Estimation of caries experience by multiple imputation and direct standardization

Valid estimates of caries experience are needed to monitor oral population health. Obtaining such estimates in practice is often complicated by nonresponse and missing data. The goal of this study was to estimate caries experiences in a population of children aged 5 and 11 years, in the presence of nonresponse and missing data. Four estimation methods are compared. Each method makes implicit assumptions about the processes that caused the nonresponse and the missing data. Three of the four methods are based on unrealistic assumptions about the missing data and underestimate caries experience. Under the missing at random assumption, multiple imputation in combination with direct standardization corrects for the deficiencies of current methodology. In the presence of missing data and nonresponse, we recommend a combination of multiple imputation and direct standardization to obtain correct estimates of caries experience.