1

electronic Journal of Health Informaticshttp://www.ejhi.net2009; Vol 4(1): e

The electronic Journal of Health Informatics is an international journal committed to scholarly excellence and dedicated to the advancement of Health Informatics and information technology in healthcare. ISSN: 1446-4381© Copyright of articles is retained by authors; originally published in the electronic Journal of Health Informatics (http://www.ejhi.net). This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 2.5 License (http://creativecommons.org/licenses/by-nc-sa/2.5/au).

A Model Driven Approach to Care Planning Systems for Consumer Engagement in Chronic Disease

ManagementAbizer Khambati1, Jim Warren2, John Grundy1,2, John Hosking2

1Department of Electrical and Computer Engineering, University of Auckland, Auckland, New Zealand

2Department of Computer Science, University of Auckland, Auckland, New Zealand

AbstractThis work demonstrates a model-driven approach to the development of care plan systems, amena-ble to: (a) a flexible and extensible definition of care plan scope; and (b) deployment of care planviewing and tracking functionality to a wide range of physical computing devices. The approachutilises a care plan domain model from which guideline implementers formulate care plan tem-plates aligning to specific clinical guidelines. A clinical end user would subsequently constrainthat template (e.g., selecting a subset of available activities and specific targets) to create a careplan instance for an individual patient. An XML care plan visualisation definition created using theMarama tool is transformed to OpenLaszlo script from which Shockwave Flash objects can becompiled, creating Flash applications that run on a variety of hardware for both clinical andpatient users. The approach is illustrated with respect to an overweight and obesity guideline.

Keywords: Health care planning, Mobile Health Applications, Model-driven Engineering, Electronic Health Records

1. IntroductionThere is obvious potential for bene-

fit from the provision of electronicsupport for chronic disease manage-ment. Warren et al. [1] demonstratedthat the intent of chronic disease man-agement guidelines could be accu-rately modelled in a decision supporttool for care planning. More recently,in New Zealand, the PREDICT CVD/Diabetes tool for cardiovascular riskassessment and management is nowwidely deployed and is contributingto improved documentation andunderstanding of cardiovascular dis-ease risk factor [2,3].

E.H. Wagner’s Chronic Care Modelenvisages ‘productive interactions’between an informed, activatedpatient and a prepared, proactivepractice team [4]. Such a model man-dates the engagement of both thehealth provider and the health con-sumer in creation, monitoring andmaintenance of a shared care plan.Successful systems have emerged forcomponents of the Wagner process:e.g., (a) the Ferret system has beeninstalled in more than 50 communityclinics in Queensland, predominantlyin the northern areas, to provide elec-tronic care plans and chronic diseasemanagement support for a range of

common chronic conditions [5]; and(b) the STOMP system has demon-strated significant improvement insmoking cessation via a txt messagingbased dialog direct to consumers [6].With the emerging success of individ-ual systems, the question is how toachieve more general platforms tosupport dissemination and tracking ofindividually tailored care plans, andhow to encompass both healthcareproviders and health consumers in thescope of a single software solution.

We describe a prototype care plandesign environment utilising a con-cept of “Domain-specific Visual Lan-guages” to enable health practitioners

| electronic Journal of Health Informatics 4(1): e

2

to model, tailor and deploy chronicdisease care plans on mobile devices.A generic care plan for managementof a chronic disease is specifiedincluding key health activities, per-formance metrics, and assessments. Ageneric care plan is then tailored to anindividual patient’s situation, specify-ing particular health activity intervals,duration, reviews, targets and metrics.A mobile application is then gener-ated completely automatically fromthis tailored care plan description anddeployed to the patient’s mobile PDA

or Smart Phone. Data is periodicallyuploaded from the device for moni-toring by the health care provider andto facilitate patient/provider dialogue.

2. MethodsWe have developed a domain

model for care plans where care plansconsist of performance measures(e.g., BMI), activities (e.g., jogging,or taking a specific medication) and,recursively, of sub-care plans. Activ-

ities have supporting attributes, nota-bly schedules and instructions.Figure 1 provides a simplified repre-sentation of the care plan domainmodel. A domain-specific language(DSL) has been created using theMarama tool suite [] with an associ-ated visual editor which permits pro-duction of XML definitions of careplans. Figure 4 shows an example of ageneric care plan constructed usingthe DSL editor.

Figure 1. Key aspectsof the care plan domain model.

Our model of care planning callsfor a two-stage application of the careplan DSL: (1) at a knowledge engi-neering stage, a care plan template iscreated to suit the requirements of anevidence-based clinical practiceguideline; (2) subsequently, when a

template is applied to a patient, themodel can be edited to tailor the careplan for instantiation (e.g., to includeindividualized instructions, eliminatespecific activities as being irrelevantand so on). At this second stage a spe-cific timeframe for the duration of the

care plan and its review schedulewould be introduced.

A second modelling tool permitsthe instantiated, but still abstract, careplan to be mapped to a suitable graph-ical user interface. This step allows

| electronic Journal of Health Informatics 4(1): e

3

for the wide variation in form factorsof mobile devices that the plan couldbe deployed to for patient use. Theresulting concrete care plan may thenbe downloaded to the patient's pre-ferred device (PDA, phone, etc) and

executed using a generic Flash orDHTML based interpreter (we haveused OpenLaszlo [] to develop this).We examine use of this approach tocare plans on a conventional PC sys-tem for a GP tailoring and reviewing

a care plan; and for reviewing andlogging notes to adhere to specificactivities on a smart phone device fora patient.

| electronic Journal of Health Informatics 4(1): e

4

Figure 2. Care plan design and implementation process.

Figure 2 depicts the care plandevelopment and implementationprocess. Chronic disease care plansare reusable and can be updated overtime to improve treatment practice.Individual tailored care plans can bemodified over time or even differentversions of the care plan for the samepatient developed and changed overtime. The interface for the mobilecare plan application can be tailored

over time to better support thepatient’s needs.

3. ResultsTo illustrate the care plan model-

ling, tailoring and user interfaceimplementation approach we use aUS National Heart Lung and BloodInstitute guideline for managing over-

weight and obese adults [9] as thesource for formulating a care plantemplate. Figure 3 provides a simpli-fied illustration of some of the careplan template components for theoverweight and obesity guidelinebased on our care plan domain model.In this example, a periodic review ofthe obesity care plan is to be under-taken. The management plan includesperiodic glucose measurements, twoalternative exercise activities (jog-

| electronic Journal of Health Informatics 4(1): e

5

ging and swimming) and a Reductilpharmacotherapy treatment plan. Theglucose measurement activity con-

sists of a set of instructions for using apricking needle, a testing meter, and a

gluco testing card. Readings are to betaken every six hours

Figure 3. Some of the care plan template components for a care plan based on an overweight and obesity guideline [9].

To create mobile care plan applica-

tions based on the guideline requirescare plan visualisation definitions tobe created. Figure 4 shows part of adiabetes care plan model specificationfor the example in Figure 3 using ourDSL editor. The result of this specifi-cation is a generic care plan defini-tion, which can be tailored for aparticular patient and then trans-formed to OpenLaszlo script. This

script can then be compiled to an enduser application. A second visual lan-guage is used to specify how theitems in the care plan are representedby elements in a mobile device userinterface.

Figure 5 shows screen shots of acare plan instantiation screen used bya care coordinator on a desktop PC.This allows a generic care plan to betailored for individual patient use. Inthis example the patient will use jog-

ging as their physical exercise activ-ity, every two days for 20 minutes.The swimming activity in the genericplan has been removed (by deselect-ing it) as being inappropriate for thispatient.

| electronic Journal of Health Informatics 4(1): e

6

Figure 4. Screen shot of part of a diabetes monitoring care plan: a data collection activity for the collection of blood glucose sugar measurements.

Figure 5. Screen shot of part of a diabetes monitoring care plan being tailored in our care plan instantiation application.

| electronic Journal of Health Informatics 4(1): e

7

Figure 6 shows use of the secondDSL used to specify the user interfacefor the generated care plan applica-tion. The target user for this DSL iscurrently an interaction designer. Thespecification in Figure 6 is a genericone representing the hierarchicalstructure of an individual health careplan to patients. It maps the datastructure for an instantiated health

care plan into an aggregated set ofmenu lists.

Figure 7 shows screen shots fromthe running mobile obesity care planapplication that was generated fromthe tailored care plan specifications.The general care plan view specifiedin Figure 6 is shown instantiated withthe obesity management care plandata at the top of Figure 7. Figure 7also shows a list of scheduled Glu-

cose measurements (left), a specificone opened and values entered (mid-dle). Instructions if needed are pro-vided (right).

Figure 6. Specification of the main screen for the patient application.

| electronic Journal of Health Informatics 4(1): e

8

Figure 7. Screen shots of the end-user Flash application compiled from OpenLaszlo.

4. DiscussionWe have described a staged domain

specific modeling approach to thedevelopment of care planning sys-tems. This approach allows a growinglibrary of care plan templates to bemapped to a broad range of devices.This in turn supports the distributionof care plans between providers andpatients, and the opportunistic use ofdesktop or mobile devices, with readyextension to new devices as theyemerge.

There are in fact a large range oftechnology options for the key stagesof our architecture (including Micro-soft technologies). The specific lan-guages and editors of our solution arejust one possibility for achieving amodel-driven approach to care plans;however, we contend that anyapproach should have similar stages.

Use of this approach in a healthcaresystem as found in New Zealand orAustralia would require integrationwith existing software infrastructure,notably the practice management sys-tems employed by General Practition-

ers (GPs). It is likely that GPs wouldfavour care plan template selectionand tailoring using their familiartools, requiring compatible compo-nents to be implemented in the popu-lar commercial software (or for thatsoftware to embed the care plan com-ponents). Further user testing isrequired for both healthcare providerand health consumer users, and itmust be recognised that model gener-ated user interfaces may lack the flairof purpose-built screens.

We envision the most direct path-way for initial use of the approachproposed herein to be applicationsconcerning relatively simple activitytracking and reminder applications.Obvious applications include weightmanagement, diabetes monitoringand management and activity promo-tion (as illustrated); one’s imaginationcan easily extend to other self-man-agement contexts, particularly for arelatively able user, or to be followedby a family member of someone lessable. Fairly simple online educationintegrates easily (as Activity Instruc-tions). Activity Routines, for which

we have only illustrated (and onlyimplemented) simple periodic timers,could be extended to invoke on spe-cific trigger conditions of arbitrarycomplexity to yield care plans thatextend to contingencies. Muchgreater exploration of use cases, how-ever, is needed to determine just howextensible the proposed care planmodel is. Moreover, the present pro-posal is best considered as a way ofarchitecting consumer engagementapplications, rather than as a one-size-fits-all solution for a completeintegrated care system.

An important direction for this sortof work is to achieve alignment withkey international standards. Of mostrelevance would be establishment ofcorrespondence of the individual careplan template concepts to SNOMEDCT. Success of Hrabak et al [HrabakK, Campbell J, Tu S, McClure R,Weida R. Creating interoperableguidelines: requirements of vocabu-lary standards in immunization deci-sion support. Proceedings,MEDINFO 2007: 930-4.] in mappingautomated guideline components to

| electronic Journal of Health Informatics 4(1): e

9

SNOMED is promising in this regard.Also relevant is the alignment of thecare plan in a larger sense to HL7,both in terms of alignment of thedomain model to the HL7 ReferenceInformation Model and of instancesbased on care plan templates to HL7Clinical Document Architecture(CDA).

5. ConclusionWe have developed a tool for mod-

eling care plans for chronic diseasemanagement using a domain-specificvisual language. These generic careplans can be tailored for specificpatients, taking into account patient-specific exercise, medication, meas-urement and informational needs. Amobile device application interface isthen specified using a second visuallanguage, describing how care planelements will be realized in a mobiledevice application. An applicationgenerator is used to completely auto-matically generate a mobile devicecare plan application. The care planmodel, patient-specific tailored careplan and mobile application interfacecan all be tailored and improved overtime using these high level tools.

AcknowledgementsWe acknowledge that a joint 4th

year project between Sumedh Kanadeand the first author provided impor-tant insights on the information modelfor care plans.

References1. Warren JR, Noone JT, Smith BJ, RuffinR, Frith P, van der Zwaag BJ, BeliakovGV, Frankel HK, McElroy HJ. Automatedattention flags in chronic disease careplanning. Medical Journal Australia.2001.175(6):308-12.

2. Riddell T, Jackson RT, Wells S, BroadJ, Bannink L. Assessing Mori/non-Moridifferences in cardiovascular disease riskand risk management in routine primarycare practice using web-based clinicaldecision support: (PREDICT CVD-2).New Zealand Medical Journal. 2007;120(1250):U2445.

3. Wells S, Kerr A, Broad J, Riddell T,Kenealy T, Jackson R. The impact of NewZealand CVD risk chart adjustments forfamily history and ethnicity on eligibilityfor treatment (PREDICT CVD-5). NewZealand Medical Journal. 2007;120(1261):U2712.

4. Improving Chronic Illness Care, TheChronic Care Model. http://www.improvingchroniccare.org/index.php?p=The_Chronic_Care_Model&s=2 (last accessed 20 February 2008).

5. McDermott RA, McCulloch BG,Campbell SK, Young DM. Diabetes in theTorres Strait Islands of Australia: betterclinical systems but significant increase inweight and other risk conditions amongadults, 1999-2005. Medical Journal Aus-tralia. 2007; 186(10):505-8.

6. Bramley D, Riddell T, Whittaker R,Corbett T, Lin RB, Wills M, et al. Smok-ing cessation using mobile phone textmessaging is as effective in Maori as non-Maori. New Zealand Medical Journal.2005; 118(1216):U1494.

7. Grundy, J.C., Hosking, J.G., Huh, J. andLi, K. Marama: an Eclipse meta-toolsetfor generating multi-view environments,Formal demonstration paper, 2008 IEEE/ACM International Conference on Soft-ware Engineering, Liepzig, Germany,May 2008, ACM Press.

8. Laszlo Systems Inc. Software Engi-neer's Guide to Developing OpenLaszloApplications, 2007. http://labs.openlas-zlo.org/wafflecone-nightly/docs/develop-ers/ (last accessed 20 March 2007).

9. National Heart Lung and Blood Insti-tute (1998). Clinical Guidelines on theIdentification, Evaluation, and Treatmentof Overweight and Obesity In Adults: TheEvidence Report: 228. http://www.nhlbi.nih.gov/guidelines/obesity/ob_gdlns.htm (last accessed 9 April2008).

10. Hrabak K, Campbell J, Tu S, McClureR, Weida R. Creating interoperable guide-lines: requirements of vocabulary stand-ards in immunization decision support.Proceedings, MEDINFO 2007: 930-4.

CorrespondenceProfessor Jim WarrenDepartment of Computer Science - TamakiUniversity of AucklandPrivate Bag 92019Auckland Mail CentreAuckland 1142New Zealand

Phone: +64-9-3737-599Fax: +64-9-3737-453http://www.cs.auckland.ac.nz

jim@cs.auckland.ac.nz