use of health information

Use of Health InformationUse o ea t o at oนพ.นวนรรน ธีระอัมพรพันธุ์

หลักสูตร Healthcare CIO

โรงเรียนบริหารงานโรงพยาบาล คณะแพทยศาสตร์ รพ.รามาธิบดี

17 ธันวาคม 2553

Except where citing other works

Uses of Health Information• Patient Care• AdministrationAdministration

• Claims & Reimbursements, Inventory Management’ k• Management’s Decision Making

• Quality Improvement & Organizational Learning

• Public Health• Surveillance monitoring disease reporting• Surveillance, monitoring, disease reporting

• Education• Research• Etc• Etc.

Types of Uses• Primary• SecondarySecondary

• Use of clinical data for researchf l l d f l• Use of clinical data for quality improvement

• Use of claims data for research• Use of clinical/administrative data for performance evaluation or strategic planningg p g

• Etc.

Issues for Secondary Data Uses• Subject coverage• Data definition & measurement (context)• Data quality

• Accuracy (validity) precision (reliability)Accuracy (validity), precision (reliability)• Timeliness• Relevance• Relevance• Completeness (vs. missing data)

• Data format/uniformity (standards)• Data format/uniformity (standards)• Data integration from disparate sources• Privacy• Unintended consequences of secondary data usesq y

Sample Problems with Secondary Data• High number of “cystic fibrosis” cases in a Thai hospital’s monthly/annual statisticsp y/

• A number of female outpatients with benign prostate hyperplasia (BPH)prostate hyperplasia (BPH)

• Changes in data definition or collection methods in HIS, not translated into reports

• “Upcoding” of diagnoses for claims &• Upcoding  of diagnoses for claims & reimbursements

• Adverse selection of patients

Informatics Tools• Data warehouse & Data marts

• Dimensional modelingimensional modeling• Facts, dimensions, cubes/hypercubes• Star schema Sno flake schema• Star schema Snowflake schema

• Extract Transform Load• Extract‐Transform‐Load

Image Source: SqlPac via http://en wikipedia org/wiki/Star schema andImage Source: SqlPac via http://en.wikipedia.org/wiki/Star_schema andhttp://en.wikipedia.org/wiki/Snowflake_schema

Informatics Tools• Business intelligence

• OLAPO AP• Routine & ad‐hoc reporting• “Slice & Dice” & “Drill‐Down”Slice & Dice  &  Drill Down

• Business analytics• Data mining (knowledge discovery from data)• Data mining (knowledge discovery from data)

Data Mining Methods: Classification• Example of classification using decision tree

MarStMarried

Single, Divorced

ID Home Marital Annual Defaulted

Home Owner

NOYes No

ID Owner Status Income Borrower

1 Yes Single 125K No

2 No Married 100K No

IncomeNO< 80K > 80K

3 No Single 70K No

4 Yes Married 120K No

5 No Divorced 95K YesYESNO

5 No Divorced 95K Yes

6 No Married 60K No

7 Yes Divorced 220K No

8 No Single 85K Yes

9 No Married 75K No

10 No Single 90K Yes From a teaching slide by Tan, Steinbach & Kumar at UMN (2007)

10

Data Mining Methods: Clustering• Clustering

Image Source: hellisp via http://en wikipedia org/wiki/Cluster analysisImage Source: hellisp via http://en.wikipedia.org/wiki/Cluster_analysis

Data Mining Methods: Association Rules• Association Rules

Market-Basket transactions

TID ItemsExample of Association Rules

TID Items

1 Bread, Milk

2 Bread, Butter, Milk, Eggs

{Bread} → {Milk},{Bread, Milk} → {Butter},{Bread Butter} → {Milk}2 Bread, Butter, Milk, Eggs

3 Milk, Eggs, Coke

4 Bread, Milk, Butter

{Bread, Butter} → {Milk},{Milk} → {Eggs}

5 Milk

Adapted from a teaching slide by Tan, Steinbach & Kumar at UMN (2007)

Data Mining Methods: Anomaly Detection• Anomaly/Outlier Detection

3 5

4

3

3.5

2.5

1.5

2

1

0.5

From a teaching slide by Tan, Steinbach & Kumar at UMN (2007) Outlier Score

CLASS EXERCISECLASS EXERCISE

Suggest 1 “question” related to gg qhealth care that can be answered with each of the data mining methodseach of the data mining methods

Take‐Home Message• Data are “gold”• Tools are useful but we need to ask good questionsTools are useful, but we need to ask good questions• Be aware of data issues

• Data collected for 1 purpose and used for another• Context/definition issues/• Data quality issues• Data format issues• Data format issues• Privacy issues

• Needs expertise on best practices in data warehouse architecture and data analysisy