Estimating Heterogeneous Treatment Effects with Real-World Health Data – A Scoping Review of Machine Learning Methods

Key Points

• Heterogeneous treatment effects estimation is gaining traction in real-world data analysis, emphasizing diverse outcomes.
• Tree-based models dominate current machine learning approaches, while customized conditional average treatment effect methods are gaining interest.
• This scoping review assesses various machine learning methods for estimating treatment effects in health economics research.
• Improved evaluation standards and transparent reporting are essential for these models to be trusted in healthcare studies.