[Estimation on the individual treatment effect among heterogeneous population, using the Causal Forests method]

Area of Science:

• Biostatistics
• Medical Informatics
• Health Services Research

Background:

• Estimating individual treatment effects is crucial for personalized medicine.
• Heterogeneous patient populations present challenges in treatment response prediction.
• The Causal Forests (CF) method offers a potential solution for estimating treatment effects.

Purpose of the Study:

• To evaluate the effectiveness of Causal Forests (CF) in estimating individual treatment effects.
• To identify characteristics of heterogeneous patient populations using CF.
• To analyze real-world data for right heart catheterization (RHC) treatment effects.

Main Methods:

• Conducted four computer simulation schemes to test CF under varying treatment effect environments.
• Applied CF to real-world data from patients undergoing right heart catheterization (RHC).
• Assessed the consistency and distribution of estimated individual treatment effects.

Main Results:

• Simulation results demonstrated CF's accuracy in estimating individual treatment effects.
• Real-world data analysis indicated positive individual treatment effects for RHC in most patients.
• Lower predicted 2-month survival probability and albumin levels were associated with reduced mortality risk post-RHC.

Conclusions:

• Causal Forests (CF) is an effective tool for estimating individual treatment effects.
• CF can assist patients in making informed decisions about receiving specific treatments.
• Identifying patient subgroups with differential treatment responses is feasible with CF.