Do Diagnostic and Procedure Codes Within Population-Based, Administrative Datasets Accurately Identify Patients with...

Area of Science:

• Health Informatics
• Cancer Research
• Epidemiology

Background:

• Administrative datasets often use diagnostic and procedural codes.
• These codes may inaccurately identify specific patient populations.
• This can lead to misclassification bias in research.

Purpose of the Study:

• To measure the accuracy of existing coding algorithms for identifying patients with rectal cancer resections (RCR) using administrative data.
• To assess the reliability of administrative data for RCR cohort identification.

Main Methods:

• A previously published coding algorithm was used to create an RCR cohort from administrative data at a single institution.
• The accuracy of this cohort was validated against a gold standard reference population.
• A systematic literature review identified studies using similar methods and their reported accuracy.

Main Results:

• The administrative data identified 1131 RCRs, while the gold standard identified 821 RCRs.
• The coding algorithm showed moderate accuracy (sensitivity 89.5%, specificity 99.9%) but poor positive predictive value (64.9%).
• Of 18 studies using similar methods, only one reported on cohort accuracy.

Conclusions:

• Using diagnostic and procedure codes to identify RCR in administrative data can introduce misclassification bias due to low positive predictive value.
• It is essential to report the accuracy of RCR cohorts derived from population-based administrative datasets.