EFFECT OF NON-RESPONSE ON RISK RATIOS IN A CARDIOVASCULAR-DISEASE STUDY

In a population-based cardiovascular disease study with a response rate of 82.1%, we obtained information on cardiovascular risk factors and diseases in both respondents and non-respondents. In general participants were a somewhat biased subset of the target population in that they tended to overrepresent the 'worried well', i.e. those with risk factors but without disease. This response bias represents independent bias in both risk factor and disease distribution. This 'combined bias' produced minor to moderate errors in risk ratio calculations in our study. However, greater error in risk ratios could occur in a study with a similarly biased respondent sample and a lower response rate. © 1979.