Machine Learning Detects Heterogeneous Effects of Medicaid Coverage on Depression
Previous research has demonstrated that Medicaid coverage reduces the risk for developing depression among recipients, but the question is who benefits most from coverage. Using a tool called machine learning causal forest to analyze data from the Oregon Health Insurance Experiment, a research team that includes senior author Dr. Yusuke Tsugawa, associate professor in residence of medicine at David Geffen School of Medicine at UCLA and of health policy and management at the UCLA Fielding School of Public Health, found that people who were older and had more physical or mental health conditions at baseline were likeliest to experience the highest benefit. Causal forest is a machine-learning algorithm that estimates the effects of an intervention on an outcome based on the person’s characteristics, allowing for personalized predictions of how each individual will benefit from a given treatment. The study found that providing Medicaid coverage to individuals with high predicted benefit generated 2.4 times greater reduction in depression prevalence at the population level, with substantially lower cost, compared with providing coverage to everyone. The approach could be useful not just for depression but “in treatment effects across the full spectrum of individual-level demographic and health characteristics.” The researchers conclude that the machine learning causal forest approach could improve the effectiveness and efficiency of insurance expansion. Read the article in the American Journal of Epidemiology.