Detecting Disease from Administrative Data: What Diseases Can We Really Detect and Implications for Clinical Decision Support
Public Deposited- Last Modified
- February 28, 2019
- Creator
-
Johns, Ellis B.
- Affiliation: School of Information and Library Science
- Abstract
- Objective: To identify rule sets based on administrative data to detect the IOM priority conditions (Persistent Asthma, Diabetes, Heart Failure, COPD, Hypertension, MI, Stroke, & Hyperlipidemia) as well as obesity and depression, and also to evaluate each rule set's applicability to clinical decision support (CDS) Design: Administrative Code based rule sets generated by an iterative process and evaluated by direct comparison with a chart review based gold standard. Data sources: Medicaid claims data from North Carolina Health Information Exchange compared with electronic medical records from the Duke University Health System. Results: 10 diseases were included in our study. Of these 10, only 2 diseases, hypertension (HTN) and diabetes mellitus (DM) produced adequate rule sets for clinical decision support. HTN had PPVs ranging from 0.89 to 0.94, and DM had PPVs ranging from 0.91 to 0.94. The rule sets for the other diseases all had PPVs less then 0.90, and were thus not considered to be useful for clinical decision support. Because not everyone using disease detection algorithms has a CDS focus, we have included the performance of all our generated rule-sets in tabular format. Conclusions: Several factors were closely correlated with a rule set's ability to accurately detect the presence of a particular disease. Perhaps the greatest of these factors was the granularity of the underlying code system. For the 2 diseases which had very specific ICD codes (i.e. HTN and DM), a simple and accurate rule set was created.
- Date of publication
- April 2009
- Subject
- DOI
- Resource type
- Rights statement
- In Copyright
- Advisor
- Losee, Robert
- Degree granting institution
- University of North Carolina at Chapel Hill
- Language
- Extent
- 30
- Deposit record
- abb1808a-3405-4d7b-866b-f35807dc08b2
- Access right
- Open access
- Date uploaded
- March 6, 2013
Relations
- Parents:
This work has no parents.