COMPARISON OF METHODS FOR THE SCORING AND STATISTICAL-ANALYSIS OF SF-36 HEALTH PROFILE AND SUMMARY MEASURES - SUMMARY OF RESULTS FROM THE MEDICAL OUTCOMES STUDY

Physical component summary (PCS) and mental component summary (MCS) me asures make it possible to reduce the number of statistical comparison s and thereby the role of chance in testing hypotheses about health ou tcomes. To test their usefulness relative to a profile of eight scores , results were compared across 16 tests involving patients (N = 1,440) participating in the Medical Outcomes Study. Comparisons were made be tween groups known to differ at a point in time or to change over time in terms of age, diagnosis, severity of disease, comorbid conditions, acute symptoms, self-reported changes in health, and recovery from cl inical depression. The relative validity (RV) of each measure was esti mated by a comparison of statistical results with those for the best s cales in the same tests. Differences in RV among scales from the Medic al Outcomes Study 36-Item Short-Form Health Survey (SF-36) were consis tent with those in previous studies. One or both of the summary measur es were significant for 14 of 15 differences detected in multivariate analyses of profiles and detected differences missed by the profile in one test. Relative validity coefficients ranged from .20 to .94 (medi an, .79) for PCS in tests involving physical criteria and from .93 to 1.45 (median, 1.02) for MCS in tests involving mental criteria. The MC S was superior to the best SF-36 stale in three of four tests involvin g mental health. Results suggest that the two summary measures may be useful in most studies and that their empiric validity, relative to th e best SF-36 scale, will depend on the application. Surveys offering t he option of analyzing both a profile and psychometrically based summa ry measures have an advantage over those that do not.