Measuring general health status in dementia: practical and methodological issues in using the SF-36

The objective of this study was to examine the measurement properties of the Medical Outcomes Study (MOS) 36-Item Short-Form health survey (SF-36) in a representative sample of dementia patients. A total of 138 subjects with dementia were evaluated with the SF-36. They were recruited from 16 centers, including 7 university hospital centers. Dementia severity, demographic variables, co-morbidity and functional limitations were also measured. The measurements obtained were compared by source of information (patient, family proxy, and care provider proxy). Thirteen patients refused to complete this questionnaire. Help from the interviewer was necessary in 72.8% of the cases. The average completion time was 18.8 minutes. Missing data exist in only 5.6% of the cases on average, except among patients with severe dementia (Mini Mental State Examination <10). With regard to reliability, internal consistency was acceptable to good; Cronbach's a ranged from a low of 0.59 to a high of 0.92 across subscales (median 0.75). Test-retest intraclass correlation coefficients were moderate to good (range 0.51-0.81) except for Role emotional (0.17), Bodily pain (0.49) and Mental health (0.45). For patients with MMSE > or = 15, test-retest coefficients were better (range 0.53-0.90). Intraclass correlation coefficient suggests that proxies are a poor substitute for obtaining a patient's perspective of his/her health status. In conclusion, the SF-36 is unsuitable for severe dementia, because severity of disease mainly affected the feasibility, acceptability, and reproducibility of the instrument. This study, however, confirms that it is possible to question subjects with mild to moderate dementia on their health status.