OBJECTIVES: The inclusion of machine learning based approaches in healthcare applications has increased at a rapid pace over the past years; however, there is limited understanding of how machine learning intersects with health economics and outcomes research (HEOR). We conducted a systematic review of machine learning and HEOR to understand the intersection of the two fields, identify gaps, and suggest future directions.

METHODS: A systematic search was conducted in Medline and Embase databases for the period between January 2004 and September 2021. The title and abstract screening were performed by two independent reviewers per PRISMA guidelines. We included studies that mentioned both machine learning and cost-effectiveness/health economics.

RESULTS: Of the 861 articles screened, 45 articles were included in the review. The included articles fell into six major categories: 1) improving clinical pathway and operational efficiency (33%), 2) predicting disease outcomes and treatment success (20%), 3) screening, detection, or diagnosis of a disease (30%), 4) predicting high-cost utilizers in insurance claims (4%), 5) improving methods used in health economic modeling (11%), and 6) predicting who will benefit from a digital intervention (2%). The supervised learning methods such as Random Forest, Support Vector Machines, and Logistic Regression were the most commonly utilized machine learning methods. Image analysis and electronic health records were the most commonly used data source.

CONCLUSIONS: Machine learning has a potential to influence HEOR but the current applications remain limited in scope and depth. Future studies can explore emerging areas including digital interventions, wearable technologies and social media data.