The effect of response bias on the odds ratio.

The effect of response bias on odds ratio results was determined based on data from a population-based cardiovascular disease survey. The study subjects consisted of 5000 adult residents of a predominantly white, upper-middle class community. Information from 60% of the 1100 non-respondents was obtained by telephone. Consistent patterns of participation associated with risk factors and diseases under study were found. A simple error term was developed to convert the odds ratio for respondents to the odds ratio for the target population using individual cell response rates. This error term demonstrates that the response patterns found tended to minimize the error in odds ratio calculations for respondents. Only by obtaining relevant information on non-respondents can investigators accurately estimate response bias and its effects on the odds ratio.