To use or not to use propensity score matching?

Area of Science:

• Epidemiology
• Biostatistics
• Observational Studies

Background:

• Propensity score matching (PSM) is a common technique to address confounding in observational research.
• Recent concerns question PSM's efficacy due to potential postmatching covariate imbalance.
• This has sparked debate regarding the appropriate use of PSM in statistical inference.

Purpose of the Study:

• To critically review the empirical and theoretical evidence supporting and challenging the use of PSM.
• To re-evaluate the property of equal percent bias reduction and its practical adaptations.
• To investigate the impact of caliper width on PSM-induced biases and population differences.

Main Methods:

• Comprehensive review of existing literature on propensity score matching.
• Theoretical re-examination of bias reduction properties.
• Simulation study to assess the influence of caliper width on matching quality and bias.

Main Results:

• PSM possesses desirable statistical properties when applied correctly.
• Inadequate caliper selection can lead to significant biases in matched samples.
• The choice of caliper width impacts the balance between matched and target populations.

Conclusions:

• The debate should shift from 'whether to use PSM' to 'when and how to use PSM effectively'.
• Proper implementation, particularly appropriate caliper selection, is key to maximizing PSM benefits.
• Guidance is provided for optimal application of PSM in observational studies.