Selection and Nonresponse Bias in Maternal Health Research Using the Electronic Health Record


Dr. Julie Bower, College of Public Health, Dept of Epidemiology
Rank at time of award: Assistant Professor
Dr. Randi Foraker, College of Public Health, Epidemiology
Rank at time of award: Assistant Professor


Electronic Health Records (EHRs) are becoming a new source of participant recruitment. In particular, EHR patient portals – also known as personal health records – can be a great way of reaching a diverse population while at the same time allowing the researcher to target certain groups of high study interest. However, there are many concerns with the bias/sample selection associated with recruiting via patient portals (e.g., characteristics of users versus non-users) and until we are clear as to the bias introduced in this process, interpretation of data garnered from the EHR will be limited for population studies. This proposal will allow our team to begin to study this issue as a research problem. We propose a pilot study to quantify the discordancy with regard to demographic and clinical characteristics of a study sample of pregnant women in active prenatal care compared to the general population.
Accessing diverse populations of pregnant women for social and behavioral sciences research remains a challenge. From a public health perspective, understanding the strengths and limitations of using the EHR and personal health record for population research will inform future study design and interpretation of research findings utilizing these rich data sources. Patient portals can be leveraged for collection of patient reported outcomes data that are not otherwise available in the EHR, improving the applicability of using the personal health record for public health research. This project will position our team well to compete for federal funding for social, behavioral, and health services research focused on maximizing the EHR as a resource for conducting population research.



This project is relevant to public health because there is growing interest in identifying new sources for participant recruitment for population-based research. The electronic health record (EHR) may be particularly useful for maternal and child health research recruitment because pregnant women and mothers of infants in the U.S. are highly unlikely to be health system non-users. Although clinical populations are believed to be biased samples of the general population, the EHR may actually serve as an ideal alternative source for participant recruitment of representative samples of certain populations. Further, the EHR contains not only detailed clinical information on this population, but also extensive demographic information that can be used to estimate and account for the bias introduced by research study non-participants using this recruitment approach.