The reverse propensity score to detect selection bias and correct for baseline imbalances

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Observational studies use propensity scores to manage baseline imbalances.
• Randomized trials often face systematic baseline imbalances, necessitating ad hoc methods.
• Existing methods for handling imbalances in randomized trials can challenge study conclusions.

Purpose of the Study:

• To develop a systematic approach for evaluating medical interventions amidst baseline imbalances in individually randomized trials.
• To introduce and define the reverse propensity score.
• To demonstrate the utility of the reverse propensity score for detecting and correcting selection bias.

Main Methods:

• Defined the reverse propensity score as the conditional probability of receiving a specific treatment given prior allocations and the randomization procedure.
• Applied the reverse propensity score to individually randomized trials with allocation concealment.
• Demonstrated detection and correction of systematic baseline imbalances using the reverse propensity score.

Main Results:

• The reverse propensity score provides a systematic method for addressing baseline imbalances in randomized trials.
• This approach allows for the identification of selection bias.
• The reverse propensity score enables the correction of systematic baseline imbalances.

Conclusions:

• The reverse propensity score is a valuable tool for enhancing the validity of randomized trial findings.
• It offers a standardized approach to managing baseline imbalances, unlike previous ad hoc methods.
• This method improves the reliability of evidence for medical interventions, even with randomization imperfections.