Feasibility and Acceptability of Implementing a Digital Cognitive Assessment for Alzheimer Disease and Related Dementias in Primary Care

INTRODUCTION

A missed or late diagnosis of Alzheimer disease or a related dementia (ADRD) can lead to poor health outcomes including worse behavioral and psychological symptoms, mismanagement of chronic comorbidities, increased care utilization, and missed opportunities to participate in advance care planning and ADRD research such as clinical trials.^1-5 For family caregivers, a late diagnosis of ADRD increases caregiver stress and burden and decreases opportunities to prepare for caregiving.¹ In primary care, a lack of workflows and care processes that support timely detection and diagnosis of ADRD limits older adults’ ability to be assessed for new pharmacologic treatments⁶ and increases the disparities of ADRD diagnosis and treatment by race and socioeconomic status.^5,7

Most Americans with ADRD receive a majority of their health care from primary care physicians (PCPs) and have limited access to specialists for diagnosis and management of ADRD.⁸ Despite requirements for screening/detection of cognitive impairment during the Medicare annual wellness visit, less than one-third of older adults report receiving an assessment of cognition by their PCP,^9,10 and it is estimated that more than one-half of older primary care patients who meet diagnostic criteria for ADRD never receive a diagnosis, with high rates of undetected cognitive impairment in underrepresented minoritized groups and socially vulnerable older adults.^3,11,12 When ADRD is detected earlier, PCPs can evaluate for potential causes of cognitive impairment, both reversible and syndromic, and facilitate advance care planning, improve patient and caregiver preparedness related to disease progression and prognosis, connect patients to home and community-based services, and initiate behavioral and pharmacologic treatment.^13,14

Research on the risks and benefits of early detection of ADRD in primary care has shown that patients do not experience adverse psychological reactions, such as depression or anxiety, from screening yet benefits are difficult to measure when the screening program is not coupled with an evidence-based diagnostic assessment that is feasible and acceptable to primary care patients, their family members, and clinicians^15,16 Interventions to improve the detection and diagnosis of ADRD have focused on improving clinician knowledge and confidence in making a diagnosis and the general public’s knowledge regarding ADRD.^13,17-22 Whereas knowledge of ADRD has improved, the effect of these informational interventions on timely ADRD diagnosis remains unknown.^17-22 In addition, patients have shown a high level of acceptance of screening for cognitive impairment in primary care.²³

With the data showing worse outcomes for patients with late diagnoses and the emergence of monoclonal antibody therapies indicated for those with mild cognitive impairment (MCI) or mild-stage Alzheimer disease,^24,25 the importance of testing strategies for early identification of patients in primary care with or at risk of developing ADRD is more critical than ever. Yet, no clinical practice guidelines or standard workflows exist to implement cognitive screening in routine primary care for care, treatment, or research recruitment. Therefore, the development, implementation, and evaluation of pragmatic processes to identify patients with ADRD in the primary care setting is essential.

From June 2022 to May 2023, Indiana University and Indiana University Health comprised 1 of 7 sites globally to participate in the Davos Alzheimer’s Collaborative Healthcare System Preparedness Project Early Detection Program as a clinical demonstration initiative to evaluate and measure the feasibility and acceptability of implementing a digital cognitive assessment (DCA) into routine primary care for older primary care patients.

METHODS

RESULTS

During the 1-year project period, there were 17,894 primary care encounters across the 7 clinics (Figure 1). Of those encounters, 16,708 (93%) were for patients who were eligible to be approached for a DCA. For those eligible encounters, 1,808 (10.8%) cognitive screenings were completed. For the eligible encounters in which a patient was approached, 3,727 (22.3%) patients declined to complete the DCA. In addition, during 9,232 (55.3%) eligible encounters, either clinicians declined to have the patient complete the DCA during that clinic visit (for various clinical reasons) or that encounter was deemed out of scope by the predetermined DCA implementation workflow developed and adapted by the clinic.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Feasibility and Acceptability of Digital Cognitive Assessment Implementation

DCA = digital cognitive assessment; PCP = primary care physician.

^a For each encounter, the PCP could decline to have the patient complete the DCA, and as different clinics rolled out the DCA screening, some encounters were defined as out of scope on the basis of clinic-specific workflows and perceived system constraints.

^b Missed: patient was eligible to complete the DCA but was not approached.

^c Incomplete: patient started the DCA but did not complete enough of the test to produce a score.

Because this was a pragmatic study, some patients were approached more than once to complete the DCA if they had >1 encounter during the 12-month study period or had screened borderline on their first assessment. Therefore, the 1,808 assessments represent 1,722 unique patients; 76 completed the DCA 2 times, and 5 completed it 3 times. For these patients, their first completed assessment with a score was used to summarize patient characteristics by DCA result.

A comparison of eligible patients who completed a DCA vs those who did not showed that those who completed a DCA were significantly younger (73.5 [SD 6] years vs 74.7 [7.4] years; P <.001), lived in neighborhoods with lower Area Deprivation Index (51 vs 54; P <.05), and varied by race (P <.001) (Table 2). After controlling for clinic, age, and race, significant differences remained for those who completed a DCA vs those who did not (Table 3).

Among the patients who completed a DCA, just under one-half (762; 44.3%) were unimpaired, 628 (36.5%) were borderline, and 236 (13.7%) were impaired (Table 4). An additional 96 patients (5.6%) completed the DCA, but the data elements needed to score the results, such as audio and video data collected during administration, were not of adequate quality or quantity for analysis and were considered inconclusive.

As noted in Table 4, statistically significant sociodemographic and clinical differences were found between patients who screened not impaired, borderline, and impaired. Patients with impaired or borderline results were significantly older (76.2 [7.2] years vs 74.2 [6.2] years vs 72.1 [5.0] years; P <.001), had fewer years of education, were less likely to report their race as White, were more likely to live in neighborhoods with a higher Area Deprivation Index, and were more likely to have a higher Charlson Comorbidity Index score. In addition, patients whose DCA indicated cognitive impairment were more likely to report being a current or former smoker and/or have a diagnosis of hypertension, chronic kidney disease, diabetes, congestive heart failure, or obesity.

All usual standard of care follow-up options for diagnostic assessment, including orders for laboratory work, imaging, and referrals to specialists, were available to PCPs for patients who were screened. In addition, PCPs had the ability to refer patients with a DCA result of impaired or borderline to the BHN for additional assessment. Orders and referrals for testing and specialists were measured within 90 days after the first completed DCA. Among the 236 patients with a DCA result of impaired, 85 (36%) had a laboratory order for a vitamin B12 test, 84 (35.6%) had an order for a TSH test, and 40 (16.9%) had an order for any imaging of the head or neck (Table 5). Twenty-eight (11.9%) were referred to neurology, 8 (3.4%) to geriatrics, 5 (2.1%) to neuropsychology, and 0 to psychiatry. One hundred and forty-eight (62.7%) of the patients who screened as impaired were referred to the BHN. Five (2.1%) patients received a new diagnosis of ADRD, and 12 (5.1%) received a new diagnosis of MCI in their EHR. Six (2.5%) were started on a new prescription for donepezil, galantamine, rivastigmine, tacrine, memantine, or memantine/donepezil combination therapy.

Among the 628 patients with a DCA result of borderline, 139 (22.1%) had a laboratory order for a vitamin B12 test, 188 (29.9%) had an order for a TSH test, and 48 (7.6%) had an order for any imaging of the head or neck. Twenty-five (4%) were referred to neurology, 4 (0.6%) to geriatrics, 3 (0.5%) to neuropsychology, and 1 (0.2%) to psychiatry. Two hundred and ninety-nine (47.6%) of the patients who screened as borderline were referred to the BHN. Four (0.6%) patients received a new diagnosis of ADRD, and 9 (1.4%) received a new diagnosis of MCI in their EHR. Three (0.5%) were started on a new prescription for donepezil, galantamine, rivastigmine, tacrine, memantine, or memantine/donepezil combination therapy.

DISCUSSION

Primary care clinicians are at the front lines of identifying people who may have cognitive impairment due to MCI or ADRD. The evidence supporting early detection of ADRD in primary care is growing. For example, there are models of care that show improved patient and caregiver outcomes with early detection,³⁵ and the window of eligibility for new disease-modifying therapies is narrow—and for patients in the earliest stages of cognitive impairment.^6,24 In this large, pragmatic clinical demonstration project in 7 diverse primary care clinics, we found that during all primary encounters for patients aged ≥65 years, a cognitive assessment was completed 11% of the time. This measure of feasibility was lower than we hypothesized, especially given the number of encounters for patients aged ≥65 years and who did not have a diagnosis of ADRD. In instances when eligible patients were offered a DCA as part of the visit, they refused only 22% of the time, indicating a willingness to undergo cognitive assessment. Notably, this level of patient refusal for cognitive assessment is lower than reported values from randomized controlled trials in primary care settings that require informed consent to undergo routine cognitive evaluation. In addition, this was a clinical demonstration project of population-level screening, for which all patients were eligible, and the DCA cost was covered by the project. It is unknown how many patients would have declined a DCA if they had a copay or out-of-pocket cost for the DCA, or if the refusal rate would have been lower if a risk-stratification method was used to identity patients with subjective complaints or others who are most at risk of having undetected cognitive impairment.

Clinician engagement was much lower than anticipated, with greater numbers of clinicians declining to have their patient complete a DCA or by establishing workflows that deemed some encounters out of scope for screening. These results imply that at least one-third of eligible patients who are approached are willing to undergo a DCA if it is built into their routine primary care, yet clinicians have hesitation regarding whether the assessment can be incorporated into existing workflows within the scheduled visit.

We used agile science as the methodology for implementation and included input from clinicians, patients, and other clinic stakeholders regarding what screening assessment to use and what adaptations to clinic workflows would be most conducive to implementing the screening. A guiding theory for this approach is that for any implementation to be successful, there must be demand—by the patient, clinician, or system. Given that there were 19.5% of encounters that the clinician declined to have the patient screened, and 35.7% of encounters that the clinics deemed to be out of scope based on the clinic-specific workflows that they established, there appears to have been tension between the potential benefits of early detection for patients and that of a primary care system that is unprepared for widespread implementation of cognitive assessments in routine workflows. Details regarding the barriers to implementation are noted in our companion study,²⁶ yet it is worth noting that the study team was actively engaged with the clinics throughout the 12-month project, which is not consistent with a typical initiative introduced into a primary care setting. This resulted in the team being highly responsive regarding adapting how the screening was implemented and responding to clinicians’ concerns (eg, creating the role of the BHN to assist clinicians with follow-up pathways after a DCA result of borderline or impaired). Completion numbers might therefore have been greater if this initiative were rolled out as standard of care, with payment to the practices, and aligned with primary care quality measures for older adults. Demand for widespread cognitive assessments will likely not increase until detection, as well as accessible evidence-based diagnostic assessments and longitudinal care for those with a diagnosis, are supported by payers and as a measure of quality primary care for older adults.

This study’s finding of 13.7% of patients testing as impaired is slightly greater than estimated rates of ADRD in populations aged ≥65 years. The most recent Alzheimer’s Association Facts and Figures⁵ report estimates that 11% of people aged ≥65 years have ADRD. Given that our results are based on screening alone and not diagnoses, we believe these values are generally comparable. In addition, 68% of Medicare-age Americans have ≥2 comorbidities³⁶ and take an average of 5 prescriptions per year.^37,38 Not surprisingly, patients who were older and who had more cardiometabolic comorbidities were more likely to screen impaired or borderline for cognitive impairment. It is difficult to differentiate from this study how many patients have cognitive impairment as a result of Alzheimer disease or another dementia etiology or other reasons not related to neurodegenerative disease. This was also a motivation for the introduction of the BHN role—to assist PCPs with risk-stratifying patients for follow-up and who might be best served by a referral to a specialist for a more thorough cognitive evaluation or a focus on comorbidity management and deprescribing.

Primary care physicians are ideally positioned to detect cognitive impairment, initiate diagnostic evaluations, and implement ADRD risk-reduction strategies. However, PCPs cannot do it all on their own. Health systems, health system leaders, payers, policy makers, professional societies, and other stakeholders must be involved to address barriers to implementing cognitive assessments and providing follow-up care. Examples of these barriers include lack of guidelines on cognitive assessment tools and workflows that support early detection, no or limited reimbursement for PCPs to assess cognition outside of Medicare annual wellness visits, and insufficient time, skills, and support for diagnostic assessment.

CONCLUSION

Digital cognitive assessments are feasible and able to be implemented as part of routine primary care but require additional resources and adaptations to workflows beyond the existing infrastructure in primary care. Without these embedded changes, clinician and patient demand for cognitive assessments will remain low. Almost one-half of patients scored impaired or borderline for cognitive impairment. Clearly, not all have ADRD, but DCAs have utility for early detection in primary care and could represent the first step of risk identification for patients who could benefit from further assessments, referrals to specialists, disease-modifying therapeutics, or other evidence-based care management interventions to improve patient and family caregiver outcomes.