Few primary care practices are designed for the timely detection of Alzheimer’s disease and related dementias (ADRD). Researchers headed by teams at Indiana University (IU) School of Medicine, Regenstrief Institute, Eskenazi Health, University of Miami School of Medicine, and Lamar University have now demonstrated a fully digital artificial intelligence (AI), zero-cost method for detecting dementia that can be scaled across primary care clinics without additional time for physicians.
The method combines the Quick Dementia Rating System (QRDS) data with a Passive Digital Marker (PDM), a machine learning algorithm that uses natural language processing to analyze data from electronic health records (EHRs). It identifies information such as memory issues, vascular concerns, and other factors linked to dementia.
In a randomized clinical trial (RCT) involving more than 5,000 patients, the researchers showed that their scalable approach combining QDRS data with the PDM increased the incidence rate of ADRD diagnosis within 12 months by a third, without requiring any additional time, labor, or effort from the clinical team.
The researchers say the breakthrough represents a major step forward in translating AI and patient-reported outcomes into everyday clinical care. By integrating scalable digital tools that operate seamlessly within existing health systems, the research team demonstrated how technology can strengthen early detection, reduce burdens on primary care teams, and improve outcomes for older adults.
“This is the most scalable approach to early detection that I know of,” commented Regenstrief research scientist Malaz A. Boustani, MD, who codeveloped the PDM. “Most early detection methods require at least five minutes of a doctor’s time and often come with licensing fees. Our dual approach, by contrast, requires zero clinician time or money.” Zina Ben Miled, PhD, a Regenstrief affiliate scientist and Lamar University professor, who codeveloped the passive digital marker tool alongside Boustani, added, “What’s powerful about this approach is that it helps level the playing field. By embedding these tools directly into the electronic health record, we can reach patients who might otherwise be overlooked—ensuring that everyone, regardless of background or resources, has the same opportunity for early detection and care.”
Boustanii and Miled are first authors of the team’s published paper in JAMA Network Open, which is titled “Digital Detection of Dementia in Primary Care: A Randomized Clinical Trial.” In their paper, the investigators concluded, “This randomized clinical trial found that the combined approach was effective at scale for the early detection of ADRD in primary care settings. This is an important feature in busy primary care settings that can benefit both the health care system and patients.”
Detection of ADRD is a challenge in primary care settings, the authors wrote. “… more than 50% of older adults in primary care never receive a formal and timely diagnosis.” The limited time that primary care clinicians are able to spend with patients, the need to focus on the health problems that brought the patient to the clinic, as well as the stigma of Alzheimer’s disease and dementia, are major reasons for the lack of recognition of the condition.
Moreover, the team continued, the use of digital or paper versions of cognitive performance tests carried out by clinicians present with scalability and sustainability challenges in primary care. “Commonly available approaches to ADRD detection in primary care face a major barrier in that they depend on clinicians performing the data collection (e.g., through direct interview or testing).” And while the FDA has recently approved a blood-based biomarker for detecting Alzheimer’s disease, there are no biomarkers for detecting other ADRDs.
More encouragingly, the researchers suggested, “Patient-reported outcome (PRO) approaches, the growth in data captured by various electronic health record (EHR) systems, and the advances in machine learning algorithms may overcome such challenges while avoiding additional time requirement from clinicians for data collection.”
In a pragmatic real-world randomized clinical trial of more than 5,300 patients from primary care practices, the investigators evaluated a dual approach combining the QDRS, a 10-question patient-reported tool, and an AI tool, called a PDM. “The QDRS is a patient-reported outcome measure, while the PDM is a machine learning algorithm that uses EHR data,” the scientists explained. “Both can be embedded in the EHR for ease of use. The participants had no diagnosis of mild cognitive impairment, dementia, or severe mental illness.
The trial, conducted at nine Eskenazi Health Center federally qualified health centers in Indianapolis, embedded the QDRS and passive digital marker directly into the Epic EHR. The system automatically invited patients aged 65 and older to complete the short QDRS survey through their patient portal, while the PDM algorithm continuously analyzed existing clinical data to flag patients at risk. Results appeared automatically in the clinician’s EHR inbox, prompting further evaluation only when necessary, requiring no extra time, staff, or manual screening.
The study found that combining these two tools increased the rate of new Alzheimer’s and related dementia diagnoses by 31% compared with usual care, over 12 months, all without requiring additional clinician time or costly testing. “This approach benefits from directly identifying patterns of health from the EHR indicative of early ADRD while mitigating known limitations of EHR data with important PRO-reported information,” the team stated. “In this RCT, we found that using the QDRS coupled with the PDM could represent a very effective approach for the detection of ADRD by nudging the primary care clinicians to complete diagnostic cognitive assessments following a positive screen result.”
The PDM has been in development for more than 10 years at Regenstrief by research scientist Boustani. “Building on more than 50 years of innovation in digital health data science and machine learning, this passive digital marker developed at the Regenstrief Institute is now open source,” said Regenstrief and IU School of Medicine faculty member Boustani. “In keeping with Regenstrief’s tradition of open medical record methodology, there’s no licensing fee—just the basic cost of deploying it, similar to how you would deploy any app. Any healthcare system with an electronic health record and the right personnel can implement it. It is zero cost and requires no clinician time.”
Beyond increasing detection, the combined digital approach also led to a 41% increase in follow-up diagnostic assessments, such as neuroimaging and cognitive testing, suggesting earlier and more accessible dementia care for populations traditionally underserved by the healthcare system.
“The Quick Dementia Rating System was designed to empower patients and families to report cognitive changes easily and quickly,” said James E. Galvin, MD, a professor of neurology and director of the Comprehensive Center for Brain Health at the University of Miami Miller School of Medicine. “When used with digital tools like the Regenstrief passive digital marker, we can bring early detection to scale—efficiently and effectively.”
“This work represents the next phase of our half-century legacy at Regenstrief—using data, innovation, and compassion to transform healthcare delivery,” said Boustani, who is the lead author of the Digital Detection of Dementia in Primary Care clinical trial. “We’ve shown that it’s possible to bring the power of AI and patient-reported outcomes directly into the clinic—seamlessly, affordably, and at scale.”