A0424
Title: Bayesian analysis of verbal autopsy data using probit model with age- and sex-dependent association between symptoms
Authors: Tsuyoshi Kunihama - Kwansei Gakuin University (Japan) [presenting]
Abstract: Verbal autopsy surveys have been used for understanding distributions of deaths by cause, which is fundamental public health information, in low-resource settings without well-organized vital statistics systems. A new Bayesian approach which extracts the information of distributions of causes of death is developed from verbal autopsy data by taking into account its feature that associations between symptoms vary over the age and sex of individuals. Using gold-standard verbal autopsy data from the Population Health Metrics Research Consortium, we assess the performance of the proposed method by comparing it with existing approaches in this literature. Further, we evaluate the importance of predictors based on information-theoretic measures.