Estimating deaths due to other infectious diseases in Thailand 1996-2009 based on 2005 Verbal Autopsy data

Abstract

Background Vital registries in Thailand lack credibility because they often contain errors in deaths reported to be due to infectious disease. In this study, the term other infectious diseases excludes tuberculosis, septicemia and HIV Objective The objective of this study was to model other infectious disease deaths based on 2005 Verbal Autopsy (VA) data and then to use that model to estimate other infectious disease deaths in Thailand from 1996 to 2009. Methods Logistic regression models were used to model other infectious disease deaths from the 2005 VA dataset. The full model comprises province, gender-age group, and death certifi cate cause location group as factors. This model was then used to predict other infectious disease deaths from the vital registration database of the province in which the Verbal Autopsy was conducted. Proportions in the remaining provinces were predicted from spatial interpolation based on province coeffi cients. Results The model without demographic covariates was preferred for comparison to full models by the area under the ROC curve and AIC. The most misclassifi ed other infectious diseases deaths were those registered as septicemia and digestive. The confi dence intervals from a statistical full model provided the estimates and their degree of precision. Conclusion Demographic covariates are not essential to this study for accurately correcting causes of other infectious disease deaths. The model in this study can be used to more accurately estimate mortality data where their national vital registration data are of poor quality.