Non-literacy biased, culturally fair cognitive detection tool in primary care patients with cognitive concerns: a randomized controlled trial

Dementia affects approximately one in nine older Americans and is frequently underdiagnosed in primary care, particularly among historically marginalized groups such as Black and Hispanic/Latino populations. Despite numerous cognitive assessment tools, multiple barriers at the instrument, patient, clinician and health-system levels impede routine detection and follow-up care. Many existing tests are lengthy, clinician-administered, culturally biased, or lack clear guidance for next steps following assessment. The study evaluates whether the 5-Cog paradigm—a brief, culturally fair, non–literacy-biased cognitive assessment paired with an EMR-embedded clinical decision tree—improves dementia-related care actions in older primary care patients who present with cognitive concerns, with a focus on populations experiencing health disparities. The primary research question is whether 5-Cog increases timely diagnoses and management actions (diagnosis, investigations, medications, or specialist referrals) compared with an active control within 90 days of the visit.

Prior work indicates that over half of dementia cases are missed in primary care, with greater underdiagnosis among Black and Hispanic populations. Many cognitive tools are not culturally adapted, often validated in predominantly white samples, and may be long, require clinicians, or fail to provide actionable guidance. The authors previously reported that MoCA cutoffs derived in mostly white samples were too high for diverse primary care populations. Studies in primary care have separately examined brief assessments, physician education, and dementia care management, but these components have seldom been integrated, and evidence that detection tools improve downstream care is limited. The US Preventive Services Task Force concluded that evidence was insufficient to assess benefits and harms of universal screening in asymptomatic adults; early case detection based on cognitive concerns was not addressed. Trials of educational interventions alone have not consistently improved diagnosis or management. Non-randomized clinic implementations pairing testing by non-physician staff with education have shown increases in actions among test-positive patients. Disparities in dementia care actions exist, with Hispanic patients often receiving fewer actions than Black or white patients. This context motivated development of the 5-Cog paradigm integrated with EMR decision support to address cultural fairness, feasibility, and actionability.

Design: Single-blind randomized controlled trial (ClinicalTrials.gov NCT03816644) conducted at one urban primary care clinic in Bronx County, NY, serving patients experiencing health disparities. Enrollment occurred from 29 May 2019 to 15 September 2022; primary outcome follow-up ended 15 December 2022 and secondary outcome follow-up ended 15 May 2023. IRB approval obtained; consent procedures changed from written to oral (telephone) after COVID-19 suspension, with no other protocol changes. Participants: Inclusion criteria: age ≥65, cognitive concerns, scheduled clinic appointment, English or Spanish speaking. Exclusions: prior MCI or dementia diagnosis, nursing home residence, or sensory impairments precluding assessment. Of 4,538 screened, 1,258 consented; 1,201 randomized (599 intervention; 602 control). Baseline: mean age 72.8 years (range 65–98), 72% women; 48.7% Black, 55.5% Hispanic/Latino; 40.4% did not graduate high school; all lived in socioeconomically disadvantaged neighborhoods; 56.7% tested in English and 43.3% in Spanish. Outcome ascertainment rate: 100% in both arms. Randomization and blinding: Block randomization stratified by sex and age (<80 vs ≥80) using SAS PROC PLAN; allocation 1:1. Staff were blinded to the next assignment until allocation. Investigators, assessors, and statisticians were blinded to assignments and not involved in recruitment or interventions. Intervention (5-Cog): Administered pre-PCP visit by trained testers and comprised: Picture-based Memory Impairment Screen (PMIS), Motoric Cognitive Risk (MCR) syndrome (cognitive complaints plus slow gait), and Symbol-Match. The 5-Cog result was ‘positive’ if any component was abnormal. An EMR note summarized results (positive/negative) and included a decision tree with recommended follow-up actions based on global dementia care guidelines. A paper token alerted PCPs to review the EMR note. Cutoffs prioritized sensitivity to minimize false negatives in this high-risk group. PCPs were free to use clinical judgment and were not required to follow recommendations. Control (active): Time- and contact-matched Short Assessment of Health Literacy (SAHL; English/Spanish) and grip strength (Jamar dynamometer). Positive/negative results were documented with EMR notes and relevant recommendations, without overlap with 5-Cog assessments. PCPs were similarly not required to follow recommendations. Procedures: Pre-trial workshops educated clinic staff on study procedures and dementia care; periodic newsletters provided updates. All 18 PCPs (16 physicians, 2 nurse practitioners) participated. Outcomes: Primary outcome was a composite of improved dementia care actions within 90 days documented in EMR with a cognitive indication: new PCP diagnoses of MCI or dementia; laboratory tests; imaging; new dementia medication prescriptions; or specialist referral for dementia evaluation. Only orders with cognitive indications were counted; other non-standard cognitive labels (e.g., ‘memory loss’) were not counted as diagnoses but could qualify as indications for tests, treatment, or referrals. Secondary outcome: healthcare utilization (all-cause ED visits and hospitalizations) within 12 months. Safety: Research staff documented any adverse events; an independent blinded assessor reviewed EMRs for harms (medical symptoms, stress, anxiety) potentially related to the intervention. Statistical analysis: Prespecified plan (SPSS v29). Intention-to-treat analysis. Group comparisons used chi-squared tests and multivariable logistic regression adjusted for age, sex, and education. Discriminant validity of 5-Cog was assessed via sensitivity and specificity to PCP diagnoses of MCI/dementia in the intervention arm; in a subsample, performance was compared against MoCA (<17). Prespecified subgroup sensitivity analyses examined sex, ethnicity, language, and education. Post-hoc analyses compared pre- vs post-COVID suspension enrollees. Chronic illness was not included as a covariate due to missing data.

- Enrollment and baseline: 1,201 participants randomized (599 5-Cog; 602 control); mean age 72.8 years; 72% women; 48.7% Black; 55.5% Hispanic/Latino; 40.4% <high school; 100% in socioeconomically disadvantaged neighborhoods; English 56.7%, Spanish 43.3%. - Primary outcome (composite improved dementia care within 90 days): 18.5% in 5-Cog vs 6.8% in control; adjusted OR 3.43 (95% CI 2.32–5.07), P<0.001. - Components (5-Cog vs control; adjusted ORs): • New MCI or dementia diagnoses: 10.5% vs 2.0%; OR 6.48 (3.41–12.31), P<0.001. • Imaging ordered: 6.5% vs 1.5%; OR 4.80 (2.29–10.06), P<0.001. • Laboratory tests ordered: 12.2% vs 2.0%; OR 7.64 (4.05–14.39), P<0.001. • New dementia prescriptions: 1.0% vs 0.3%; OR 3.24 (0.63–16.65), P=0.15 (not significant). • Specialist referral: 10.0% vs 4.7%; OR 2.38 (1.49–3.80), P<0.001. - Actions by test result within 5-Cog arm: 43.8% of 5-Cog positive (106/242) received dementia care actions vs 1.4% of 5-Cog negative. - Timing of diagnoses: In 5-Cog arm, 89% of new MCI/dementia diagnoses occurred on day of visit; in control, 92% occurred on day of visit. - Diagnostic performance: In 5-Cog arm, 5-Cog had sensitivity 96% and specificity 71% for PCP diagnoses of MCI/dementia. In a subsample (n=295) agreeing to additional testing, 5-Cog had sensitivity 70% and specificity 71% for detecting MCI/dementia using MoCA <17. - Secondary outcomes (12 months): No significant differences in hospitalizations (15.2% 5-Cog vs 16.7% control, P=0.49) or ED visits (16.3% vs 15.2%, P=0.61). - Safety: No study-related adverse events or EMR-documented harms. - Sensitivity analyses: Significant benefit of 5-Cog across subgroups (adjusted ORs): • Male: OR 2.21 (1.13–4.35); Female: OR 4.23 (2.60–6.86). • Black: OR 3.57 (1.90–6.70); Hispanic: OR 2.94 (1.83–4.73). • English: OR 4.26 (2.29–7.93); Spanish: OR 2.94 (1.76–4.91). • <High school: OR 3.56 (2.16–5.88); ≥High school: OR 3.20 (1.71–5.97). • Pre-COVID suspension: OR 4.02 (2.20–7.36); Post-suspension: OR 3.11 (1.86–5.22).

The trial demonstrates that a brief, culturally fair cognitive assessment integrated with EMR-based decision support (5-Cog) substantially increases dementia-related care actions in diverse, socioeconomically disadvantaged primary care populations presenting with cognitive concerns. The findings address a key evidence gap not considered by the USPTF, which focused on universal screening of asymptomatic adults. By triggering assessment upon cognitive concerns and providing actionable recommendations, 5-Cog improved new diagnoses, testing, and referrals without increasing healthcare utilization over 12 months. While PCPs used 5-Cog to inform decisions, they relied on clinical judgment and did not act on all positive results, highlighting that decision support complements but does not replace clinician discretion. The markedly low action rates in control and test-negative cases are consistent with routine practice and underscore the added value of the EMR-embedded decision tree. The paradigm’s brevity, non–physician administration, bilingual implementation, and cultural adaptation likely contributed to its acceptability and effectiveness in historically marginalized groups, where disparities in dementia detection and care are pronounced. The results support scaling 5-Cog within primary care (e.g., Medicare annual wellness visits) and its role in triaging patients amid emerging Alzheimer’s therapies, with potential adaptability to other cognitive tests and settings.

In a single-blind RCT involving older primary care patients with cognitive concerns in an underserved, racially/ethnically diverse urban setting, the 5-Cog paradigm significantly increased dementia care actions (diagnoses, investigations, referrals) compared with an active control, with no increase in hospitalizations or ED visits and no intervention-related harms. This provides an evidence base for integrating brief, culturally fair cognitive detection with EMR decision support to improve dementia care in primary care. Future work should assess clinical effectiveness and implementation at scale (ongoing pragmatic cluster RCT NCT05515224), evaluate economic impact, adapt the approach to various clinical contexts and cognitive tools, and investigate its utility in asymptomatic populations and in broader, international settings.

- Consent process changed post-COVID suspension (from written to oral), although baseline characteristics and outcomes were similar pre- and post-suspension. - MCI/dementia classifications were based on PCP diagnoses rather than comprehensive neuropsychological evaluation, reflecting real-world practice but potentially introducing diagnostic variability. - The composite outcome included selected dementia care actions; other quality indicators were not assessed. - Substantial non-reporting of race limited detailed subgroup comparisons. - Findings do not generalize to asymptomatic older adults; the utility of 5-Cog in this group remains to be determined. - Detailed psychological interviews were not conducted and may have missed patient stress or anxiety related to diagnosis. - Primary outcome window was restricted to 90 days; although most actions occurred on the visit day, later actions may have been missed. - Single-center US study limits generalizability; replication in other populations and settings is needed. - A planned covariate (chronic illness) could not be included due to incomplete data.