Search

The problem of missing values has increasingly being recognized in epidemiology. New methods allow for the analysis of missing data that can provide valid estimates of epidemiological quantities of interest. The GISSI-Prevenzione study was aimed to reliably assess the long-term relationship between the consumption of foods typical of the Mediterranean diet and the risk of mortality amongst 11,323 Italians with prior myocardial infarction. Food intake frequencies were recorded repeatedly over the 4.5 years of follow-up and missing values affected each food variable at increasing rates over the course of the study. Comparisons were made between the results obtained from the analysis of the complete data and those obtained after imputing the missing data with simple imputation methods and with various implementations of the multiple imputation (MI) method. MI appeared to best address the issue of missing data on the food intake frequencies, preserving the observed distributions and relationships between variables whilst producing plausible estimates of variability. Given its theoretical properties and flexibility to different types of data, MI is more likely to provide valid estimates, compared to complete data analysis and imputation by simple methods, and is thus worthy of wider consideration amongst epidemiological researchers. [ABSTRACT FROM AUTHOR]