Developing Electronic Health Record (EHR) Strategies Related to Health Center Patients' Social Determinants of Health

Rachel Gold, PhD, MPH; Erika Cottrell, PhD, MPP; Arwen Bunce, MA; Mary Middendorf, BS; Celine Hollombe, MPH; Stuart Cowburn, MPH; Peter Mahr, MD; Gerardo Melgar, MD


J Am Board Fam Med. 2017;30(4):428-447. 

In This Article


This work was conducted at OCHIN, a nonprofit community-based organization that centrally hosts and manages an Epic© EHR for >440 primary care CHCs in 19 states; it is the nation's largest CHC network on a single EHR system. Socioeconomic risks of patients in OCHIN member CHCs are clear from SDH data that are already collected: 23% are uninsured and 58% are publicly insured, 25% are nonwhite, 33% are of Hispanic ethnicity, 28% are primarily non–English speakers, and 91% are from households living <200% below the federal poverty level (among patients with available data).

The processes described here constituted the first phase of a pilot study designed to develop EHR-based tools that CHCs could use to systematically identify and act on their patients' SDH-related needs. We call these the "SDH data tools."

With the goal of creating SDH-related workflows that parallel clinical referral processes, we began with the assumption that addressing patients' SDH needs require 5 key steps: (1) collecting SDH data; (2) reviewing patients' SDH-related needs; (3) identifying referral options to address those needs; (4) ordering referrals to appropriate services; and (5) tracking outcomes of past referrals. This assumption was based on team members' knowledge of the CHC workflows used to refer patients to specialty medical care.

We also considered the following factors:

  • CHCs are federally required to collect certain SDH measures from the IOM list, including race/ethnicity, tobacco/alcohol use, and depression. Our SDH data tools had to incorporate these data, without requiring duplicate data entry.

  • CHCs have varying staffing structures, resources, and workflows. To accommodate this, SDH data tools should be accessible to various team members (eg, front desk, medical assistants, community health workers, behavioral health staff).

  • SDH tools should use existing EHR-based functionalities to facilitate their adoption. Table 2 describes the options we initially considered to address each of these 5 steps.

  • Many CHCs already identify or address SDH needs using ad hoc methods. Some may already have mechanisms for tracking local resources, such as a 3-ring binder or files on a shared drive; some use online resources (eg, United Way 2–1–1, local department of human services). We sought to incorporate existing resources into our SDH referral tools.

We recruited 3 OCHIN CHCs in Oregon and Washington as pilot sites and project partners. We also engaged OCHIN's Clinical Operations Review Committee (CORC)—a group of CHC clinicians who collectively review proposed changes to their shared EHR—in all process steps. We conferred with leaders from PRAPARE, Kaiser Permanente (KP), Epic, and other national SDH experts (see the Acknowledgments). These stakeholders were asked to discuss 3 overarching questions.

1. Which SDH Domains Should Be Included?

The CORC reviewed the IOM-recommended SDH domains and the wording for each domain, additional questions or alternate wording from PRAPARE and KP's SDH screening tools, and other domains currently collected in OCHIN's EHR that were not in the IOM/PRAPARE recommendations. Based on these options, they chose which patient-reported SDH measures to include and the specific wording for each included domain. Geocoded domains were not considered, as the CORC felt they were not readily actionable. The pilot CHCs were present at most of the SDH-related CORC meetings.

2. How Do Care Teams Want to Collect, Review, and Act on Data on Patients' SDH Needs Within the EHR?

We asked CORC members whether and how their clinics monitor patients' SDHs and what the SDH-related EHR tools should include. We presented options for how the SDH data could be collected and summarized using existing EHR structures, and we considered how existing tools aligned with the 5 key steps described above. We then mocked up a set of SDH data EHR tools and proposed workflows for using them. We presented the mock-ups and draft training materials to the CORC over multiple meetings, and to each of the pilot CHCs at staff meetings. We asked diverse CHC staff for critical feedback on the draft tools, suggestions for and potential barriers to collecting and acting on SDH data using the tools, and how best to train CHC staff in their use. Our team's Epic programmer attended these meetings to provide real-time input about the technical feasibility of any suggestions. The SDH data tools were revised based on the feedback received, and the pilot CHCs' various workflows and staff structures were considered. The revised tools were presented to the CORC (in person) and the study sites (via webinar) to verify that the revisions addressed requested changes.

This review and refinement process aligns with best practices for technology development,[47] for example, user participation and prototyping.[48–55] Evidence shows that for technology to be used effectively and as intended, end users must find it easy to use and must perceive that the technology will improve efficiency.[56–58] Therefore, we sought input from end users in order to increase the probability that the tools would be used.[47] The EHR tools were then built in OCHIN's testing environment, an off-line, internal "copy" of the EHR, and tested by an OCHIN quality assurance analyst.

3. How Can Care Teams Ensure That Patients Receive Up-to-Date Referrals?

The CHCs hoped to avoid referring patients to local resources that were not currently accepting new clients (service agencies sometimes close enrollment because of demand) or that had limitations about who could be assisted (eg, some services are not open to persons with past felonies). We discussed the options and approaches for identifying resources described above. We also conferred with colleagues at KP who were considering similar choices, and we spoke with representatives from organizations that create databases of community resource information (eg, United Way 2–1-1, Health Leads, and Purple Binder) to understand those options. The 3 pilot clinics then identified 3 to 5 prioritized SDH domains for which they wanted a list of community resources; based on these preferences, we provided lists of local resources for housing, food, transportation, social isolation, and intimate partner violence.


Participants from our study clinics consisted of primary care providers (n = 3), medical assistants (n = 5), clinic managers (n = 3), community health workers (n = 4), behavioral health staff (n = 2), nurses (n = 5), referral specialists (n = 3), EHR specialists (n = 3), and medical directors (n = 2).


The development process took 10 months. Five 1-hour meetings with the CORC were held over the course of 6 months in order to reach consensus on which SDH domains to include and how the tool would function. The pilot sites were then given 6 weeks to test the tools for functional errors.