Concurrent validity estimates for the accuracy of administrative data case definitions for uveitis in Ontario, Canada

This study aimed to validate administrative data case definitions for identifying individuals with uveitis within health administrative data and to assess case definition performance in different age groups.

Among 200,386 eligible patients in a large, province-wide, multisite primary care electronic medical record data repository in Ontario, Canada, 1 020 individuals were selected for medical review based on uveitis-related search terms. After medical record review, 431 cases were identified as having uveitis, in addition to 10,000 nonuveitis cases, which were used to make up the validation reference sample. After the reference standard was linked with administrative data, the performance characteristics (sensitivity and positive predictive value [PPV]) of case definitions using combinations of data sources and time windows were assessed overall, and stratified by age groups.

Within the validation sample, the median age of identified uveitis patients was 55 years, with the majority falling within the 26-64 age group. Females represented 59.9% of cases. Family physicians (72.2%) and ophthalmologists (60.6%) were the most common sources of diagnosis documentation, and anterior uveitis was the most prevalent classification (64.0%). At least 1 diagnosis code for uveitis had the highest sensitivity at 74.0% (95% CI: 69.9-78.2) with a PPV of 54.3% (95% CI: 50.2-58.3). PPV increased with more physician diagnosis codes but was accompanied by decreased sensitivity. Case definitions demonstrated variations in sensitivity and PPV across different age groups.

This study identified administrative data case definitions for identifying uveitis, with varying sensitivity and PPV for a range of research purposes and data availability.