Social and addiction-related needs are common, clinically consequential, and too often unidentified and unaddressed during the emergency department (ED) visit. The broad goals of the Molina Lab are to reduce disparities in healthcare access, mitigate bias in treatment, and integrate social, addiction, and clinical care in emergency settings—without adding burden to an already overstretched ED workforce.
We do this through:
- Informatics-based research to understand how social, addiction, and clinical data are captured and used in the EHR
- Human-centered design to build EHR-based tools that support evidence-based, whole-person care
- Clinical trials testing interventions that improve access to prevention and treatment
- Artificial intelligence (including large language models) to improve ED workflow efficiency and decision-making
- Qualitative research to identify implementation barriers and design equitable solutions with and for vulnerable populations
Leveraging Large Language Models to Improve ED Workflow
Many of the most important clinical and social signals in emergency care live in unstructured documentation—creating missed opportunities for timely, equitable care. We study how large language models (LLMs) can reduce cognitive load and unlock actionable information from routine ED notes.
Using LLMs to identify opioid use disorder (OUD) from ED notes
In this preprint, we demonstrate that an LLM-based approach can identify OUD using emergency clinician documentation and improves upon structured EHR phenotypes, underscoring the value of unstructured text for detecting undertreated conditions.
Reference URL: https://www.medrxiv.org/content/10.64898/2025.12.17.25342510v1
Physician- versus LLM-generated ED summaries
This preprint evaluates how LLM-generated summaries compare with physician-written summaries, with implications for documentation efficiency, information prioritization, and workflow support in high-acuity care settings.
Reference URL: https://www.medrxiv.org/content/10.1101/2025.08.13.25333609v3
Clinical Trials to Improve Preventive Care Access in Emergency Settings
The ED is a critical point of access for underserved patients—but moving from “opportunity” to “action” requires scalable, rigorously tested interventions.
COVID-19 booster vaccine messaging to increase vaccine uptake in EDs
In this cluster randomized clinical trial, we tested pragmatic strategies to increase updated COVID-19 vaccine uptake after ED visits, generating evidence on what works—and what is insufficient—when deploying scalable messaging approaches in real-world emergency settings.
Reference URL: https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2840135
Understanding Social Drivers of Health Documentation & Screening in US Emergency Departments
Equitable care requires that social risk be visible in clinical systems. We examine how often SDOH are documented and screened in ED practice nationally, and what structural gaps persist.
Social Risk Factor Documentation in Emergency Departments
Using national data, we found that documentation of ICD “social Z codes” in ED encounters is very low, raising concerns that patients’ social needs may remain invisible to systems designed to identify inequity and target interventions.
Reference URL: https://pubmed.ncbi.nlm.nih.gov/36210245/
Screening and Response for Adverse Social Determinants of Health in US Emergency Departments
In this national study of ED-based screening and response policies, we identify a substantial gap between the growing recognition of SDOH and the real-world infrastructure needed to screen and respond—highlighting opportunities for scalable, equity-centered implementation strategies.
Reference URL: https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2833176