The SF-36 arthritis-specific health index (ASHI) I. Development and cross-validation of scoring algorithms

An arthritis-specific health index (ASHI) for the SF-36 Health Survey was d eveloped by studying its responsiveness to changes in clinical indicators o f arthritis severity. Longitudinal data from 1,076 patients participating i n four placebo-controlled trials were analyzed. All had at least a 6-month history of moderate to severe osteoarthritis or rheumatoid arthritis of the knee or hip. All had undergone a washout period of 3 to 14 days before bas eline assessment to bring about a flare state in osteoarthritis or rheumato id arthritis symptoms. Their average age was 60 years and 72% were female. Change scores for the eight-scale SF-36 health profile (acute version) and five arthritis-specific measures of disease severity (knee pain on weight b earing, time to walk 50 feet, physician global evaluation of symptom severi ty and impact, patient global evaluation of symptom severity and impact, an d pain intensity visual analogue scale) were computed by subtracting scores before treatment from scores at two-week follow-up. Canonical correlation methods were used to derive weights for changes in SF-36 scales to score a single index (ASHI) that maximized its correlation with changes in the set of five clinical measures of arthritis severity. The weights used to score the ASHI were cross-validated in a 25% holdout group (N = 144) from the fir st two osteoarthritis trials and in two additional osteoarthritis and rheum atoid arthritis trials (N = 530). Only one SF-36 canonical variate (ASHI) c orrelated significantly (F = 4.69, P < 0.0001) with the clinical canonical variate that served as the "criterion" measure of change in the severity of arthritis. Changes in the ASHI and clinical canonical variate were substan tially correlated in the developmental sample (r = 0.628, P < 0.0001) and o n cross-validation (r = 0.629, P < 0.0001). The clinical canonical variate correlated highly (r = 0.75-0.88) with changes in all but one of the five c linical measures (50-foot walk; r = 0.41). The pattern of correlations betw een changes in SF-36 scales and the ASHI indicated that ASHI is primarily a measure of bodily pain (r = 0.92) and other aspects of physical and role f unctioning and well-being (r = 0.69 for Pole-Physical, r = 0.68 for Physica l Functioning, r = 0.52 for Social Functioning, and r = 0.51 Vitality). The patterns of correlations between SF-36 scales and the ASHI were very simil ar across developmental and cross-validation samples. This research demonst rates the feasibility and generalizability of a single ASHI scored from cha nges in responses to the SF-36 Health Survey. The generic SF-36 health prof ile, which has already been shown to be useful in comparing arthritis with other diseases and treatments, can also be scored specifically to make it m ore useful in studies of osteoarthritis and rheumatoid arthritis.