Problems in using health survey questionnaires in older patients with physical disabilities. The reliability and validity of the SF-36 and the effectof cognitive impairment

Reliability and validity of the SF-36 Health Survey Questionnaire was asses sed in older rehabilitation patients, comparing cognitively impaired with c ognitively normal subjects. The SF-36 was administered by face-to-face inte rview to 314 patients (58-93 years) in the day hospital and rehabilitation wards of a department of medicine for the elderly. Reliability was measured using Cronbach's alpha (for internal consistency) on the main sample and i ntraclass correlation coefficients on a test-retest sample: correlations wi th functional independence measure (FIM) were examined to assess validity. In 203 cognitively normal patients (Mini-Mental State Examination greater t han or equal to 24), Cronbach's alpha scores on the eight dimensions of the SF-36 ranged from 0.545 (social function) to 0.933 (bodily pain). The rang e for the 111 cognitively impaired patients was 0.413-0.861. Cronbach's alp ha values were significantly higher (i.e. reliability was better) in the co gnitively normal group for bodily pain (P=0.003), mental health (P=0.03) an d role emotional (P=0.04). In test-retest studies on a further 67 patients, an intraclass correlation coefficient of 0.7 was attained for five out of eight dimensions in cognitively normal patients. and four out of eight dime nsions in the cognitively impaired. Only the physical function dimension in the cognitive ly normal group attained the criterion level (r>0.4) for con struct validity when correlated with the FIM. In this group of older physic ally disabled patients, levels of reliability and validity previously repor ted for the SF-36 in younger subjects were not attained, even on face-to-fa ce testing. Patients with coexistent cognitive impairment performed worse t han those who were cognitively normal.