SF-36 scores vary by method of administration: implications for study design

Background Previous research suggests that people respond differently to health status measures when data are collected by interview or self completion of a questionnaire. The objective of this study was to determine whether SF-36 health status scores differ systematically by method of administration. Method A randomized cross-over study was carried out on 210 new attenders at general medicine, endocrinology, gastroenterology and urological out-patient departments. The outcome was the difference in SF-36 profiles comparing clinic based interviews with self completion at home by the same subjects. Results For seven of the eight variables of the SF-36 scores were lower in the self assessment, the differences being statistically significant in four of the eight comparisons. The largest differences were in role limitations due to emotional problems (difference 14.74, 95 per cent confidence interval (CI) 7.76-21.7) and social function (difference 7.21, 95 per cent Cl 3.19-11.23). Conclusions Clinic based interviews systematically exaggerate health status compared with self assessment. The difference is sufficiently large to underestimate the effectiveness of health service interventions when a clinic based preintervention and postal self completed follow-up design is used, unless adjustment is made for this systematic bias.