How many covariates should be used for propensity score matching?

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

Why should I use as many covariates as available?

The number of covariates used for propensity score matching depends on the specific study and the availability of relevant data. In general, it is important to include as many covariates as possible that are believed to be associated with both the treatment and the outcome of interest. This helps to ensure that the groups being compared are as similar as possible and reduces the risk of bias.

In randomized studies, it is hoped that randomization will balance differences between groups, even for variables that researchers are unaware of. In a non-randomized study, many parameters can differ between subjects in the treatment and the control group.

Propensity score matching can reduce variability between groups, but this is true only for variables that the researcher is aware of and has included as covariates (matching variables).

This is why it is important to include as many covariates as possible.

In the medical literature, the number of covariates was 7 on average (range 3 to 16) for one systematic review in the field of neurology [1], and a median of 12 (range 3 to 41) for another systematic review in the field of surgery [2].

Which covariates should I use?

Please refer to the following article for more information:

Which covariates should be used to match patients in Propensity Score Matching?


[1] Karim, Mohammad Ehsanul, Fabio Pellegrini, Robert W Platt, Gabrielle Simoneau, Julie Rouette, et Carl de Moor. « The Use and Quality of Reporting of Propensity Score Methods in Multiple Sclerosis Literature: A Review ». Multiple Sclerosis Journal 28, no 9 (août 2022): 1317‑23.

[2] Lonjon, Guillaume, Raphael Porcher, Patrick Ergina, Mathilde Fouet, et Isabelle Boutron. « Potential Pitfalls of Reporting and Bias in Observational Studies With Propensity Score Analysis Assessing a Surgical Procedure: A Methodological Systematic Review ». Annals of Surgery 265, no 5 (mai 2017): 901‑9.

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