An evolutionary algorithm for the direct optimization of covariate balance between nonrandomized populations.

Area of Science:

• Biostatistics
• Health Informatics
• Epidemiology

Background:

• Confounding bias in nonrandomized patient comparisons necessitates robust matching methods.
• Propensity score (PS) matching is a common technique, but model misspecification can lead to residual bias.
• Matching quality is typically assessed by post-matching covariate balance.

Purpose of the Study:

• To develop and evaluate a novel matching approach that directly optimizes covariate balance.
• To compare the performance of this new method against traditional PS matching.
• To explore the impact of different balance metrics on matched population properties.

Main Methods:

• An evolutionary algorithm was designed to directly optimize arbitrary covariate balance metrics.
• The proposed method was tested on a large simulated dataset (275,000 patients, 10 covariates).
• The algorithm was applied to match clinical trial patients (250) to electronic health record data (160,000+ patients, 101 covariates).

Main Results:

• The evolutionary algorithm outperformed PS matching in achieving specified covariate balance.
• The method demonstrated comparable performance to linear integer programming optimization techniques.
• The approach supports arbitrary balance metrics, including nonlinear functions, offering flexibility.

Conclusions:

• Directly optimizing covariate balance with an evolutionary algorithm provides superior matching quality compared to PS methods.
• This approach offers a flexible and effective tool for constructing external control arms and reducing bias in observational studies.
• The implementation is available in Python for broader application.