Vanderbilt Health researchers are developing an artificial intelligence system that could help patients with early-stage Alzheimer’s disease reach newly approved treatments faster. Supported by an 18-month, $600,000 grant from Eli Lilly and Company, the project will create an AI triage agent embedded directly within Vanderbilt Health’s electronic health record (EHR). The system is intended to identify missing information, organize clinical histories and prioritize referrals before patients reach a neurology clinic, potentially reducing delays in a treatment pathway that can currently take weeks or longer.
The project focuses on monoclonal antibody therapies for patients with amyloid-positive Alzheimer’s disease who meet the clinical and safety criteria for treatment. These drugs are designed to slow the progression of cognitive decline during the early stages of the disease, but access involves a complex sequence of medical and administrative steps. A patient may first discuss memory concerns with a primary-care physician or geriatrician, then require specialist evaluation, cognitive testing, brain imaging, confirmation of amyloid pathology and insurance authorization before the first intravenous infusion can be scheduled. Each handoff creates an opportunity for incomplete documentation or delays.
The Vanderbilt team’s initial AI tool will operate at the point where a primary-care or geriatrics clinician refers a patient to neurology for cognitive concerns. Rather than making an autonomous diagnosis or treatment decision, the system will analyze information already contained in the patient’s EHR and generate a structured summary for the receiving team. It will identify details that are commonly absent from referrals, such as relevant medical history, cognitive symptoms, prior testing, medication information or imaging status. It will also recommend whether a referral should be handled as a priority or through the standard workflow.
Clinicians will remain responsible for the final decision at every stage. They will be able to accept, modify or override the AI-generated recommendation, a design intended to keep the technology within a supervised clinical workflow. The system’s role will be administrative and organizational as much as clinical: it will help transform fragmented data into a concise referral picture, allowing specialists and referral coordinators to determine more rapidly what must happen next. By reducing the time spent searching through lengthy records and requesting missing information, the researchers hope to accelerate the evaluation of patients who may be eligible for therapy.
The project is led by You Chen, PhD, Associate Professor of Biomedical Informatics, with Amalia Peterson, MD, Assistant Professor of Neurology, and Sean Huang, MD, Assistant Professor of Medicine and Biomedical Informatics, serving as co-principal investigators. Peterson said that the treatment timeline often begins well before a patient arrives at a specialty clinic, making early referral quality a critical factor. Huang emphasized that primary-care and geriatric clinicians are frequently the first to recognize cognitive changes and need practical tools to communicate the relevant information completely and efficiently.
Before deploying the triage agent, the researchers will map how patients currently move through Vanderbilt Health’s dementia-care pathway. Using a cohort of more than 5,300 patients, the team plans to apply AI methods to reconstruct care timelines from EHR data. These timelines may include referral dates, appointments, diagnostic procedures, imaging, laboratory work, authorizations and infusion scheduling. By comparing the intervals between these events, investigators will search for points at which patients routinely wait and determine whether delays arise from missing documentation, scheduling capacity, unclear responsibility between teams or insurance-related requirements.
The analysis will be combined with input from neurologists, geriatricians, primary-care clinicians, referral staff and other operational personnel. This human-centered step is important because an EHR record captures events but does not always explain why a delay occurred. A missing test, for example, might reflect an unavailable appointment, uncertainty about which clinician should order it or a requirement that was not clearly communicated. Understanding these operational causes will allow the researchers to design an AI agent around actual workflow problems rather than simply adding another alert to an already crowded EHR environment.
The pilot evaluation will measure whether the system reduces the time from diagnosis to the patient’s first infusion. Researchers will also need to examine whether the agent’s recommendations are accurate, whether clinicians use or override them, and whether the tool performs consistently across different patient groups and referral sources. Because the technology will rely on clinical records, its performance could be affected by incomplete or unevenly documented data. Monitoring these limitations will be essential to ensure that a tool intended to speed access does not inadvertently prioritize some patients while delaying others.
The Vanderbilt proposal also outlines two possible extensions of the technology. A second AI agent could analyze magnetic resonance imaging scans and prepare Alzheimer’s-related safety reports for radiologists to review. Such reports might help organize findings relevant to treatment eligibility and monitoring, although the radiologist would retain responsibility for interpreting the images. A third agent could assemble the documents needed for insurance authorization and track the status of each request. Together, the proposed tools would address different stages of the pathway, from the first referral through imaging, payer review and infusion planning.
The grant builds on Vanderbilt’s ongoing collaboration with Lilly, including a separate $1 million-funded project led by Chen that examines gaps in obesity care. The Alzheimer’s initiative reflects a broader effort to use EHR-integrated AI not only for diagnosis, but also for improving the delivery of care. If the triage agent successfully shortens waiting times without compromising clinical oversight, the researchers intend for its design to be transferable to other health systems. The project’s central test will be whether carefully supervised automation can turn a complicated, fragmented treatment pathway into a faster and more coordinated experience for patients and their families.
Subject of Research: AI-enabled triage within electronic health records to accelerate Alzheimer’s disease referrals and treatment access
Article Title: Vanderbilt Researchers Build EHR-Based AI Tool to Speed Access to Alzheimer’s Treatments
References: Vanderbilt Health researchers; Eli Lilly and Company grant information
Keywords: Alzheimer’s disease, artificial intelligence, electronic health records, dementia care, monoclonal antibody therapy, neurology, clinical workflow, healthcare innovation
