JMIR Publications

ABSTRACT

Background:

Many U.S. healthcare providers have adopted electronic health records (EHR). But there is no universal standardized format for documenting social and behavioral determinants of health in EHR’s structured data, and extracting data from unstructured EHR data requires time-consuming and subjective methods such as chart review for identification of patients with high social risk, which is not a feasible approach to screening a large population of patients. The U.S. healthcare system has started addressing these issues by assessing and optimizing the documentation and collection of social and behavioral determinants in EHRs on smaller datasets and for subpopulations of patients. Still, there is a need for a comprehensive assessment of documentation, data collection, and presentation in EHRs and development of an investigator-friendly report of domains addressing social and behavioral determinants of health in EHRs of major healthcare systems.

Objective:

To examine the availability and characteristics of social and behavioral determinants of health in the EHR of a multi-level academic healthcare system with linked ambulatory provider networks in Baltimore, Maryland, providing service to inner-city Baltimore and across the state.

Methods:

We analyzed the EHR’s structured data, represented information collected between January 2003 and June 2018 from ~5.4 million unique patients, to determine the existence of key information on social and behavioral determinants of health. We defined social and behavioral determinants as characteristics of patients and communities where they live. Therefore, we assessed the availability of patients’ address and location among other determinants. Patients had various social and behavioral risk factors and different socioeconomic status. We also explored the use of natural language processing (NLP) to determine the availability of a selected number of variables addressing social and behavioral determinants of health in the EHR’s unstructured data for 1,188,202 unique patients captured between July 2016 and May 2018.

Results:

We identified an address or zip code data for 95% of ~5.4 million patients, ethnicity for 50%, at least one race for 90%, a preferred language for 49%, information regarding alcohol use for 9.08%, and smoking status for 32% of patients. Using ICD-10 diagnoses codes, we identified 35,171 patients (0.64%) with information related to social connection/isolation, 10,433 patients (0.19%) with housing issues, and 3543 patients (0.06%) with income/financial resource strain. Of 1,188,202 unique patients, 2.6% had at least one clinical note containing phrases referring to social connection/isolation, 3% included housing issues, and 1% had mentions of income/financial resource strain.

Conclusions:

Variables addressing social and behavioral determinants of health, except demographic data, were not regularly collected for the patient population. Healthcare providers should assess the availability and characteristics of data addressing these determinants of health in their structured and unstructured EHRs and modify their workflows to improve documentation, data collection, and extraction.