Why Propensity Scores Should Not Be Used for Matching

Abstract

We show that propensity score matching (PSM), an enormously popular method of preprocessing data for causal inference, often accomplishes the opposite of its intended goal—thus increasing imbalance, inefficiency, model dependence, and bias. The weakness of PSM comes from its attempts to approximate a completely randomized experiment, rather than, as with other matching methods, a more efficient fully blocked randomized experiment. PSM is thus uniquely blind to the often large portion of imbalance that can be eliminated by approximating full blocking with other matching methods. Moreover, in data balanced enough to approximate complete randomization, either to begin with or after pruning some observations, PSM approximates random matching which, we show, increases imbalance even relative to the original data. Although these results suggest researchers replace PSM with one of the other available matching methods, propensity scores have other productive uses.

Interprofessional Teamwork Innovation Model (ITIM) to promote communication and patient-centred, coordinated care

“…Matching was done using Mahalanobis distance35 between patient discharges. The Mahalanobis distance is the ‘distance’ between observations/cases, based on the standardised measurements of the variables used in the distance calculation.…”

Comparison of the ability of double‐robust estimators to correct bias in propensity score matching analysis. A Monte Carlo simulation study

“…Contrary to the property of covariate balance which is measurable, there is no means to verify that the double‐adjustment models are correctly specified. Because the estimate depends on the model's specification, researchers might be tempted to report the results, which fit their “favorite hypothesis.” As model dependence can be removed by covariates balancing, King and Nielsen recommend the use of other matching methods, such as coarsened exact matching . Indeed, optimal balance avoids the need for double‐adjustment and, thus, eliminates bias related to model dependence.…”

“…Recently, King and Nielsen showed that, in contrast with coarsened exact matching, matching on the PS was suboptimal for balancing every covariate perfectly across the groups . Propensity score matching approximates a completely randomized controlled trial, which, by balancing covariates in average, is likely to be concerned by residual imbalance if the sample size is small .…”

Proton pump inhibitors and gastric cancer: association is not causation

“…More dramatic risk reductions have been reported by other authors after adjusting for confounders in similar studies 3. Cheung et al conducted a propensity-matched analysis as this can partially adjust for unknown and known confounders, but this approach is only as good as the underlying dataset and it can have many inherent biases 4. In this case, the approach is likely to be flawed as the HR actually increases with the trimmed results and is similar to the univariate analysis.…”