Impact of differential response rates on the quality of data collected in the CTS physician survey.

Survey administrators face trade-offs between expending additional survey resources to maximize response rates versus using fewer resources and accepting lower response rates. Using data from the Community Tracking Study's Physician Survey, we examined how survey estimates and data quality changed as additional respondents completed the survey. Results showed that improvements in response rates over the range examined (i.e., up to 65%) did not change estimates appreciably nor affect data quality. As long as these results are not overstated to imply that extremely low response rates are credible, this study may permit researchers to disseminate interesting results in peer-reviewed journals even when the response rate falls slightly short of current standards. It must also be emphasized, however, that we were unable to measure the nonresponse effect of those who were never interviewed. Achieving a response rate significantly above 65% might have changed the survey results appreciably.