Latent Class Models for Prevalence Estimation of Emerging Diseases Using Verbal Autopsy Data

• Activity Number: 417 - Statistical Methods for Discovering Latent Structures in High-Dimensional and Complex Data
• Type: Topic Contributed
• Date/Time: Wednesday, August 10, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Zehang Richard Li*
• Companies: UCSC
• Keywords: data shift; dependent binary data; quantification learning

Abstract

A widely-used tool to obtain information on causes-of-death when a medically certified cause-of-death is not available is verbal autopsies (VA). VA involves a structured questionnaire administered to family members or caregivers of a recently deceased person. During public health emergencies when new diseases emerge, VAs are usually the only feasible tool to gather information on cause-of-death in many low- and middle-income countries. Existing statistical methods for analyzing VAs, however, are not suitable for such situations. We propose a novel latent class model framework to estimate the prevalence of a new disease using limited VAs collected under an informative selection process. We will also extend the framework to monitor the dynamics of mortality rates over time.