A methodology for small area prevalence estimation based on survey data

BACKGROUND: Brazil conducts many health surveys to provide estimates by national level, macro-regions, states, metropolitan regions and capitals. However, estimates for smaller areas are lacking due to their high cost. The Health Vulnerability Index (in Portuguese, Índice de Vulnerabilidade em Saúde, IVS) is a measure that combines socioeconomic and environmental variables in the same indicator and allows for the analysis of the characteristics of population groups residing in census tracts, grouping them into four health risk areas (low, medium, high and very high risk) in addition to showing inequalities in the epidemiological profile of different social groups. This index was developed by the Municipal Health Secretariat of Belo Horizonte to guide health planning. OBJECTIVE: The aim of the study is to produce a methodology for obtaining reliable estimates for tobacco smoking in small areas for which the IVS was not designed. METHODS: The Vigitel dataset from 2006 to 2013 was used to obtain estimates of the prevalence of smokers based on the IVS employing small area estimation methods that use data from a larger domain to obtain estimates in smaller areas. For indirect estimates, the covariates included were sanitation, housing, education, income, and social and health factors. Post-stratification weights were used according to the IVS based on the population of the 2010 census. RESULTS: From 2006 to 2009, 16.2% (95% CI: 13.6–14.8%) of the adult population in Belo Horizonte were smokers, and 14.8% (95% CI: 14.0–15.6%) were smokers between 2010 and 2013. The very high-risk population maintained a high prevalence over the same period of 21.1% (95% CI: 17.1–25.0%) between 2006 and 2009 and 20.8% (95% CI: 17.0–24.6%) between 2010 and 2013, while in the low-risk group, the prevalence in the same period fell from 14.9% (95% CI: 13.7–16.2%) to 11.8% (95% CI, 10.6–13.1%). CONCLUSIONS: The present study identified differences in the profile of smokers by the IVS in the city of Belo Horizonte. While the smoking prevalence declined in richer areas, it remained high in poor areas. This methodology can be used to produce reliable estimates for subgroups with greater vulnerability in small areas and thus subsidize the formulation, monitoring and evaluation of public health policies and programmes aimed at smoking.