Reliability, validity and responsiveness of the short form-36 health survey: findings from the women’s health study of Accra, Ghana

Measuring the health status of individuals or a population is an important first step in the assessment of health care needs and the evaluation of the impact of health interventions. Reliable population-wide statistics on the health status are scarce especially in developing countries including Ghana. This paper assesses the validity and reliability of a standard instrument widely used in high income countries, the Short Form-36, as a tool to evaluate health status among women age 20 years and older in urban Ghana. The study draws on a representative sample of 2’814 respondents from the Women’s Health Study of Accra. We used multi-trait, multi-item matrix methods, Cronbach’s alpha estimates, half-split reliability tests and principal component analysis to evaluate the validity of SF-36 in the Ghanaian context. The response rate and the rate of completion for each health domain was over 97%. Item-level validation satisfied the rigorous psychometric criteria for internal consistency and reliability was above the suggested 0.40 criterion for all SF-36 scoring items apart from one item under Vitality sub-scale. The normed SF-36 scores showed wide variability and acceptable internal consistency (Cronbach’s alpha >0.70). Following simple orthogonal rotation two principal components were identified and interpreted as Physical and Mental health components, identical to those found in other study settings. These components accounted for 62.6% of the reliable variance in scale scores across the study population. The absence of strong correlations between the various eight SF-36 sub-scales, demonstrate that each scale measures a distinct concept of health. The study provides evidence that SF-36 is suitable for the assessment of health status among women in Accra and, by extension, suitable for assessment of the health status of women in urban areas of Ghana. Key words Introduction Reliable, objectiv