Replacement in Propensity Score Matching

Modified on Sat, 21 Jan 2023 at 09:01 PM

The default parameters for PSM: no replacement

The most common approach to propensity score matching is to use a 1-to-1 matching. Meaning that every subject from the treatment group is matched with one subject from the control group.

This can be very effective if the groups have the relatively same size and if there is enough "overlap", ie if it possible to match patients with similar characteristics in both groups.

In some cases, however, this method does not provide good enough matching.

Why 1-to-1 matching may be problematic?

In some cases, 1-to-1 matching can lead to poor results with propensity score matching. Here are some of these situations:

  1. A small sample size: it is hard to find controls when there are few patients available for matching
  2. A large number of covariates: finding similar subjects can be difficult if there are many parameters for which the two groups must be similar, especially in the case of small sample sizes
  3. High variability among subjects from the treatment group

In these cases, your methodology score may be impacted with a low mark for the "% of included patients" panel.

How can replacement help in these cases?

Replacement is a technique allowing to match the same control patient to several patients in the treatment group. If a control subject has characteristics close to several subjects in the treatment group, it will be matched to several of these treated patients.

Replacement is often used in combination with oversampling which is another technique that allows matching several patients to the same treated patient.

Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select atleast one of the reasons
CAPTCHA verification is required.

Feedback sent

We appreciate your effort and will try to fix the article