive. Called Power Bank from Sony company, with power capacity is approximately 3000mAh. This battery has a larger size than those found in the market, but for prototyping purposes responds well to the desired.

•

Microcontroller and Communication Module: The company Particle IO has a tool named Spark core, whose usefulness is the prototyping in the Internet of Things area. This kit consists of an ARM Cortex M3 micro controller and a SimpleLinkTM CC3000 Wi-fi module that enable data acquisition and transmission over the Internet through the Wi-Fi 802.11 b/g protocol. IV.

R ESULTS

As a result of this work, we have a prototype Wearable Computer containing temperature, position, and heartbeat sensors, all integrated inside a Neoprene belt with is fits to improve a baby’s thermal sensation. These connected with a micro-controller access the Internet through Wi-Fi, powered by a battery, and are sent to an Internet server. The belt weighs approximately 200 grams, the microcontroller itself, being coupled to the Wi-Fi card, measures

35

around 3 centimeters, which together with the sensors contained in the range provide a good size for the application. The belt is made of a comfortable material, which makes its prolonged use possible. One of the critical points of the Designed Wearable Computer is the battery, since lighter batteries usually have a very limited range, and larger batteries are large and uncomfortable to be worn by a baby. In order for the user to have access to the developed Wearable Computer data, an application in the Android Platform was design, in order to read the current data accurate by the sensors, these are present in the Ubidots server, as well as verify a history of reading this data, these are displayed in a graphic. The figure 3 displays the application’s home screen, which contains all the important information to avoid browsing between screens. The instantaneous values of the internal temperature, which represents the temperature of the baby, the external temperature representing the temperature of the environment, the amount of heart rate per minute and the position of the baby are displayed on the screen. Also in graph form the reading history is displayed. This graph is dynamic and interactive, and the user can click and drag to see older values. A verification of errors in the system can be implemented, this verification makes it possible to determine if the readings should occur, they occurred. It is very useful because often the system performs incorrect readings, and the parameters passed to the application appear to be correct, however there may be an inconsistency in this data, generating a false interpretation by the user. Also a parameter screen has been developed, it can be accessed by the user through a menu. In this screen, you can register critical parameters, such as in which situations the user would like to receive notification. As an example, we can cite body temperature above or below a range considered as normal temperature, the same for heart rate. The user can also be informed if the baby moves in the crib (preventing a newborn from falling over, for example), or information such as room temperature too high or too low, or Wearable Computer Critical / Wearable battery Computer without sending information to the server in a longer time than the registered, so the user knows that the connection as Wearable Computer has been lost. The figure 4 displays an incoming notification. This message identifies the sender with the name of the application as well as the text of the problem (in the example, Non-standard heartbeats). By clicking on the message, the user is directed to the main screen of the application. Upon receiving the notification, in addition to the sound effect, a vibration effect of the device is also performed. The tests were carried out in a controlled environment, in an adult human being, and in a baby. They demonstrated the operation of the device and provided an analysis of what should be improved or improved for greater usability. The battery has a system life of approximately 20 hours, and its

ISBN: 1-60132-459-6, CSREA Press ©

36

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Fig. 3.

Main Screen of the application in a Android Device.

battery life is approximately 4 to 5 hours. It was noticed the need to be implemented a system to check the charge of this battery, warnings of the state of this charge and the needed to recharge it. The purpose of the work was to develop a wearable computer in which a baby’s vital signs were conditioned and transmitted to an android device. Everything was made using low-cost hardware as well as free software. Larger, more accurate tests will be performed in the future, since the first part of the experiments, part of this more technical, operated as expected. The system was designed to be as dynamic as possible, and the architecture was validated proving that it is possible to develop it, as well as adding more sensors to it. More sensors would enable a check of extra vital signs, or even sounds and an abnormality check.

Fig. 4.

Screen upon receipt of a notification after moving the belt.

V.

C ONCLUSION

Smart sock 2 [14] is the name of a device found on the market with similar functions. It is a sock capable of measuring the heart rate, blood oxygenation rate, and transmitting this data to a celualr application. However its price is around 299 dollars. This system, made with free hardware / software has a cost of approximately 40 dollars, and adds more vital signs in the measurement This work has integrated free/open-source hardware and software to help prevent a problem that affects thousands of children every year: sudden death, as well as helping parents monitoring vital newborn data, and also inform the user about the environment where the baby is and about the operation of Wearable Computer itself. The results of this study could be used by parents who

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

are concerned about the health status of their children, health professionals who wish to keep a register of physiological parameters about their patients in order to study and decides what is best. The project was developed in order to facilitate the integration of new hardware resources such as sensors, facilitating their presentation and monitoring by the application.

37

[5]

[6] [7] [8]

In addition to submitting the data sent by Wearable Computer, there is the possibility of notification, if any data is not in the normality standards registered by the user. During development, some limitations have been encountered and can be circumvented in future releases, as the limitations of the free version of Ubidots server (in the free version the number of users and sensors is limited). For a commercial version, one should use the paid version of Ubidots. Another limitation is the size of the battery used, for a commercial version, one should study the possibility of a smaller battery, even at a higher cost, for better comfort of the baby. Another improvement that must be deployed on the platform concerns security and fault tolerance mechanisms, preventing Wearable Computer from stopping sending data to the server and sending this data in encrypted form.

[9] [10] [11] [12] [13]

[14]

B. B. Randall, D. S. Paterson, E. A. Haas, K. G. Broadbelt, J. R. Duncan, O. J. Mena, H. F. Krous, F. L. Trachtenberg, and H. C. Kinney. Potential Asphyxia and Brainstem Abnormalities in Sudden and Unexpected Death in Infants. Pediatrics, 132(6):e1616–e1625, 2013. World Health Organization. Under-five mortality, 2016. Center for Disease Control and Prevention. Sudden Unexpected Infant Death and Sudden Infant Death Syndrome, 2015. Deborah Carvalho Malta, Elizabeth Franc¸a, Daisy Xavier De Abreu, H´elio De Oliveira, Rosane Aparecida Monteiro, Luciana M. V. Sardinha, Elisabete Carmen Duarte, and Gulnar Azevedo E Silva. Atualizac¸a˜ o da lista de causas de mortes evit´aveis (5 a 74 anos de idade) ´ por intervenc¸o˜ es do Sistema Unico de Sa´ude do Brasil. Epidemiologia e Servic¸os de Sa´ude, 20(3):409–412, 2011. SISTEMA UNICO DE SAUDE. Painel de Monitoramento da Mortalidade Infantil e Fetal. 2016. Dra Magda and Lahorgue Nunes. S´ındrome da Morte S´ubita do Lactente Epidemiol´ogicos , Fisiopatologia e Prevenc¸a˜ o. Medicina, 2005. [Online]. StatCounter GlobalStats, 2017. [Online]. Statista, 2017. R. Pereira. Socially aware computing - computac¸a˜ o socialmente consciente. In http://web.inf.ufpr.br/rpereira/courses/ci751-interacaohumano-computador/arquivos/socially-aware-computing/. Acessado em 26 de novembro de 2016., nov 2016. Owletcare. owletcare, 2017.

However, these limitations did not impede the testing and validation of the architecture. The acquisition of data using the sensors was implemented through a micro-controller, which in turn allowed the transmission of data to the server. The development of the application took into account factors such as usability, dynamic and enabled the visualization of the vital signs to read in the sensors. Also for the development of the abdominal belt were used light and small, that low energy consumption designed to IoT. As a increase of the project, it is desired to add more sensors, such as blood pressure and oxygenation rate of the blood, increasing the usability of the device and opening a wider range in the monitoring of vital signs based on normality standards. Also find a solution for the authentication of users, allowing greater privacy in the data recorded on the server and also a provision of this data on different platforms, such as an Internet site for example, so that the data can also be monitored by a health professional. The innovation proposed in this work may help the field of the monitoring of vital signs in newborns, it is a cheap solution that can be used by parents and health professionals/universities to increase the amount of information for the study of sudden death. R EFERENCES [1]

Alexandre Renato, Rodrigues De Souza, Francisco Cesar, and Campbell Mesquitta. Monitoramento de Sinais Vitais. 2(1):1–32, 2013. [2] Andr´e Luis Macedo, Marcelo Oliveira De Moraes, Tha´ıs Almeida, and Brito Fernandes. Computac¸a˜ o Ubiqua e Computac¸a˜ o Verde, 2007. [3] K. Van Laerhoven, A. Schmidt, and H.-W. Gellersen. Multi-sensor context aware clothing. Proceedings. Sixth International Symposium on Wearable Computers,, pages 49–56, 2002. [4] J. Melorose, R. Perroy, and S. Careas. Definition of ”Wearable Computer”. Statewide Agricultural Land Use Baseline 2015, 1, 2015.

ISBN: 1-60132-459-6, CSREA Press ©

38

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Smartphones application in Ophthalmology - Keratoconus detection Rafael S. Ishibe1 , Maximiliam Luppe1 , and Jean Jacques Groote2 1 Dept. of Eletrical Eng. and Computing, University of São Paulo, São Carlos, SP, Brazil 2 Moura Lacerda Center University, Ribeirão Preto, SP, Brazil Abstract— This article describes the development of a smartphone application with the goal to detect and analyze the keratoconus using images captured by a keratometer based on LED ring. This application detects LED reflection in the patient eye and calculates the curvature around the center of the cornea using an interpolation algorithm. The reliability was verified using high degree of aberration images and ellipses images with known parameters to test the software. Comparing real parameters of the simulated images with the obtained by the software the absolute errors are very small. For images of a data bank the software detects all LEDs images and calculates the ellipse axes and inclination angle. For very high degree of aberration the detection is problematic. Keywords: Smartphone, keratometer, android, mobile, healthcare

1. Introduction Researchers and professionals that work with medical ophthalmic instruments are improving methods and measurement techniques in optical systems, data analysis, diseases diagnostics and improving the surgery procedures on patients [1]. New equipment for improving health care does not always need sophisticated and expensive computer systems, but inexpensive ones that most of us have in our pocket: the smartphones [4]. The smartphones have become integrated into out lives. The devices have evolved incredibly during the last years, an evolution that can contribute to develop applications in the ophthalmology medical field [2]. Smartphones have many sensors embedded on such as cameras, GPS, gyroscope, accelerometer and internet access, allowing the development of sophisticated applications as the processor are getting faster. However, developing real time application is still a challenge. Researchers and professionals are developing library’s with fast algorithm to improve the processing time [3], an example is the image processing that has various types of applications, but it’s extremely heavy to process. An application that use the image processing on the ophthalmology field is the keratometer equipment. The Troutman Keratometer is an equipment to measure human cornea anterior curvature. This equipment considers the human eye as a specular convex mirror; a ring with LEDs project light into patient eye, a camera capture the image of the light reflected from the eye and a software calculates the curvature of the cornea at specific directions. The keratometer used during the tests has 36 LEDs.

2. Method and Goal The project goal is to develop a smartphone application to detect keratoconus in patients. The software is implemented in Android Operational System using the JAVA platform, with OpenCV library to manipulate images data. The proposal is to develop an application to be used on remote area to collect information and transmit it to a doctor or specialist to diagnose the problem.The program identifies and locate the keratometer LED’s projection, performs the Lagrange interpolation, and send the processed data to be analyzed.

2.1 Lagrange Interpolation Lagrange interpolation is in most cases the method of choice for dealing with polynomial [5]. Given a set of n + 1 distinct interpolation points kj , j = 0, ..., n, together with corresponding numbers fj . Assuming that the nodes are Q real, let n denote the vector space of all polynomials of degree at most n. The Q problem addressed is that of finding the polynomial p  n that interpolates f at the points xj [5]. The formula finds a continuously solution with the data captured in the image. The Lagrange form [6] are: Qn n X k=0,k6=j (x − xk ) (1) p(x) = fj lj , lj = Qn k=0,k6=j (xj − xk ) j=0 This interpolation algorithm is used to calculate the major axis, minor axis, angle and dioptre of the circle image projected onto the patient eye.

2.2 Image Processing After capturing the image projected onto the patient eye using the LEDs keratometer [Fig. 1], an 7x7 pixels mask is used to find the area where the sum of intensity have higher value, this area is the more probable place of a LED projection. This area is used as a new mask, and have twice the size as the preview mask. The new mask is used to find the other LEDs projection. A convolution with the new mask is made, to calculate the difference between the new mask and each region in the image. The absolute value of this difference and the position of each area are stored in a list sorted in descending order of intensity. Then, the last N elements on the list are selected after verifying if these elements are not too close to each other in order to avoid overlapping. The next step is to apply the Lagrange interpolation using the points found to calculate the segment between two points in polar coordinates.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Fig. 1: Image with projected LEDs.

Lagrange interpolation used two posterior points and two anterior to calculate the segment. After finding the segment the program calculates the major axis, minor axis and the angle. The program is tested using images with all the parameter known, an absolute error is calculated comparing the parameters obtained by the program and the real ellipse parameters, furthermore a set of irregular image is used to analyze the reliability. The data compared are the ellipse angle, major axis and minor axis, the images with known parameters have 480 X 360 resolution, the major axis length is between 100 and 150 pixels, the minor axis has length superior than 100 and lesser or equal than the major axis and the angle is between 0 and 170.

3. Result The parameters obtained by the software developed were compared with 54 images with known parameters. The major axis and minor axis error was below 4 pixels in both cases. Analyzing the angle data, it was observed that when the minor axis is closer to the major axis, higher is the angle error calculation. Images with higher degree of aberration also results in difficulties to detection of the LEDs projections. This leads to mismatch’s in the LEDs projection detection and Lagrange interpolation. Images with higher degree of aberration the detection is problematic, the program developed does not find the LEDs projection correctly, because the images has too much noise. Since the LEDs detection is problematic the interpolation is incorrectly.

39

Fig. 2: Detection of 36 LEDs and Lagrange Interpolation.

4. Conclusions The project is still in development, but the results found in the tests were satisfactory, the algorithm locates the LEDs reflected and the Lagrange interpolation implemented is precise enough to avoid significant errors in case of missing LEDs. However, images with high cases of eye aberration, the detection still have some problem to locate and calculate the segment between two points, this aberration can cause overlapping LEDs projection or the image has too much noise. Points to be worked in the future are, implement an algorithm to find an ellipse with the points calculated with the Lagrange interpolation, reduce the processing time, improve the algorithm, and to create a database in cloud to store the information captured in the field. This database should store basic information about the patient as, eye image and the parameters major and minor axes, dioptre, inclination angle, them allow doctors to access the data and make a diagnosis. Other improvements for this work are to use the keratometer ring with continuous light and to develop a corneal topographer for the equipment to map the curvature in a large area of the cornea.

References [1] Liliane Ventura, “Corneal astigmatism measuring module for slit lamps” , Physics in Medicine and Biology, vol. 51, pp. 3085–3098, 2006. [2] R. K. Lord, V. A. Shah, A. N. San Filippo, R. Krishna, “Novel Uses of Smartphones in Ophthalmology”, American Academy of Ophthalmology, Ophthalmology, vol. 117, pp. 1274–1274, e3, 2010. [3] C. M. Iyomasa, L. Ventura, J. J. De Groote, “Software for keratometry measurements using portable devices”, BiOS, International Society for Optics and Photonics, pp. 75502B-75502B, Feb 2010. [4] A. Kailas, C. C. Chong, F. Watanabe. “From mobile phones to personal wellness dashboards.”,IEEE pulse 1.1 pp. 57–63, 2010.

ISBN: 1-60132-459-6, CSREA Press ©

40

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[5] Berrut, Jean-Paul, and Lloyd N. Trefethen. “Barycentric lagrange interpolation.”, SIAM review 46.3 pp. 501-517, 2004. [6] J. L. Lagrange, Leçons élémentaires sur les mathématiques données à l’École Normale en 1795. ID., Oeuvres completes, a cura di J.A. Serret, Gauthier-Villars, Paris, vol. 7, pp. 271–271, 1867.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

41

IoT Support for Dementia Patients Short Research Paper

Bernd Müller Faculty of Computer Science, Ostfalia University, Wolfenbüttel, Germany

Abstract— We report on the preliminary findings of an ongoing research project aimed to detect departure of dementia patients of care departments of hospitals or residential care homes for the elderly. The project is based on mobile beacons and some IT infrastructure to detect moving beacons respectivly their carriers. The motivation behind the project is to verify whether usage of standard IoT hardware can reduce costs of health systems. Keywords: Beacon, dementia, patient monitoring, Java, BLE

1. Introduction Beacons are small IoT devices sending simple identification messages. Location based indoor services are possible based on this beacon signals. The architecture of such location based services mainly uses stationary beacons and mobile devices to detect the beacon signals to compute the location of the mobile device. In the project we use mobile beacons weared by dementia patients to detect whether they depart from their area of care. In section 2 we introduce beacons. Section 3 gives an overview of the project, namely the possibilities of location determination, used hardware like Raspberry PI and mobile beacons and implementation characteristics for Linux and Java based BLE access. We clonclude with an honest assessment of the project results so far.

compared to iBeacons but for the purpose of this research roughly the same. In opposite to iBeacon Eddystone is not patent-registered and open source. Because of the predicted commercial possibilities in the last years beacon manufacturer spring up like mushrooms. In our projects we are using beacons from different manufacturers, namely Estimote [3], Blukii [4], Beaconinside [5] and Onyx Beacon [6] to name just a few. The mentioned commercial possibilities and use cases are unlimited for example in retail, advertising, logistics, event management. Usually, information about stationary beacons are received by mobile devices and used for locating the device respectivly the device owner. Beside usual retail, merchandising and advertising applications in shopping malls some very big installations get considerable public interest, for example the Super Bowl 50 or some German out-ofhome and online advertising company. At the Super Bowl 50 [7] 2000 beacons were installed throughout the stadium to enable an app to help people to find their way through the stadium to their seats, to find the nearest bathroom or to order hot dogs. Ströer, one of the leading providers of out-of-home and online advertising in Germany announced to install 50,000 beacons [8] mainly at airports and railroad stations to enable location based and personalized advertising in this locations.

3. The Project

2. Beacons Basics

3.1 Wandering of Dementia Patients

iBeacon is a patent-registered protocol and also a trademark introduced 2013 by Apple Inc. for indoor positioning [1]. It is based on Bluetooth Low Energy (BLE) and mainly consists of a universally unique identifier (UUID), sent repeatedly. An indoor positioning system with information about the location of different beacons and received beacon signals of a mobile device can compute the position of the mobile device. Usually, this is done with trilateration algorithms. BLE is supported on iOS 7 and later and on Android devices since version 4.3 (API 18). Because location based services based on GPS are very common and commercially successful the corresponding indoor alternative based on beacons seems to be a field of commercial profit, too. It is not astonishing that Google invented a competitive protocol named Eddystone [2], also a trademark in 2015. It is a slightly enhanced protocol

Wandering of dementia patients is a general risk in geriatric as well as medical care. Because of loss of appraisal capabilities of dementia patients with respect to risks for themselves as well as risks for others dementia patients schould not leave the area of care. Solutions to prevent unwanted department of dementia patients exists already. They are primarily based on GPS [9] for outdoor or proprietary radio technology for indoor. Low priced and mainstream IoT based solutions are still missing.

3.2 Municipal Hospital and University Cooperation The municipal hospital of Wolfenbüttel, Germany [10] operates a care department where some portion of patients suffer from dementia. The municipal hospital in cooperation with Ostfalia University works on a beacon based system to

ISBN: 1-60132-459-6, CSREA Press ©

42

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

common room

Fig. 1: Footprint of Hospital Floor

detect unwanted department of dementia patients on which we will report on in the following.

LE. In some sense the Raspberry PI takes the part of mobile devices in case of alternatives 1 and 2 of section 3.3.

3.3 Beacon Detection and Location Alternatives

3.6 BLE and Java

Usually, stationary beacons and mobile devices receiving beacon signals are used for the location based systems mentioned in section 2. However, there are of course different alternatives to build systems to track persons at least partially: 1) fixed beacons and approaching mobile device detection 2) fixed beacons and leaving mobile device detection 3) mobile beacons and fixed BLE receivers Because of some building constraints of the respective hospital unit and of course no affinity of elderly and demented people to mobile devices the third alternative was used.

3.4 Building Constraints The footprint of the respective hospital unit floor is depicted in figure 1. All patent rooms are solid brick-built. Only the corridor has a suspended ceiling. WiFi is not available inside the hospital. Thefore the only possible way to install some beacon based location system without constructional modifications was the third alternativ of section 3.3. In figure 1 the corridor is marked with three cricles near the staircases. These marks indicate the location of three Raspberry PIs which are connected via Ethernet to the hospital network.

At the moment there is no Java support for Bluetooth. Therefore we are using some workaround. To access BLE on Linux you have to use some GATT (Generic Attributes [12]) compatible software usually BlueZ [13]. With the help of the programs hciconfig, hcitool and hcidump a file is created with all received Bluetooth signals appended to the end. With Java’s Watch Service, introduced in Java 7, the lines written by BlueZ are read into our beacon detector software.

3.7 The mobile Beacon Beside the normal beacons there are also mobile versions available. We use one from Blukii [4] which mimics a wristwatch in the hope that demented people resist to remove it. The beacon is shown in figure 2.

3.5 Raspberry PI and BLE The Raspberry PI [11] is a small but powerfull computer running a Debian based Linux. Model 2 has no Bluetooth device build in. In order to use Bluetooth some Bluetooth dongle has to be used. We started with this configuration but moved to Model 3 after market launch. Model 3 has Bluetooth built-in support. The same holds true for Bluetooth

Fig. 2: Blukii Mobile Beacon

3.8 Server Application The Java applications running on the Raspberry PIs and mentioned in section 3.6 detect beacon signals in the file

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

written by BlueZ. If it is a beacon to observe this information is send to an Java EE application via REST. The application server we use is WildFly 10. Depending on the kind of signal (approaching/leaving) client software at the nurses’ station issues an alarm.

4. Conclusions The detection of mobile beacons with Raspberry PI computers is very reliable. However, because of some geometric constraints in the Hospital it is not possible to determine if the received signal is by a patient simply moving around or trying to leave the care department. We reported on a ongoing research project. The interim result is negative because it was not possible to detect departure of dementia patients in a relieble way. In the next step we will investigate on how to install Raspberry PI computers inside each patient room and detect departure of patients by geofencing algorithms, i.e. the 2 alternative of section 3.3.

43

References [1] [2] [3] [4] [5] [6] [7]

iBeacon. https://en.wikipedia.org/wiki/IBeacon Eddystone. https://developers.google.com/beacons/eddystone Estimote. http://estimote.com/ Blukii. http://www.blukii.com/index_en.html Beacon Inside. https://www.beaconinside.com/ Onyx Beacon. http://www.onyxbeacon.com/ Super Bowl 50 Has a Sweet App and Crazy-Fast Internet. URL: goo.gl/o7a0eo [8] Nationwide beacon infrastructure in Germany. URL: goo.gl/N05vhu [9] 10 Lifesaving Location Devices for Dementia Patients. URL: goo.gl/gS9dJ5 [10] http://www.klinikum-wolfenbuettel.de/ (in german) [11] https://www.raspberrypi.org/ [12] Generic Attributes (GATT) and the Generic Attribute Profile. URL: goo.gl/9KCDTs [13] http://www.bluez.org/

ISBN: 1-60132-459-6, CSREA Press ©

44

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Undergraduate Experience Developing a Medication Reminder App Suhair Amer and Steven Sebastian Department of Computer Science Southeast Missouri State University, Cape Girardeau MO 63701 Abstract – This paper describes the process of implementing and testing a taking medication reminder app. The medication app utilizes a simple interface that is easy for the audience, mostly of elderly. Functions to enter and change medications are easy and quick to use. This project was completed for an undergraduate Human Computer Interaction course taught at our institution. The student had around one month to design, implement and evaluate the system.

audience to navigate and use. Functions to enter and change medications should be quick to use. Examples of usability goals are: • The medication app should include functions to store a medication name, dosage, and time to take medication. • Functions of the buttons on the medication app should be obvious to the user. • The buttons on the medication app should only require one click to activate functions. • Movement between functions on the medication app should flow logically (i.e., buttons should be arranged, based on a logical order of medication app functions)

Keywords: medication app, reminder

1 Introductions Not taking medication when scheduled reduces the effectiveness of a treatment and imposes a financial burden on health care systems [2][11]. In the USA alone, it is estimated that $100 billion each year is spent including the cost of hospital and nursing home admissions [9]. Even motivated people can forget which results in unintentional non-adherence [8]. For example, around one million unwanted pregnancies each year results from nonadherence [6] and forgetfulness [3][7]. One trend focuses on reminders, which alerts people to take their medication at a specified time [1][10]. However, it is important to understand that time-based tasks are more difficult to remember than tasks related to routine actions [5] and that medication regimens can be easily incorporated into a daily routine. This support can be provided by technology. With the increasing popularity of smartphones, people now have access to thousands of health-related applications (“apps”) [4] that could help them remember their medication.

2 Analysis and Design In an undergraduate human computer interaction course, the student developed an app that reminds users, mostly elderly, to take their medications. The medication app should utilize a simple interface that is easy for the

Examples of use experience goals are: • The medication app should not frustrate the user. • The medication app should not illicit a negative response from utilizing it (for instance, make the user feel stupid or patronized by the app). With regard to the Conceptual Model, the medication reminder app is like a specialized digital planner (specialized to track medication only). Dates for medications could be arranged in a calendar-like format, with medications displayed for certain times for certain days on the calendar. Clicking on a certain day would display the times of all medications needed to be taken on that day. The system checks the times entered for the current day using the device’s timer. If the times match, the medication reminder app displays a reminder to the user (either for each medication at that time or one reminder for all medications needed to be taken at one time). Times for medications entered into the app would be organized by day rather than month (the interface would show all the medications for a particular day instead of a particular month).

3 Implementation 3.1.

Consider interface design issues

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

The app needs to maintain access to the device’s timer. Therefore, it will be running at all times in the background). In addition, to make the app effective, it should be installed on a device that the user will be using frequently. Therefore, the ideal interface involves the use of the Mobile and Multimedia interface types. While WIMP can be used, this type suggests that the device would be a desktop or laptop computer, which may not be as readily accessible as a mobile device.

45

obtain information about specific problems that participants noticed), and questionnaire (to obtain quantitative data about the effectiveness of the medication app and get the general consensus on app effectiveness). For observation tasks, The participant will enter a medication approximately 2 minutes after the current time. The participant will modify the added medication to change the dosage to 2 pills instead of 1. The participant will click on the ‘taken medication’ button for the modified medication. The participant will add another medication and then remove it.

• • •

3.2.

Implementation related issues

The system is implemented using C#. Rather than storing dosage as mg (text input), it was stored as number of pills to avoid confusion and mistakes. The app will only show the medications that are currently being taken although more medications are stored.

4 Evaluation and results 4.1.

Determine evaluation goals

The focus of the evaluation is to determine the medication app’s ease of use. The goals of the evaluation are the following: • Find out whether medications in the medication app can be entered, modified, removed, and used easily by the users. • Find out whether the functions of the medication app are easily understood from the interface. • The evaluation should assess the overall usability of the medication app.

•

With regard to the interview component, The participant will be asked to recommendations to improve the app.

•

•

•

Identify the practical issues

Explore the questions

From evaluation goals, the following questions can be derived: • Does the user understand what a function will do just from reading button names? • Does the user have any difficulties with the processes of entering, modifying, or removing a medication? • Does the app prevent duplicate entries or otherwise invalid entries? • Does the app update to the next time correctly after taking medication? • Is the correct information displayed?

4.3.

any

For the questionnaire component, The participant will select his/her age group, gender, and whether the participant has used a medication reminder app before. The participant will assess the ease of understanding the different interfaces (main interface, add medication interface, modify medication interface), selecting one option out of five (strongly disagree, disagree, in between, agree, strongly agree). The participant will assess the functions of the medication app for ease of use and understanding, selecting one option out of five (strongly disagree, disagree, in between, agree, strongly agree).

•

4.4. 4.2.

offer

Choose the evaluation method

The evaluation method appropriate for the goals and questions listed above would be a usability test. Evaluation methods used will be observation (to test if the functions are usable and easy to understand), interview (to

Practical issues with the evaluation mainly involve the time allowed for the evaluation and the ideal participants for this type of evaluation. The app mainly focuses on the elderly as the primary audience. However, with only a week to conduct the evaluation, participants will not be limited to the elderly for the study. Additionally, an ‘in the wild’ evaluation method would have been unrealistic to implement. Therefore, the evaluation will have to be conducted in a controlled environment.

4.5.

Decide how to deal with ethical issues

The main ethical issue is related to ensuring that the data collected could identify participants and jeopardize their careers. If the information collected could identify a participant, an informed consent form is completed before we can conduct the evaluation legally. However, the data collected from the observation (mainly, if the participant encountered problems), interview (mainly, recommendations to improve the medication app), and

ISBN: 1-60132-459-6, CSREA Press ©

46

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

questionnaire (age group, gender, experience with medication reminder apps, etc.) does not identify a participant. Therefore, an informed consent form is not required for this particular evaluation. The ethical issue is dealt with by ensuring that collected information cannot be used to identify a participant. In addition, to avoid the problem of damaging someone’s device, testing will be performed using the developers’ own device.

4.6.

Running the system

Figures 1, 2, and 3 show screen shots of the interface of the app. In the main form (interface), the list of medications is displayed in a text box. Four buttons are available for adding a medication, modifying a medication, removing a medication and “I have taken this medication”. In the add medication form (interface) the user will enter the name of medication, number of pills, how many times, start date and start time. Once finished, the user can submit the changes or cancel. The start date is in calendar format where the user can select the day, date and year.

Figure 3 MedAlarm Modify Medication Form Interface

4.7.

Evaluate the collected data

The participants varied widely by age group, and were evenly distributed by gender. The data obtained from the evaluation relates to the usability of the app such as problems with reading text on the app or functions not working as intended, and all data collected points to the goals of the evaluation. Therefore, the evaluation method seems to be valid. As the environment used was the evaluator’s computer in a controlled setting, this method does have low ecological validity. The evaluator was in charge of recording problems that users experienced with the medication app, so there may be some bias in the data collected through this evaluation method.

Figure 1: MedAlarm Main Form Interface

The scope of the evaluation was concerned with the usability of the app, in terms of its interfaces and functions.

4.8.

Analyze the data

Most of the ratings for the interfaces and functions were in the ‘Strongly Agree’ category for ease of use and understanding, so, despite glitches with certain functions and issues with reading text, the medication app is generally easy to use and understand. These ratings do not seem to be skewed by gender, as the participants were evenly distributed by gender. The participants were somewhat weighted in the 20-30 age group and elderly age groups (6070 and 70+). However, the distribution is even overall.

Figure 2: MedAlarm Add Medication Form Interface

Problems observed by the participants include: time glitch in the ‘Modify Medication’ function, font size being too small for particular portions of the app, notification message too large for high numbers of

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

medications, list update function interfering with other functions on the app, etc. However, the most consistently noticed problem was the aforementioned time glitch. This suggests that the time glitch could cause considerable losses in usability (one participant even pointed out that it was not easily noticeable) and should be resolved (high priority).

47

• • •

4.9.

Interpret the data

Judging from the high concentration of results in the positive categories for interface and functions being easy to use and understand, the medication app is, overall, easy to use and understand. However, participants have also noted problems with the app itself (like the time glitch, small font size, list update function interfering with other functions, etc.). The problems observed, apart from the time glitch and list update function overriding other functions, could be resolved with little modification to the code. Most errors within the program seemed to be minor and did not seriously detract from the usability of the app. While the problems certainly need to be fixed, the medication app still seems to retain the impression that it is easy to use and understand.

• •

•

•

4.10.

Present the data

The age groups of the participants varied widely: three participants were in the 20-30 group, one participant was in the 30-40 group, one participant was in the 40-50 group, one participant was in the 50-60 group, two participants were in the 60-70 group, and two participants were in the 70+ group. Given that the interfaces and functions generally received positive ratings for ease of use and understanding, this suggests that the medication app is easy to use, virtually regardless of age group. Genders were split evenly with collected data (five males, five females), suggesting that gender does not impact understanding how to use the medication app. Ten subjects were asked to participate in evaluating the system. The following is a summary to the observation component of collecting the data: • The participant added a medication for the next day and noticed that the medication was not displayed. As the app was supposed to show medications to take in the current day, this was normal. • When the participant changed the time of the medication in the ‘Modify Medication’ function, the time moved to the next day, even though the current day would have been valid (this is not normal behavior). • The participant noticed that the ‘Remove Medication’ function prompts the user before removing a medication.

•

•

• •

The participant kept pressing ‘Delete’ to “remove” the incorrect elements of the time and add the correct elements of the time. Deleting was not necessary. The participant noticed that the time moved to the next day when modified (when it should not have) and asked why the app did that. The participant had minor difficulties reading the text in many parts of the interface, including elements of the “list of medications to take today” and the text box that displays messages to the user. The participant noticed the aforementioned time glitch noticed by the other two participants. The participant noticed that the function sometimes did not activate when selected and was confused. After explaining that the function did not run because the list was updating the time until time to take, the user wondered if this feature in the list is necessary. The participant noticed that the app did not notify the user when a particular medication was supposed to be taken. After being told that the app was not running at the time and, thus, could not notify the user to take a medication, the participant wondered why the app had to be running at all times to function properly. The participant entered multiple medications with the same time to take. When the time to take arrived, the participant noticed that the notification message sent seemed excessively long (10+ medications being displayed one by one). The participant noticed the time glitch in the ‘Modify Medication’ function and suggested that the function should be fixed, so the glitch does not occur (could easily be missed, resulting in the app tracking the medication tomorrow instead of today). The participant understood why the app only showed the medications for the current day, but noted that, if an incorrect medication was entered for a different day than the current day, it couldn’t be removed until that day. The participant liked that the time “until time to take medication” was displayed, but wondered why clicking on a function button did not always call the function. The participant had some difficulty reading the text within the medication list.

The following is a summary to the interview component of collecting the data: • The participant suggested including an additional list that shows all of the medications added into the medication app. • The participant suggested adding decimal dosages (adding dosages by 0.25 of a pill and incrementing by 0.25 when increasing dose). • The participant suggested adding a list that showed all the medications that a user has taken in the current day.

ISBN: 1-60132-459-6, CSREA Press ©

48

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

• •

• •

• • •

•

Several participants noted a glitch in the Modify Medication function where the time moved to the next day when it was not supposed to. The participant suggested increasing the font size on many parts of the interface, including the text box that displays messages, start time, start date, dosage, times to take per day, and medication name (the labels are large enough, though). The participant noted that the value of “the time until time to take” interferes with the other functions of the app. The participant suggested having the program run in the background to ensure that the app can notify users even if no forms are displayed. The user is unlikely to keep the form open at all times. The participant suggested displaying a message to check the list of medications when several medications have to be taken at the same time. The participant suggested fixing the time glitch in the ‘Modify Medication’ function and adding an additional list that shows all medications stored in the app. The participant suggested fixing the app to ensure that clicking on a function button will always call the function (i.e., the list updating function should not override medication app functions). The participant suggested increasing the font size of the text in the medication list.

Figures 4 to 10 summarize the results of the questionnaire component of collecting the data: 8 6 4 2 0

Main Interface Easy To Understand

Figure 4: Questionnaire results regarding if the main interface is easy to understand. 10 8 6 4 2 0

Add Medication Interface Easy To Understand

Figure 5: Questionnaire results regarding if adding medication interface is easy to understand.

8 6 4 2 0

Modify Medication Interface Easy To Understand

Figure 6: Questionnaire results regarding if modify medication interface is easy to understand. 8 6 4 2 0

Add Medication Function Easy To Use and Understand

Figure 7: Questionnaire results regarding if adding medication function is easy to use and understand. 8 6 4 2 0

Modify Medication Function Easy To Use and Understand

Figure 8: Questionnaire results regarding if modifying medication function is easy to user and understand. 10 8 6 4 2 0

ISBN: 1-60132-459-6, CSREA Press ©

Remove Medication Function Easy To Use and Understand

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 9: Questionnaire results regarding removing medication function is easy to use and understand. 8 6 4 2 0

Medication Taken Function Easy To Use and Understand

Figure 10: Questionnaire results regarding if taking medication function is easy to use and understand.

5 Conclusion The student applied the concepts studied in the human computer interaction course to develop a medication reminder app. The student was required to design, implement and evaluate the system within a month. According to the participants that tested (evaluated) the system, the developed medication app was easy to use and to understand.

6 References [1] Haynes, R. B., Ackloo, E., Sahota, N., McDonald, H. P., Yao, X. Interventions for enhancing medication adherence. Cochrane database of systematic reviews, CD000011 (2008). [2] Hughes, D. A., Bagust, A., Haycox, A., and Walley, T. O. M. The impact of non-compliance on the osteffectiveness of pharmaceuticals: a review of the literature. Health Economics, 10, May (2001), 601–615.

49

http://mashable.com/2012/09/26/smartphones-healthcareinfographic/ [5] Park, D. C., & Kidder, D. P. Prospective memory and medication adherence. In M. Brandimonte, G. O. Einstein, & M. A. McDaniel (Eds.), Prospective memory: Theory and applications. Lawrence Erlbaum Associates (1996), 369– 390. [6] Rosenberg, M. J., & Waugh, M. S. Causes and consequences of oral contraceptive non-compliance. American Journal of Obstetrics and Gynecology, 180, 2 Pt 2 (1999), 276–279. [7] Smith, J., Oakley, D. Why Do Women Miss Oral Contraceptive Pills? An Analysis of Women’s SelfDescribed Reasons for Missed Pills, Journal of Midwifery & Women’s Health, 50, 5 (2010), 380–385. [8] Unni, E. J., & Farris, K. B. Unintentional non-adherence and belief in medicines in older adults. Patient education and counseling, 83, 2 (2011), 265–268. [9] Vermeire, E., Hearnshaw, H, Van Royen, P., and Denekens, J. Patient adherence to treatment: three decades of research. A comprehensive review. Journal of clinical pharmacy and therapeutics, 26, 5 (2001),331–342. [10] Vervolet, M., Linn, A. J., van Weert, J. C. M., de Bakker, D. H., Bouvy, M. L., and van Dijk, L., The effectiveness of interventions using electronic reminders to improve adherence to chronic medication: a systematic review of the literature, Journal of the American Medical Informatics Association: JAMIA, 19, 5 (2012), 696–704.

[11]

WHO. Adherence to long-term therapies: Evidence for action. World Health Organization (2003).

[3] Jones, R. K., Darroch, J. E., and Henshaw, S. K. Contraceptive Use Among U.S. Women Having Abortions in 2000-2001. Perspectives on Sexual and Reproductive Health, 34, 6 (2002), 294– 303. [4] Laird, S. (2012). How smartphones are changing healthcare. Retrieved 11/07/2013, from

ISBN: 1-60132-459-6, CSREA Press ©

50

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Digital Filter Array Optimization for Directivity Pattern

S. S. Jarng

Y. J. Kwon

Dept. of Electronics Eng. Chosun University Gwangju, South Korea [email protected]

Hearing Impairment & Rehabilitation Res. Ins. Algorkorea Co. Ltd. Seoul, South Korea [email protected]

D. S. Jarng Dept. of Computer Science Biola University Los Angeles, CA, USA [email protected]

Abstract—Two microphone arrays are used for producing a specific directivity pattern by a time delay between the two microphones. Directional hearing aids are examples of the two microphone arrays application. In this paper, the conventional time delay method was replaced with a digital filter method for directional digital hearing aids. CSR8675 multimedia digital signal processing hybrid was used for the main platform of the suggested digital filter method instead of well-known specific digital hearing aid DSP hybrids. The results of the suggested filter method were well matched with those of the time delay method.

II. PHASE DIFFERENCE BY DIGITAL FILTERS Figure 1 shows the time delay method in the diagram. One front microphone is separated from the other rear microphone with D distance. Both microphones are located in the same directional axis. The output of the rear microphone signal is delayed with T and is subtracted from the output of the front microphone, and the result, x(t), is converted to a digital signal, x[n]. n = integer discrete number.

Keywords-CSR8675; Digital Signal Processing; Digital Filter; Optimization; Directional Hearing Aids; Phase Difference

I.

INTRODUCTION

One of hearing aid main functions is directivity. This implies that voices from some specific direction sound louder than other voices from other directions. The directionality can be realized by arrays of omnidirectional microphones. For example, if two microphones of the same sensitivity are located in the direction of one axis, the directivity can be produced by delaying and subtracting of signals between the two microphone outputs [1]. This conventional time delay method requires a precise time delay and signal subtraction control units. When the distance between two microphones is changed, the delay time should also be changed. The scale of the delay time is some microseconds in unit. Well-known DSP hybrids such as Ezairo5920 or Ezairo7110 are equipped with this time delay module for digital hearing aid application purpose, but they are expensive [2]. CSR8675 Bluetooth (BT) hybrid studied in this paper is cheap and has many Bluetooth features, but it doesn’t have the time delay module between two microphones in stereo channels (right channel and left channel) [3]. In the paper the conventional time delay method was replaced with digital filters for realizing the directivity of CSR8675 hybrid IC chip.

Figure 1. Directionality by time delay method.

Figure 2 shows the digital filter method. Both front and rear microphone outputs are converted to digital signals, x1[n] and x2[n], and they are passed through digital filters, DF1 and DF2, resulting y1[n] and y2[n]. Then y1[n] is subtracted from y2[n], y[n]= y1[n] - y2[n].

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

51

a unit magnitude and 2fT phase difference along the frequency.

Figure 2. Directionality by digital filter method.

Figure 3 shows the 2nd order IIR (Infinite Impulse Response) digital filter layout. There are 5 coefficients; a1, a2, b0, b1, b2. y[n] is computed from x[n] as follows: y[n]=-a1y[n-1]-a2y[n-2]+b0x[n]+b1x[n-1]+b2x[n-2] (1)

Figure 4. Directivity of time delay method, =1(cardioid).

Figure 4 shows the cardioid directivity patterns of the time delay method. Three frequencies, 500Hz(continuous), 2kHz(dotted), 4kHz(dashed), were considered. The maximum sensitivity is figured as 40[dB] while the center is 0[dB]. 0o indicates the front direction while 180o is in the rear. And figure 5 shows the cardioid directivity pattern of the digital filter method. a11, a12, b10, b11, b12 and a21, a22, b20, b21, b22 are optimally computed as follows: 1.0, a11= 1.2962, a12= 0.3496, a10= 1 -1.0 b 0=-0.3496, b11=-1.2962, b12=

and

1.0, a21= 0.8770, a22=0.1012, a20= -1.0 b20=-0.1012, b21=-0.8770, b22= Figure 3. 2nd order IIR digital filter layout.

Digital signal filtered by either DF1 or DF2 changes in its magnitude and phase response as a function of frequency depending on the values of the filter coefficients. If DF1 and DF2 are optimally designed so as to produce the same magnitude response, but the constant phase difference response between DF1 and DF2, the resulting y[n] in figure 2 should be the same as x[n] in figure 1.

Figure 4 and figure 5 show almost the same cardioid directivity patterns with different methods. The digital filters are noticed as symmetric, that is, a0=-b2, a1=-b1, a2=-b0. Therefore the filter coefficients’ optimization can be done with only 4 parameters instead of 10.

There are 10 digital filter coefficients for the two digital filters: a11, a12, b10, b11, b12 and a21, a22, b20, b21, b22. We applied Nelder-Mead optimization method in order to compute the optimal filter coefficients producing the same time delay or phase difference effects [4-6]. III.

PHASE DIFFERENCE BY DIGITAL FILTERS

The sound transmission speed is set up as 343 m/sec at 20oC. And D = 9mm, the sampling frequency is 16kHz. Cardioid directivity pattern was considered, so that T should be 26.239s [7]. The filter coefficients are optimized to have

Figure 5. Directivity of digital filter method, =1(cardioid).

ISBN: 1-60132-459-6, CSREA Press ©

52

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 8. Maximum sensitivity frequency response comparison between Two mic. and three mic. arrays. Figure 6. Frequency response of the maximum sensitivity, =1(cardioid).

Figure 6 shows the frequency response of the maximum sensitivity magnitude for the same cardioid directivity patterns. The time delay result is drawn by continuous lines while the digital filter method is done by rounds. Both results are very close.

IV.

CONCLUSIONS

This paper compared directivity patterns of conventional time delay method with suggested digital filter method. Both results showed almost the same results for a specific time delay value of cardioid pattern. The 2nd order two digital filters are optimized to produce the same phase difference as the delay time. The optimally computed coefficients indicated that only 4 coefficients are good enough instead of 10 parameters. We’ll continue research for other directivity patterns and find how the filter coefficients are correlated. This work would be further useful for scanning directional beam pattern not only for 2D but also for 3D in medical imaging technology.. ACKNOWLEDGMENT This study was supported by research fund from the ministry of commerce, industry and energy (MOCIE Korea) 2015 core medical device commercialization technology development project (smartphone controlled 64 channel digital hearing aid with 6dB voice SNR (Signal to Noise Ratio) improvement: project number 10054678) in 2015.

Figure 7. Directivity of digital filter method with three microphone arrays, =1(cardioid).

REFERENCES Figure 7 shows the directivity patterns of the filter method with three microphone arrays. The distance between two adjacent microphones is 11mm. The beamform of the three microphone arrays looks narrower than that of the two microphone arrays. But the maximum sensitivity decreases more rapidly for lower frequency (6dB/oct. to 12dB/oct. in Figure 8).

[1]

[2]

[3]

[4] [5]

[6]

You J. Kwon, Soon S. Jarng, “Directivity pattern simulation of the ear with hearing aid microphones by BEM”, Acoustical Society of Korea 2014 Spring Conference, Vol. 23/1, pp. 361-366, 2004. EZAIRO 7110 HYBRID, http://onsemi.com, Semiconductor Components Industries, LLC, Publication Order Number: E7110/D, 2013. BlueCore® CSR8675 BGA, Qualcomm Technologies International, Ltd., CS-333028-PS, http://www.csr.com/sites/default/files/csr8675_bga.pdf, 2015. J. A. Nelder, R. Mead, “A simplex method for function minimization”, Computer Journal, Vol. 7, pp. 308-313, 1965. W. H. Press, B. P. Flannery, S. A. Teukolsky, W. T. Vettering, Numerical Recipes in C, Cambridge University Press, Cambridge, UK, 1988. Math Works, Matlab, The Math Works, Natick, MA, 1994.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[7]

Soon Suck Jarng, Carl Swanson, Frank Lee and Joseph Zou, "64 Bands Hybrid Noise Reduction and Feedback Cancellation Algorithm for Hearing Aid", International Journal of Control and Automation, Vol.7, No.1, pp.427-436, 2014.

ISBN: 1-60132-459-6, CSREA Press ©

53

54

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

SESSION HEALTH INFORMATICS, HEALTHCARE AND PUBLIC HEALTH RELATED SYSTEMS Chair(s) TBA

ISBN: 1-60132-459-6, CSREA Press ©

55

56

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

57

An HL7 v2 Platform for Standards Development and Testing S. Martinez1 and R. Snelick1 1 National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA

Abstract – Development of healthcare data exchange standards has long been problematic, plagued with ambiguous and inconsistent requirement specifications. This situation leads to potential misinterpretation by implementers, thus limiting the effectiveness of the standard and creating artificial and unnecessary barriers to interoperability. Likewise, the ability to test implementations effectively for conformance to the standards is hindered. The current approach of standards development and test plan creation relies on word processing tools, meaning implementers must read and interpret the information in these documents and then translate it into machine-processable requirements and test assertions. This approach is error prone—a better methodology is needed. We present a set of productivity tools in an integrated platform that allow users to define standards and test plans that result in machine-processable artifacts. A testing infrastructure and framework subsequently uses these artifacts to create conformance testing tools automatically. We present and demonstrate the utility of a platform for developing standards, writing test plans, and creating testing tools. This end-to-end methodology is illustrated by describing a case study for the HL7 v2.6 Vital Records Death Reporting Implementation Guide. Keywords: Healthcare Data Exchange Standards; Healthcare Information Systems; Interoperability; Standards Development; Testing.

1

Introduction

For 30 years, HL7 Version 2 (v2) has been the predominant standard used for the exchange of healthcare administrative and clinical data. Healthcare information systems use the HL7 v2 protocol to develop standardized interfaces to connect to and exchange data with other systems. HL7 covers a broad spectrum of domains including Patient Administration, Laboratory Orders and Results, and Public Health Reporting. The base HL7 v2 standard [1] is a framework that contains many message events, and for each event it provides an initial template (starting point) that is intended to be constrained for a specific use case. The application of constraints to a message event is referred to as profiling [2,3]. For example, the ADT (Admit, Discharge, Transfer) A04 (Register a Patient) message event is a generic template

for communicating information about a patient. The base message template is composed of mostly optional data elements. For a given use case, e.g., Vital Records Death Reporting (VRDR) [4], the message template is “profiled”. That is, elements can be constrained to be required, content can be bound to a set of precoordinated codes, and so on. The base message event (e.g., ADT A04) that has been constrained for a particular use (e.g., VRDR) is referred to as a conformance profile. An implementation guide is a collection of conformance profiles organized for a particular workflow (e.g., report, revise, or cancel a death report). In this example, three conformance profiles exist each with different message events, one for report, revise, and cancel. To date, HL7 v2 implementation guides have been created using word processing programs, which has resulted in ambiguous and inconsistent specification of requirements. This practice has hindered consistent interpretation among implementers, which has created an unnecessary barrier to interoperability. We present an end-to-end methodology and platform for developing standards (implementation guides), writing test plans, and creating testing tools in the HL7 v2 technology space [3,5]. The platform includes three key foundational components:  A tool to create implementation guides and conformance profiles  A tool to create test plans, test cases, and associated test data  A testing infrastructure and test framework to build testing tools A key to the approach is that the “normal” process of creating implementation guides, test plans, and testing tools is “reversed”. Instead of creating requirements using a natural language and subsequently interpreting the requirements to create test plans and test assertions, the requirements are captured with tools that internalize the requirements as computable artifacts. Figure 1 illustrates the methodology. Domain experts develop use cases, determine the message events that correspond to the interactions in the use cases, and then proceed to define the requirements. Using the NIST

ISBN: 1-60132-459-6, CSREA Press ©

58

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Fig. 1. NIST HL7 v2 Standards Development and Testing Platform

methodology, they accomplish these tasks by entering this information into the Implementation Guide Authoring and Management Tool (IGAMT). During this process, the domain experts constrain the message events according to the requirements needed by the use case. Section 2 will elaborate more on this process and on the details of how the requirements are constrained. The output of IGAMT is a set of artifacts that are represented in Word, HTML, and XML formats. The complete implementation guide, including the narrative and messaging requirements, can be exported in Word or HTML. Such formats are suitable for ballot at standards development organizations such as HL7 or IHE (Integrating the Healthcare Enterprises). Each conformance profile can be exported as XML. The XML format contains all of the messaging requirements in a machine processable representation, which is the most important aspect of IGAMT since the XML conformance profiles have many uses including message validation, test case and message generation, and source code generation. The XML conformance profile and/or the internal IGAMT model are imported by the Test Case Authoring and Management Tool (TCAMT). TCAMT is used to create targeted test cases for interactions (profiles) defined in the implementation guide. The output is an additional set of constraints in an XML format. The entirety of the output generated from IGAMT and TCAMT is called a “resource bundle”. The NIST platform includes a testing infrastructure of common utilities used for testing, such as a message validation engine, along with a testing framework that provides various testing tool components, such as a communication framework and a profile viewer. Testing Tool instances are then created using the testing infrastructure and framework components as well as the resource bundle output generated from IGAMT and TCAMT. The NIST platform in essence allows end users to create conformance testing tools by means of a set of productivity tools. This streamlined approach can greatly reduce today’s problems with conformance test tools for standards. These problems include: there are too few of these tools, they are expensive to build, they are not

dynamic for local refinements, and their time to market is protracted. Additionally, the platform provides value through enforcing consistent and rigorous rules for requirements specifications. The remainder of this paper explains the NIST platform in more detail in the context of a real-world case study. The layout of the VRDR implementation guide is presented, and we describe how IGAMT is used to capture the messaging requirements. Next, an explanation of how a set of targeted test cases are created in TCAMT is provided. Finally, the resulting VRDR test tool is presented. The goal is to inform the reader about the ease with which HL7 v2 implementation guides, test cases, and testing tools can be created using the NIST platform compared to the current laborious methods used today.

2

IGAMT

IGAMT is a tool used to create HL7 v2.x implementation guides that contain one or more conformance profiles. The tool provides capabilities to create both narrative text (akin to a word processing program) and messaging requirements in a structured environment. Our focus in this paper is on the messaging requirements. IGAMT contains a model of all the message events for every version of the HL7 v2 standard. Users begin by selecting the version of the HL7 v2 standard and the message events they want to include and refine in their implementation guide. For example, the message events ADT^A04, ADT^A08, and ADT^A11 are used to create 14 conformance profiles in the VRDR implementation guide. Each message event is profiled (constrained) to satisfy the requirements of the use case. Rules for building an abstract message definition are specified in the HL7 message framework, which is hierarchical in nature and consists of building blocks generically called elements [1]. These elements are segment groups, segments, fields, components, and subcomponents. The requirements for a message are defined by the message definition and the constraints placed on each data element. The constraint mechanisms are defined by the HL7 conformance constructs, which include usage, cardinality, value set, length, and data

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

type. Additionally, explicit conformance statements are used to specify other requirements that can’t be addressed by the conformance constructs. The process of placing additional constraints on a message definition is called profiling. The resulting constrained message definition is called a conformance profile (also referred to as a message profile). An example of a constraint is changing optional usage for a data element in the original base standard message definition to required usage in the conformance profile. IGAMT provides, in a table format user interface, the mechanisms to constrain each data element at each level in the structure definition. The rows of the table list the data elements according to the structure being constrained (segments, fields, and data types). The columns list the conformance constructs that can be constrained for a data element, including the binding to a value set. One key philosophy of IGAMT is the capability of creating reusable building blocks. These lower level building blocks can be used to efficiently create higher level constructs. The building blocks include data type flavors, segment flavors, and profile components. A base data type can be constrained for a particular use; the resulting data type is called a data flavor (or data type specialization). A given base data type may have multiple data type flavors. These flavors can be saved in libraries and reused as needed. A similar process applies to creating segment flavors. A profile component represents a subset of requirements that can be combined with other profiling building

59

blocks. One such example is the definition of a profile for submitting immunizations. The CDC creates a national level profile. However, individual states may have additional local requirements that can be documented in a profile component. Only the delta between the national and local requirements is documented in the profile component. Combining the national level profile and the state profile component yields a complete (implementable) profile definition for that particular state. This design provides a powerful and effective approach to leveraging an existing profile. A utility for creating and managing value sets is also provided. Specific value sets can be created and bound to data elements. Value set libraries can also be developed for reuse.

3

VRDR Use Cases

The Vital Records Death Reporting (VRDR) Implementation Guide (IG) [4] was developed to support the transmission of death-related information from the health care provider’s electronic health records (EHR) to the jurisdictional vital records offices (JVRO) and to the National Statistical Agency (NSA) [6]. Five use cases/workflows are identified to describe the transmission of data: Provider Supplied Death Information (PSDI), Jurisdiction Death Information (JDI), Void Certificate Reporting (RVCA), Coded Cause of Death (CCODA), and Coded Race/Ethnicity (CREIA) [4]. The use cases require three message events: ADT^A04, ADT^A08 and ADT^A11. A given use case has more than one interaction; in total, 14 interactions are needed.

Fig. 2. Vital Records Death Reporting Interactions

ISBN: 1-60132-459-6, CSREA Press ©

60

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Fig. 3. VRDR Profiles and Design

Figure 2 shows the 14 interactions supporting the VRDR use cases. The type of information being exchanged determines how each interaction (message) is constrained. For the PSDI use case, the ADT^A04 is constrained for reporting about a person’s death, the ADT^A08 message is constrained for updates to the report, and the ADT^A11 message is constrained to cancel the report.

The approach used in the development of the A04 profiles is followed in the creation of the A08 and A11 profiles. The base message events are loaded and constrained to develop the PSDIA11_V1.0 and PSDIA08_V1.0 profiles. These profiles are then leveraged to create the remaining profiles sharing the same message event using the IGAMT cloning capability.

Each interaction is assigned a unique profile identifier; e.g., “PSDIA04_V1.0” is a profile identifier for the “Send Patient Death Information” interaction (ADT^A04 interaction in Figure 2). The same three message events (ADT^A04) are employed across various use cases, however, the context in which they are used is different; therefore, the set of constraints applied are different, each resulting in a unique conformance profile (for the same base ADT^A04 event). The content is defined by the set of initiating and responding systems.

Profiling at the value set, segment, field, and data type (component) levels is followed, and it can be achieved in any order, thus taking advantage of the IGAMT capability that allows for the creation of reusable building blocks. The value set library can be created using the IGAMT built-in mechanism for loading value sets from HL7 tables and the CDC PHINVADS (Public Health Information Network Vocabulary Access and Distribution System) value sets [7].

Figure 3 shows the conformance profile-building approach for the VRDR profiles and the group of profiles that share the same trigger event. The A04 message event is loaded and constrained to create the common A04 profile. From there, the PSDIA04_V1.0 profile is created. The PSDIA04_V1.0 profile is used in building all of the profiles associated with the A04 event, and the corresponding message-level constraints are added to each profile. Constraints at the message level include segment and group usage, cardinality, and any additional requirement in the form of conformance statements.

VRDR data type flavors are built from the HL7 v2.6 data type library. They are defined using the constraints specified in the IG, such as length, usage, and value set binding and any constrains in the form of conformance statements. Occasionally, depending on how the data type is going to be used, the IG defines more than one data type specialization for a base data type. In the case of the “HD (Hierarchic Designator Assigning Authority)” data type, an additional flavor called HD_AA is defined. This flavor is used when an OID (Object Identifier) is assigned to designate an assigning authority.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

The VRDR segment flavor are created from the HL7 v2.6 segment library using the length, data type, cardinality, usage, value set binding, and conformance statement constraints defined in the IG. In some instances, it is necessary to create additional segment flavors to indicate constraint deltas among the profiles. For example, the constraints in the PV1 segment are applied to every profile, therefore only one PV1 flavor is created. In the case of the PID segment, the constraints are different in each use case; therefore, a segment flavor is created for each use case, and that flavor can be reused in the respective profiles. As shown, a key feature in IGAMT is the capability to create precise object definitions and to use (and reuse) the objects as building blocks to create higher level objects, such as segment and conformance profiles.

4

TCAMT

TCAMT is a tool used to create HL7 v2.x test plans that contain one or more (typically many) test cases. A test case can consist of one or more test steps. A test step can be an HL7 v2.x interaction or a manual step such as visually inspecting the contents of the system under test’s (SUT’s) display screen. Each test case and test step can consist of a test description, pre- and post-conditions, objectives, evaluation criteria, and additional notes and comments. Test steps for an HL7 v2.x interaction contain an HL7 v2 message (that is, specific data) that is in alignment to an XML conformance profile created from IGAMT. TCAMT allows domain experts to create test cases that target certain scenarios and capabilities. Using these test cases provides context, which expands the scope of testing beyond just the constraints in the conformance profile. Without context, a validation tool cannot test a message exhaustively to all requirements specified in the implementation guide. For example, elements with “required, but may be empty (RE)” usage or elements with “conditional usage (C)” cannot be assessed without targeted tests. A message that is validated against the requirements of a conformance profile without any provided context is called “context-free testing”. A message that is validated against the requirements of a conformance profile and with a provided context is called “context-based testing” [3]. The test cases provide context, and TCAMT is a tool that allows users to create the test cases. A key feature in TCAMT is its use of the conformance profiles created in IGAMT as a foundation. The message definition and requirements are available to the TCAMT user based on information that was entered into IGAMT. Then, the TCAMT user provides the data associated with each message element of interest. TCAMT also allows

61

the user to enter additional assertion indicators based on what they want to test. For example, for an element with a usage of “RE”, the user might provide data that are expected to be entered into the sending system for the element, and the user also might select an assertion indicator. There are several assertion indicators that could be selected, for example, “presence”. In this case, if the user provides test data and the indicator of “presence”, an additional assertion (constraint) is generated by TCAMT and is provided to the validation. For elements with “RE” usage, the element must be supported by the SUT, but in a given message instance the element may not be populated. For this construct, the tester wants to ensure that the implementation has, in fact, included support for the element. In a context-free environment, the absence of data in a message is not a conformance violation for elements with “RE” usage. However, in the example test case described above, data were provided, and an assertion for the presence of the data was selected. Now, when a message created for this test case is validated, the additional assertion triggers the check for the presence of data for this element. This method is one way to determine support for the element. Via TCAMT, the user can create an unlimited number of test cases and test a broad spectrum of requirements. Other assertion indicators can be used to test for specific content or for the non-presence of an element. Additionally, test data can be provided to trigger conditional elements. In other instances, support for certain observations may need to be ascertained. In such cases, test data for specific observations (e.g., cause of death, date/time pronounced dead, etc.) are provided, requiring the message instance to contain an OBX segment for that observation. TCAMT provides the mechanisms to conveniently and consistently create test cases. Output from TCAMT provides the additional constraints that are interpreted by the validation engine.

5

VRDR Test Plan

The VRDR test plan consists of a set of scenarios and test cases that emulate real-world events and workflow. The scenarios are designed to target specific requirements that are not easily testable in a context-free environment. The goal of the VRDR test plan is to provide a set of test cases that collectively tests the spectrum of requirements defined in the VRDR implementation guide. Therefore, for each interaction in support of a use case, test scenarios and test cases are needed. Figure 4 shows an excerpt of the test plan. A scenario for the Provider Supplied Death Information (PSDI) is indicated that contains three test cases and associated test

ISBN: 1-60132-459-6, CSREA Press ©

62

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

steps. Test steps have a 1-to-1 relationship to an HL7 v2 message (interaction), and each message is bound to the requirements in its corresponding conformance profile.

Fig. 4. VRDR Test Plan Excerpt

For each test case, a real-world story is given along with specific test data that coincide with the test story. The test data provide a known data set that can be used to create additional assertions (beyond those provided in the conformance profile). This approach is the principle behind context-based testing. Each test step interaction contains a message that is associated with its corresponding conformance profile. TCAMT provides a productivity mechanism to create the test messages using the underlying structure provided in the conformance profile. Once the test message (and therefore test data) has been created, additional assertions can be specified that align with the testing goals.

Table 1 shows two important examples of how this process works. TCAMT facilitates specific assertions by allowing the test plan designer to assign assessment indicators to the data elements. The combination of the provided test data and assessment indicator generates the assertions. The PDA-2.9 example shows a case where the Death Location Description element is constrained with “RE” usage, which indicates that the element must be supported but data may not always be available. To test support for this element, the test case provides test data (“Mercy Hospital”) and the assessment indicator is set to “Presence”. These settings will generate an assertion that makes the presence of the PDA-2.9 element required. The OBX-3.1 example shows a case where an observation (an OBX segment) is expected in which the observation is “Cause of Death”. Here, the value “694539” is provided, which is a LOINC (Logical Observation Identifiers Names and Codes) code that indicates a “Cause of Death”. By explicitly requiring the “Cause of Death” observation be included in the message, the testing is ensuring that the SUT can support this observation. In this example, only one element in this segment is shown, but typical testing scenarios will have a coordinated set of assertions for the set of elements in the OBX segment. For example, OBX-3.3 would assert that the content of this element is “LN” to confirm that the code “69453-9” is in fact drawn from the LOINC code system. Creating test cases that target specific capabilities, such as “sending the Cause of Death observation” is an important aspect of testing and a key incentive for conducting context-based testing. Without this level of specificity, assessment of systems is limited. Only a few examples have been provided here to give the reader a sense of the sorts of items that can be tested. However, this approach expands the test space significantly. Other aspects that can be tested include cardinality, length, value set constraints, conditional elements, specific content, workflow, and functional requirements.

Table 1. Context-based Assertion Examples

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

6

Infrastructure and Framework

NIST has built an HL7 v2.x testing infrastructure and framework to aid in the process of creating conformance testing tools. The testing infrastructure provides a set of services utilized by the test tool framework to build specific instances of tools. A test tool can be specific for a particular domain, or it can be general-purpose. The general-purpose tool is a NIST-hosted web application where a user can upload conformance profiles and test plans to create a test tool. The conformance test tool essentially is generated as a by-product “for free” once the validation artifacts have been created. This liberates the domain experts from the tool building process. Alternatively, the framework can be leveraged, customized, and installed locally. Using the framework, developers can choose to create domain specific or general-purpose web application conformance test tools, access the validation via web services, or incorporate validation via JAR (Java Archive) files or source code. Regardless of the use, the NIST platform can significantly improve the quality of implementation guides, assist in the creation and maintenance of test plans, expedite the stand-up of a validation tool, and, overall, reduce the cost and time of the entire process.

7

VRDR Test Tool

A VRDR conformance testing tool is built using the testing infrastructure and framework, the IGAMTproduced conformance profiles, and the TCAMTproduced test plan. The test tool is a web-based application (see [8] to access) that supports both contextfree and context-based validation. In addition to performing message validation, the tool provides a browse-able view of the requirements for each conformance profile. In the context-based mode, the test story, test data, and an example message are provided for each test step. In the context-free mode, the user simply selects the conformance profile to validate against and imports the message. The validation is performed automatically and a report is given. In the context-based mode, the user selects the test step and imports the message to validate. The test tool sets the validation to the conformance profile linked to the test step, performs the validation, and provides a report. In both modes, a tree structure of the message is shown on the left panel of the validation screen and can be used to inspect the content of individual data elements.

8

63

platform includes three key foundational components: (1) a tool to create implementation guides and conformance profiles; (2) a tool to create test plans, test cases, and associated test data; and (3) a testing infrastructure and test framework to build testing tools. We demonstrated the approach by creating a test tool for the HL7 v2.6 Vital Records Death Reporting use case. Requirements were captured in IGAMT and exported as conformance profiles. TCAMT was used to create a set of test cases based on the conformance profiles. A conformance test tool was created by combining the validation artifacts with the testing infrastructure and framework.

9

References

[1] Health Level 7 (HL7) Standard Version 2.6, ANSI/HL7, October 2007, http://www.hl7.org. [2] Principles for Profiling Healthcare Data Communication Standards. R. Snelick, F. Oemig. 2013 Software Engineering Research and Practice (SERP13), WORLDCOMP’13 July 22-25, 2013, Las Vegas, NV. [3] Healthcare Interoperability Standards Compliance Handbook. F. Oemig, R. Snelick. Springer International Publishing Switzerland, ISBN 978-3-319-44837-4, December 2016. [4] HL7 Version 2.6 Implementation Guide: Vital Records Death Reporting, Release 1. Draft Standard for Trail Use. August 2016. http://www.hl7.org. [5] NIST Resources and Tools in Support of HL7 v2 Standards. http://hl7v2tools.nist.gov/ [6] CDC National Vital Statistics System: http://www.cdc.gov/nchs/nvss/evital_standards_intiative s.htm [7] CDC Public Health Information Network Vocabulary Access and Distribution System (PHIN VADS); https://phinvads.cdc.gov/. [8] NIST Vital Records Death Reporting (VRDR) Conformance Testing Tool; http://hl7v2-vr-r2testing.nist.gov

Summary

We presented an end-to-end methodology and platform for developing standards, writing test plans, and creating testing tools in the HL7 v2 technology space. The

ISBN: 1-60132-459-6, CSREA Press ©

64

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Infrastructure for Health Care Simulation: Recommendations from the Model for Telecare Alarm Services Project B.F. Smaradottir1, R.W. Fensli1, E.S. Boysen2, and S.G. Martinez3 1 Department of Information and Communication Technology, University of Agder, Grimstad, Norway 2 SINTEF, Oslo, Norway 3 Department of Health and Nursing Science, University of Agder, Grimstad, Norway

Abstract - In Norway, a recent health reform targeting coordination and continuity of care has urged municipalities to manage telecare alarms related to an increasing number of welfare- and sensor technologies in citizens’ homes. In this context, the research project “Model for Telecare Alarm Services” has the aim to study the organization and operation of existing telecare alarm services and identify new service models of the future. In the project, simulation of health care services was used when key informants from municipalities, end-users, research partners and industry tested different models for future telecare alarm services. This paper presents the technical and physical infrastructure of a clinical laboratory environment for simulation of health care services, with recommendations based on experiences from the project. Keywords: Health Care Simulation, Technical Infrastructure, Telecare, User-centered Design

1

Introduction

Telecare technology is used to support communication between citizens in home environment and health care services [1]. This type of technology is considered an important remedy to cope with significant challenges of an increasing number of older people due to societal demographic changes [2]. An overall goal is to enable people with physical or medical limitations to live self-dependent in their own home as long as possible [3]. In Norway, a health reform [4] was adopted in 2012 to improve the continuity of care and collaboration across the traditional health care services. Services that for several years were carried out in hospitals were transferred to municipalities. Today, municipal healthand social care services are responsible for a 24/7 service in emergency primary health care. In addition, due to an increased number of telecare- and sensor technologies in citizens’ homes, many municipalities are preparing for establishment of a service for management of alarms. In many cases, this requires re-organization of the already existing health and care services. In this context, the research project Model for Telecare Alarm Services aimed to explore, evaluate and propose

models for telecare alarm services. 18 municipalities, 2 research institutions and 1 industry partner participated in the research project from 2015-2017, where the aim was to study how existing telecare alarm services in Norwegian municipalities were organised and operated. In an early phase, workshops were organised together with health care professionals, operators at existing telecare alarm services and representatives from patient organisations with focus on challenges in already existing services and user needs. In addition, critical factors were identified for the design of new models for future telecare alarm services [5][6][7][8]. Later in the project, the industry partner (Imatis) developed an information and communication technology (ICT) system prototype with a smartphone application for telecare alarm services. During the development process, three simulations with participants from the project’s partners were run in a clinical laboratory environment, with a home-based alarm scenario involving new technology handled by telecare alarm service operators and municipal home nursing service. During the preparation and the execution of the simulations in the clinical laboratory, the research team reflected on the infrastructure and lessons learned that were considered useful for future simulations of health care and related technologies. This paper presents recommendations for a technical and physical infrastructure for simulation in a clinical laboratory environment based on the experiences from the Model for Telecare Alarm Services project. The following two research questions (RQs) were addressed in this study: RQ1: What technical and physical infrastructure is suitable for simulation of telecare alarm services? RQ2: What are the lessons learned that could be transferable for simulation of other health care contexts? Following this introduction, an overview of related research is presented. In the next section, the technical and physical infrastructure for simulation in clinical environment will be described. Later, the discussion reflects on lessons learned from carrying out the simulations during the project. Finally, the conclusions are drawn regarding the characteristics of a technical infrastructure for simulation in clinical laboratory environments.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

2

Research background

3.2

User-based simulation refers to the participation of endusers in an evaluation and test of an application in a clinical laboratory or hospital-like environment [9][10][11]. Participants are asked to perform a role-play and do tasks described on a task-list, or to use a system in a clinical environment, while being observed and recorded. The goal is to identify system flaws, cause of errors or difficulties in the use of the system, but also analyzing how the introduction of new technology influences the existing clinical workflow. Measurements are performed on time for task solving and errors. The aim is to provide a better understanding of the interaction between end-user groups, clinical workflow and technology involved when defining and executing tasks and accessing information. This kind of simulation is recommended when a test in real clinical environments is unsuitable for legal, ethical or privacy reasons that may negatively affect the protection of patients [9]. Therefore, a simulation of a health care environment is important for creating a realistic scenario generally safe for health professionals and patients. In addition, the use of actors is recommended for some situations, such as in the patient-like role [10][11][12] Simulations are most often made in clinical laboratory settings, which have the strength of providing a controlled environment for the variables studied. A weakness of the laboratory environment can in some cases be the influence of observation on the behavior of the participants, such as the Hawthorne effect [13].

3

The health care simulations

Three simulations with task-based scenarios of models for telecare alarm services were executed in the Clinical Laboratory of the University of Agder (UiA) in Norway during April, September and November of 2016. The simulations were led by a research team consisting of people with health and technology background. Between 16 and 25 people from the project’s partners participated in the simulations.

3.1

Telecare alarm technology

In the first simulation, digital mock-ups were used with graphical sketches managed by the test leader, to create the scenario and simulate the workflow for the handling of a telecare alarm. The moderators used a chat-channel to manage time-synchronization between actual information exchanged between the participants in the different test rooms. In second and third simulations, participants used a technology under development provided by the industry partner consisting of a smartphone application and prototype of an ICT system for alarm service operators.

65

Simulation scenarios

The scenarios tested, as role-play in the laboratory, included triggering of an alarm by a patient at home and the following interaction with the telecare alarm service based on different models for how to operate the service. The contact between the telecare alarm service operator and the response team for handling the alarm was also part of the scenarios, taking into account the organization of home nursing services, as well as the telecare alarm service. The scenarios also included scheduling and visiting a patient’s home when necessary due to medical reasons. In that case, the status of the situation was reported back to the telecare alarm service operator. The information flow between the operator and the response team was made of electronic messages that represented the tasks to be executed and transmitted through mobile devices (e.g., smartphone and/or tablet).

3.3

Technical and physical infrastructure

The Clinical Laboratory facilities used in the simulations consisted of 3 separate test rooms and 1 observation room. The technical and physical infrastructure for the simulations is illustrated in Figure 1. The Test room 1 represented the patients home, Test room 2 the telecare alarm service and Test room 3 the office of the municipal home nursing services. In each test room, there was a separate recording camera source. In the observation room, the simulation was followed simultaneously on 4 monitors, one for each camera source and one monitor merging and showing all sources simultaneously. Between the Test room 3 (where the smartphone application was tested) and the observation room there was a one-way mirror that allowed the observers to closely follow the simulation process. In the Test room 1, the patient (played by different participants throughout the different scenarios) used a Safemate geolocation alarm device [14] to trigger an alarm. The device was equipped with phone connection capability and allowed the voice communication between operator and patient. In the Test room 2, the ICT system for telecare alarm service was accessed and used in a laptop PC, but also shown on a Smart board display on the wall. In the Test room 3, the home nursing service used a smartphone or a tablet device for accessing the test application. The observation room had a desktop PC connected to four monitors, allowing the observers to remotely follow the technology interactions and work processes between the test rooms. The zooming and operation of the fixed cameras was made in the observation room for recording purposes. The observation room and the test rooms were connected with a dedicated segment of the secured LAN infrastructure of the Centre for eHealth at UiA, making use of VLAN technology.

ISBN: 1-60132-459-6, CSREA Press ©

66

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 1. The technical and physical infrastructure for the simulations. This connection was also used for the IP-based streaming of video and audio signals from the test rooms to the observation room. The recordings from the audio-visual sources were merged into one file in F4v video format, using Wirecast [15] as a capture software tool. The purpose was to ease the retrospective data analysis, having just one file including multiple video perspectives with a single audio channel.

3.4

Test room 3: • Samsung smartphone with the Telecare Alarm Service application activated (and a tablet device) • Fixed Camera: SONY BRCZ330 HD 1/3 1CMOS P/T/Z 18x Optical Zoom (72x with Digital Zoom) Colour Video Camera. • Sennheiser e912 Condenser Boundary Microphone. • Landline phone communication.

Technological materials

For replicability and information purposes, the technological material used during the simulations is presented below grouped by rooms. Test room 1: • Safemate geolocation device. • Fixed Camera: SONY BRCZ330 HD 1/3 1CMOS P/T/Z 18x Optical Zoom (72x with Digital Zoom) Colour Video Camera. • Sennheiser e912 Condenser Boundary Microphone. • Landline phone communication. Test room 2: • • • • •

Laptop for the Telecare Alarm Service ICT system. Smart board display 65" Portable Camera: SONY HXR-NX30 Series. Logitech 886-000012 Boundary Microphone. Landline phone communication.

Observation room: • Stationary PC: Mac Pro • Monitor: 3x HP Compaq LA2405x • 27" Mac Monitor • Streaming: 2x Teradek RX Cube-455 TCP/IP 1080p H.264. • Software Wirecast 4.3.1. • Landline phone communication.

3.5

Simulation procedure

The scenarios were performed as a group simulation with interaction between the test rooms by use of technology. Each scenario had a description of a context and a situation to be handled. One moderator from the research team and the participants in groups of 3-5 people were placed in each test room, also in the observation room. The scenario was repeated at least once and for each repetition the roles of the participants within each group were swapped. When the next scenario was tested, the participants also changed test room,

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

so each group played the different roles assigned to the simulation. Each scenario had assigned roles, with a separate task list for each role. The moderator asked the participants to think aloud [10][16] and speak freely during the simulations.

3.6

Ethical considerations

This research study was approved by the Norwegian Centre for Research Data [17] with project number 44494. All participants received oral and written information about the project and they signed a consent form.

4

Discussion

This paper has presented recommendations for a technical and physical infrastructure for simulations in a clinical laboratory environment based on the experiences from the research project Model for Telecare Alarm Services. The preparation and the execution of the simulations led to a series of reflections and lessons learned by the research team that are considered useful for future simulations of health care services and associated technologies. The two research questions (RQs) formulated at the beginning of this paper are answered below based on the study. RQ1 asked about a suitable technical and physical infrastructure for simulations of health care. An infrastructure suitable for the simulations of health care would be one that firstly allows flexibility in interaction between participants and technology within a clinical simulated environment for the related work processes. Secondly, the infrastructure should allow the research team to collect high-quality data for an effective retrospective analysis under more dynamic and cognitively demanding conditions than individual testing in usability laboratory, crucial to avoid interference and distortion of the simulation results. Thirdly, it is recommended that the audio and video sound from each test room are collected both in separate files and as one synchronized file, to allow a detailed retrospective analysis from each test room, but also the simultaneous interactions between the test rooms involved. In addition, the data should be collected through multimodal channels (e.g., video and audio), having the necessary tools to synchronise audio and video signals with sufficient quality and avoiding network latencies. RQ2 asked about lessons learned that can be transferable for simulations of other health care contexts. One lesson learned is that the technology is necessary for the interaction between participants but it is not the only focus when simulating the workflow of health care specific processes [11]. The use of low fidelity software, such as mock-ups, early prototypes and systems under development together with enduser groups simulating current and future health work processes, provides a useful insight on how the technology would impact work processes in a real clinical setting. This is a relevant factor to consider due to user acceptance of new technology [18]. As simulations are usually more difficult to perform due to ethical and legal issues in real clinical

67

environments with patients and health care professionals, the clinical laboratory represents the environment where health professionals and patients interact with each other and through the technology. Moreover, the fact of having health care professionals and representatives from end-user groups (e.g., patients, close relatives and volunteers) to experiment in different workflow processes gives an understanding of how the workflow and corresponding procedures can be optimized, and the potential impact that the technological solutions can have in real settings. Another lesson learned was that groupdebriefing participants after each tested scenario, was very useful and provided information for subsequent system development and modelling of telecare services. Finally, due to the difficulties of recruiting participants and the discomfort of having to unnecessarily repeat simulation sessions, a redundancy in data collection is strongly recommended using two or more independent sources of data storage to avoid accidental data loss. This study on the technical and physical infrastructure for simulation of health care had some limitations, such as including data from only one research project. However, several simulation sessions were made within the project, which provided rich experiences regarding the technical and physical infrastructure. The simulations informed the development of prototypes from early low-fidelity mock-ups to wireframes and functional prototypes. The empirical research data from the user workshops and simulations regarding clinical workflow and functionality of telecare alarm technology under development are not in the scope of this paper, as the main focus is the technical and physical infrastructure for simulations in clinical laboratory environments.

5

Conclusion

This study was conducted as part of the research project Model for Telecare Alarm Services, with the aim to provide experiences on technical infrastructure for simulations in clinical laboratory environments. The main contribution of this study lies on the descriptions of a proposed and tested technical and physical infrastructure for simulations and the sharing of lessons learned that are transferable to other health care services research projects. Health care technology is widely used by multiple user groups and when designing, testing and evaluating such technology, there is a specific need to balance the interface design and functionality on the one hand and taking into consideration how the technology would adapt to existing or proposed clinical workflow on the other. Simulations in a clinical environment are essential to analyse not only the interface design of the technology, but also the interactions between end-users, devices and their impact in associated work processes. These simulations are enabled by a laboratory environment, where the research team has full control over all steps of the simulation scenario, including tasks and interactions between the test participants and the technology used. Low-fidelity technology prototypes allow for low-cost scenario simulations where end-user can

ISBN: 1-60132-459-6, CSREA Press ©

68

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

influence the development of the technology and associated work processes. The simulation infrastructure provided sufficient control over the factors involved and at the same time provided the flexibility to dynamically adjust the environment for adequate data collection. The results presented are in line with other studies on simulation in clinical environments [10][11]. Future research agenda of the authors includes testing of the physical and technical infrastructure with other types of scenarios, also focusing on procedures and methodology of simulations in clinical environments.

6

Acknowledgement

The research team would like to thank all participants and informants for their disinterested contribution in the study. Financial support was provided by the Regional Research Fund of Agder [19] in Norway with Grant number 247605 and the work on technical infrastructure by AustAgder Innovation- and Competence Fund [20] (2013-2017) ref. 2012/42.

7

References

[1] A. Kinsella. “Home telecare in the United States”; J Telemed Telecare, 4(4)., 195-200, 1998, doi: 10.1258/1357633981932226 [2] World Health Organization. “Telemedicine in Member States”; Global Observatory for eHealth series Vol. 2., 2010. [retrieved: May, 2017]. Available from: http://www.who.int/goe/publications/goe_telemedicine_2010. pdf [3] A.M. Schülke, H. Plischke and N.B. Kohls. “Ambient Assistive Technologies (AAT): socio-technology as a powerful tool for facing the inevitable sociodemographic challenges?”; Philos Ethics Humanit Med, 5(8)., 2010, doi:10.1186/1747-5341-5-8 [4] Norwegian Ministry of Health and Care Services. “The Coordination Reform, Proper treatment – at the right place and right time”; Report No. 47 (2008-2009) to the Storting. [retrieved: May, 2017].]. Available from: https://www.regjeringen.no/contentassets/d4f0e16ad32e4bbd8 d8ab5c21445a5dc/no/pdfs/stm200820090047000dddpdfs.pdf [5] R. Fensli, T. Vatnøy, I. Svagård and E.S. Boysen. “Evaluation of organizational models for response centres for Telecare services of the future”; Int J Integr Care 16(5)., S13, 2016, doi:dx.doi.org/10.5334/ijic. 2561 [6] I. Svagård, E.S. Boysen, R. Fensli and T. Vatnøy. “Response Centres for Telecare Services: Needs and vision of the future (in Norwegian. Responssentertjenester i helse- og omsorgstjenesten: Behov og fremtidsbilder) World Health Organization World Health Organization”; Sub-report 1-2016

in the Model for Telecare Alarm Services Project. SINTEF A27689, 2016. [7] E.S. Boysen, I. Svagård and D. Ausen. “A study of telecare alarms in seven municipalities. When and why are the alarms triggered? (in Norwegian Studie av utløste trygghetsalarmen i syv kommuner. Når og hvorfor utløses trygghetsalarmene?)”; Sub-report 2-2016 in the Model for Telecare Alarm Services Project. SINTEF A27757, 2016, ISBN: 9788214061291 [8] K. Askedal and S. Sjaavaag: “Municipal Response Centres for Telecare Services. Mapping and recommendations for establishments (in Norwegian Kommunal responssentertjeneste. Kartlegging og anbefaling for etablering)”; Statement report. Municipality of Kristiansand, Norway, 2016. [9] D. Svanæs, O.A. Alsos and Y. Dahl. “Usability testing of mobile ICT for clinical settings: Methodological and practical challenges”; Int J Med Inform, 79(4)., e24-e34, 2010, doi:10.1016/j.ijmedinf.2008.06.014 [10] A.C. Li, J.L. Kannry, A. Kushniruk, D. Chrimes, T.G. McGinn, D. Edonyabo and D.M. Mann. “Integrating usability testing and think-aloud protocol analysis with “near-live” clinical simulations in evaluating clinical decision support”; Int J Med Inform, 81(11)., 761-772, 2012, doi:10.1016/j.ijmedinf.2012.02.009 [11] E. Borycki and A. Kushniruk. “Identifying and preventing technology-induced error using simulations: Application of usability engineering techniques”; Healthc Q, 8(Sp)., 99-105, 2005, doi:10.12927/hcq..17673 [12] B.F. Smaradottir. “The steps of user-centered design in health information technology development: Recommendations from a PhD research study”; In 2016 International Conference on Computational Science and Computational Intelligence (CSCI), 116-121, IEEE, doi: 10.1109/CSCI.2016.0029 [13] J.G. Adair. “The Hawthorne effect: A reconsideration of the methodological artifact”; J Appl Psychol, 69(2)., 334-345, doi:http://dx.doi.org/10.1037/0021-9010.69.2.334 [14] S. Martinez, B. Smaradottir, T. Vatnøy and M. Bjønnes. “Usability evaluation of a geolocation technology: Safemateperspectives on Norwegian municipal location-based alarm system”; In press at 2nd IEEE Workshop on ICT Solutions for eHealth, part of the 22nd IEEE Symposium on Computers and Communications (ISCC2017), 3-6 July 2017 in Heraklion, Crete, Greece. [15] Wirecast. [retrieved: May, 2017]. Available from: http://www.telestream.net/wirecast/overview.htm

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[16] M. W Jaspers. “A comparison of usability methods for testing interactive health technologies: methodological aspects and empirical evidence”; Int J Med Inform, vol. 78(5)., 340353, 2009. [17] The Norwegian Centre for Research Data. [retrieved: May, 2017]. Available from: http://www.nsd.uib.no/personvern/en/index.html [18] B.F. Smaradottir, S. Martinez, E. Holen-Rabbersvik and R. Fensli. “eHealth-extended care coordination: development of a collaborative system for inter-municipal dementia teams: A research project with a user-centered design approach”; In 2015 International Conference on Computational Science and Computational Intelligence (CSCI), 749-753, IEEE, doi:10.1109/CSCI.2015.79 [19] Regional Research Fund Agder. [retrieved: May, 2017]. Available from: http://www.regionaleforskningsfond.no/prognett-rffhovedside/RFF_in_English/1253976860326 [20] Aust Agder Innovation and Competence Fund. [retrieved: May, 2017]. Available from: http://www.aaukf.no/

ISBN: 1-60132-459-6, CSREA Press ©

69

70

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Examining Locking Down of Electronic Medical Records Ryan Carr, Sudip Chakraborty, Leah Johnson, Jasmin Miravete, and Joshua Vega Department of Computer Science, Valdosta State University, Valdosta, GA, USA

Abstract – In the rapidly changing world tools used for everyday life are changing even faster. Health care is one of the areas in our daily life that is changing so that health care professionals can coordinate better care for their patients. Hence, out of many ways, the medical field is shifting from paper records to electronic medical record (EMR) to achieve the above goal. However, the urge for digitization of health records has placed the confidentiality, integrity, and availability of personal sensitive information (health records) at risk to be compromised. In this paper, we examine the security related to EMR and the role of U.S. federal government’s push to digitization of health records. We examine HIPAA and ARRA in the context. We present examples of threats and attacks, with statistical data, related to EMR. We also discuss potential damage in a breach and, finally, examine the ways to secure it. This paper is our attempt to highlight the importance of security concerns in digitization of health records. Keywords: Health Records, EMR, Security, Privacy, HIPAA, ARRA, Digitization

1

Introduction

A patient’s medical record is an invaluable asset to all involved entities, including health insurer and pharmacy, in health care services. It is undeniable that a prompt access to such records is of high importance, especially for physicians and health care providers. The traditional form of recording and storing these records can satisfy this demand only to a limited extent. While it still works within close proximities (within a clinic, or across two services in a short distance), it is not sufficient for exchange of, or discussion on such records over larger distances. Traditional paper format of storing health records is not conducive to exchange of it in a time-sensitive manner, especially in a collaborative environment where the physicians or health care providers are not within close proximities. Therefore, digitization of health care records becomes a natural choice. Consequently, we have observed tremendous growth in software for creation, maintenance, storage, and exchange of electronic medical records (EMRs). These alleviated issues related to storage, access, and exchange of health records in traditional format. It is also found to be time and cost effective as duplication, storage, and exchange is faster and cheaper. However, with all these benefits, EMRs bring unique challenges that are not much pertinent in traditional format of health records. The major concern that rise is who has access to it and how secure is it? These issues are

of great importance as health records are very sensitive private information that should not be accessed by anyone other than a selected set of entities (physician, health care provider, insurer, and patient). In the traditional format the record is created and stored within a specific facility and can be accessed by certain entities involved in the facility only. It is also expected that the records will be stored in secure manner (under lock and key in a proper cabinet). With EMR these expectations erode as digital records are stored on computers and exchanged over the Internet. Though a standard securing mechanism can be placed, a suspicion about the sufficiency to maintain confidentiality and privacy of the records can rise. This is especially due to lack of control on the communication over the Internet. Also, enforcement of proper access control and authorization become more complex compared to the traditional format. The entities in this context must rely on the software, which are developed by entities not involved in the health care environment. The efficacy can be questioned even further if such EMRs are stored on Cloud storage that are often under the control of third parties. Hence, there is a strong need for examining privacy & security concerns related EMR, keeping track of incidents or events targeting breach of such security, designing & deployment of appropriate security measures, and creating legislation to address deterrence and violations. In the light of these objectives, especially the last one, in this paper we examine the state of securing EMRs. We explore the origin of EMRs popularity in recent times and role of legislation like ARRA, HIPAA in its growth. It is found that HITECH part of ARRA has a major role to play for increase in adoption of EMRs. We highlight security concerns and potential damage resulting from a breach of security. We present the data available to demonstrate severity of the issue. We also suggest few basic breach prevention mechanisms that can be applied to systems involving EMRs. The rest of the article is organized as follows: Section 2 gives a detailed overview of the legislation forming ARRA and HIPAA. Section 3 identifies existing security concerns. Section 4 discusses potential damages that can be done with compromised information and provides examples of actual breaches. Section 5 proposes a variety of ways to further improve security of sensitive medical information. Finally, we conclude the work in Section 6.

2

Government Role in EMRs

The first half of the last decade observed a big surge in use of EMRs in the USA. More and more physicians started enjoying the benefits of the system. According to the Office of the National Coordinator for Health Information Technology percentage of physicians using EMR for e-prescriptions rose

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

from 7% to 66% in 5 years. Figure 1 illustrates county–wise distribution of percentage of physicians using EMR for above stated purpose.

71

2.2

HIPAA

In 1996, Congress realized that technological advancements could make it increasingly difficult to secure the privacy of health information. Consequently, the Health Insurance Portability and Accountability Act of 1996, or HIPPA, was enacted. This piece of legislation required that the Department of Health and Human Services (HHS) is to establish a set of national standards for electronic healthcare transactions, health identifiers, and security. The HHS published two main rules: The Privacy Rule and the Security Rule [8]. 2.2.1 HIPAA Privacy Rule

Figure 1. Percentage of participating physicians e-Prescribing through EMRs in USA [16]

2.1

HITECH under ARRA for EMRs

The last decade showed an economic instability whose effect was observed throughout the world including the United States. In an attempt to revitalize the economy, the U.S. Congress responded to this economic weakness by enacting a multitude of stimulus packages. The American Recovery and Reinvestment Act of 2009 (ARRA) was one such stimulus package that aimed to lessen the effects of the recession on American citizens [1], [2]. Under the ARRA, the Health Information Technology for Economic and Clinical Health Act (HITECH) made a push towards Electronic Medical Records, to cut down on government spending and health care costs. To enforce their push for EMRs, the government offered certain incentives for eligible facilities. Essentially, these incentives offered any medical facility “incentive payments”, both to physicians and to the hospitals themselves. To continue receiving these incentives, health care providers had to provide “meaningful use” of EMRs [3]. Meaningful use is defined as using electronic health records to: (i) improve quality, safety, efficiency, and reduce health disparities, (ii) improve care condition, and (iii) maintain privacy & security of health information [4]. These incentive payments were given out from 2011 to 2016, and gradually decreased every year, allowing EMR users to earn up to $44,000 dollars [5]. Another motivating factor for the switch to EMRs is the penalty incurred should a facility not comply. The penalty for any medical personal who didn’t utilize or show meaningful use of EMRs by 2015 was a 1% reduction in Medicare reimbursements received. That percentage increased 1% each year to a maximum of 5% [6]. Thus, since then, EMR use by physicians has steadily increased to over 83% [7]. Another reason for increased EMR use is due to the convenience they offer. The EMR allows patients’ medical records to move with them digitally, and it gives the health facilities an overall view of a patient’s health. Figure 1 shows just how widely used EMRs have become since ARRA was enacted to the year 2013 [14].

The Privacy Rule is a nation-wide set of standards that protects all “individually identifiable health information” held or transmitted by a covered entity or its business associate, in any form or media whether electronic, paper, or oral. The Privacy Rule calls this information “protected health information” (PHI). This protected information includes: -

Individual’s past, present and future physical or mental health condition. The provision of health care to the individual. The past, present, or future payment for the provision of health care to the individual [10].

Under the Privacy Rule, clients are given the right to get a copy of their medical records, have corrections made to it, receive notification about how their medical information is being used or shared, to decide whom they want their information shared with, and to get a report for why and when it was shared. Additionally, the Privacy Rule even limits who can share and receive your medical information. Medical information is only shared if it is necessary for the client’s health; it is needed for payment; the client explicitly names friends or family members who can receive it; the information is used to ensure public’s health; and it can be shared with law enforcement [9]. 2.2.2 HIPAA Security Rule The Security Rule provides the set of standards for ensuring the protection of confidentiality, integrity, and availability of electronic PHI. The main goal of the Security Rule is to protect the privacy of the individual while allowing the covered entity to implement the appropriate policies, technology, and procedure. Because of this, the Security Rule is designed to be flexible and scalable. The Security Rule is composed of four sub rules; all covered entities must: 1. 2. 3. 4.

Ensure the confidentiality, integrity, and availability of all electronic PHI they handle. Identify and protect against reasonably anticipated threats to the security of PHI. Protect against reasonably anticipated use or disclosures from unauthorized individuals. Ensure compliance within the workforce [11].

ISBN: 1-60132-459-6, CSREA Press ©

72

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

2.2.2 HIPAA Overall The Privacy Rule and Security Rule of HIPAA apply to health plans, health care clearing houses, and any health care providers who transmit health information in electronic form about transactions for which the Secretary of HHS has adopted standards under HIPAA. Due to revisions made by HITECH, even business associates are covered and are required to follow HIPAA regulation. To ensure that business associates comply, they must sign business associate agreements. Moreover, subcontractors of business associates may also be required to comply with HIPAA regulation. In addition to the previously listed PHI, HIPAA also protects information regarding an individual’s: (i) Name, (ii) Specific Dates (birth, admission, discharge, death, etc.), (iii) Phone number, (iv) Social Security Number, (v) Medical Record Numbers, (vi) Photographs, and (vii) City info, ZIP, and other geographic location identifiers. The exceptions of HIPAA apply to the people who are not required to abide by HIPAA regulations include, but are not limited to, life insurers, employers, workers’ compensation carriers, schools and school districts, certain state agencies, law enforcement, and many municipal offices [9].

3

Figure 3. Percentage of hospitals using e-PHI The above increasing trends tally with information presented in Figure 1 and gives idea about the complexity & size of the problem, as EMRs have become substantially more commonly used. It becomes even more concerning when we observe that since 2013, the number people affected by breaches of their PHI via hacking and other IT methods has increased exponentially, which is depicted in Figure 4.

Security Concerns

Even with HIPAA in place, there is always a level of security risk. Security breaches regarding medical information is becoming a common form security issue. Health and medical companies are becoming more lucrative targets of hackers and, as a 2014 report showed, account for nearly 43% of reported breaches [12]. According to data released by Office for Civil Rights, Dept. of Health and Human Services of USA, there is growth in number of breaches of health-related information [22]. The following figure shows the number of PHI breaches between 2009 – 2016, where each breach has affected at least 500 individuals.

Figure 4. Number of individuals affected by breach

Figure 2. Number of PHI Breaches Also, according to joint ONC/AHA report (September 2016), 95% of the hospitals now allow patients to access e-PHI for viewing [21]. This is four times increase over a four-year period, as shown in Figure 3.

The above figure shows that there is a very sharp rise in number of individuals affected by security breaches of health records in 2015. While non-IT/hacking related incidents (for example, theft, loss, improper use or disposal etc.) were the major contributor to affect individuals between 2010 – 2014, hacking or IT-related incidents contributed more than 99% to the total number of individuals affected in 2015 [14]. This forced us to examine further and identify the source of the security breaches. We compared the sources of health information breaches in 2010 and with that of 2015. The pie charts in Figure 5a and 5b show the percentage of individuals affected by information breaches in different sources like desktop computers, laptops, portable devices, emails, paper documents, network servers, and other forms of breaches, in the year 2010 and 2015 respectively. In 2010 majority of the individuals

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

73

•

affected were related to other form breaches and by information breaches from laptop.

•

Budget: because there is a scarcity of IT experts, their salary is significantly higher; thus, healthcare facilities cannot afford to hire more. Outdated Technology: Healthcare facilities are usually behind other industries in adapting new technologies, which results in them outdated technology [12].

Level of impacts of each of the above on secure management of EMRs can be investigated separately. However, we did not include that analysis within the scope of the current article.

4

Figure 5a. Percentage of individuals affected by different sources of breaches in 2010 However, it is observed that in 2015 majority of the individuals (95% of who were affected in some form of health information breaches) were affected by security breaches in network servers, as evident from Figure 5b.

Figure 5b. Percentage of individuals affected by different sources of breaches in 2015 This rise in percentage of individuals affected by network related breaches can be attributed to increase in use of EMR and their storage on various networked storages (like web servers and Clouds) and exchange over public networks like the Internet. Therefore, it is of utmost importance to store the PHIs securely on the servers and use encrypted/secure channels for sharing these sensitive records.

3.1 Contributing Factors There are several reasons why these numbers are so high. As discussed above, rise in use and exchange of e-PHIs in recent years is one of the main contributors, but there are other factors that really make impacts are: •

Lack of trained IT experts in the field of security, especially, secure management of e-PHIs.

Potential for Damage

The potential for damage varies greatly when it comes to PHI breaches. The breaches can come from mainly two sources, outside attackers and insiders (individuals involved in creation and maintenance of the EMRs). First, let’s examine the potential damage that can be caused by outsiders:

4.1 Impact of Outsiders Attack There are basically two possibilities when a hacker or cyber-attacker gains access to an individual’s medical record – either they cause loss of identity of the victim resulting in financial issues of varying degrees (small to massive), or they cause little or no damage at all. For example, if a medical care employee left their personal laptop in their vehicle, and on the laptop that individual had sensitive, unencrypted medical information in one of the folders. If someone broke into the car and stole the device the security of those files has been compromised, technically, because the criminal now has potential access to this sensitive information; however, this does not necessarily mean that the criminal will use, or can use, or even successfully access, those files on the computer. It is quite possible that the criminal was not after the data on the laptop, rather he was after the laptop itself. In such a scenario, no damage was caused through the breach of that information. On the other hand, attackers who are intentionally after this information can use the information for identity theft, credit card theft, blackmail, and even to obtain prescription medication or services on the victim’s behalf. Following we highlight some concrete examples of breach incidents of this sort happened in the USA. In June 2014, in Montana, hackers could breach the security of EMRs that compromised approximately 1.3 million people’s Social Security Numbers and other personal information. To compensate the damage, individuals whose information may have been compromised were given free credit monitoring and identity protection insurance. These “free” services were covered by a twomillion-dollar state insurance policy regarding cyber security [20]. In another incident, later in the same year, administrators at Clay County Hospital in Illinois received a notification demanding a ransom be paid to protect the PHIs. The hackers threatened to release the PHI information gathered from the hospital. The hospital notified the FBI and a formal investigation was conducted. No one was arrested and no information was released as the ransom letter had stated [18].

ISBN: 1-60132-459-6, CSREA Press ©

74

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

4.2 Example of Insiders Attack PHI is not only susceptible to outsider attacks. Employees of a healthcare provider can also compromise sensitive information. As a concrete example, at Howard University Hospital in USA, a medical technician obtained 40 medical records and sold them for a total of $2,100. The person was found guilty of violating HIPAA, which has a maximum sentencing of 10 years in prison [19].

4.3 Role of Healthcare providers The healthcare providers have a significant role in the context of potential damage in an EMR data breach. They are the primary target of these breaches. From the data, available from [22], Figure 6 shows nearly 78% of healthcare data breaches were targeted to healthcare providers in the year 2016, which is a rise in 7% points compared to that of 2015. Figure 7. Percentage of individuals aware of medical identity theft based on sources of information [18]

5

Figure 6. Major targeted entities in healthcare data breaches This imposes greater responsibilities to healthcare providers in protecting individual’s health related information. In accordance to HIPAA, when there is a breach of patient PHI, healthcare providers are required to notify the US Department of Health and Human Services, all victims affected, and give a press release should the breach affect 500 or more people. If there is a breach that affects less than 500 people, healthcare providers are only required to include that information on an annual report to the government and to notify all affected individuals. However, in a survey of victims affected by medical identity theft, only about 25% of them were actually notified. Figure 7 shows the sources through which individuals came to know about medical identity theft related to them. This shows that the healthcare providers could not meet the expectation of notifying individuals about potential or actual breaches related to their accounts. One of the reasons for this failure could be the cost, as the average cost to correct implications caused by medical identity theft is $13,500 [18]. This lead to the conclusion that to reduce the overhead incurred due to breaches (especially medical identity theft), there is a need for stronger security mechanisms to protect electronic medical records.

Ways to Mitigate Breaches

In this section, we briefly discuss the possible preventive mechanisms that can be adopted to alleviate the security breach issue of EMRs. There is no single security mechanism to achieve that, rather there are several approaches that the healthcare facilities can take to mitigate the number of security breaches and strengthen existing security and HIPAA regulations. The following is a suggestive list of approaches that include: • Investing in well-educated and well-trained IT experts • Periodical security risk analysis, vulnerability assessments, and testing • Separation of data Let us elaborate on the above approaches.

5.1 Investing in IT-experts Many IT firms hire experts based on experience rather than education. This causes many young but talented IT professionals, like fresh graduates, to be overlooked because they have relatively less number of years of experience. Regular training seminars for current employees would also be of great benefit. This training should be directed towards staying up to date with new technology or recent security breaches, discussing topics such as: how a security breach happened and how it could have been prevented. Healthcare facilities may also benefit by building stronger relationships with employers and Universities to better train future graduates. Expanding internship and training programs to educate young professionals is another solution [12].

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

5.2 Periodic Security Risk Analysis & Testing Healthcare facilities should conduct an annual HIPAA security risk analysis. This essentially checks for new vulnerabilities that may have arisen from changes made throughout the year, such as new employees, system deployments, and IT enhancements. Many breaches are caused by theft of a portable devices containing unencrypted information, as supported by Figure 5. Because of this, simply encrypting data-at-rest would cut down largely on the number of e-PHI breaches. This also includes hardware inside offices. Any data that is being stored should be encrypted so sensitive information cannot be retrieved [13]. To further strengthen ePHI security, healthcare facilities could conduct more frequent vulnerability assessments and penetration testing to search for potential security risks and liabilities [13].

5.3 Separation of Data Another good security procedure is keeping guest and professional data separate by using different wireless internet network access for each. If this is not done, then the potential for holes in security and breaches go up. The access keys for these networks should be changed regularly. Finally, access to sensitive data should be carefully monitored and regulated. Access should only be allowed on a need-to-know basis and in accordance with HIPAA guidelines [17]. However, none of the above approaches is alone sufficient to provide desired level of security for sensitive electronic medical records. Our observation is, a combination of variety of approaches, as appropriate, should be adopted and deployed to countermeasure the threat of security breaches of EMRs.

6

Conclusions

In this paper, we examine the issues and importance of securing electronic medical records (EMR) of individuals. We investigated the role of government legislations in the context and the security concerns. We discussed the trend in security breaches, potential sources and major targets in these breaches, and the consequences (extent of potential damages). After careful examination, we draw the following conclusions: the use of EMRs (or, EHRs) has increased in recent years, particularly due to the incentives offered to participants who can demonstrate meaningful use of it. Another factor for the rise in usage is the penalties incurred by not implementing and showing meaningful use. Finally, one of the major aspects is the convenience it provides for clients and medical professionals. U.S. Congress, anticipating a need for security of medical data, enacted HIPAA, which was further revised into its current version by the ARRA and HITECH Act. HIPAA provides an infrastructure of security guidelines and regulations that are intended to secure an individual’s PHI. It is composed primarily of two major rules: The Privacy Rule, which protects an individual’s sensitive medical information, and the Privacy rule, which protects the confidentiality, integrity, and availability of that information. Even with HIPAA in place, EMRs and EHRs are increasingly becoming

75

a bigger target for hackers. Challenges that are making these quite susceptible to breaches include lack of IT experts involved in the healthcare domain, difficulty in hiring desired number of experts due to budget constraints, and latency in using state-of-the art security technologies for protection of PHI. Should an individual’s medical and health records be compromised, the financial damage done can be devastating, which is why it is important to increase security by utilizing more preventative measures. Some of which are as simple as: encrypting data-at-rest, keeping professional wireless networks separate from public ones, and running frequent tests and diagnostics to check for potential security vulnerabilities.

7

References

[1] M. Rouse. (2010, March). ARRA (American Recovery and Reinvestment Act of 2009. Available: http://whatis.techtarget.com/definition/ARRA-AmericanRecovery-and-Reinvestment-Act-of-2009 [2] Congressional Budget Office. (2012, February). Estimated Impact of the American Recovery and Reinvestment Act on Employment and Economic Output form October 2011 Through December 2011. Available: www.cbo.gov/sites/default/files/cbofiles/attachments/02-22ARRA.pdf. [3] Bisk. (2016, June). Federal Mandates for Healthcare: Digital Record-Keeping Requirements for Public and Private Healthcare Providers. Available: https://www.usfhealthonline.com/resources/healthcare/electro nic-medical-records-mandate/#.WDvlafkrLIU. [4] HealthIT. (2015, February). Meaningful Use Definition and Objectives. Available: https://www.healthit.gov/providers-professionals/meaningfuluse-definition-objectives [5] S. Gokak. (2014, June 1). Avoiding the 2015 Medicare EHR Incentive Program Penalty. Available: http://bulletin.facs.org/2014/06/avoiding-the-2015-medicareehr-incentive-program-penalty/ [6] Myemr360. (2011). Federal EMR Electronic Medical Records Mandate 2014/2018 Deadline. Available: http://myemr360.com/emr-mandate-2014 [7] B. Monegain. (2015, September). More than 80 Percent of Docs Use EHRs. Available: http://www.healthcareitnews.com/news/more-80-percentdocs-use-ehrs/ [8] Office for Civil Rights. HIPAA for Professionals. Available: http://www.hhs.gov/hipaa/for-professionals/ [9] Office for Civil Rights. Your Rights Under HIPAA. Available: http://www.hhs.gov/hipaa/forindividuals/guidance-materials-for-consumers/

ISBN: 1-60132-459-6, CSREA Press ©

76

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[10] Office for Civil Rights. Summary of the HIPAA Privacy Rule. Available: http://www.hhs.gov/hipaa/forprofessionals/privacy/laws-regulations/ [11] Office for Civil Rights. Summary of the HIPAA Security Rule. Available: http://www.hhs.gov/hipaa/forprofessionals/security/laws-regulations/. [12] T. Cannon. (2015). The Root of the Problem: How to Prevent Security Breaches. Available: https://www.wired.com/insights/2015/02/the-root-of-thesecurity-problem/ [13] D. Manos. (2014, February). 5 Ways to Avoid Health Data Breaches. Available: http://m.healthcareitnews.com/news/5-ways-avoid-healthdata-breaches

[20] L. Zuckerman. (2014, June). Montana Health Record Hackers Compromise 1.3 Million People. Available: http://www.reuters.com/article/us-usa-hacker-montanaidUSKBN0F006I20140625. [21] HIPAA Journal (2016, September). Number of Hospitals Sharing ePHI. Available: http://www.hipaajournal.com/sharing-of-health-data-withpatients-95-of-hospitals-offer-ephi-access-3597/. [22] HIPAA Journal (2017, February). Largest Healthcare Data Breaches. Available: http://www.hipaajournal.com/largest-healthcare-databreaches-of-2016-8631/.

[14] Office of the National Coordinator for Health Information Technology. (2016, February). 'Breaches of Unsecured Protected Health Information,' Health IT QuickStat #53. Available: https://dashboard.healthit.gov/quickstats/pages/breachesprotected-health-information.php [15] Office of the National Coordinator for Health Information Technology (2014, April). 'Percent of Hospitals Able to Send and Receive Secure Electronic Messages Containing Patient Health Information to and from External Sources,' Health IT Quick-Stat #27. Available: https://dashboard.healthit.gov/quickstats/pages/FIG-HospitalCapability-Secure-Electronic-Messaging.php [16] Office of the National Coordinator for Health Information Technology (2014, February). 'Percent of Physicians e-Prescribing through an Electronic Health Record,' Health IT Quick-Stat #17. Available: https://dashboard.healthit.gov/quickstats/pages/FIG-PercentPhysicians-eRx-through-EHR.php. [17] WMX (2016, February). How to Prevent Data Breaches in Healthcare Organizations. Available: https://www.mdcyber.com/blog/how-to-prevent-databreaches-in-healthcare-organizations/. [18] A. Sankin. (2015, April). The Real Reason Hackers Want Your Medical Records. Available: http://kernelmag.dailydot.com/issue-sections/features-issuesections/12688/identity-theft-medical-records-healthcare/. [19] FBI. (2012, June). Former Howard University Hospital Employee Pleads Guilty to Selling Personal Information About Patients. Available: https://archives.fbi.gov/archives/washingtondc/pressreleases/2012/former-howard-university-hospital-employeepleads-guilty-to-selling-personal-information-about-patients.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

77

Targeted Data Swapping and K-Means Clustering for Healthcare Data Privacy and Usability Kato Mivule [email protected] Department of Computer Science Norfolk State University

Abstract – Healthcare data regulations and laws require that entities protect patient privacy when managing healthcare records that contain personally sensitive and revealing information. However, during the process of implementing data confidentiality on the healthcare records, the usability of the privatized healthcare data reduces as sensitive information is removed or transformed. Realizing equilibrium between the need for healthcare records privacy and the usability of the privatized dataset is problematic. In this study an analysis of healthcare records privacy and usability using targeted data swapping and K-Means clustering is done. Focus is placed on the healthcare data with the main problem being, how can such data be transacted securely, confidentially, while retaining the data usability. The question being probed is, can healthcare data such as, the heart disease dataset be securely and confidentially shared while maintaining data usability. Keywords: Healthcare Data; Privacy; Usability; K-Means

I. INTRODUCTION Healthcare data policies necessitate that organizations handling healthcare related data safeguard patient privacy when transacting in healthcare records. The basic feature in such a data privacy process is to strip the information of any sensitive or revealing patient data. However, during this privacy procedure the usability of the privatized healthcare data decreases as useful information that might be sensitive is suppressed or altered. Further compounding this problem is that realizing equilibrium between the need for healthcare records privacy and the usability of the privatized dataset is intractable [1]. Attention in this study is placed on the healthcare data. The key problem being examined is how can healthcare data get transacted securely and confidentially between a patient and a healthcare entity while maintaining the usability of such data. In this study, an investigation of data confidentiality and usability using targeted data swapping and K-Means clustering unsupervised learning as an assessment is done [2]. Rather than focus on data swapping of every record in the dataset, attributes with sensitive information are selected for data swapping, then the records of that particular patient x are swapped with patient y, making if difficult to reconstruct the original data.

II. BACKGROUND Data swapping is a data privacy algorithm proposed by Dalenius and Reiss (1978), that consists of an swapping of items in a variable from the same dataset while conserving the original statistical traits of the data in that variable [3] [4]. Data swapping techniques maintain original statistical traits of data and is approvingly employed as a data privacy technique by the US Census Bureau [5]. The 2k data swapping model: The following is the definition for the 2k basic data swapping procedure [6]: Given an N x V data matrix: Where N is the total number of records, the symbol V is the total number of variables, where Xj is the jth variable in the data matrix, the symbol ith is the ith record of Xj such that xi = (xi1, xi2,…,xiv), and the symbol k representing the elementary swaps. Then the True Swap is the complete exchange of every value in the variable. Therefore 2k data swapping can basically be described as an exchange of 2k values in relation to the k elementary swaps and or a random selection of two records i and j from a variable and swapping them [6]. The Swap Rate is then the swap rate is then the fraction of N records to be swapped = 2k/N. Swap Pair: The swap pair is then the pair of values to be swapped = (i, j). The Post 2k data swapping scenario is best shown as follows [7]: If the values i and j of a variable X1 are swapped then the post swap condition for the ith and jth values will be (xj1, xi2,…xiv) and (xi1, xj2,…xjv ). K-Means (KNN): is an unsupervised machine learning method that groups similar values together in cluster by computing the distance between any values and a central point k, and grouping all items that are close to that central point, k. The k value is tunable; in other words, the user can select the number of k central points for which they expect values to accumulate around. K-Means employs the Euclidean distance in computing the distances between values and a central mean value and is formally articulated in the following formula [1]: 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑥, 𝑦 =

! !!!(𝑥!

− 𝑦! )!

(1)

The Davis Bouldin index (DBI): The DBI is a metric used to gauge how well a clustering algorithm performs. DBI index is formally noted as follows [2]: 𝐷𝐵𝐼 =

! !

! !!! 𝐷!

Where 𝐷! = 𝑚𝑎𝑥 𝑅!,! !:!!!

ISBN: 1-60132-459-6, CSREA Press ©

(2) (3)

78

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

And 𝑅!,! =

!! !!!

(4)

!!,!

The symbol Ri,j is a measure of how decent the clustering is. The symbols Si and Sj are the observed distance inside each cluster. The symbol Mi,j is the distance among clusters. III. METHODOLOGY The following steps are followed in the implemented heuristic. (i) In the first phase of the heuristic, personal identifiable information (PII) and any sensitive patient data are removed from the dataset. (ii) In the second phase, attributes with sensitive data are chosen and selected for the data swapping. (iii) In the third phase, a 2k random swapping is done until a true swap is achieved. Random swapping involves arbitrarily selecting values for swapping, with a goal of concealing all records in the dataset [8]. However, in this study, swapping is targeted at specific attributes. (iv) In the fourth phase, K-Means clustering is used to gauge how well the privatized dataset compares to the original. In this case, the Davies Bouldin Index (DBI) is used to examine the performance of the clustering.

case, 13 attributes were selected for data swapping. The Patient ID was not selected for swapping. The notion here being that the records of patient x is swapped with those of patient y. The other factor is that data swapping has to be applied on a case-by-case basis. In this study, all the 13 features are necessary in determining the outcome if the patient has an indication of heart disease or not. A privatized dataset was then generated after the data swap. The next phase done in the experiment was to test for data usability – how usable would such datasets be to any other medical or research entity. K-Means algorithm was implemented using RapidMiner environment [10]. In the next section, an exploration of preliminary results from the study is given. V. RESULTS The results as shown in Figure 1, depict the K-Means clustering results of the original heart disease dataset before applying the data-swapping algorithm for privacy.

Figure 1: K-Means clustering results of the original data

Figure 1: The Targeted Data Swapping Model

IV. EXPERIMENT In the experiment done in this study, the heart disease dataset from the UCI repository is used [9]. The dataset contains 14 attributes and 303 instances. The 14 attributes include features such as the age, sex, chest pain type, resting electrocardiographic results, serum cholesterol, fasting blood sugar, resting blood pressure, among others [9] used to capture the heart state of the patient and predict the diagnosis of heart disease using the label number attribute; with the class values 1, 2, 3, and 4, with 0 indicating no presence of heart disease and 1, 2, 3, and 4, indicating presence and level of risk, 4 being the highest. The first step was to search for and remove any PII information. The next step was to select the attributes for the data swap. In this

Figure 2: K-Means clustering before data swapping – 12 attributes

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | The x-axis in Figure 1 represents the four clusters that are unlabeled and would correspond to the four classes (1 to 4) in the number label attribute, indicating the presence or absence of heart disease. The y-axis in this case indicates the age used as a basis to cluster and show which patients might or might not have heart disease. Each dot in the cluster represents the value or number of items in that cluster as shown more elaborately in Table 1. Figure 2 corresponds to Figure 1 and shows the clustering results before data swapping is implemented.

79

cluster 3 with fewer items than in the original data as shown in Figure 2. Table 1 shows the cluster model performance results.

Figure 5: Distribution of items in clusters before privacy

Figure 3: K-Means clustering results after data swapping

Figure 6: Distribution of items in clusters after privacy

Figure 4: K-Means clustering results after data swapping – 12 attributes

Each dotted color represents an attribute value used in the determination of the clustering process. In Figure 3 and 4 correspondingly, the clustering results of the privatized data after data swapping is presented. It can be shown clearly that the number of data items in cluster 3 of the privatized data is reduced as further elaborated in Table 1. Only five dots are shown in cluster 3. This corresponds to Figure 4 that shows

Figure 5 and Figure 6 further highlight this point by showing the distribution of items in the original clusters as compared with the clusters from the privatized dataset. For instance, cluster 1 has 96 items in the original data, while cluster 1 has only 5 items in the privatized data. This further demonstrates that data swapping distorts the statistical distribution of data. Additional, the study was interested in how efficiently the K-Means clusters. It is clear in Table 1 that the clusters contains 96, 96, 55, and 56 items each for cluster 0, 1, 2, and 3, respectively before data swapping. However, after aplying data privacy using data swapping algorithm, the number of items in each cluster changes to 78, 5, 93, and 127 for clusters 0, 1, 2, and 3 respectively. This would be an

ISBN: 1-60132-459-6, CSREA Press ©

80

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 | indication that while data swapin might have provided a layer of privacy protection for the patient record, the distortion is noticed in the privatized result. In this case, cluster 1 had originally 96 items but is reduced to 5. This could be problematic for a researcher or medical practioner using the privatized dataset. The original value could have indicated a group of patients with indicatio of heart disease while the privatized data could indicate fewer. Futhermore, the performance vector shown in Table 1 shows the variation bewteen the average within centriod distance in the clusters. For instance, cluster 1 shows that this value is 61.071 in the original dataset while in the privatized dataset – that is, after data swapping, shows a growth to 115.56. Yet again the results indicate the intractability bewteen privacy needs and usability. In such cases, trade-offs would need to be considered. As data privacy researchers have noted, perfect privacy can only be attained by disseminating nothing at all which would indicate no usability; on the otherhand, perfect data usability can be achievd by disseminating data exactly as it is received, yet this would indicate no privacy [11].

should be combined with initial data privacy sanitization such as, the removal of PII and other sensitive data to reduce chances of reconstruction attacks. Yet still data swapping of items in big datasets is conducive that it makes if difficult for an attacker to reconstruct the full identity of a patient record in such big datasets without prior information. Nevertheless, as the results indicated, data swapping distorts the values in the privatized dataset. Finding equilibrium between data privacy needs and data usability requirements, necessitates trade-offs, and remains a challenge in need for further study and research. Future works include pursuing investigative studies in various healthcare records and how such data can be privately shared while maintaining a satisfactory level of usability. ACKNOWLEDGMENT A special thanks to the Department of Computer Science at Norfolk State University for assistance with this research. REFERENCES [1] [2] [3] [4]

[5] [6] Table 1: Cluster Performance Results

Futhermore, the results shown in Table 1 indicate the Davies Bouldin Index (DBI) that shows how well the KMeans performed in clustering. A lower DBI indicates better clustering, while a higher DBI value indicates poor clustering. The DBI value is normalized bewteen 0 and 1. In Table 1, it is shown that the DBI results for the original dataset before applying data swapping is 0.092 which interestingly is higher than the DBI resulst for the privatized dataset after applying data swapping, at a low of 0.082. Due to the distortion of the data after data swapping, it was expected that the DBI value of the privatized data might be higher. However, the results show that it is important that other factors be considered when analyzing the performance of the datasets.

[7] [8] [9]

[10] [11]

K. Mivule, “An Investigation Of Data Privacy And Utility Using Machine Learning As A Gauge,” Bowie State University, 2014. K. Mivule and B. Anderson, “A study of usability-aware network trace anonymization,” in 2015 Science and Information Conference (SAI), 2015, no. February, pp. 1293–1304. S. P. Dalenius, T., Reiss, “Data-swapping: A technique for disclosure control.,” J. Stat. Plan. Inference, vol. 6, no. 1, pp. 73– 85, 1982. Tore Dalenius and Steven P. Reiss., “Data-swapping: A technique for disclosure control (extended abstract).,” in American Statistical Association, Proceedings of the Section on Survey Research Methods., 1978, pp. 191–194. and J. M. Fienberg, Stephen E., “Data swapping: Variations on a theme by Dalenius and Reiss.,” in Privacy in statistical databases, 2004, pp. 14–29. W. E. Winkler, “Masking and re-identification methods for public-use microdata: Overview and research problems.,” in In Privacy in Statistical Databases, 2004, pp. 231–246. S. P. Reiss, “Practical data-swapping: the first steps.,” ACM Trans. Database Syst, vol. 9, no. 1, pp. 20–37, 1984. P. J. Lavrakas, Encyclopedia of Survey Research Methods. Thousand Oaks, California: SAGE, 2008. R. Lichman, M. Janosi, Andras. Steinbrunn, William. Pfisterer, Matthias. Detrano, “Heart Disease Dataset - UCI Machine Learning Repository.” UCI Machine Learning Repository, p. [http://archive.ics.uci.edu/ml], 2013. R. Hofmann, Markus., Klinkenberg, Rapidminer: Data Mining Use Cases and Business Analytics Applications. Chapman & Hall/CRC., 2013. C. Dwork, “Differential Privacy,” Autom. Lang. Program., vol. 4052, pp. 1–12, 2006.

VI. CONCLUSION Preliminary results from this study indicate that data privacy could be implemented on healthcare data using data targeted data swapping. With the exponential rise of big data, data swapping becomes a suitable data privacy mechanism for implementing confidentiality. However, this

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

81

A Study on the regulatory oversight of Direct-to-consumer Genetic Testing in USA Jiantao Ding 1,2, Eunjoo Huisung Pacifici 3 Graduate student, School of International Pharmaceutical Business, China Pharmaceutical University; Nanjing,Jiangsu Province, 211198, China ; email address: [email protected] 2 Visiting Scholar, International Center for Regulatory Science, University of Southern California, USA; email address: [email protected] 3 Associate Director, International Center for Regulatory Science, University of Southern California; Los Angeles, California, 90089-9004, USA; email address: [email protected] 1

Abstract - In recent years, direct-to-consumer genetic testing via multiple media was forming a growing market and attracted more consumers to predict risks of diseases or other healthcare and understanding ancestral origins for individuals, families, or populations. It has affected not only on healthcare industry, but also on entire national potential economy. This paper reviewed relevant regulatory affairs and provided pharmaceutical business decisions with future development considerations. It gave a ongoing essay on the past and present regulatory oversight of Direct-to-consumer Genetic Testing in the United States. Through overall discussion on the industry, companies, regulatory history, and future overlook, the paper suggested that our government and society be working together toward a reasonable and fair approach to regulation that can give consumers confidence in direct-to-consumer genetic testing and facilitate progress in personalized healthcare. Keywords: direct-to-consumer genetic testing, healthcare industry, regulation, government regulation oversight

1

Direct-to-consumer Genetic Testing Industries

The purpose of genetic testing (GT) includes predicting risks of diseases, screening newborns, directing clinical management, identifying carriers, and establishing prenatal or clinical diagnoses or prognoses in individuals, families, or populations[1]. Direct-to-consumer (DTC) genetic testing (GT) refers to genetic tests that are marketed directly to consumers via television, print advertisements, and the Internet mostly today[2]. As a prime mover, diagnosing genetic diseases drives DTC GT preponderantly to growing public awareness. Meanwhile, healthcare and understanding ancestral origins play an instigation part on DTC GT in personal ongoing life. With gaining prominence over the past couple of years, DTC GT production from many companies shows the major source of revenue in healthcare market and contributes to our economy as an indicator of the relevant industries. DTC GT provides only one piece of information about a person’s body system but applies in a broad range,

from single-gene disorders to complex, multifactorial diseases. Family medical history, lifestyle changes, and the other genetic and environmental factors also result in a person’s risk of developing potential diseases but in many cases are not mentioned by DTC GT. In 2001, the Secretary’s Advisory Committee on Genetic Testing recommended that the Food and Drug Administration (FDA) be involved in the review of all new GT regardless of how they were formulated and provided. In 2008, the Secretary’s Advisory Committee on Genetics, Health, and Society recommended that FDA address all GT using a risk-based approach. In 2010 and afterwards, the expenses of GT solutions dropped sharply in prices and the market is mostly shared by a few players. The use of genetic data is yet not explored in depth, however both the demand of increasing researches and the solutions of growing availability will pave way for upcoming improvements around the business context. Meanwhile, high risks to the market growth posed by improper health advices and decisions sans involvement of physician assistance and the associated repercussions, like vulnerability to misleading results from unproven or invalid diagnosis. Another challenge that may hold back the pace of this market is potential risk of invasion of genetic privacy through unauthorized use of consumer data. According to the report data[3], DTC GT market was valued at US$ 70.2 Mn in 2015, and is expected to reach US$ 340 Mn by 2022, expanding at a Compound Annual Growth Rate (CAGR) of 25.1% from 2016 to 2022. Andelka M. Phillips from University of Oxford provides a brief introduction to the DTC GT industry and the issues it raises for the law [4].

2

Direct-to-consumer Genetic Testing Companies

The DTC GT market is highly concentrated, with a few companies offering testing solutions, including 23andMe[5], GeneByGene[6], Genetrainer[7], Myriad Genetics[8], Genecodebook Oy[9], MD Revolution[10], and so forth. Many others work in relevant businesses, like Ancestry[11] providing DTC DNA tests for genealogy purposes, and Navigenics[12] staffing a physician and offering genetic counseling[13]. Though there are more than 700 tests

ISBN: 1-60132-459-6, CSREA Press ©

82

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

available, only a handful are approved by the FDA. For GT industry, DTC GT is an emerging market segment. The companies were eager to advertised and popularize GT through a variety of channels, such as the Internet-based social media. The FDA has been aware of these companies marketing to consumers for years. In 2006, Government Accountability Office (GAO) investigated DTC testing and found that most of the diagnostics were nutritional genetic tests to assess what kinds of foods individual consumers should eat and dietary supplements they should take. The FDA followed up with the companies. FDA, CDC, and FTC published a cautionary statement on DTC GT. After that time, more and more DTC GT companies subsequently came into the market. Since 2007, the FDA’s Center for Devices and Radiological Health (CDRH) have met with the companies continuously and had a better understanding of what the companies were actually doing or going to do. Initially their business models were not clear and well-designed. Moreover, in many cases the link between the genetic results and the risk of developing a disease has not been well-established. Recently, companies more aggressively marketed DTC GT. Pathway Genomics corporation (PGC) was poised to offer its Genetic Health Report by using a saliva collection kit DTC GT[14] which was announced to be sold at many of Walgreens' nearly 7,500 stores nationwide[15][16]. 23andMe was marketing DTC GT online and partnering with retailing industry companies. AncestryDNA [17] was partnering with Amazon.com. Based on the Federal Food, Drug, and Cosmetic Act, from 2003 on, the FDA has removed lots of GT companies. Six years ago, the FDA informed to confine a few test-manufacturing companies and restrict their marketing behaviors, such as PGC; Knome, Inc. which was taken over by Tute Genomics in 2015, then acquired by PierianDX in 2016; Navigenics; deCODE Genetics headquartered in Reykjavik, Iceland[18]; 23andMe; and Illumina Inc.[19]. They generally have not lodged data on the analytical and clinical validity of their tests to the FDA for approval or clearance. Inter alia, many companies that have left the DTC GT market are an indication that hyped products and unrealistic expectations may not create the expected ROI. Further regulatory oversight may well make it impossible for DTC GT companies to operate by using the same business model in the future[20]. Presently, some companies who stopped their Internet-based DTC GT delivery but yet continued the DTC marketing are working through the public healthcare system. This system should deter collaboration with retailing industry companies while offering tests sans clinical utility and be used for testing implementation.

3

Direct-to-consumer Genetic Testing Regulatory History

The growing market for DTC GT may provide a way for consumers to be aware of their ancestral origins and explore, advise them to undertake a precautionary preparation for their healthcare, and promote knowledge of genetic

diseases. Therefore, a variety of DTC GT range from alleles testing linked to breast cancer to mutations testing linked to cystic fibrosis. DTC GT benefits encompass respectively promotion of proactive healthcare, approachability to consumers, and privacy of genetic information obtained. Possible attached risks of DTC GT are the lack of governmental regulation, the potential misinterpretation of genetic information, issues related to testing minors, privacy of data, and downstream expenses for the public healthcare system[ 20]. As far as GT and its data use are concerned nationwide, the Genetic Information Nondiscrimination Act prevents health insurance companies from refusing insurance coverage to a healthy consumer if his or her genetic predisposition to developing a disease in forthcoming years. At the same time, the legislation prevents employers from using individuals’ genetic information when making hiring, firing, job placement, or promotion decisions[21]. The legislation was passed by the United States Senate on April 24, 2008 and was signed into law by President George W. Bush on May 21, 2008[22]. It took effect on November 21, 2009. If GT is only designed for use in establishing clinic diagnoses, directing management of cures, mitigating risks of diseases, and other treatment activities, it is subject to FDA oversight. For GT testing, there currently are two ways used in clinical patients’ management, as is the case for other In Vitro Diagnostic tests (IVD). One is through the development of a test by labs for use only by those labs themselves. The other is through development of a business purposed test kit by an IVD device manufacturer for distribution to labs. The agency has some of genetic factors authorized from national governmental regulatory departments over these products and has the relevant tests approved. The former usually is called lab-developed tests (LDT). The FDA has the authority to regulate IVDs as it does all LDTs. FDA oversight of an IVD is rooted in the possibility of an erroneous test result. In 1976, the device law was passed. It now makes the FDA exercise enforcement discretion. In early years, tests made by labs were used as simple and easy-to-use ones for rare diseases’ diagnoses and lowered prognostic risks. The accuracy of the results had much to do with interpreter’s intelligence than to the test design technically. And now, most genetic tests being offered are LDTs. The FDA’s oversight of GT has generally been stared at commercial test kits. Also, for the use of IVDs as well as LDTs, The FDA’s oversight is focusing on making both of them a legitimate, consistent, and impartial approach to GT and ensuring their safety and innovation. By 2010, 353 U.S. labs offering GT were formally listed but actually more than 700 [1]. Making significant improvements of regulation requires a sense of urgency and strong leadership. DTC GT Companies should comply with the Heath Insurance Portability and Accountability Act (HIPAA) and maintain the privacy of all individuals’ genetic data and disclose their privacy protection policies.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

4

Future Overlook

As an advertisement for a BRCA-predictive GT for breast cancer stated: “There is no stronger antidote for fear than information”[23], the FDA has observed the following problems with DTC GT and stressed on relevant solutions in forthcoming years: Faulty data analysis, Exaggerated clinical claims, Fraudulent data, Lack of traceability/change control, Poor clinical study design, Unacceptable clinical performance, and so forth. The FDA insists on that DTC GT used for consumer care should have the same assurances of safety and efficacy regardless of being manufactured for distribution to labs or created for use in a lab. Prior to market, by the review of moderate and high risk, analytical validity and clinical validity of DTC GT are well-evaluated and ensured in light of their intended use. It presents standalone data assessment reports to backup analytical and clinical claims for those DTC GT. Non-validated tests produce higher occurrence of erroneous results and generate wrong diagnoses and poor treatment decision-making. Therefore, on the one hand, a reasonable regulation demands a pre-market review to ensure test labeling containing the test claims, the supporting data, the interpreted process, and the limitations. On the other hand, the post-market surveillance and enforcement tools from regulation perspective are required to keep tests safety and efficacy. Regulations would greatly influence the DTC GT market with exerting governmental power. In the current regulatory frameworks, many companies run their businesses in USA and confront the country regulatory oversight deficiencies in the future. As PGC and Walgreens planed to join forces to sell DTC GT, the FDA decided to investigate the DTC GT companies with discretion. The report Direct-to-consumer genetic tests: Misleading test results are further complicated by deceptive marketing and other questionable practices by the GAO is still expected to show its efficacy[24]. The FDA does require that most LDTs be reviewed for clinical validity [1]. With no regard to regulation clauses changed, it is always anticipated that regulatory oversight will be enhanced in upcoming years. For regulatory framework, its principles should be developed, updated, and improved by task forces including experts in regulation, representatives from the DTC GT industry, specialists in clinical disciplines, and molecular geneticists in biological science, genetic counselors in healthcare fields, and so on. From perspective of ISO standardized context, self-regulation of the DTC GT market is crucial to global GT standards promotion for commercial operators. The regulation establishment may not only protect consumers from hurts, but also expedite the industry growth. Additionally, federal, state & local governments will be working toward a reasonable and fair approach to regulation that can give individuals and organizations confidence in the DTC GT and facilitate progress in personalized healthcare. DTC GT are commonly used to improve the detection and treatment of diseases earlier, which results in market expansion gradually and regulation requirement urgently.

83

Moreover, it is necessary for extensive regulatory means made by the industry experts to validate the data analysis of market size. According to GlobalData[25], regulatory annualized revenues data on the United States Genetic Testing market involved value in dollars, upcoming years’ forecast, and the key market players’ profiles, like Hologic, Inc.[26], Transgenomic, Inc.[27] , Bio-Rad Laboratories, Inc.[28] and PerkinElmer, Inc.[29]. Meanwhile, for the sake of market evaluation, there is a need to segment them into different groups, including acquired gene, inborn gene and other GTs, even for global key companies operating within the United States Genetic Testing market, like F. HoffmannLa Roche Ltd.[30]. By meeting the requirements of regulation framework, those companies in the relevant industries should setup business strategic goals, be aligned with business market-entry and market expansion strategies, identify the key market segments posed for strong growth, and construct competition mechanism against the market in the near future..

5

Acknowledgement

This paper was supported by 2016’ University Graduate Students’ Practice Innovation Projects of Jiangsu Province, China (SJZZ16_0110).

6

References

[1] Jeffrey Shuren, “Direct-to-Consumer Genetic Testing and the Consequences to the Public”, https://www.fda.gov/newsevents/testimony/ucm219925.htm, July, 2010, retrieved on 2017-05-12 [2] NIH. “What is direct-to-consumer genetic testing?”. https://ghr.nlm.nih.gov/primer/testing/directtoconsumer, retrieved on 2017-04-18 [3] Credence Research. “Direct-to-Consumer Genetic Testing Market By Channel, Business Model - Growth, Share, Opportunities & Competitive Analysis, 2016-2022”. http://www.credenceresearch.com/report/direct-to-consumergenetic-testing-market, May 2016, retrieved on 2017-04-06 [4] Andelka M. Phillips, https://www.scl.org/articles/3160direct-to-consumer-genetic-testing-a-brief-introduction, August, 2014, retrieved on 2017-05-12 [5] https://www.23andme.com/, retrieved on 2017-05-10 [6] https://www.genebygene.com/pages/company, retrieved on 2017-05-10 [7] https://www.genetrainer.com/#about, retrieved on 201705-10 [8] https://myriad.com/, retrieved on 2017-05-10

ISBN: 1-60132-459-6, CSREA Press ©

84

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[9] http://www.finbio.net/en/bioinformatics-itservices/item/408-genecodebook-oy, retrieved on 2017-05-13 [10] http://mdrevolution.com/, retrieved on 2017-05-13

[24] United States Government Accountability Office.” Direct-to-consumer genetic tests: misleading test results are further complicated by deceptive marketing and other questionable practices”, http://www.gao.gov/products/gao10-847t, Jul, 2010, Retrieved on 2017-04-08

[11] https://www.ancestry.com/, retrieved on 2017-05-10 [12] acquired by Life Technologies in 2012, then acquired by Thermo Fisher Scientific in 2014 [13] Heidi C. Howard, Pascal Borry. “Personal Genome Testing: Do You Know What You Are Buying?” http://www.tandfonline.com/doi/abs/10.1080/1526516090289 4005. June, 2009, retrieved on 2017-04-06

[25] GlobalData. “United States Genetic Testing Market Outlook to 2021”, http://www.marketresearchstore.com/report/united-statesgenetic-testing-market-outlook-to-2021-40936, August, 2015, retrieved on 2017-05-15 [26] http://www.hologic.com/, retrieved on 2017-05-15 [27] http://www.transgenomic.com/, retrieved on 2017-05-15

[14] https://www.pathway.com/skinfit/#tab-23721-0, retrieved on 2017-05-15 [15] http://www.cleveland.com/business/index.ssf/2010/05/w algreen_to_hold_off_selling_g.html, retrieved on 2017-0515 [16] https://www.bioprocessonline.com/doc/pathwaygenomics-to-offer-retail-genetic-test-0001, retrieved on 201705-15

[28] http://www.bio-rad.com/, retrieved on 2017-05-15 [29] http://www.perkinelmer.com/genetics/, 2017-05-15

retrieved

[30] http://www.roche.com/about/our_purpose.htm, retrieved on 2017-05-15

[17] https://www.ancestry.com/dna/ , 2017-05-15 [18] https://www.decode.com/company/ , 2017-05-15 [19] https://www.illumina.com/company/about-us.html, 2017-05-15 [20] Borry, P., Cornel, M.C. & Howard, H.C..”Where are you going, where have you been: a recent history of the direct-to-consumer genetic testing market”. Journal of Community Genetics. https://link.springer.com/article/10.1007/s12687-010-0023z#CR7. Oct, 2010, Retrieved on 2017-04-06 [21] Statement of Administration policy. “Executive Office of the President, Office of Management and Budget”, https://www.genome.gov/pages/policyethics/geneticdiscrimin ation/saponhr493.pdf, April 2007, Retrieved on 2017-04-08 [22] Keim, Brandon. "Genetic Discrimination by Insurers, Becomes a Employers Crime". https://www.wired.com/2008/05/the-genetic-inf. May, 2008, Retrieved on 2017-04-08. [23] Sarah E. Gollust, Sara Chandros Hull, Benjamin S. Wilfond. "Limitations of Direct-to-Consumer Advertising for Clinical Genetic Testing," JAMA.2002; 288: 1762-1767, http://jamanetwork.com/journals/jama/articleabstract/195392#REF-JSC20258-14. October, 2002, Retrieved on 2017-04-08.

ISBN: 1-60132-459-6, CSREA Press ©

on

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Mobile Rescue Management for Medical Emergencies: Introducing a Novel Application Framework Martin Zsifkovits, Madalina Stefania Toader, Stefan Wolfgang Pickl Department of Computer Science, Universitaet der Bundeswehr Muenchen, 85577 Neubiberg, Germany e-mail: [email protected], [email protected], [email protected],

Abstract: In medical emergency situations, prompt and quickly first-aid provided to patients is essential and might make the difference between life and death. Especially in rural areas it is rather difficult to reduce “non-therapy-time”, as ambulances need to service large areas. One promising approach to overcome this issue is the training of volunteers as first responders. These trained people are equipped with some general equipment for first aid and can be called to emergencies in their vicinity until the ambulance arrives. However, an appropriate alarming system is key for the success of this approach. In a prior project we identified that there is no unique alarming system in use, which causes several shortcomings. In the paper at hand we introduce the implemented mobile application FRICS that optimizes the communication between the control center and medical first responders. Keywords: Mobile Rescue Management, First Responder, Control Center, Reachback

1. Introduction Mobile applications, designed for smartphones and tablets are gaining increasing importance in various fields, also in medicine. The prompt availability over the app market, the steady availability, as well as the connection to devices that are regularly in use are only some of their advantages. Besides many other fields, mobile applications are also available for medical support. Information on pharmaceutics, processing of health information such as pulse, blood pressure, mobility, or emergency applications that send messages for help in emergency cases are only some of the examples. Thus, most of the applications are focusing on one side only, meaning the consumer or patient. In the article at hand we introduce a new mobile application that allows for fast communication between trained medical first responders and the control center. In a previous study presented in [1], we introduce the topic of medical first responders in Austria and discuss the need of better coordination. The issue hereby is that individual, regional control centers follow their own rules and inform only first responders from their own district, regardless of their current position. The alarming procedure reaches from phone calls to sms-messages and mainly does not allow for

ISBN: 1-60132-459-6, CSREA Press ©

85

86

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

“offline modes”. The therewith coming problems are described and discussed in detail in [1]. In general, the novel application FRICS (First Responder Information and Communication System) aims at position-based alarming and equal distribution of information in real time. This would reduce needless actions from various first responders at the same time and keep the control center well informed. Furthermore, navigation and positioning for first responders is considered, as well as feedback to the control center. Based on the positive feedback on this previous work from various practitioners, we developed the mobile application FRICS in a Beta phase. This implemented framework is presented in this article. Furthermore, we discuss potential further fields of application of this framework. The reminder of this article is structure as follows: In chapter 2 we present a brief taxonomy on existing solutions for managing rescue operations and health issues. In chapter 3 we present the implementation of our solution FRICS and we conclude the article in chapter 4.

2. Taxonomy In the field of mobile devices and applications, various innovations from different field are brought to the market in very short time intervals. Before we introduce our approach, we present a brief taxonomy on existing solutions for mobile devices from the healthcare domain that might also be relevant for the domain at hand. [2] introduced a PDA-based system used for collecting and handling emergency medical care data for improving the effectiveness of rescue operations, while [3] introduced a mobile device with functions for emergency situations. The device is provided to send/receive operations of mobile radio communication, characterized with a global positioning system receiver, a memory cell for store information, and an emergency key for commanding the mobile phone. There is also a method of transmitting SOS signals in a mobile telecommunication terminal that provides a menu to select one of the SOS phrases upon entry into a SOS service mode [4]. Another device is used as a pervasive healthcare gateway that collects data via Bluetooth/Wi–Fi and uploads the data to the back-end server in desired format. Therefore, devices like ECG, pulse oxygen meter, body fat analyzer, or blood pressure monitor are used [5]. MobileMap is a low-cost mobile collaborative application used in emergency situations to complement the radio communication system [6]. It was considered to extend the location-based services with a specific mobile decision support functionality, as decision support capabilities are limited to navigation support and database querying with no analytic evaluation of the attractiveness of alternative destinations being offered [7]. In this context, also an improvement regarding navigation using a mobile map-application was introduced that can be facilitated with sensor-context aware adaptations, like automatic map position and alignment, as well as with intuitive gesture-based control [8]. For rural areas, an integrated GIS and GPS-based system was introduced in order to meet the unique challenges of first responders by providing those with critical route and site specific information [9]. Furthermore, communication between first

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

responders can be improved using MyFlare, a communication app for mobile devices. Here, preconfigured messages are sent to a preconfigured set of emergency contacts. The application can also be configured to capture additional information, such as GPS location of the device, photo, video, or sound recording, in preconfigured time intervals [10]. A similar concept was used in the field of firefighters and other emergency first responder training with the purpose to allow first responders to receive and visualize text messages, iconic representations and geometrical visualizations of a structure as transmitted by the incident commander from a computer or other device, either on scene or at a remote location [11]. Medical first responders can exchange information about patient’s states with emergency department practitioners using an inter-organizational system [12]. Another internal patient monitoring system was developed that integrates vital signs sensors, location sensors, ad-hoc networking, electronic patient records and web portal technology to allow remote monitoring of patient vital-sign status [13]. Improvements in communication were made for first responders including fire, police, border patrol, emergency medical service, safety etc. – communication between multiple on-scene agencies and various command and control personnel by automatically providing positioning information as well as other sensor information [14]. First responders collaborate also with civil engineers using an information-technology-based collaboration framework that facilitates disaster response operations by accessing critical building information [15]. For general communication in disaster cases, an app was designed that allows a public service agency to provide alert information to first responders – an automatic call back from first responders triggers a voice call launching a location fix on the current location of the first responder. Delivery confirmation is received and driving directions with map images are sent to the first responder based on their current location and desired destination for response [16]. The communication between entities was the focus of even more projects, e.g. a communication server using a mediator unit combining potentially non-compatible communication systems [17], dual-use technology and build-in architectural and protocol redundancy for radio interoperability issues [18], or inter-working electronic media that operate under disparate protocols, and most particularly to improving access to emergency call centers from phone services that employ networks, such as the Internet, for packet-based data transmission [19]. An advanced hybrid satellite and terrestrial system for emergency mobile communications, quickly deployable and dynamically adaptable to any type of natural disasters and location was proposed as a potential solution to support communications in medical emergency situations [20]. Especially for tracking indoor operations of first responders, shoe-mounted sensors were presented [21]. The application of this technology is especially beneficial, when there is to GPS available and all dates are transmitted to the command and control system via a local infrastructure Wi-Fi network [22]. Another approach for tracking persons in indoor environments is presented by [23], where the system assumes no existing infrastructure, no pre-characterization of the area of operation and is designed for spectral compliance [23].

ISBN: 1-60132-459-6, CSREA Press ©

87

88

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

For the field of communication, detection, and navigation, the presented taxonomy represents the basis of our work. Based on this existing knowledge, we implemented the proposed solution in [1] for FRICS.

3. Implementation The main purpose of this application is to be used as a supplement to emergency rescue services. Following [1] this case is about trained volunteers who should arrive earlier at the scene than the rescue service crew, provide the first aid to the patients and so decrease the preclinical “non-therapy-time”. The service was mainly introduced for rural areas, where ambulances have to caver large geographic areas and therefore face longer drives. As the Android operating system has a current market share of over 81%, [24] the demonstrator of the tool at hand was implemented as an Android application in Java. For the server we used the Apache server with JEE-RESTful Webservice. This also offers information service and processing information that are generally independent of the hardware platform. We used the REST (Representational State Transfer) architecture, a resource-oriented architectural model for creating web services. The approach is different to web services. Technically, the REST architecture is described by some basics: resource, URI, representation, and uniform interface. The architecture enables function calls, method invocation, remote procedure calls and other messages are understood by a particular server or a small subset of components in the architecture. We chose to use a JAX-RS implementation (Jersey) because it is extensible, allowing easy addition of new features and uses Google Protocol Buffers. Both systems are designed for high performance and reduce CPU used during serialization and data non-serialization. As a database storage we used HeidiSQL – an open source front-end tool [25]. The communication chain between the application and the server is illustrated in Figure 1.

Figure 1: Communication chain

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

89

After downloading the application, the first responder is asked to register himself. This ensures that the application can only be used by professionals only and is not misused by not authorized people. After login in, the user is guided to the “Welcome” page, where he can set is status to online and receive alerts. This user diagram is illustrated in Figure 2. In case of an emergency, the GPS positions of online first responders are scanned within a predefined (Euclidian) distance. An alarmed responder gets more information on the case when desired and can then respond to the case with the decision of going there or not. After accepting the case, this information is processed in the control center and the status for every other first responder is updated. The notification diagram is shown in Figure 3.

Figure 2: User Diagram

Alerts are shown to the first responder in every mode of the phone. Thus, the FRICS app does not have to be open. The alert pops up and guides the user to the app. After accepting the case, the first responder gets further information on the case, such as the address and the degree of injury. An automated navigation to the scene of action is offered. In the case of no positive feedback from any first responder within the geographical vicinity, the distance towards next available first responders is enlarged as long as a shorter arrival time compared to the ambulance can be expected. The graphical representation of different states in the app are shown in Figure 4. Figure 3: Notification diagram

ISBN: 1-60132-459-6, CSREA Press ©

90

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 4: Representation of various states in FRICS

4. Conclusion In the paper at hand we present the implementation of a novel mobile application (FRICS) that improves the communication between medical first responders and the control center strongly. Furthermore, the application allows for position-based alerting and navigating. The solution was evaluated based on a theoretical concept presented in [1] and is now available as a Beta phase solution. The implementation plan thereby considered various shortcomings of available tools and platforms and considered concerns of data security. The logic of the communication between applications and the control center was kept rather simple and the navigation within the application was designed userfriendly. The implemented application was tested in various instances and represents a promising solution for the information, communication and navigation of medical first responders. The presented taxonomy in this paper implies that such an application might be extended to general disaster management or might be transferred to firefighters, mountain rescue services or police officers. For an in-depth analysis of these future potentials, further research is needed.

References [1] M. Zsifkovits, S.-H. Cheng, F. Waldner, K. Heidenberger, Smartphone-Based Coordination Support for the Austrian Medical First Responder System, in: Emerging Intelligent Data and Web Technologies (EIDWT), 2012 Third International Conference on, 2012, pp. 140–146. [2] L. Chittaro, F. Zuliani, E. Carchietti, Mobile devices in emergency medical services: user evaluation of a PDA-based interface for ambulance run reporting, in: Mobile Response, Springer, 2007, pp. 19– 28. [3] Y. Weng, Mobile phone device with function of emergency notification, Google Patents, 2003. [4] S.-B. Hong, Device and method of transmitting SOS signals in mobile telecommunication terminal, Google Patents, 2000. [5] Mobile healthcare infrastructure for home and small clinic, ACM, 2012.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

[6] Á. Monares, S.F. Ochoa, J.A. Pino, V. Herskovic, J. Rodriguez-Covili, A. Neyem, Mobile computing in urban emergency situations: Improving the support to firefighters in the field, Expert systems with applications 38 (2) (2011) 1255–1267. [7] C. Rinner, Mobile maps and more–extending location-based services with multi-criteria decision analysis, in: Map-based mobile services, Springer, 2008, pp. 335–352. [8] L. Arhippainen, T. Rantakokko, M. Tähti, Navigation with an adaptive mobile map-application: user experiences of gesture-and context-sensitiveness, in: Ubiquitous Computing Systems, Springer, 2005, pp. 62–73. [9] F.A. Needham, R.L. Anundson, G.R. Smith, K.F. Goschen, Responder route and site-specific critical data system, Google Patents, 2007. [10] B.E. Haimo, D.A. Kahn, Emergency communications mobile application, Google Patents, 2013. [11] M. Bastian, J. Ebersole, D. Eads, Method for displaying emergency first responder command, control, and safety information using augmented reality, Google Patents, 2002. [12] Process improvement and consumer-oriented design of an inter-organizational information system for emergency medical response, IEEE, 2011. [13] Improving patient monitoring and tracking in emergency response, 2005. [14] R. Burkley, C. Mason, G. Taras, C. Curran, I. Carreto, J. Cronin, First responder communications system, Google Patents, 2007. [15] F. Peña-Mora, A.Y. Chen, Z. Aziz, L. Soibelman, L.Y. Liu, K. El-Rayes, C.A. Arboleda, T.S. Lantz Jr, A.P. Plans, S. Lakhera, Mobile ad hoc network-enabled collaboration framework supporting civil engineering emergency response operations, Journal of Computing in Civil Engineering (2010). [16] M.A. Titus, J. Pohutsky, First responder wireless emergency alerting with automatic callback and location triggering, Google Patents, 2009. [17] K.L. Blossom, L.C. Foss, P.E. Leuba, N.D. Pereira, T.K. Som, Mediator based architecture for first responder interoperability systems (FRIS), Google Patents, 2009. [18] B.S. Manoj, A.H. Baker, Communication challenges in emergency response, Communications of the ACM 50 (3) (2007) 51–53. [19] S. Keagy, Apparatus and method for interfacing packet-based phone services with emergency call centers, Google Patents, 2008. [20] Advanced hybrid satellite and terrestrial system architecture for emergency mobile communications, 2008. [21] Omnidirectional pedestrian navigation for first responders, IEEE, 2007. [22] K.V. Hari, J.-O. Nilsson, I. Skog, P. Händel, J. Rantakokko, G.V. Prateek, A prototype of a firstresponder indoor localization system, Journal of the Indian Institute of Science 93 (3) (2013) 511– 520. [23] J. Duckworth, D. Cyganski, S. Makarov, W. Michalson, J. Orr, V. Amendolare, J. Coyne, H. Daempfling, D. Hubelbank, H. Parikh, WPI precision personnel locator system–evaluation by first responders, Proceedings of ION GNSS,(Fort Worth, Texas) (2007). [24] Jason Hahn, Android claims 81.5% of the global smartphone OS market in 2014, iOS dips to 14.8%Read more: http://www.digitaltrends.com/mobile/worldwide-domination-android-and-iosclaim-96-of-the-smartphone-os-market-in-2014, available at http://www.digitaltrends.com/mobile/worldwide-domination-android-and-ios-claim-96-of-thesmartphone-os-market-in-2014/ (accessed on December 16, 2015). [25] M.S. Toader, Smartphone Based Mobile Application for Medical Emergency Situations. Bachelor Thesis, Craiova, 2015.

ISBN: 1-60132-459-6, CSREA Press ©

91

92

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

What do College Undergraduates Know about Zika and What Precautions Are They Willing to Take to Prevent its Spread? Michele E. Miller1, William L. Romine1, Megan A. Rúa1, Environmental Sciences Program, Dept. Of Biological Sciences, Wright State University, 3640 Colonel Glenn Hwy, Dayton OH 45435

1

Abstract – We describe the development and validation of a Zika knowledge assessment and a survey for measuring precautions about Zika. We use these to explore conceptions Midwest United States college undergraduates have concerning Zika, compliance with precautions to prevent Zika, how likely college undergraduates are to take different precautions, and analysis of sentiments about Zika prevention. We found that the median undergraduate is likely to comply with all precautions except using an insect fogger and avoiding casual contact. Specific sentiments tended to accompany different levels of compliance. For example, undergraduates with high compliance expressed concerns rooted in fear and morality. Keywords: Zika mitigation, health informatics, quantitative analysis, knowledge, misconceptions Regular Research Paper

1 Introduction There are currently no self-report assessments for determining what conceptions university undergraduates have about Zika or the likelihood of taking precautions against Zika. Since Zika’s R0 value (the number of new cases one case generates over its infectious period) for different countries ranges from 2.0-6.6 [1], it is essential to understand compliance with precautions in an effort to mitigate the spread of Zika. This is especially important in university settings where crowded conditions prevail in all countries. The purpose of this study was to develop and validate two instruments, the “Zika Knowledge Assessment” (ZiKA) [15] and the “Zika Prevention Survey” (ZiPS) [15], and use these to explore three questions: 1) what conceptions do college undergraduates have concerning Zika; 2) how likely are college undergraduates to take precautions to prevent the spread of Zika; and 3) what sentiments about Zika prevention accompany different levels of compliance?

2 Literature Review

2.1 Zika Death due to Zika is rare and infection is often asymptomatic [2]. When people do have symptoms, they typically last a few days to a week and are mild. However, two serious conditions, Guillain-Barre syndrome and microcephaly, have been linked to Zika. The Zika virus is spread through vectors but can also be transmitted through contact with human fluids. The mosquito species that spread Zika, Aedes aegypti and Aedes albopictus, most aggressively bite during the daytime, but can also bite at night. Those most at risk of getting Zika include people that perform sexual acts with people that have Zika, fetuses of infected mothers, and people that live in or travel to an area with current Zika transmission who have been bitten by a mosquito. There is currently no medicine or vaccine to prevent or treat Zika. However, transmission can be prevented by abstaining from sexual activity or using protection, wearing long-sleeved shirts and long pants, staying in places with air conditioning, staying in places that have door and window screens, using an insect fogger, sleeping under a mosquito net, and using insect repellants.

2.2 Related Work Public conceptions of Zika have been explored through supervised machine learning, survey, and topic modeling methodologies. Dredze, Broniatowski, and Hilyard [3] explored misconceptions about the Zika virus vaccine using Twitter. They identified tweets making pseudoscientific, erroneous claims including that a mosquito larvicide, pyriproxyfen, is responsible for microcephaly even though scientists have found no link between microcephaly and larvicide. The second erroneous theory is that microcephaly is caused by side effects of existing vaccines, and by blaming Zika, pharmaceutical companies profit by having the opportunity to create new vaccines to sell. Another study [4] used a self-report survey methodology to measure the level of knowledge on symptoms, epidemiology and transmission of Zika in Colombia before and after a symposium on Zika in June-

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

93

July 2015. Knowledge improved significantly after the symposium. A study by Glowacki et al. [5] collected tweets during a live discussion about Zika on Twitter hosted by the CDC. Using topic modeling, they found some prevalent themes included symptoms, sexual transmission, consequences for infants and pregnant women, spread, and virology of Zika. A study by Miller et al. [13] used a combination of natural language processing and machine learning techniques to determine what people were tweeting about Zika. Tweets in each disease category (treatment, transmission, prevention, and symptoms) were then examined using latent Dirichlet allocation (LDA) to determine the five main tweet topics for each disease characteristic. The five topics for prevention were Zika control, money need, Zika prevention, legislative bill, and research. Vectors (mosquitoes), sexual transmission, infants, spread, and sports were the topics for transmission. Treatment topics were lack of treatment, Zika testing, vaccine development, blood testing, and lab test development. Finally, the five topics for symptoms were Zika symptoms, brain defects, confirmed defects, Zika is scarier than thought, and reports of Zika emergence and mortality. Finally, a study by Muppalla et al. [14] used text-based features, extracted with N-grams and Parts of Speech taggers, to build a better classifier to detect Zika related tweets from Twitter. They found that a simple logistic classifier was able to detect Zika tweets with 92% accuracy. Our assessment includes questions covering these topics to ascertain what misconceptions university undergraduates may have concerning these topics as well as compliance with precautions.

Validity of items for measuring the constructs of knowledge about Zika and precautions against Zika was evaluated based on mean squares fit with the Rasch model. We used infit, which is information-weighted to reduce the influence of outliers, and outfit, which is more outlier-sensitive. Expected values for these measures is 1.00. With respect to validity, we were primarily concerned with items with mean squares fit values above 1.30; such misfit indicates that the item favors undergraduates lower on the latent scale, implying a potential validity concern [6]. Precision of undergraduate and item locations along the latent scale was quantified with the Rasch reliability index, which ranges between 0 (no precision) and 1 (perfect precision).

3 Experiments and Analysis

3.4 The Zika Prevention Survey (ZiPS)

3.1 Study Context

The analysis of ZiPS focused on compliance with nine precautions (Table 1). Undergraduate self-reports were gathered using a Likert survey methodology (Table 1). An open-ended question asking undergraduates to explain reasons for compliance or non-compliance was also included.

One hundred fifty-eight undergraduates in a general education college biology class in the Midwestern United States participated in the study. Of those, 23% were male. Sixteen percent of the undergraduates were enrolled in STEM majors.

3.2 The Rasch Validity Model The Rasch model was used to provide a philosophical framework for validity of the items on both the ZiKA and ZiPS assessments. The Rasch model proposes that the likelihood of an undergraduate answering an item in the affirmative should be proportional only to the difference between the undergraduate’s and item’s locations on the latent scale. Like many latent variable models, Rasch models assume that all of the variables measure a single latent dimension, and that the items are independent after accounting for that latent dimension.

3.3 The Zika Knowledge Assessment (ZiKA) Analysis of ZiKA focused on fourteen multiple choice items along with their Certainty of Response (CRI) values. Four degrees of certainty were used in this assessment: “Complete Guess” was coded as 0, “Uncertain” was coded as 1, “Certain” was coded as 2, and “Very Confident” was coded as 3. A correct answer was diagnosed when an undergraduate made a correct answer selection and indicated a degree of confidence above guessing on the item [7]. An incorrect answer was coded as “0”. If a person got an answer correct but the CRI indicated guessing, the score was changed from a “1” to a “0” since guessing indicates that they do not actually know this information. Coded answers were then analyzed using the Rasch model.

Undergraduate reports were coded as follows: extremely unlikely=0, unlikely=1, likely=2, and extremely likely=3. The Rasch Rating Scale model (RSM) was used to model compliance with Zika prevention and difficulty of precautions on a common latent scale (in log-odds, or “logit,” units). The RSM was used to provide validity and reliability evidence for the assessment as well as to generate a model that is useful toward predicting the extent of college undergraduates’ compliance with precautions against Zika (Figure 2). Using qualitative open response data, we explored how undergraduates’ sentiments on Zika prevention related to their compliance levels.

ISBN: 1-60132-459-6, CSREA Press ©

94

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

Figure 1: Person-item map for concepts about Zika tested in the ZiKA assessment. Undergraduates (denoted by “#”) located above the item are predicted to have mastered the associated concept; those located below the item have not mastered the associated concept.

ISBN: 1-60132-459-6, CSREA Press ©

Int'l Conf. Health Informatics and Medical Systems | HIMS'17 |

3.5 ZiKA Validity Accuracy and precision were of primary concern when validating both ZiKA and ZiPS. For ZiKA, accuracy was checked by having a virology expert and a science education researcher evaluate the assessment for correct content. Items which both experts agreed were accurate and important with respect to prevention of Zika were retained for data collection and analysis. Rasch analysis indicated that all items fit well with validity expectations (0.77