Computing risk ratios from data with complex survey design

Edward C. Norton, Nathan W. Carroll, Morgen M. Miller, Kasey Coyne, Jason J. Wang, Lawrence C. Kleinman

Research output: Contribution to journalArticlepeer-review

3 Scopus citations


We demonstrate how to apply regression risk analysis to compute risk ratios and risk differences from logistic regression models using complex survey data. We validate the use of regression risk analysis for complex survey design. First, we derive formulas for adjusted risk ratios (ARRs) and adjusted risk differences (ARDs) adjusted for weighting, stratification, and clustering. Then we use Monte Carlo data to show why correcting statistics for complex survey design is important. We show how to calculate and interpret ARRs using a publicly available data set with a binary outcome. Regression risk analysis can be applied to complex survey data to calculate correct ARRs and ARDs from logistic regression.

Original languageEnglish (US)
Pages (from-to)3-14
Number of pages12
JournalHealth Services and Outcomes Research Methodology
Issue number1-2
StatePublished - Jun 2014
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Health Policy
  • Public Health, Environmental and Occupational Health


  • Complex survey design
  • Econometrics
  • Logistic regression
  • Regression risk analysis
  • Risk ratios


Dive into the research topics of 'Computing risk ratios from data with complex survey design'. Together they form a unique fingerprint.

Cite this