Validation of a surveillance case definition for arthritis

Abstract

Objective: To assess whether self-reports of chronic joint symptoms or doctor-diagnosed arthritis can validly identify persons with clinically verifiable arthritis.

Methods: The Behavioral Risk Factor Surveillance System (BRFSS), a telephone health survey, defines a case of arthritis as a self-report of chronic joint symptoms (CJS) and/or doctor-diagnosed arthritis (DDx). A sample of health plan enrollees aged 45-64 years and >/= 65 years with upcoming annual physical examinations were surveyed by telephone using the 2002 BRFSS CJS and DDx questions. Based on responses (CJS+, DDx-; CJS-, DDx+; CJS+, DDx+; CJS-, DDx-), respondents were recruited to undergo a standardized clinical history and physical examination for arthritis (the gold standard for clinical validation). Weighted sensitivities and specificities of the case definition were calculated to adjust for sampling.

Results: Of 2180 persons completing the telephone questionnaire, 389 were examined; of these, 258 met the case definition and 131 did not. For those examined and aged 45 to 64 years (n = 179), 96 persons had arthritis confirmed, of whom 76 met the case definition. Among those examined and aged >/= 65 (n = 210), 150 had arthritis confirmed, of whom 124 met the case definition. Among those without clinical arthritis, 45 of 83 of those aged 45 to 64 years and 40 of 60 of those aged >/= 65 did not meet the case definition. For those aged 45 to 64 years, the weighted sensitivity of the case definition in this sample was 77.4% and the weighted specificity was 58.8%; for those aged >/= 65, the sensitivity was 83.6% and specificity 70.6%. CJS+ had higher sensitivity and lower specificity than DDx+ in the younger age group; CJS+ and DDx+ behaved more comparably in the older age group.

Conclusion: The case definition based on self-reported CJS and/or DDx appeared to be sensitive in identifying arthritis, but specificity was lower than desirable for those under age 65 years. Better methods of ascertaining arthritis by self-report are needed. Until then, a change in the surveillance case definition for arthritis appears warranted.