Learning Optimal Treatment Rules

Machine learning combined with decision modelling can improve estimation of costs and health outcomes, according to Dr. Noemi Kreif and Dr. David Glynn of the University of York. When assessing health policies such as pharmaceutical interventions, some individuals benefit more from the treatment than others and some might even be harmed. Existing approaches do not always sufficiently account for this variation.

In a recent CHDS seminar, Dr. Kreif and Dr. Glynn discussed their work that uses new machine learning tools to better address these concerns. They integrate machine learning estimates of heterogeneous causal effects in a decision model, using SPRINT (Systolic Blood Pressure Intervention Trial) as a case study. With this method, they are able to improve the estimation of costs and health outcomes at the population level, make predictions at the individual level, and formulate optimal treatment allocation rules under resource constraints.

To be added to the email list for future CHDS seminars, please write to chds@hsph.harvard.edu.

Learn more: Read the article, Machine Learning Methods in Health Economics and Outcomes Research-The PALISADE Checklist: A Good Practices Report of an ISPOR Task Force
Learn more: Explore Chapter 16 of the Handbook of Research Methods and Applications in Empirical Microeconomics: Machine learning for causal inference: estimating heterogeneous treatment effects
Learn more: Visit Harvard’s Public Impact Analytics Science Lab

Related news: Clinical Effectiveness with Steven Pearson
Related news: COVID-19 Policy Modeling with Jagpreet Chhatwal