Background/Purpose:

In a predominantly universal single-payer health system, Canadian health administrative data are a valuable tool and increasingly used for research. Few studies have rigorously evaluated the accuracy of administrative data for identifying patients with rheumatoid arthritis (RA). The aim of this study was to validate administrative data algorithms to identify RA in the Canadian province of Ontario. 

Methods:

We performed a retrospective chart abstraction study among a random sample of 450 patients (unselected by diagnoses), from 18 rheumatologists. Using rheumatologist-reported diagnosis as the reference standard, the RA and non-RA patients were then linked to administrative data to validate different combinations of physician billing diagnoses (P), hospitalization diagnoses (H) and pharmacy (drugs dispensed) data (Rx) to estimate sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) in differentiating RA from non-RA patients.

Results:

149 rheumatology patients were classified as RA and 301 as non-RA based on our reference standard definition. Most patients were female (77% RA cases; 65% non-RA cases) and the mean (SD) age for RA cases and non-cases was 62 (14) and 58 (17) years, respectively. Among non-RA cases, the most prevalent diagnoses were osteoarthritis (45%), seronegative spondyloarthropathies (19%) and connective tissue diseases (19%). Test characteristics of selected algorithms tested are reported in Table 1. Overall, using any physician-billing algorithms, sensitivity was very high (94-100%). Specificity and PPV were modest to excellent and increased when algorithms required multiple RA claims or claims by a specialist. There was a slight increase in sensitivity and decrease in specificity and PPV as the observation window for multiple billing diagnoses increased from 1 to 2 years. The addition of RA drugs (disease-modifying agents, biologics, or systemic steroids) in our algorithm had little impact on sensitivity but decreased both specificity and PPV.

Conclusion:

This study has demonstrated the accuracy of administrative data algorithms for identifying RA. We found that for RA patients that have seen a rheumatologist, physician-billing algorithms are highly sensitive in identifying these patients. Our findings suggest that pharmacy data do not improve the accuracy in identifying RA. One potential limitation is that our sample was drawn from rheumatology clinics, and thus our estimates may not be generalizeable on the population level. However, ongoing work to validate these algorithms in a random sample of 7500 patients from the general population is being done to further support our findings.