P Marsh1, R Carlisle, A J Avery. 1. Division of General Practice, University of Nottingham, University Hospital.
Abstract
BACKGROUND: The best way for practices to determine the health status of patients living in areas with different socioeconomic characteristics is unclear. AIMS: To see how much SF-36 health status varies between electoral wards, how much of this variation can be explained by census-derived Jarman and Townsend scores, and compare the performance of census scores with direct socioeconomic information. METHOD: A postal questionnaire survey of 3000 randomly selected 18 to 75-year-olds residing in 15 electoral wards and registered with two urban practices. RESULTS: The response rate was 73%. Only two of the eight SF-36 domains were significantly associated with Jarman scores, whereas seven domains were associated with the Townsend score. Of the four socioeconomic variables derived directly from the survey, unemployment showed the weakest association, housing tenure was associated with seven domains, and car ownership and low income were associated with all eight. Income explained between 47% to 71% of the variation across the eight domains. CONCLUSION: The most accurate predictions about health status were made from direct socioeconomic information. Nonetheless, the association between Townsend score and health status was strong enough to be of practical importance. This study cautions against assuming the Jarman score of a population has a clear relationship with its health status.
BACKGROUND: The best way for practices to determine the health status of patients living in areas with different socioeconomic characteristics is unclear. AIMS: To see how much SF-36 health status varies between electoral wards, how much of this variation can be explained by census-derived Jarman and Townsend scores, and compare the performance of census scores with direct socioeconomic information. METHOD: A postal questionnaire survey of 3000 randomly selected 18 to 75-year-olds residing in 15 electoral wards and registered with two urban practices. RESULTS: The response rate was 73%. Only two of the eight SF-36 domains were significantly associated with Jarman scores, whereas seven domains were associated with the Townsend score. Of the four socioeconomic variables derived directly from the survey, unemployment showed the weakest association, housing tenure was associated with seven domains, and car ownership and low income were associated with all eight. Income explained between 47% to 71% of the variation across the eight domains. CONCLUSION: The most accurate predictions about health status were made from direct socioeconomic information. Nonetheless, the association between Townsend score and health status was strong enough to be of practical importance. This study cautions against assuming the Jarman score of a population has a clear relationship with its health status.