Background/Purpose: Identification of RA cases in administrative healthcare databases is used to estimate disease frequency, healthcare utilization and cost for RA. However, the optimal methodology for achieving this is unclear. Our aim was to examine and validate a variety of decision rules which can be applied to administrative databases to identify patients with RA.

Methods: The study was conducted at a single academic medical center and utilized administrative health care data from a geographic area of approximately 1 million people who had access to a universal healthcare system. A retrospective cohort study was performed through the Population Health Research Unit at our institution and utilized data from existing administrative databases. These included information on hospital discharges and physician billings over a 10 year period. Each RA study subject was matched by age and gender to randomly selected control subjects in the same datasets but without a diagnosis of RA or other inflammatory arthropathies. A total of 7 decision rules, some derived from previous studies, were applied to the administrative data to identify RA cases. The sensitivity, specificity, overall accuracy, positive (PPV) and negative (NPV) predictive values of these rules was compared to the diagnosis of a rheumatologist in the academic medical center as determined by chart review.

Results:

Conclusion: The performance of decision rules for the identification of RA cases in administrative healthcare databases in variable and should be considered when comparing data across studies. This variability may also be used to advantage in study design when, for example, either sensitivity or specificity is the most critical issue for different population health research questions.