ppt

11

Incorporating Data Mining Incorporating Data Mining Applications into Clinical GuidelinesApplications into Clinical Guidelines

Reza SherafatReza SherafatDr. Kamran SartipiDr. Kamran Sartipi

Department of Computing and SoftwareDepartment of Computing and SoftwareMcMaster University, CanadaMcMaster University, Canada{sherafr, sartipi}@mcmaster.ca{sherafr, sartipi}@mcmaster.ca

Computer-based Medical Systems (CBMS Computer-based Medical Systems (CBMS ’06)’06)June 22, 2006June 22, 2006

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 22

OutlineOutline

Decision making based on data mining Decision making based on data mining resultsresults

Data and knowledge interoperabilityData and knowledge interoperability Knowledge management frameworkKnowledge management framework Tool implementationTool implementation ConclusionConclusion

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 33

Decision MakingDecision Making

Practitioners face critical questions which Practitioners face critical questions which requires decision making:requires decision making:

– The cause of a symptomThe cause of a symptom

– Drug prescriptionDrug prescription

– Treatment planningTreatment planning– Diagnosis of a diseaseDiagnosis of a disease– … … (many more)(many more)

Clinical Decision Support Systems (CDSS)Clinical Decision Support Systems (CDSS)– Computer programsComputer programs

– Provide online and Provide online and patient-specific assistance patient-specific assistance to health care professionals to health care professionals to make better decisionsto make better decisions

– Clinical knowledge is stored Clinical knowledge is stored in a knowledge-basein a knowledge-base

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 44

Data Mining ApplicationsData Mining Applicationsin Health carein Health care

Patient

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 55

Decision LogicDecision Logic

IFIF

the patient has had a heart stroke and is the patient has had a heart stroke and is above 50 above 50

THENTHEN

his health condition should be monitored!his health condition should be monitored!

Condition

Action

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 66

Decision Logic (cont’d)Decision Logic (cont’d)

Decision making logic:Decision making logic:

– Logical expressionsLogical expressions ‘‘If-then-elseIf-then-else’’ structures structures

– Test for conditionsTest for conditions– Trigger actionsTrigger actions

if ( (patient.age > 50) && if ( (patient.age > 50) && (patient.previous_heart_stroke == true) (patient.previous_heart_stroke == true) ))

then …then …

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 77

Data Mining Decision LogicData Mining Decision Logic Data miningData mining

– Analysis and mining of data to extract hidden facts in Analysis and mining of data to extract hidden facts in the datathe data

– The extracted facts are represented in a data The extracted facts are represented in a data structure called structure called “data mining model”“data mining model”

TrainingTraining vs. vs. ApplicationApplication of a data mining model: of a data mining model:– Training the model: Building the modelTraining the model: Building the model– Application of the mode: interpreting for specific Application of the mode: interpreting for specific

patient datapatient data

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 88

Data Mining Decision Logic (cont’d)Data Mining Decision Logic (cont’d) ClassificationClassification: mapping data into predefined classes. : mapping data into predefined classes.

(e.g., whether a patient has a specific disease or not)(e.g., whether a patient has a specific disease or not)

RegressionRegression: mapping a data item to a real-valued : mapping a data item to a real-valued prediction variable. (e.g., prediction variable. (e.g., planning treatments.)planning treatments.)

ClusteringClustering: To identify clusters of data items. (e.g., to : To identify clusters of data items. (e.g., to cluster patients based on risk factors.)cluster patients based on risk factors.)

Association RuleAssociation Rule MiningMining: to find hidden associations in : to find hidden associations in the data set (e.g., how different patient data are related the data set (e.g., how different patient data are related based on shared relations such as: “specific diseases”, based on shared relations such as: “specific diseases”, “patients habits”, or “family disease history”.)“patients habits”, or “family disease history”.)

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 99

Data Mining Decision Logic (cont’d)Data Mining Decision Logic (cont’d)

An example of regression model An example of regression model [source:Otto,Pearlmen][source:Otto,Pearlmen]

Vmax

Doppler AVA

AVR not recommendedAVR recommended

AI severity

≥4m/s3-4m/s

≤ 3m/s

≤ 1 cm2 ≥1.7 cm21.1-1.6 cm2

2-3+ %100%100

%66

0-1+

%100

%100

%88

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1010

Application of Data Mining ResultsApplication of Data Mining Results

Predictive Model Markup Language (Predictive Model Markup Language (PMMLPMML):):– XMLXML based specification based specification– Meta modelMeta model: Define the data structure of the model: Define the data structure of the model– Different types Different types of data mining models (clustering, of data mining models (clustering,

classifications, …)classifications, …)– ExtendableExtendable for model specific constructs for model specific constructs

Share, access, exchange PMML documentsShare, access, exchange PMML documents

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1111

Proposed Health Care Proposed Health Care Knowledge Management FrameworkKnowledge Management Framework

Guideline modeling

Knowledge Extraction

Guideline Execution

Phase 1: Phase 1: Build the data mining modelsBuild the data mining models

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1212

Proposed Health Care Proposed Health Care Knowledge Management FrameworkKnowledge Management Framework

Data and knowledgeinteroperability

Knowledge Extraction

Guideline Execution

Phase 2: Phase 2: Encode data and knowledgeEncode data and knowledge

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1313

Proposed Health Care Proposed Health Care Knowledge Management FrameworkKnowledge Management Framework

Data and knowledgeinteroperability

Knowledge Extraction

Knowledge Interpretation

Phase 3: Phase 3: Apply the knowledge for specific Apply the knowledge for specific patient datapatient data

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1414

Knowledge

Data and Knowledge InteroperabilityData and Knowledge Interoperability

HL-7 Reference Information Model (HL-7 Reference Information Model (RIMRIM))– A general high level health care data model A general high level health care data model

Clinical Document Architecture (Clinical Document Architecture (CDACDA))– An XML-based standard for defining structured templates for An XML-based standard for defining structured templates for

clinical documentsclinical documents

Standard Terminology Systems (Standard Terminology Systems (UMLS, SNOMED CT, UMLS, SNOMED CT, etc)etc)– Standard clinical vocabulary setsStandard clinical vocabulary sets

Predictive Model Markup Language (Predictive Model Markup Language (PMMLPMML))– An XML-based standard for representing data mining resultsAn XML-based standard for representing data mining results

Guideline Interchange Format 3 (Guideline Interchange Format 3 (GLIF3GLIF3))– A clinical guideline definition standardA clinical guideline definition standard

Data

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1515

Tool ImplementationTool Implementation A guideline execution engine based on GLIFA guideline execution engine based on GLIF Logic modules apply data mining models and Logic modules apply data mining models and

are accessed through web services technologyare accessed through web services technology Provides additional information to help guide the Provides additional information to help guide the

flow in the guideline.flow in the guideline.

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1616

ConclusionConclusion Data mining results can be used as a source of Data mining results can be used as a source of

knowledge to help clinical decision making.knowledge to help clinical decision making.

We described an approach to apply different types of We described an approach to apply different types of data mining models in CDSS.data mining models in CDSS.

We used PMML and CDA for knowledge and data We used PMML and CDA for knowledge and data representation.representation.

A tool is developed that can interpret and apply the A tool is developed that can interpret and apply the mined knowledge.mined knowledge.

We envision a future that data mining analysis results We envision a future that data mining analysis results are seamlessly deployed and used at usage sites.are seamlessly deployed and used at usage sites.

Integrating data mining applications into clinical guidelinesIntegrating data mining applications into clinical guidelines 1818

Questions and CommentsQuestions and Comments