A Development Methodology for a Stroke Rehabilitation Monitoring

Application

Pilar Mata

, Craig Kuziemsky

and Liam Peyton

Faculty of Engineering, University of Ottawa, 800 King Edward Avenue, Ottawa K1N 6N5, Canada

Telfer School of Management, University of Ottawa, 55 Laurier Avenue East, Ottawa K1N 6N5, Canada

Keywords: Clinical Performance Monitoring, Application Development Methodology, Stroke Rehabilitation

Monitoring, Performance Measurement, Healthcare.

Abstract: The capabilities of mobile devices (e.g. flexibility, portability, and the ability to retrieve information

quickly) have been leveraged for the development of clinical performance monitoring applications. In this

paper we assess the suitability of a methodology for development of clinical performance monitoring

applications to support stroke rehabilitation. We use a case study, with two use cases of patients recovering

from stroke events, to design a monitoring application at a conceptual level and compare it to other clinical

performance monitoring applications.

1 INTRODUCTION

Healthcare information technology (HIT) has the

potential to enhance care delivery by providing

timely access to data that can be used to deliver

patient centred care (IOM, 2012). A particular

benefit of HIT is the ability to monitor care delivery

across providers and settings (Xu et al., 2015).

Mobile devices can be particularly valuable for

monitoring care delivery. The capabilities of mobile

devices (e.g. flexibility, portability, and quick

information retrieval) have been leveraged in the

development of clinical performance monitoring

applications for practice profiling and community

care (Ferenchick et al., 2010; Ferenchick and

Solomon, 2013; Chamney et al., 2014).

Monitoring care delivery across diverse care

providers and settings requires defining

measurement goals and then consolidating data from

fragmented data sources to monitor the goals

(Vincent et al., 2014). For example, a clinical

process can involve multiple healthcare actors (e.g.

doctors, nurses, therapists, pharmacists) that

generate and store data in heterogeneous systems

(e.g. electronic health record systems, paper based

charts). Also, the same process may have different

workflows across providers, generating inconsistent

data. The overarching challenge is that data needed

to monitor a process may not be available at the

right time using existing data sources.

Developing performance monitoring applications

is a bounding problem as one needs to define the

objectives and goals to be monitored and then

identify and integrate the data needed for monitoring

to occur. There is also a need for user involvement

in the design of such applications, owing to the high

rate of failure in the implementation of HIT (Avison

and Young, 2007; Novak et al., 2012) that occurs

from a disconnect between clinicians and the HIT

development team (Avison and Young, 2007).

The predominant question is how to define the

goals and metrics to enable collecting the right data,

at the right time, for the right metrics. Much of the

existing work in this area has focused on reactive

responses to problems where issues are identified

after the fact (Kuziemsky, 2015). A better solution

would be to pro-actively identify and manage data

collection and integration issues so they can be dealt

with in real-time. However, doing that requires a

method that is robust enough to define and obtain

the necessary metrics and data but is flexible enough

to enable goals or metrics to be adapted as needed

(Vincent et al., 2014).

In this paper we assess the suitability of a

methodology for development of clinical

performance monitoring applications to support

stroke rehabilitation. We use a case study, with two

use cases of patients recovering from stroke events,

to design a monitoring application at a conceptual

400

Mata, P., Kuziemsky, C. and Peyton, L.

A Development Methodology for a Stroke Rehabilitation Monitoring Application.

DOI: 10.5220/0005785104000405

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 5: HEALTHINF, pages 400-405

ISBN: 978-989-758-170-0

level. Finally, we compare the development of the

stroke monitoring application against two other

clinical performance monitoring applications we

developed in previous research. We conclude with a

discussion of the implications for designing clinical

monitoring applications to support different types of

monitoring in healthcare delivery.

2 BACKGROUND

Stroke rehabilitation is the care processes that occur

after a patient has been stabilized from a stroke

event. Early rehabilitation interventions may

positively impact rehabilitation outcomes (i.e.

recovery of functional disabilities) and an integrated

multidisciplinary approach is key to ensure optimal

recovery (Duncan et al., 2005). Depending on the

severity of the event, the rehabilitation team may

include more than one care provider (e.g. physical

therapist, speech therapist, occupational therapist,

physician, nurse, or pharmacist). Family members

may also be involved in the rehabilitation process

(Gresham et al., 1997). Regular communication

between the care team on patient’s progress towards

common goals can positively impact patient's

rehabilitation outcomes (Cifu and Stewart, 1999).

An individualized stroke rehabilitation plan is

designed for each patient and it includes specific

rehabilitation goals and targets for each of the

exercises. The goals are defined in agreement with

the patient, family and care team (Gresham et al.,

1997). Monitoring progress of the patient’s plan

towards meeting rehabilitation goals and exercise

targets is important in order to identify gaps and

make adjustments as needed (Gresham et al., 1997).

One of the most commonly measurements used to

assess patient rehabilitation progress is the

Functional Independence Measure score (FIM)

(Duncan et al., 2005; Brown et al., 2015).

HIT is one way to increase quality of patient care

through efficient coordination and deployment of

resources in the community (Chukmaitov et al.,

2014). Recent studies have explored the use of

mobile applications in clinical settings to support the

provision of better care (Baarah et al., 2012;

Ferenchick and Solomon, 2013).

A Clinical Performance Monitoring Application

(CPMA) is a type of Business Intelligence

application (Chamney et al., 2014) that collects and

integrates data from various data sources in order to

compute metrics to instantiate goals related to the

performance of a particular clinical task or process.

In previous research, we developed a methodology

for development of CPMAs (Mata et al., 2015). Two

applications were used to show proof of concept of

the methodology: the Standards and Indicators

Dashboard (SAID) (Mata et al., 2014) and the

Resident Practice Profile (RPP) (Chamney et al.,

2014).

One core aspect of the development

methodology is to leverage user-centred design

methods to ensure user acceptance and adoption.

User centred design involves “users for a clear

understanding of user and task requirements,

iterative design and evaluation, and a multi-

disciplinary approach” (Vredenburg et al., 2002).

3 DEVELOPMENT

METHODOLOGY FOR CPMA

The development methodology for CPMA is a user-

centred design approach that engages users,

developers and project champions in an iterative

process of application modelling, implementation

and evaluation. Figure 1 depicts the three main

phases in the development methodology.

Figure 1: CPMA Development Methodology.

The first phase in the methodology is the

modelling of goals, metrics and data sources. This

phase is led by clinical and technical experts with

the aim to understand the clinical process and define

goals for monitoring. Next, they identify and define

adoption criteria to ensure adoption and acceptance

of the technology (refined later during the evaluation

phase). Analysis of the metrics, linked to goals and

data sources used to compute the metrics, is carried

out to provide meaningful insights on the clinical

process. The point is to define who will collect what

data, how and when, in order to compute the metric.

During the implementation phase, the clinical

process or task monitored is mapped to a star-

A Development Methodology for a Stroke Rehabilitation Monitoring Application

401

schema database model (Kimball, 2013) that is

optimized for fast querying and reporting. Data

related to the clinical process or task monitored is

mapped to a fact table. Clinical dimensions are

mapped to dimensional tables that represent the

attributes used to compute metrics. Each

dimensional table is linked to a control in a form

and, values in dimensional tables supply the values

for each of the controls in the form. Same values are

used for grouping, labelling and filtering metrics in

graph and chart reports.

The evaluation phase embraces user-experience

walkthroughs and think-aloud sessions in order to

understand the context and thought processes of the

users as they use the application. Clinical and

technical experts analyse the feedback obtained from

these sessions to generate checklists for

development of the application. Checklists are often

the result of trade-offs made between user needs and

adoption criteria and limitations and constraints of

the technology. This tension drives creative

solutions and innovations in user interface design.

The cycle of Model, Implement, and Evaluate is

repeated until no significant innovations and barriers

to user acceptance and adoption are identified during

the evaluation phase. The end of a cycle is reached

when clinical users and the development team are in

sync and only minor adjustments are required.

4 CASE STUDY: STROKE REHAB

PROGRAM

In our case study, we first define two use cases that

help us conceptualize the application development

by following the development methodology for

CPMA described in section 3.

Table 1: Use Case 1. Rehabilitation Plan - Betty.

Goal Therapy Target Metrics

Increase

Mobility

(ADLs)

Physical-

walking

75% # Steps (Day

Day

n-2

) > 10

Physical-

Treadmill

75% Maximal heart

rate <= beats of

predicted

maximum +20

Improve

retrieval

of words

Retrieval of

words

90% # Words

Retrieved (Day

Day

n-1

) >5

For the first use case, we have patient Betty. She

suffered a severe stroke event and after she was

stabilized from the event, the team at the acute care

unit in the hospital assessed her condition. Her

diagnosis included disabilities in more than one area

and the care team recommended her to be admitted

to the In-Patient Rehabilitation Unit. Physical

therapy and speech therapy were included in her

rehabilitation plan. The therapists at the In-Patient

discussed rehabilitation goals with Betty and her

family members. Table 1 shows goals, metrics by

therapy and, expected rehabilitation outcome targets

for each of the therapies. Information in table 1 is

just an example as metrics can change frequently as

the patient progress in her rehabilitation program.

Table 2: Use Case 2. Rehabilitation Plan - John.

Goal Therapy Target Metric

Increase

Mobility

(ADLs)

Physical-

walking

90% # Steps (Day

Day

n-1

) > 20

Improve

retrieval of

words

Retrieval

of words

50% # Correct

pictures selected

(Day

-Day

n-1

)

>10

The second use case refers to patient John. He

suffered a moderate stroke and was also assessed by

the team at the acute care unit in the hospital. The

team determined he had physical and speech

deficiencies. Table 2 depicts John’s rehabilitation

plan in terms of goals, metrics by therapies, and

rehabilitation outcome targets. Although Betty and

John’s rehabilitation plans are similar in terms of the

therapies they both require, the metrics for each of

their therapies vary, and the rehabilitation outcome

expectations before discharge are also different.

Therefore, the stroke rehabilitation monitoring

application needs to have sufficient flexibility to

accommodate different goals, metrics and targets

that drive the plans.

Following the development methodology for

CPMA described in section 3, we review the

modelling, implementation and evaluation phases

for designing a Stroke Rehabilitation Monitoring

Application.

4.1 Model

Modelling goals and metrics for a stroke

rehabilitation application is complex. The first step

is to understand goals and adoption criteria.

Rehabilitation programs require the collaboration of

a multidisciplinary team of healthcare providers and

information needs from each of the providers need

to be integrated seamlessly into one single

application that reports the overall progress of the

patient and the effectiveness of the care team.

HEALTHINF 2016 - 9th International Conference on Health Informatics

402

Goals can be defined at two levels. First, goals

related to the overall rehabilitation program (all

patients) that provides insights on performance of

the healthcare team. Second, we need to define goals

that provide insights on individualized rehabilitation

plans. Definition of metrics linked to goals at the

patient level is complex as rehabilitation goals are

tailored to specific needs of each patient by therapy.

For example, in our case study, at the patient level,

both Betty and John’s rehabilitation plan include

physical walking therapy. However, the number of

steps each patient is expected to take and frequency

of the therapy varies (10 steps for Betty every two

days and 20 steps for John daily). Also, the metrics

and benchmarks can vary as the rehabilitation

progresses.

Many data sources are needed as multiple

technologies (e.g. fitness tracking bands (i.e. FitBit),

speech apps) are used to support rehabilitation plans.

In our case study, we assumed that patients

performing the same exercise use the same

technology. This way we can standardize data

formats and define a set of pre-defined values for

each exercise by therapy that allows us to track

progress of therapies at both the patient and program

level.

4.2 Implement

The development methodology leverages the use of

QuickForms (Baarah et al, 2014), which is an

application framework optimized to collect data

directly into a reporting database. The database

model is a multi-dimensional model, i.e. star-

schema, with one fact table and multiple dimensions

linked to the fact table. Table 3 depicts the database

configuration of the Stroke Rehabilitation

application.

Individualized metrics can be assigned to each

patient in table Exercise_Multi and personalized

reports can be generated from the data that show

patient progress. In Table Exercise, we assign

standardized repetition values by exercise that is

used for reporting metrics on exercises at a program

level. Finally, by setting individual therapy goals in

table Patient, we can report on the overall progress

of the patient.

4.3 Evaluate

The evaluation of the Stroke Rehabilitation

application is complex given the multidisciplinary

nature of stroke rehabilitation which involves the

collaboration of multiple actors, all of whom are

candidates for data collection. Therefore, we need to

select at least one user representing each of

disciplines (e.g. physical therapist, speech therapist,

occupation therapist, pharmacist, and physician) as

part of the evaluation process. In addition, the

application is intended to empower patients and

family members/caregivers in decision-making and

enactment of a rehabilitation plan. Patients in a

stroke rehabilitation program have multiple needs,

which will require a careful selection of user

representatives that can participate in the evaluation

sessions. To understand how the application will

impact and be received by the various actors we use

a variety of usability evaluation approaches

including think-aloud and walk-throughs (Kushniruk

et al, 2013).

Table 3: Stroke Rehabilitation Database Schema.

Clinical

Dimension

Table Type Attributes

Rehabilitation

Plan

Rehabilitation_

Progress

Fact Exercise_Multi_ID; Exercise_Summary; Date; Therapist_ID;

Patient_ID; Therapy_ID; Age_ID; Gender_ID; Severity_ID;

Facility_ID

Tracking Date Dimensional Date; Day; Month; Year; Week

Therapist Dimensional Therapist_ID; Name; Email; Type

Patient Dimensional ID; Physical therapy target; Occupational therapy target;

Recreational therapy target; Speech therapy target; Pharmacist

therapy target

Therapy Dimensional Therapy_ID; Label

Demographics Age Dimensional Age ID; Age label

Gender Dimensional Gender ID; Label

Severity Dimensional Severity ID; Label

Facility Dimensional Facility ID; Name

Care Exercise Dimensional Exercise ID; Therapy ID; Domain; Category; Repetitions

Exercise_Multi Fact Exercise_Multi_ID; Exercise_ID; Metric; Completed

A Development Methodology for a Stroke Rehabilitation Monitoring Application

403

5 DISCUSSION

In this paper we have described our work in progress

research at developing a CPMA for stroke

rehabilitation.

A key contribution from our work is an

understanding of how CPMAs design for stroke

rehabilitation differs from CPMA design in other

domains. We compare the development of the

Stroke Rehabilitation application against the two

other CPMA’s we developed in previous research,

RPP and SAID (see Table 4). In terms of modelling,

RPP and SAID are much less complex as both

applications can be defined based on one single

generic performance model with the same set of

goals and metrics for all users. The clinical process

for RPP and SAID is clearly defined. For the Stroke

Rehabilitation application, there is no one single

clinical pathway but rather goals and metrics are

defined based on each patient's specific plan. In the

case of RPP and SAID, attributes to compute metrics

come from a set of predefined values. The multiple

clinical pathways in the stroke rehabilitation

program require consideration of multiple metrics

and possible values for the attributes used to

compute the metrics.

The implementation phase for the Stroke

Rehabilitation application also introduced new

challenges. Metrics are defined according to each

patient rehabilitation plan, which dramatically

increases the number of attributes required to

include in the application. As reports and forms are

linked together via the same reporting database, the

configuration of the reports is also complex.

Multiple are the reports that can be generated, and

the values that can be used to group and labelling

data in the reports. RPP and SAID, both have one

generic performance model, and attributes to

compute metrics can be easily mapped to forms.

The evaluation phase for the Stroke

Rehabilitation application is complex as it involves

multiple users with heterogeneous information

needs, including the patient. This is the first

application we have designed that brings the patient

into the evaluation group. Users of RPP and SAID

are homogeneous in terms of information needs, and

a smaller group of users with a smaller set of user

requirements suffices for evaluation purposes.

Table 4: CPMA Development Methodology - App Comparison.

Methodology RPP SAID Stroke Rehabilitation App

Model

Understand

Goals & Adoption

Criteria

Generic goals (program

curriculum).

Homogeneous users.

Generic goals (outcomes

of care). Homogeneous

users.

Individualized goals (Patient

rehabilitation goals).

Heterogeneous users.

Define Metrics For

Goals

One generic model. One generic model. Customized model for each

patient.

Identify Data

Resources Who/What

Homogeneous users.

Well-defined data

needs.

Homogeneous users.

Well- defined data needs.

Heterogeneous users. Highly

variable data needs.

Data sources are varied

Technologies as a driver

(monitoring and empowerment)

Implement

Map Clinical

Dimensions to

Reporting Database

Well-defined clinical

dimensions.

Well-defined clinical

dimensions.

Well defined clinical

dimensions.

Configure App Forms Fixed form attributes. Fixed form attributes. Variable form attributes

depending on patient’s

rehabilitation plan.

Configure App

Reports

Common metrics. Common metrics. Highly variable metrics.

Evaluate

User Experience

Walkthroughs, Think

Aloud

Homogeneous users. Homogeneous users. Heterogeneous users.

Technical and

Clinical Experts

Checklists

Residents

Programs directors

Administrative (training,

case manager)

Clinical (nurses)

Patient

Clinical (physicians, therapists,

caregivers)

HEALTHINF 2016 - 9th International Conference on Health Informatics

404

6 CONCLUSION AND

LIMITATIONS

Some adjustments to our development methodology

for CPMA are needed for development of

applications to support stroke rehabilitation

programs. The first consideration is due to the

variable clinical pathways that can be followed in a

stroke rehabilitation program, which adds

complexity to the definition of the performance

model -no one single model can be defined-.

Second, the implementation of the application is

challenging as the configuration of forms and reports

must be flexible to tailor to the specific information

needs of each patient. Third, this is the first

application that targets the patient as a user. As

multiple are the clinical pathway that can be

followed during rehabilitation, multiple are the

information needs for each patient. A careful

selection of patient representatives is key to ensure

success during the evaluation phase.

This paper presents our in-progress research on

the development of a stroke rehabilitation

application following a specific methodology for

development of CPMAs. We acknowledge our

analysis is limited in that we have not yet developed

a prototype to evaluate proof concept of our

approach.

ACKNOWLEDGEMENTS

We acknowledge funding support from the Mitacs

Accelerate program, NSERC and IBM.

REFERENCES

Avison, D., & Young, T. (2007). Time to rethink health

care and ICT? Commun.ACM, 50(6), 69-74.

Baarah, A., Mouttham, A., & Peyton, L. (2012).

Architecture of an event processing application for

monitoring cardiac patient wait times. International

Journal of Information Technology and Web

Engineering (IJITWE), 7(1), 1-16.

Brown, A. W., Therneau, T. M., Schultz, B. A., Niewczyk,

P. M., & Granger, C. V. (2015). Measure of functional

independence dominates discharge outcome prediction

after inpatient rehabilitation for stroke. Stroke; a

Journal of Cerebral Circulation, 46(4), 1038-1044.

Chamney, A., Mata, P., Viner, G., Archibald, D., &

Peyton, L. (2014). Development of a resident practice

profile in a business intelligence application

framework. ICTH-2014, Halifax, Canada.

Chukmaitov, A., Harless, D. W., Bazzoli, G. J., Carretta,

H. J., & Siangphoe, U. (2014). Delivery system

characteristics and their association with quality and

costs of care. Health Care Management Review, 40(2),

92-103.

Cifu, D. X., & Stewart, D. G. (1999). Factors affecting

functional outcome after stroke. Archives of physical

medicine and rehabilitation, 80(5), S35-S39.

Duncan, P. W., Zorowitz, R., Bates, B., Choi, J. Y.,

Glasberg, J. J., . . . Reker, D. (2005). Management of

adult stroke rehabilitation care. Stroke; a Journal of

Cerebral Circulation, 36(9), e100.

Ferenchick, G. S., Foreback, J., Towfiq, B., Kavanaugh,

K., Solomon, D., & Mohmand, A. (2010). The

implementation of a mobile problem-specific

electronic CEX for assessing directly observed

student-patient encounters. Medical Education Online,

15.

Ferenchick, G. S., & Solomon, D. (2013). Using cloud-

based mobile technology for assessment of

competencies among medical students. PeerJ, 1, e164.

Gresham, G. E., Duncan, P. W., & Stason, W. B. (1997).

Post-stroke rehabilitation DIANE Publishing.

IOM (Institute of Medicine). (2012). Health IT patient

safety: Building safer systems for better care.

Washington DC: The National Academic Press.

Kimball, R. (2013). The data warehouse toolkit,(3rd ed.).

Indianapolis, Ind. Wiley.

Kushniruk, A., Nohr, C., Jensen, S., & Borycki, E. M.

(2013). From usability testing to clinical simulations.

Yearbook of Medical Informatics, 8, 78-85.

Kuziemsky, C. E. (2015). A model of tradeoffs for

understanding health information technology

implementation. Studies in Health Technology and

Informatics, 215:116-28.

Mata, P., Chamney, A., Viner, G., Archibald, D., &

Peyton, L. (2015). A development framework for

mobile healthcare monitoring apps. Personal and

Ubiquitous Computing, 19(3-4), 623-633.

Novak, L., Brooks, J., Gadd, C., Anders, S., & Lorenzi, N.

(2012). Mediating the intersections of organizational

routines during the introduction of a health IT system.

European Journal of Information Systems, 21(5), 552-

569.

Vincent, C., Burnett, S., & Carthey, J. (2014). Safety

measurement and monitoring in healthcare: A

framework to guide clinical teams and healthcare

organisations in maintaining safety. BMJ Quality &

Safety, 23(8), 670-677.

Vredenburg, K., Mao, J., Smith, P. W., & Carey, T.

(2002). A survey of user-centered design practice.

Proceedings of the SIGCHI Conference on Human

Factors in Computing Systems, 471-478.

Xu, B., Xu, L., Cai, H., Jiang, L., Luo, Y., & Gu, Y.

(2015). The design of an m-health monitoring system

based on a cloud computing platform. Enterprise

Information Systems, 1-20.

A Development Methodology for a Stroke Rehabilitation Monitoring Application

405