Machine Learning for Effect Estimation in International Health

• Activity Number: 588
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Sherri Rose*+
• Companies: Harvard Medical School
• Keywords: robust estimation ; matched cohort ; epidemiology ; machine learning ; causal inference

Abstract

The impact of chronic disease on prosperity outcomes, such as poverty, has not yet been determined in many resource-limited settings. Due to a lack of health-systems focus on chronic disease, there is a preventable load of premature mortality from chronic disease. Our new statistical work was driven by data available in a health and demographic surveillance system in Bangladesh, linked with a novel economic impact survey collected via a cohort matched design. Matching based on exposure in cohort studies is not implemented as frequently as other types of matching, such as in case-control designs. A new method for robust statistical machine learning in matched cohort studies will be presented. The findings from our study will contribute to isolating the extent of the causal effect between chronic disease and poverty in Bangladesh, quantifying the potential maximum benefit of interventions, such as improved health services, to reduce death from chronic disease.