Methods for estimating subgroup effects in cost-effectiveness analyses that use observational data.

Area of Science:

• Health Economics
• Econometrics
• Biostatistics

Background:

• Cost-effectiveness analyses (CEA) require subgroup-specific estimates for patient decision-making.
• Propensity score (PS) matching and inverse probability of treatment weighting (IPTW) are used to address selection bias in nonrandomized studies.
• Effective covariate balance between treatment groups is crucial for the validity of PS matching and IPTW.

Purpose of the Study:

• To compare the performance of genetic matching (GM), PS matching, and IPTW in estimating subgroup effects within CEA.
• To evaluate the robustness of these methods to propensity score misspecification.
• To assess the impact of different methods on cost-effectiveness estimates for specific patient subgroups.

Main Methods:

• A case study involving a CEA of drotrecogin alfa (DrotAA) for severe sepsis patients (n=2726).
• A simulation study comparing GM, PS matching, and IPTW under correct and misspecified propensity score models.
• Evaluation metrics included covariate balance, bias, and root mean squared error (RMSE) of incremental net benefits.

Main Results:

• In the case study, GM demonstrated superior covariate balance compared to PS matching and IPTW.
• For a high-risk subgroup, cost-effectiveness probabilities varied significantly: 30% (IPTW) to 90% (PS matching and GM).
• In simulations, GM showed better balance, reduced bias, and more precise estimates, particularly when the PS was misspecified.

Conclusions:

• GM, PS matching, and IPTW can yield unbiased cost-effectiveness estimates if the PS is correctly specified and IPTW weights are stable.
• GM offers greater robustness to propensity score misspecification compared to PS matching and IPTW.
• GM is a valuable tool for reliable subgroup analyses in CEA, especially when potential model misspecification is a concern.