Causal Discovery Maps Observational Medical Research
A scoping review published in JMIR Medical Informatics in 2026 systematically maps applications of machine-learning causal discovery methods in observational medical research, searching six databases through May 2025 and including 72 eligible studies from 2,296 records. The review finds constraint-based algorithms most common (53%), predominant clinical applications in mental health and chronic disease, and recurring challenges such as unmeasured confounding, limited sample sizes, and lack of standardized validation frameworks.
Key Points
- 1Reports that 72 studies (3.1% of 2,296) applied causal discovery in medical contexts
- 2Highlights constraint-based methods dominate (52.8%), with mental health and chronic disease emphasis
- 3Warns about unmeasured confounding, small samples, and missing validation, urging standardized evaluation frameworks
Scoring Rationale
Comprehensive, peer-reviewed synthesis across medical domains, but modest novelty and persistent methodological limitations limit direct translational impact.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems