In a study published in the AMA Journal of Ethics, researchers explored the role of social and behavioral data in precision medicine research.1

Electronic health records (EHRs) can offer information on social and behavioral data, which can aid research investigating genetic and social factors across health disparities; for example, factors such as substance use and eating habits inform some of the risk associated with preventable premature deaths in the United States. Brittany Hollister, PhD, and Vence L. Bonham, JD, from the National Human Genome Research Institute at the National Institutes of Health, discussed potential biases in collecting, using, and interpreting EHR-based data in precision medicine research.

Continue Reading

Current collection of behavioral and social data by precision medicine researchers is increasingly done using EHR data, as opposed to self-report methods such as surveys. However, extraction and use of EHR data poses challenges of inconsistencies or inaccuracies. Another challenge is determining what data are included or excluded from EHRs, and the consequences of using data collected through biased methodologies. The National Academy of Medicine addressed some of this in recommendations for the systematic capture of behavioral and social measures.2 They recommended intentional collection of structured social environment data, as well as the development of a plan by the National Institutes of Health to include social and behavioral data in EHRs. The current inconsistencies in collecting social and behavioral data pose difficulties to use in precision medicine research, but with improved collection methods these difficulties could be amended.

Related Articles

Another obstacle of using EHRs is patient participation. The willingness, or lack thereof, of potential study participants to allow ongoing access to their EHRs has caused some concern for precision medicine research programs. To remedy this, it is crucial that researchers engage individuals as partners, not merely as human subjects. Potential participants should understand how their data might be used and any risks associated with privacy or data use.

Another factor researchers should consider when using EHR-derived social data is the social and historical biases in the collection of the data. For example, data used to build models of predictive policing are based on existing police activity, which show a trend of overpolicing in minority neighborhoods. This can skew results from models because the data used for the models are skewed. In clinical settings, EHR data can contain biases for several reasons, and these should receive due consideration from researchers.

Dr Hollister and Mr Bonham concluded that the inclusion of structured social and behavioral data in EHRs will allow for a more holistic perspective of health. However, researchers must carefully use the data so as not to perpetuate existing injustices in healthcare based on biases in the collection and analysis of EHR data.


  1. Hollister B, Bonham VL. Should electronic health record-derived social and behavioral data be used in precision medicine research? AMA J Ethics. 2018;20(9):E873-E880.
  2. Institute of Medicine. Capturing social and behavioral domains and measures in electronic health records: phase 2. Washington, DC: National Academies Press; 2015.