VALIDATION DATA

Cronbach’s alpha >0.70

Internal consistency estimates were >0.90 for physical functioning and bodily pain scales in at least 2 studies and for general health in at least 1 study

ICC: 0.55 on mental health, 0.66 on role emotional, 0.44 on role physical, 0.03 on vitality scales.

There is a high proportion of measurement error on the SF-36 questionnaire in rheumatology, with large SDs for one-time administration (up to or exceeding the mean score) and large variations (change of up to 200% from the initial score) in the SF-36 scores over a one-week test–retest period. In orthopedic surgery, minimal detectable change at an individual level ranged from 22% (general health) to 97% (role physical) of the total score range.

The SF-36 was developed out of the RAND Corporation Health Insurance Experiment. The initial study measured 40 health concepts, 8 of which were selected for the inclusion into the new questionnaire. These 8 concepts were chosen to represent issues that were frequently used in health surveys and were affected by disease and treatment. Items for the questionnaire were generated following review of the content of various instruments that, at the time of SF-36 construction, were used to measure mental health and general health perceptions as well as limitations in physical, social, and emotional role functioning. No patient groups or representatives of the general population were involved in questionnaire construction. 

The dimensional structure of the SF-36 (8 first-order and 2 higher-order factors) has been questioned in several studies. For example, higher-order factor analysis of the scale scores have confirmed separation of the scale scores into mental and physical health summary scores in some rheumatic studies, but not others. First-order factor analysis has also failed to confirm the 8-dimensional structure of the SF-36 in either exploratory or confirmatory factor analysis.