027 - Optimal full matching under a new constraint on the sharing of controls Application in pediatric critical care

Health policy researchers are often interested in the causal effect of a medical treatment in situations where randomization is not possible. Full matching on the propensity score (Gu & Rosenbaum, 1993) aims to emulate random assignment by placing observations with similar estimated propensity scores into sets with either 1 treated unit and one or more control units or 1 control unit and multiple treated units. Sets of the second type, with treatment units forced to share a comparison unit, can be unhelpful from the perspective of statistical efficiency. They are often necessary to achieve an experiment-like arrangement (as measured with observed covariates), but optimal full matching on estimated propensity scores is known to exaggerate the number of many-one matches that are truly necessary, generating lopsided matched sets and smaller effect sample sizes (Hansen, 2004).

In this presentation, we introduce an enhancement of the Hansen and Klopfer (2006) optimal full matching algorithm that counteracts this exaggeration by enabling analysis to permit treatment units to share a control while limiting the number that are permitted to do so. The result is a more well-balanced matching structure that prioritizes 1:1 pairs as opposed to matches with lopsided, many-to-one configurations of matched sets.

This enhanced optimal full matching is then illustrated in a pilot study on the effects of Extracorporeal Membrane Oxygenation (ECMO) for treatment of pediatric acute respiratory distress syndrome. Within this pilot study, in which data sample size was strictly limiting, existing methods for limiting the sharing of controls have already resulted in an increased effective sample size. The present enhancement of Hansen and Klopfer's optimal full matching algorithm provides an additional boost. Results indicate an increase in effective sample size at the small expense of covariate balance. Our enhancement of Hansen and Klopfer matching algorithm provides researchers with a new tool on how to manage this bias-variance tradeoff.

Causal Inference

Propensity Score Matching

ECMO

PARDS 

Beginner

Knowledge

International Conference on Health Policy Statistics 2023