Digital Screening Tool Duo Shows Promise for ADRD Detection in Primary Care

A randomized clinical trial of 5,325 older adults in federally qualified primary health care centers found that combining a patient-reported questionnaire with a machine learning algorithm embedded in the electronic health record (EHR) increased documented dementia diagnoses by 31% compared with usual care.¹

The combined approach also led to 41% more diagnostic workups in the frontline care setting without requiring additional clinician time, according to the findings, published in JAMA Network Open.¹

Available Tools Aren't Enough

According to corresponding author Malaz A Boustani, MD, MPH, and colleagues at Indiana University School of Medicine, the primary care services provided for adults living with Alzheimer disease and related dementias (ADRD) are being “reshaped” by 3 important developments: research has shown that lifestyle modifications can reduce dementia risk by up to 45%² and there are now 2 FDA-approved disease-modifying therapies available to help slow progression of AD at early stages.³ Third, Medicare has adopted comprehensive care models for people with dementia and their families.⁴

Currently, however, more than half of older adults see in primary care practice do not receive “a formal and timely diagnosis,” the authors wrote. Traditional cognitive screening faces major scalability barriers because it depends on busy clinicians to collect data through direct testing or interviews, they continued, and while blood-based biomarkers approved for detection of AD are promising, there are no such assays for surrogate markers of other ADRD pathologies.¹

The Study

Boustani and fellow researchers randomly assigned 9 urban federally qualified health centers serving predominantly low-income, diverse populations to 1 of 3 approaches in the clinic:

• usual care with no routine dementia screening
• use of a Passive Digital Marker (PDM) alone, or
• combined use of the Quick Dementia Rating System (QDRS) plus PDM

The QDRS is a 10-question patient-reported outcome (PRO) tool that takes less than 3 minutes to complete and has 85% accuracy for detecting dementia, with a cutoff score of 1.5 on a 0-30 scale. The PDM uses a machine learning algorithm to analyze existing EHR data for early detection of ADRD, achieving 80% accuracy with a risk threshold of 59%. Both tools were embedded directly into the health system's Epic EHR and triggered clinical decision support alerts when results were positive. The trial was performed between July 2, 2022 and July 1, 2024 and enrolled adults 65 years and older without existing diagnoses of cognitive impairment, dementia, or severe mental illness.¹

The primary outcome was defined as 12-month cumulative incidence of ADRD diagnoses with a secondary outcome including any ADRD diagnostic workup, eg, laboratory tests, neuropsychological testing, or brain imaging, according to the study. The final cohort had a mean age of 71.1 years and was 62.2% women; 55% self-identified as Black or African American, and 15.5% as Hispanic or Latino.¹

Key Results

After 12 months, clinics using QDRS plus PDM had a 15.4% incidence of new dementia diagnoses compared with 12.4% in usual care clinics (adjusted odds ratio [AOR], 1.31; 95% CI, 1.05-1.64). Clinics using PDM alone showed no significant difference from usual care, with an incidence of 10.3% (AOR, 0.84; 95% CI, 0.63-1.11).¹

The investigators reported that using the combined QDRS plus PDM approach also increased completion of diagnostic assessments. More than one-third (36.7%) of participants in clinics using QDRS plus PDM underwent diagnostic workups compared with 29.0% in usual care clinics (AOR, 1.41; 95% CI, 1.12-1.77). Time-to-diagnosis analysis showed patients in the combined intervention arm received diagnoses earlier, with an adjusted hazard ratio of 1.37 (95% CI, 1.12-1.68).¹

Boustani et al emphasized that while the QDRS completion rate was only 20.7% in the combined intervention arm, the approach remained effective. Participants with positive PDM screening results had higher rates of both diagnostic assessments (41.5% vs 34.4%) and diagnoses (24.0% vs 11.3%) compared with those who screened negative, regardless of QDRS completion.¹

Trust in Machine Learning

The researchers hypothesized that the QDRS may have enhanced clinician trust in the PDM results. "Trust in machine is influenced by a combination of human factors, such as user education, past experiences, user biases, and perceptions toward automation," they wrote.¹ The finding that PDM alone was ineffective while the combined approach succeeded suggests that integrating patient-reported data with algorithmic predictions may be necessary to change clinician behavior.

Study Limitations

Among the study’s limitations Boutsani and colleagues acknowledge that the low QDRS completion rate likely reflected barriers including limited patient portal access (40% of eligible patients had active portals), language translation needs, and relatively low education levels in the study population. The single-health-system design in federally qualified health centers may limit generalizability. In addition, EHR data may have captured tests ordered for non-cognitive reasons, though randomization should have distributed this measurement error equally across study arms.

The limitations notwithstanding, the researchers state that the trial demonstrates that a scalable, low-burden approach to dementia detection is feasible in resource-limited primary care settings. The combined digital screening system required no additional clinician time for data collection while substantially increasing both diagnostic workups and documented diagnoses. As they note, the technology has the potential to addresses the critical 50% gap in diagnoses of ADRD identified in the National Plan to Address Alzheimer Disease despite the most recent review by the US Preventive Services Task Force that continues to find "insufficient evidence to recommend for or against cognitive screening in primary care."⁵

References

1. Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet Standing Commission. Lancet. 2024;404(10452):572-628. doi:10.1016/S0140-6736(24)01296-0 2. Cummings JL. Maximizing the benefit and managing the risk of anti-amyloid monoclonal antibody therapy for Alzheimer’s disease: strategies and research directions. Neurotherapeutics. 2025;22(3):e00570. doi:10.1016/j. neurot.2025.e00570 3. 2024 Alzheimer’s disease facts and figures. Alzheimer’s Association. 2024. Accessed May 5, 2025. https://www. alz.org/getmedia/76e51bb6-c003-4d84-8019-e0779d8c4e8d/alzheimers-facts-and-figures.pdf 4. Patnode CD, Perdue LA, Rossom RC, et al. Screening for cognitive impairment in older adults: an evidence update for the U.S. Preventive Services Task Force. Evidence Synthesis No. 189. AHRQ publication 19-05257-EF-1. Agency for Healthcare Research and Quality; 2019.