Efeito do plano amostral complexo em inquérito epidemiológico de saúde bucal em Minas Gerais 2012

Abstract

OBJECTIVES: The objective of this paper was to present the most important aspects of the sampling plan of the Survey of oral health conditions of the Minas Gerais’ population - SB Minas Gerais, conducted in 2012, especially regarding the design effect (deff). METHODS: Estimates were calculated for each domain defined by geographical units (Capital, Interior I and Interior II) and demographic groups by 5 and 12 years and age groups 15 to 19, 35 to 44, 65 to 74 years. Comparing these estimates with or without basic weights allows us to evaluate the effect of intraclass homogeneity and the impact of sample weights on precision. RESULTS: Overall, there is a low impact of clustering and sample weight on precision when comparing estimates with and without weighting on estimated caries prevalence. Estimates for mean, standard errors, and deff of the DMFT index in the 15-19 age group and 35-44 age group can also be considered stable for the vast majority of domains and thus compatible with the precision criteria predicted in the design. However, the highest value for the deff was obtained in the group of 35 to 44 years for municipalities of the Interior II, yet not strongly affecting the point estimate, only increasing the confidence interval amplitude and standard error. The presented deff consolidate the results obtained for averages and proportions in the three geographic domains and in four age domains (except for the adults in the Interior II domain). CONCLUSIONS: Statistical analyzes performed with the database from complex samples should take into account the sample design. Furthermore, in future epidemiological studies, when planning sampling, consideration should be given to the need to vary the deff according to age groups