Characterizing emergency clinician engagement with social drivers of health data among patients with opioid use disorder

Although emergency clinicians reviewed social drivers of health data more frequently during encounters with patients having opioid use disorder, this engagement was rare overall, did not predict the prescription of medications for opioid use disorder, and failed to address persistent racial and ethnic treatment disparities.

The Gist

Imagine the Emergency Department (ED) as a bustling, high-speed train station. Thousands of people arrive every day, many of them in crisis, needing immediate help to get back on track. Among these travelers are many people struggling with Opioid Use Disorder (OUD). For these travelers, the train station isn't just about treating a physical injury; it's also about understanding their life outside the station—do they have a place to sleep? Can they afford food? Do they have a ride home? These are called Social Drivers of Health (SDOH).

To help, the train station installed a high-tech digital dashboard (the Electronic Health Record, or EHR) that can display this "life outside" information. The researchers wanted to see: Do the station workers (doctors) actually look at this dashboard? And if they do, does it help them give the right medicine (like buprenorphine) to people with OUD?

Here is what they found, broken down simply:

1. The Dashboard Exists, But Few People Look at It

The hospital had a special "Social Wheel" and a list of codes (like a checklist) where staff could write down or see if a patient was struggling with housing, money, or food.

• The Reality: Even though this information was there, doctors only looked at it in about 1 out of every 5 visits.
• The Analogy: It's like having a GPS in your car that tells you about traffic jams and road closures, but 80% of the time, the driver just stares straight ahead and ignores it.
• The Good News: When doctors did look, they mostly checked the "Problem List" (a simple list of codes) because it was right on the main screen. They rarely looked at the detailed "Social Wheel" or free-text notes.

2. They Looked a Little More for OUD Patients

When a patient arrived with Opioid Use Disorder, the doctors were slightly more likely to check the dashboard (about 27% of the time) compared to other patients (16%).

• The Analogy: If a passenger is carrying a heavy, dangerous package (OUD), the station workers are a bit more curious about their background. But even then, they still missed the dashboard in 3 out of 4 cases.

3. Looking Didn't Always Lead to Action

The big question was: If the doctor sees the social struggles, do they give the patient the life-saving medication (MOUD)?

• The Surprising Result: No. Whether the doctor looked at the social data or not, it didn't change the likelihood of the patient getting the medication.
• The Worrying Trend: In fact, there was a strange pattern. When doctors saw that a patient had adverse social problems (like homelessness or no money), they were actually less likely to give them the medication.
• The Analogy: Imagine a mechanic sees a car has a flat tire and a broken engine. Instead of fixing it, the mechanic thinks, "Well, this car is in such bad shape, it's not worth the effort," and drives away. The doctors might have been thinking, "This patient has so many social problems, they won't be able to follow up with treatment anyway," so they didn't prescribe the medicine.

4. The Inequality Gap Remained

The study also found that race and ethnicity still played a huge role.

• The Reality: Black, Hispanic, and Asian patients were significantly less likely to get the OUD medication than White patients, regardless of whether the doctor looked at the social data or not.
• The Analogy: Even with the high-tech dashboard, the station workers were still treating different groups of travelers differently based on who they were, not just on the data they saw.

The Big Takeaway

The researchers concluded that just putting a dashboard in the station isn't enough.

If you give a doctor a list of a patient's problems (like "no home" or "no money") but don't give them a tool to fix those problems right then and there, the doctor might just get discouraged and do nothing.

The Solution?
Instead of just showing the problems, the system needs to be smarter. It should say: "This patient has no home, but here is a social worker on standby to help them find shelter, and here is a button to prescribe the medication immediately."

In short: We have the data, but we need better tools to turn that data into action. Otherwise, the doctors are just looking at the map without ever changing the route.

Technical Summary

1. Problem Statement

Despite the rapid proliferation of Electronic Health Record (EHR) tools designed to capture Social Drivers of Health (SDOH) data (e.g., housing, food security, transportation), there is a significant gap in understanding how clinicians actually interact with this data in real-world settings. Specifically:

• Lack of Behavioral Data: It is unclear whether clinicians merely document SDOH data or actively review existing data to inform clinical decisions.
• Impact on Care: There is limited evidence on whether engaging with SDOH data influences treatment outcomes, particularly for high-risk populations like those with Opioid Use Disorder (OUD).
• Equity Concerns: It is unknown if SDOH data engagement helps mitigate or inadvertently exacerbates racial and ethnic disparities in evidence-based treatments, such as Medications for Opioid Use Disorder (MOUD).

2. Methodology

Study Design & Setting:

• Type: Cross-sectional study.
• Setting: University of California, San Francisco (UCSF) Medical Center, an adult Emergency Department (ED) serving >40,000 patients annually.
• Period: January 1, 2023 – October 17, 2024 (5 years post-implementation of new SDOH tools).
• Sample: 17,103 adult ED encounters.
  • OUD Group: 5,701 encounters meeting a computable phenotype for OUD (based on ICD-10 codes, prescriptions, toxicology, and clinical notes).
  • Control Group: 11,402 non-OUD encounters, matched 1:2 to OUD encounters using propensity scores (balancing age, race, sex, insurance, zip code, etc.).

Data Sources & Measures:

• Audit Logs: The study utilized EHR audit logs to track clinician behavior (documentation and review) rather than relying on self-reporting.
• SDOH Data Types Analyzed:
  1. Structured: "SDOH Wheel" (Epic structured questions on housing, finance, transport, food, safety) and ICD-10 Z codes (Problem List).
  2. Unstructured: Social History free text and Social Work notes.
• Definition of "Engagement": Any instance where a clinician either documented new SDOH data or reviewed (hovered/clicked/opened) preexisting SDOH data during the encounter.
• Outcomes:
  1. Rates of documentation and review.
  2. Association between OUD status and SDOH engagement.
  3. Association between SDOH engagement and MOUD treatment (buprenorphine/methadone initiation).

Statistical Analysis:

• Multivariable logistic regression was used to estimate associations, adjusting for demographics (age, race/ethnicity, sex, insurance).
• Sensitivity analyses included excluding patients already on MOUD and testing clinician fixed effects (though clustering by clinician was limited by cross-classification).

3. Key Contributions

• Audit Log Granularity: This is one of the first studies to use granular audit log data to distinguish between documentation (creating new data) and review (consuming existing data) in an ED setting.
• Differentiation of Data Types: The study categorizes engagement by specific EHR fields (structured vs. unstructured) and distinguishes between "field review" (checking a field regardless of content) and "data review" (checking populated data).
• Risk Engagement Metric: Introduced a specific measure for "risk engagement"—engaging with SDOH data specifically when adverse social risks were already documented.
• OUD Context: Specifically examined the intersection of SDOH data usage and OUD care, a critical area where social determinants heavily influence treatment adherence.

4. Key Results

Engagement Rates:

• Documentation: Extremely rare (<1% of encounters).
• Review: Clinicians reviewed structured data more often than unstructured notes.
  • Problem List (Z Codes): 98.4% review rate (highest, likely due to visibility on the ED trackboard).
  • SDOH Wheel: 27.9% review rate (when data existed).
  • Social Work Notes: 16.0% review rate.
  • Social History (Free Text): 3.3% review rate.
• Overall Engagement: Occurred in 19.5% of all encounters.
  • OUD Encounters: 26.6% engagement.
  • Non-OUD Encounters: 16.0% engagement (Odds Ratio [OR] 1.91, 95% CI 1.77–2.07).

Association with MOUD Treatment:

• Overall: SDOH engagement was not significantly associated with receiving MOUD (OR 1.11, 95% CI 0.84–1.47).
• Adverse vs. Non-Adverse:
  • Reviewing adverse SDOH was associated with a non-significant trend toward lower MOUD treatment (OR 0.87).
  • Reviewing non-adverse SDOH was associated with a non-significant trend toward higher MOUD treatment (OR 1.90).
• Disparities: Significant racial and ethnic disparities in MOUD treatment persisted regardless of SDOH engagement:
  • Asian (OR 0.45), Black/African American (OR 0.71), and Hispanic/Latino/a/x (OR 0.48) patients were significantly less likely to receive MOUD compared to White patients.

5. Significance and Implications

Clinical Workflow & EHR Design:

• Accessibility Drives Usage: Clinicians overwhelmingly reviewed the Problem List (Z codes) because it is visible on the main ED trackboard without opening the full chart. Unstructured notes and free-text fields were largely ignored.
• Documentation vs. Review: The low documentation rate (<1%) suggests that current workflows do not incentivize or facilitate entering new SDOH data, but clinicians do look at existing structured data if it is easily accessible.

Equity and Bias:

• Potential for Bias: The trend suggesting that viewing adverse SDOH data correlates with lower MOUD initiation raises concerns that clinicians may use social risk data as a proxy for "non-compliance" or "lack of follow-up," leading to treatment denial.
• Persistent Disparities: The fact that racial disparities in MOUD treatment persisted even when clinicians engaged with SDOH data indicates that simply having the data available does not solve equity issues.

Recommendations:

• Coupled Interventions: The authors argue that expanding SDOH screening alone is insufficient. To improve care, EHRs must surface SDOH data at the point of decision-making and pair it with Clinical Decision Support (CDS) tools (e.g., order sets, reminders) that guide clinicians on how to address barriers (e.g., providing resources for housing or transportation) rather than using the data to withhold treatment.
• Future Research: Qualitative studies are needed to understand why clinicians might hesitate to prescribe MOUD when adverse SDOH are present, and further research should test CDS interventions to see if they can mitigate these biases.

Limitations:

• Single-site study (academic medical center) limits generalizability.
• Audit logs show access but not the depth of cognitive processing or decision-making rationale.
• Limited power to detect significant associations between specific risk engagement and MOUD due to sample size constraints in subgroups.