A growing body of evidence points to socioeconomic factors – educational attainment, housing security, access to healthful food, transportation, personal safety, etc. – as even more consequential to individual and population health than health care services.
Electronic health record systems have enabled health care organizations to begin collecting data on these social determinants, or drivers, of health. The World Health Organization’s International Classification of Diseases now includes codes for SDOH data, so data elements can be standardized. In 2024, the Centers for Medicare and Medicaid Services began requiring hospitals to screen patients for unmet SDOH needs.
But simply capturing data is not enough to make an impact. The conundrum is how to make use of the data collected in unstructured text to effect change and improve patient health.
Endeavor Health is among the health systems using artificial intelligence to turn SDOH data into actionable health information. Endeavor is the product of several mergers over the past half-decade and comprises nine community hospitals and more than 300 care locations across the greater Chicago area. Endeavor anesthesiologist and pain medicine specialist Nirav Shah recently elaborated on these efforts in a webinar led by Lauren Riplinger, chief public policy and impact officer at the American Health Information Management Association. The webinar is available for free on demand.
Shah, Endeavor’s medical director of quality innovation and clinical practice analytics and the program director of outcomes research for quality and transformation, says having insight into SDOH is critical for health systems, especially accountable care organizations and health systems that serve diverse patient populations.
Using a natural language processing program with SDOH
Endeavor has developed a natural language processing program to extract data on SDOH from unstructured text in clinical notes, empowering clinical experts to take action and connect patients with the ancillary services they need.
How did Endeavor create a trusted workflow and integrate the technology with people and processes? The emergency department seemed like the setting in which such a tool would have the greatest impact. Moreover, a screening workflow was already in place in the ED, and front-line social work staff were already deeply engaged. So Endeavor chose the ED to pilot test the NLP screening system.
The team started with a documentation strategy around five core social drivers: housing, transportation, food, alcohol use and financial resources. They formed a partnership with IQVIA division LInguamatics, which had a rules-based platform infused with machine learning on the back end that enabled Endeavor to integrate existing queries into the system. The tool’s rules-based algorithm made it easier to refine queries.
IT staff built connections between the NLP engine, the enterprise data warehouse and the enterprise-wide EHR. They identified SDOH, annotated notes, adjusted Linguamatics’ queries and optimized them for Endeavor’s patient population. Endeavor retrospectively validated queries to build trust, comparing the system’s performance to national census data and assigning accuracy metrics.
The team worked with key operational owners to design the clinical workflow. Finally, they prospectively validated the system and incorporated care team feedback in an iterative process.
ED chosen for pilot project
Screening for SDOH in the ED is challenging, according to a recent study in JAMA Network Open that found wide variation in adverse SDOH screening and documentation practices, skepticism that adverse SDOH screening and referral in the ED is valuable, and challenges regarding resources, staffing and time to screen and refer ED patients.
Nonetheless, Endeavor’s pilot project succeeded. ED social workers began to feel that they were more efficient and making a greater impact. Clinicians also gained insights into patients’ SDOH needs. Now, instead of spending 80% of their time combing through nurses’ notes to root out SDOH needs, ED social workers spend 80% of their time helping patients.
Keys to success
Shah said key components to success include the sociotechnical model of people, process and technology. Data governance is critical, which Endeavor achieved through two-stage validation. Trust between clinical and operational staff is a must for diffusion and last-mile deployment success. AI will scale at the speed of trust, and so will change management, Shah says. Leaders must think critically and ask themselves, “Is this the best way? Is this the best technology? Is this a black box? How do we build trust?”
A human was in the loop every step of the way. Data scientists and social workers met regularly to update queries. As a result, the social workers trusted the system. Additionally, the solution is not a black box, so if bias is suspected, data scientists can trace it.
Executive sponsorship is crucial, Shah says. Team members got the time and space they needed, and executive leaders were kept in the loop.
Diffusion is as important as innovation. Technology must align with the organization’s strategic interests. You must understand your tech stack and ensure the architecture is in line with what people are already using.
Starting a pilot in silo is a recipe for disaster, Shah says. Start conversations early, and build for scale from the beginning.
Patients must also trust the technology. People share their personal information in apps, online shopping sites, social media sites and so on, because the apps and platforms provide a service. Shah points out that AI-based screening for SDOH is no different. Patients need to know that the technology is being used to improve their care and health.
When SDOH data is appropriately collected, used and shared, health care teams gain insight into their patients, enabling them to take steps to improve overall health and well-being. The AHIMA Data for Better Health initiative offers tools and resources for health data professionals to better understand the importance of SDOH data and how it can be used to improve health and health care outcomes.
AHIMA is hosting its next free SDOH online webinar, Sharing SDOH Data to Improve Outcomes, on April 18 at noon CDT.
_____________________
Stay on top of health information management news by subscribing to AHIMA SmartBrief.