OBJECTIVE

To evaluate the measurement properties of an Internet-based self-administered questionnaire in ascertaining cases of hip and knee osteoarthritis (OA).

METHODS

Questionnaire data from 4269 Canadian subjects aged 45-85 were collected on hip and knee joint health including self-reported items on medically-diagnosed hip and knee OA and joint replacement. A sub-cohort of 100 subjects was recruited for clinical examination. The self-reported outcomes were evaluated using the American College of Rheumatology clinical classification criteria for hip and knee OA as the gold standard for clinical verification. Analysis was at the joint level (200 knees, 200 hips). Validity was examined using sensitivity, specificity, and predictive values; to account for correlated joints of the same subject, bootstrapping was performed to yield valid 95% confidence interval (CI's).

RESULTS

Self-reported measures for a medical diagnosis of knee OA had a positive predictive value of 86%, negative predictive value 91%, sensitivity 73% and specificity 96% for correctly identifying clinical knee OA. For hip OA, the values were 61%, 98%, 81% and 94% respectively.

CONCLUSION

Internet self-report of medically-diagnosed hip and knee OA in metro Vancouver residents correctly identified most cases and non-cases of clinical OA when compared with the ACR clinical classification criteria gold standard. In particular, specificity was very high, important in risk factor studies due to the profound effect of even small losses in specificity on the measure of association. The findings provide evidence that these questionnaire case definitions have utility for identifying hip and knee OA in community and population-based studies when the purpose is to link potential risk factors with knee and hip health.