A Logistic Regression Model for Estimating Transport Accident Deaths Using Verbal Autopsy Data

Abstract

This study aimed to create an appropriate model using verbal autopsy (VA) data to estimate transport accident deaths from vital registration data in Thailand. A sample of 9644 VA deaths was obtained from the Thai Ministry of Public Health. VA assessed transport accidents accounted for 546 deaths (5.7% of sample). Logistic regression was used to model transport accident deaths classified by 9 provinces, 16 gender–age groups, 14 combinations of vital reported cause groups, and place of death (in or outside hospital). The receiver operating characteristic curve was used to match the number of reported transport accident deaths to the number predicted by the model with sensitivity 73.8% and false positive rate 1.6%. The estimated transport accident deaths ranged from 1.68 to 2.65 times higher than the vital registration data reported according to gender–age groups.