Are propensity scores really superior to standard multivariable analysis?

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Health Services Research

Background:

• Clinical decisions often lack randomized trial evidence, necessitating reliance on non-randomized studies.
• Multivariable approaches, including propensity scores, are increasingly used to address confounding in observational research.
• Ongoing debate exists regarding the comparative utility of propensity scores versus traditional multivariable methods.

Purpose of the Study:

• To provide a practical guide for selecting between propensity score methods and standard multivariable analysis.
• To elucidate the strengths and weaknesses of both analytical approaches in the context of clinical research.
• To inform clinicians and researchers on best practices for analyzing non-randomized data.

Main Methods:

• Qualitative review of existing literature on propensity scores and multivariable analysis.
• Comparative analysis of statistical methodologies, considering advancements like bootstrap and Bayesian methods.
• Focus on practical application and decision-making for clinicians.

Main Results:

• Propensity scores offer a valuable tool for adjusting confounders, particularly in smaller datasets.
• Standard multivariable approaches like logistic regression and Cox analysis remain robust alternatives.
• The choice between methods depends on specific study characteristics and research questions.

Conclusions:

• Both propensity scores and standard multivariable analyses have distinct advantages and limitations.
• Understanding these differences is crucial for appropriate application in clinical research.
• This review aids in selecting the most suitable method for generating reliable clinical evidence from non-randomized studies.