Matching EN - pdmoerland/wiki-biostatistiek-bilingual GitHub Wiki

Matching

Deze pagina lezen in het Nederlands.

This page was translated from the Dutch-language original using AI and has not been fully reviewed. Where the two differ, the Dutch version takes precedence.

Contents

Matching patients is the deliberate pairing of patients with similar baseline characteristics in order to ensure an unbiased comparison of the outcome measure.

When is matching useful?

In the case of a case-control design in which the two groups are formed based on the outcome measure, matching is not advisable because it can introduce bias. In the case of a cohort study, matching is an option. However, it is often not necessary and can require more effort than it yields. Often, the effect of potential confounders can equally well be accounted for at the analysis stage, by including multiple predictors in a multivariable analysis model.

Cases in which matching offers a clear advantage are, for example, the following:

  • When data has been collected over a period in which treatment has clearly improved due to new insights, it is wise to select the 'control group' from the same time period.
  • When there are so few cases that including multiple variables at the analysis stage is (statistically) not possible.
  • When the two groups clearly have very imbalanced distributions of predictors that are expected to influence the outcome measure (for example, if one group is 30-50 years old and the other group is 45-60 years old).

How do I practically carry out matching?

If matching is chosen, the following structured approach can be used: For each person from the first group, select all persons from the second group who match on the important predictors. From this selection, randomly choose one (or more - more than 5 controls per case often does not provide additional benefit) person.

How can I match automatically?

From version 22 onwards, SPSS has a procedure for (propensity score) matching. In the software package R, a package called Matching is available for various types of matching.

Do I need to match the case and control group in my survey study in order to compare them?

I have collected data from a group of patients from the AMC and want to compare the outcomes of the questionnaires with those of a random sample of the population, for whom I also have that data (a kind of unmatched case-control study). The demographic characteristics of my patient population differ significantly from the control group. Am I allowed to directly compare the outcomes of the questionnaires with each other, or do I first need to match the two groups?

Since this is not a randomized study, other variables that could confound the relationship being studied indeed need to be taken into account. Matching can be an option, but it is also a disputed method. Since you apparently have already collected the data for all controls, I would advise using all controls and correcting for the imbalance in variables that you expect could confound the relationship you are studying at the analysis stage, by using a multivariable model.

As a book you could consult the following: Clinical Epidemiology - Principles, Methods and Applications for Clinical Research, Diederick E Grobbee / Arno W Hoes (Chapter 9 Case control Studies)

Additional sources on observational studies

References

Finished reading? You can go to the Overview of all topics, or to Choosing a test for help picking a suitable test or analysis. Want to learn more about biostatistics? Then follow the AMC e-learning Practical Biostatistics. Did you find something on these pages that isn't correct? Does a link not work? Or would you like to contribute to the wiki? Then get in touch with us.

This biostatistics wiki is an initiative of the former statistics helpdesk of Amsterdam UMC, location AMC. Amsterdam UMC staff can request statistical support via intranet.