Analysis of multisite intervention studies using generalized linear mixed models

Infect Control Hosp Epidemiol. 2019 Aug;40(8):910-917. doi: 10.1017/ice.2019.114. Epub 2019 Jun 21.


Multisite intervention studies have become increasingly common in infection control, for example, looking for a change in hospital infection rates after a regional policy change. The design of these studies can take various forms, from pre-post observational studies to randomized trials, in which sites are randomly assigned to the intervention or in which the intervention is sequentially introduced to different sites over time. Data collected under these settings are clustered by hospital and/or ward, consist of repeated measurements and, in some cases, exhibit temporal and/or seasonal patterns. Failure to account for these features in data analysis could well result in biased estimates of intervention effectiveness and impact on the generalizability of model results.

Publication types

  • Research Support, Non-U.S. Gov't
  • Review

MeSH terms

  • Bias
  • Cross Infection / prevention & control
  • Data Analysis
  • Linear Models*
  • Multicenter Studies as Topic*