Miller School of Medicine Researchers Build Real‑World Model of Alzheimer’s Risk
A new research platform integrates health records, social data and analytics to reveal how chronic conditions shape Alzheimer’s disease risk and progression.
As Alzheimer’s disease and related dementias continue to rise alongside an aging population, researchers are confronting a central challenge. Most patients do not experience dementia in isolation. Instead, Alzheimer’s unfolds amid a complex web of chronic conditions, social factors and real‑world health trajectories that traditional research models often fail to capture.
The Multimorbidity 3‑City Alzheimer’s Disease (M3AD) Analytic Platform aims to change that.
Investigators from the University of Miami Miller School of Medicine, working in collaboration with research teams from Columbia University and the University of Chicago, are playing a key role in building this large‑scale, real‑world research infrastructure. M3AD integrates electronic health records (EHRs), neighborhood‑level data and advanced analytics to better understand how Alzheimer’s develops, progresses and might be prevented.
“Alzheimer’s disease does not occur in a vacuum,” said Tatjana Rundek, M.D., Ph.D., professor of neurology and the Evelyn F. McKnight Chair for Learning and Memory in Aging at the Miller School. “Patients arrive with decades of vascular, metabolic and social exposures behind them. If we want to intervene earlier and more effectively, we have to understand those trajectories, not just a single diagnosis.”
Why Alzheimer’s Rarely Occurs in Isolation
Historically, Alzheimer’s research has focused on isolated risk factors or narrowly defined patient cohorts. The M3AD platform takes a different approach by embracing the presence of two or more chronic conditions. Multimorbidity affects nearly 90% of adults older than 60 and is strongly linked to faster cognitive decline, higher hospitalization rates and increased mortality.
Using longitudinal EHR data from health systems in Miami, New York and Chicago, the consortium assembled a dataset representing nearly 10 million patients, including tens of thousands diagnosed with Alzheimer’s or related dementias. Importantly, the data remain within each institution’s secure environment, protecting privacy while allowing shared analyses through federated learning methods.
For Miller School researchers, this approach reflects real clinical experience.
“We rarely see ‘pure’ Alzheimer’s,” said David Loewenstein, Ph.D., professor of neurology, psychiatry and behavioral sciences at the Miller School and a nationally recognized expert in cognitive aging. “We see patients managing diabetes, hypertension, depression and social stressors all at once. This platform finally gives us the tools to study that complexity at scale.”
Building the Multimorbidity Alzheimer’s Disease Platform
The M3AD platform harmonizes EHR data using a common data model. Researchers track diagnoses, medications, laboratory values, imaging and clinical notes over time. Natural language processing is used to extract key symptoms and cognitive indicators that may not appear in structured fields, helping identify both diagnosed and previously unrecognized dementia cases.
Crucially, the platform embeds clinical data within census‑tract–level social determinants of health, including education, income, insurance coverage and neighborhood characteristics. This allows investigators to examine how social and environmental contexts interact with medical risk factors across the life course.
Dr. Rundek praised the work of University of Miami Frost Institute for Data Science and Computing colleagues Nick Tsinoremas, Ph.D., a UM professor of computer science and health informatics, and Deepthi Puram, manager of systems and data engineering at UM, for their work on the study.
“Our group is responsible for the harmonization and curation of the data, as well as providing advance analytics and computational approaches,” Dr. Tsinoremas said. “This is a one-of-a-kind study that involves three major sites with vast amounts of data. Data harmonization for advance analytics is a major challenge that we address in this study.”
“They played a key role by assisting with data structure and organization, which is essential for achieving the study’s objectives,” Dr. Rundek said.
Rather than focusing solely on disease counts, the study examines patterns, sequencing, duration and control of chronic conditions, such as whether blood pressure or diabetes was well managed, intermittently controlled or uncontrolled over time. These trajectories may reveal windows when intervention could meaningfully alter dementia risk or progression.
Early Insights Into Chronic Conditions and Cognitive Decline
Early analyses underscore the importance of looking beyond simple comorbidity counts. Certain clusters of conditions, particularly those involving vascular, metabolic and neuropsychiatric disorders, appear to accelerate cognitive decline. Others interact in less predictable ways.
The platform also highlights the clinical burden of multimorbidity, including polypharmacy, fragmented care across multiple providers and higher rates of avoidable hospitalizations. These real‑world patterns help explain why guideline‑based, single‑disease care often falls short for patients with dementia.
“This work allows us to test prevention strategies in real life, not just in idealized trial settings,” Dr. Rundek said. “We can ask whether improving blood pressure control, smoking cessation or diabetes management at specific points in time actually changes Alzheimer’s risk years later.”
What This Means for Personalized Alzheimer’s Prevention and Care
For patients and families, the implications are significant. By identifying modifiable risk factors and high‑risk trajectories earlier, clinicians may one day personalize prevention strategies long before memory symptoms emerge. The platform also creates a trial‑ready infrastructure for evaluating new treatments across diverse, real‑world populations.
Dr. Loewenstein emphasized the translational potential.
“Understanding how cognitive decline unfolds alongside medical and social complexity helps us design care models that are more realistic and ultimately more effective for patients,” he said.
More from the Miller School
Miller School researchers at AD/PD 2026 highlighted why APOE is now a leading genetic target for Alzheimer’s treatment.
The Miller School of Medicine honors its 2026 MFA Student Research Award winners for breakthroughs in student research.
A University of Miami study shows how an ancestry‑specific ABCA7 gene deletion disrupts lipid metabolism, increasing Alzheimer’s disease risk.
University of Miami researchers link long‑term exercise, diet, and social engagement to better cognition with aging.
Tags: AI, Alzheimer's & Dementia: The Journal of the Alzheimer's Association, Alzheimer's disease, artificial intelligence, cognitive decline, dementia, Dr. David Loewenstein, Dr. Tatjana Rundek, team science, technology