Skip to main content icon/video/no-internet
icon backReturn to Entries

Overmatching

To control for potential confounders or to enhance stratified analysis in observational studies, researchers may choose to match cases and controls or exposed and unexposed subjects on characteristics of interest. If matching is superfluous or erroneous, overmatching may occur. The three main effects of overmatching are a loss of statistical efficiency, introduction of bias, and loss of financial efficiency.

Background

To reduce confounding or to enhance stratified analysis, unexposed subjects in cohort studies or controls in case-control studies may be chosen to be identical or similar to exposed subjects or cases with respect to the distribution of one or more variables. Overmatching, sometimes referred to as overmatching bias, occurs when matching is done incorrectly or unnecessarily leading to reduced efficiency and biased results. Overmatching generally affects case-control studies.

Effects of Overmatching

Loss of Statistical Efficiency

In case-control studies, if cases and controls are matched on a variable that is associated to the exposure but not the disease, chosen controls are more similar to cases than the base population in respect to the exposure. The forced similarity between cases and controls in respect to the exposure obscures the relationship between the exposure and the disease. Matching on an exposure-associated variable will cause the crude odds ratio to be closer to 1—that is, to the null value. However, when stratified by the matching variable, stratum-specific odds ratios will be unbiased. If confounding is present, bias due to matching on an exposure-associated variable will cause the odds ratio to go toward the null regardless of the direction of the confounding. The degree of information loss due to overmatching depends on the absolute correlation between the matching variable and the exposure of interest. Matching on a nonconfounder necessitates stratified analysis that would otherwise not be necessary, and it reduces study efficiency.

Introduction of Bias

If controls are matched to cases on a variable that is affected by both the exposure and the disease or is an intermediate between exposure and disease, both the crude and adjusted odds ratios will be biased. Like matching on exposure-only-associated variables, matching on an intermediate or variable affected by exposure and disease will force the odds ratios toward the null. However, unlike matching on an exposureonly-associated variable, it is not possible to get unbiased stratified measures of effect.

Loss of Financial Efficiency

Matching can lead to greater statistical efficiency by ensuring that cases will have one or more matched controls for comparison in stratified analysis. Also, matching may offer a cost benefit if the collection of exposure data from many people is very expensive. However, if the matching process is complicated and involves many matching variables, it may be difficult and costly to identify and recruit potential controls. Also, if matching is done unnecessarily, additional costs associated with recruiting further controls may incur. Potential statistical benefits and costs should be assessed prior to matching.

  • overmatching
  • odds ratio
  • efficiency
  • case-control studies
  • disease
  • bias
MichelleKirian
http://dx.doi.org/10.4135/9781412953948.n333
See also

Further Readings

Agudo, C., and Gonzales, C. A.Secondary matching: A method for selecting controls in case control studies on environmental risk factors. International Journal of Epidemiology28 (1999). 1130–1133.http://dx.doi.org/10.1093/ije/28.6.1130
Rothman, K. J., & Greenland,

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Read next

More like this

Sage Recommends

We found other relevant content for you on other Sage platforms.

close

Have you created a personal profile? Login or create a profile so that you can save clips, playlists and searches

close
close
close Signed In to profile You are signed in as:
close
Logged in Institution Institution