Investigating response bias in an information technology survey of physicians

Monitoring the diffusion of electronic health records (EHR) into ambulatory clinical practice has important policy implications. However, estimates of EHR use are typically derived from survey data and may be subject to significant response bias. The current study is a retrospective analysis testing for response bias in a large information technology survey of physicians (n = 14,921). To detect bias, respondents were compared to nonrespondents on known characteristics. Moreover, early respondents were compared to late respondents with respect to key variables in the survey that are likely to influence participation. The 4203 respondents (28.2% participation rate) did not differ demographically from nonrespondents. Response rates, by specialty, differed slightly. When comparing early and late survey respondents, no differences were detected in EHR use, length of time since EHR installation, practice size, physician age, years since medical school graduation, and years of practice in their current community. Overall, response bias was not detected using established methodologies in this mailed survey of physician EHR use. Similar surveys of physicians, even with a lower than expected response rate, may still be valid.