Objective: To assess self-reported health status (SRHS) in two cohorts of participants with radiographic knee osteoarthritis (OA) and examine the extent that differences in SRHS are due to study design.
Method: We used data from the Third National Health and Nutritional Examination Survey (NHANES-III; population-based national survey) and the Osteoarthritis Initiative (OAI; prospective cohort study). Inclusion criteria for this analysis were age 60-79 and presence of radiographic knee OA. SRHS, elicited as a five-item domain (excellent, very good, good, fair, poor), was analyzed by dichotomizing the general health status measure as "fair/poor" versus all other states. We estimated the proportion of participants in fair/poor health from each study. Propensity score methodology was used to adjust for the differences in sampling strategies between the two studies.
Results: Thirty-four percent (N = 1,608) of OAI and 29% (N = 756) of NHANES-III participants satisfied inclusion criteria. The proportion in fair/poor health was higher in NHANES-III (28%) than in OAI (5%). After adjusting for the propensity score, the proportion in fair/poor health was four times higher in NHANES-III than in OAI.
Conclusion: SRHS was substantially better in OAI than in NHANES-III. Self-selection bias may contribute to overestimation of SRHS in prospective cohort studies such as OAI.