Article Text

Download PDFPDF

Protocol
Primary care detection of cognitive impairment leveraging health and consumer technologies in underserved US communities: protocol for a pragmatic randomised controlled trial of the MyCog paradigm
  1. Rebecca Lovett1,2,3,
  2. Morgan Bonham1,2,
  3. Julia Yoshino Benavente1,2,
  4. Zahra Hosseinian4,
  5. Greg J Byrne4,
  6. Maria Varela Diaz4,
  7. Michael Bass4,
  8. Lihua Yao4,
  9. Andrei Adin-Cristian5,
  10. Stephanie Batio1,2,
  11. Minjee Kim6,
  12. Amanda Sluis7,
  13. Margaret Moran7,
  14. David R Buchanan7,
  15. Justin Hunt7,
  16. Stephanie R Young4,
  17. Richard Gershon4,
  18. Cindy Nowinski4,
  19. Michael Wolf1,3,4
  1. 1General Internal Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  2. 2Center for Applied Research on Aging, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  3. 3Psychiatry and Behavioral Sciences, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  4. 4Medical Social Sciences, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  5. 5Preventive Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  6. 6Neurology, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA
  7. 7Oak Street Health LLC, Chicago, Illinois, USA
  1. Correspondence to Dr Rebecca Lovett; r-lovett{at}northwestern.edu

Abstract

Introduction Early identification of cognitive impairment (CI), including Alzheimer’s disease and related dementias (ADRD), is a top public health priority. Yet, CI/ADRD is often undetected and underdiagnosed within primary care settings, and in health disparate populations. The MyCog paradigm is an iPad-based, self-administered, validated cognitive assessment based on the National Institutes of Health (NIH) Toolbox Cognition Battery and coupled with clinician decision-support tools that is specifically tailored for CI/ADRD detection within diverse, primary care settings.

Methods and analysis We will conduct a two-arm, primary care practice-randomised (N=24 practices; 45 257 active patients at the proposed practices), pragmatic trial among geographically diverse Oak Street Health sites to test the effectiveness of the MyCog paradigm to improve early detection CI/ADRD among low socioeconomic, black and Hispanic older adults compared with usual care. Participating practices randomised to the intervention arm will impart the MyCog paradigm as a new standard of care over a 3-year implementation period; as the cognitive component for Annual Wellness Visits and for any patient/informant-reported or healthcare provider-suspected cognitive concern. Rates of detected (cognitive test suggesting impairment) and/or diagnosed (relevant International Classification of Diseases-9/10 [ICD-9/10] code) cognitive deficits, impairments or dementias including ADRD will be our primary outcome of study compared between arms. Secondary outcomes will include ADRD severity (ie, mild or later stage), rates of cognitive-related referrals and rates of family member or caregiver involvement in ADRD care planning. We will use generalised linear mixed models to account for clustered study design. Secondary models will adjust for subject, clinic or visit-specific characteristics. We will use mixed-methods approaches to examine fidelity and cost-effectiveness of the MyCog paradigm.

Ethics and dissemination The Institutional Review Board at Advarra has approved the study protocol (Pro00064339). Results will be published in peer-reviewed journals and summaries will be provided to the funders of the study.

Trial registration number NCT05607732.

  • Primary Care
  • Aged
  • Clinical Trial
  • Dementia
  • Electronic Health Records
  • GENERAL MEDICINE (see Internal Medicine)
http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THIS STUDY

  • MyCog is a comprehensive detection strategy; cognitive testing is coupled with detailed clinical decision-making support surrounding further assessment, documentation and diagnosis, and care management.

  • MyCog is specifically designed to fit into resource-constrained, primary care work flows; it leverages increasingly low cost consumer technologies (iPads), is self-administered, takes 5–7 min to complete, and is linked to electronic health records, thereby enabling automated scoring and sharing of results.

  • MyCog can provide longitudinal trend data on cognitive performance, capturing ‘baseline’ functioning to further assess relative, in addition to normative decline.

  • The MyCog study is a ‘real-world’ pragmatic trial that will examine effectiveness and fidelity of the MyCog paradigm among geographically diverse, primarily health disparate populations.

  • Primary study limitations include restriction of data collection to a single health system and use of only electronic health record outcomes.

Introduction

Cognitive impairment (CI), including Alzheimer’s disease and related dementias (ADRD), is significant societal concerns, in both the general population, health disparate populations and predominantly among older adults. Their consequences, especially in ADRD and prodromal mild CI (MCI), are devastating for patients and their families, in terms of declining physical and mental functional status, independence and well-being.1 Prevalence estimates widely vary, from 4% to 62% of older adults who may be affected by some type of CI.2 3 A recent meta-analysis concluded the prevalence of MCI, in particular, widely varies (7–25%; 16.6% average estimated prevalence).4 A range of 487 300–910 000 new cases are estimated per year in the USA; as of 2023 6.7 million Americans 65 and older are living with ADRD.5 These numbers are expected to grow given our ageing population and recent increases in life expectancy estimates.6–9 Furthermore, racial and ethnic disparities have also long been noted in the context of ADRD.10–17 In particular, rates of ADRD are twofold higher among older black adults compared with non-Hispanic/Latino (H/L) whites, whereas H/L older adults are 1.5 times more likely to have an ADRD diagnosis compared with non-H/L whites.6 18 19 More broadly, black and H/L adults, especially at younger ages and in the presence of social inequalities (eg, poverty, low education attainment) and comorbid chronic conditions, are also more likely to report subjective cognitive decline compared with non-H/L whites.20

Despite recent advances in biological therapies that can slow disease progression, there are no definitive means to prevent or treat ADRD, once established. While ADRD currently lacks a cure, some forms of CI may be secondary to a medical condition and potentially reversible.21 For this reason, as well as the positive impact early care in patients with ADRD has in possibly reducing further decline and/or minimising its effects on patients’ daily lives, early detection efforts are needed. In addition to addressing reversible causes and initiating available therapies, with early identification clinicians can also modify existing medical treatment plans to account for impacts on patient understanding and adherence to treatment,22 23 initiate referrals to supportive community-based and social services, and provide information to patients and their families to assist in advance care planning.4 24–26 Early detection of MCI is particularly critical as it is a significant risk factor for progression to ADRD.27 Early detection also has economic benefit; according to a 2019 Alzheimer’s Association report, assuming 88% of ADRD diagnoses could be detected at an early MCI stage, the US healthcare system would save an estimated US$7 trillion in medical and long-term care costs through the case finding of at-risk individuals.8

Primary care is ideally suited for the earlier detection and management of CI/ADRD, given primary care providers’ (PCPs) regular contact and established relationships with patients. Most (94%) older adults ages 65 and older have at least annual contact with ambulatory healthcare professionals, with half of these visits occurring in primary care.28 Indeed, the Center for Medicare and Medicaid Services has supported routine cognitive screening since 2011 by including it as a requirement for the Medicare Annual Wellness Visit (AWV).29 However, primary care practices have diverse patient populations, particularly by age and health status, with limited time and resources to support CI/ADRD detection and any subsequent complex care planning and management.30–40 Clinic visit times are typically brief and care coordination services may be minimal, if any, compared with many subspecialty practices that focus their attention on a single condition.38 41 42 A recent systematic review estimated that 40%–76% of primary care physicians fail to recognise CI and almost 90% of MCI cases are missed.43–46 The issue of under detection is particularly salient in marginalised populations. Among incident cases of either MCI or ADRD, both black and H/L Medicare enrollees are significantly less likely to receive a timely diagnosis (MCI vs ADRD) compared with non-H/L whites.47 Other studies have also found that cases of CI among black and H/L adults are nearly twice as likely to go undetected.48–50

Reasons for the failure to detect CI in primary care are many, including patient/family denial, presence of comorbidities, limited time and resources, as well as training and confidence of clinicians to interpret cognitive test results, make a CI-related diagnosis, or subsequently engage in care plans and management moving forward.45 51 But an underlying problem to early detection efforts is there are no current standard, minimally burdensome, cognitive assessments that are both sensitive to early indicators of impairment and appropriate for use in resource-constrained primary care settings.44 45 52 53 Instead, providers use a variety of methods that are inconsistently applied and may be less effective. Clinicians may solely rely on patients proactively self-reporting concerns when only a third will acknowledge any cognitive problems.54 Clinicians may also conclude no CI exists based on results from a single assessment that fall within a ‘normal’ range, even though performance represents a decline from previous functioning levels.55 56 In addition, currently available detection measures may not target those abilities which show promise as early indicators of CI in laboratory settings, such as intraindividual variability (IIV) in reaction time (RT),57–59 or they may perform differently in health disparate groups due to cultural or linguistic biases, or patient factors (eg, education, test-taking experience).60

The MyCog paradigm

The MyCog assessment is a brief, iPad-based, cognitive assessment designed for primary care settings. MyCog adapted two well-validated measures derived from the National Institutes of Health (NIH) Toolbox for Assessment of Neurological and Behavioural Function: Dimensional Change Card Sort (DCCS) and Picture Sequence Memory (PSM).61 Both PSM and DCCS assess two of the first cognitive domains to show CI. DCCS is an executive function test measuring cognitive flexibility. It requires individuals to sort images based on colour or shape. Scores are based on number correct (accuracy); IIV and RT are also assessed. PSM measures episodic memory via recall of sequentially ordered pictures. A full description of the MyCog assessment’s development and validation has been previously reported.62 The MyCog assessment takes <10 min to complete and has a sensitivity of 79% and specificity of 82% in detecting CI within primary care samples.

The MyCog paradigm consists of the MyCog assessment, in addition to turnkey clinical recommendations if a CI is detected (see figure 1). The MyCog paradigm is linked to the electronic health record (EHR) and can be deployed in one of two ways: (1) whenever a patient, involved family member voices a concern or when a clinician suspects an impairment or (2) as part of routine testing during Medicare AWVs. The MyCog assessment can be completed on an iPad either in the exam room or in the waiting room, depending on individual clinic workflows. As a self-administered assessment, the MyCog test requires only minimal supervision by a medical assistant to provide necessary technical support. Once the MyCog test is completed, results are securely transmitted and autopopulate within discrete, queriable fields within the patient medical record. Depending on the EHR platform used, this can be under a screening tab, labelled ‘cognitive abilities’, and/or within a flow sheet to capture trends within future repeated tests. Trend data can inform physicians of a patient’s relative versus normative decline, further framing any testing decision-making. Both a binary, objective classification of ‘impairment detected or suspected’ or ‘no impairment detected’ will populate in the record, as well as a summary score to further guide the clinician by clarifying the extent to which a patient’s performance falls outside a normal threshold.

If a CI is detected via the MyCog test, structured guidance is provided within the screening tab, along with the test result, to provide specific instructions for clinicians following American Medical Association recommendations.63 This first includes ruling out reversible causes (depression, labs (complete blood count [CBC], electrolytes, glucose, thyroid function, vitamin B12, folate-addressing abnormalities), sensory issues (vision, hearing), polypharmacy (Beer’s list medications), inadequate control of vascular disease factors). If a non-reversible CI or ADRD is suspected, clinicians will then be recommended to counsel patients and families on beneficial behaviours (maintaining physical and mental activity), and safety. If cerebrovascular disease is suspected by history, clinicians can receive additional guidance to determine if brain imaging is needed. Subspecialty referrals (eg, behavioural neurology) at the local level are also provided to confirm diagnosis and initiate care plans, which will then be shared with primary care. After a confirmed diagnosis, local medical, allied health, and non-medical referrals may also be made depending on need (eg, physical, occupational, speech assessments, Alzheimer’s Association dementia care coordination planning, social work). Primary care follow-up will then be recommended every 6 months or earlier if a significant change in status is reported by a patient or a family member. Contacts for involved family members/caregivers will be documented/updated in the medical record; they will also be asked to participate in subsequent patient primary care visits, as well as in local caregiver education services.

If a cognitive concern is reported or suspected, but CI is not detected with the MyCog assessment, clinical recommendations will therefore be to: (1) perform a focused history (eg, changes in cognitive function, functional health status (eg, basic/instrumental activities of daily living, finances, safety), full medication review including over-the-counter medications and supplements, neurological and psychiatric symptom assessment), (2) review MyCog assessment findings with patient and family member and (3) determine patient/family preferences to proceed by following-up and reassessing patient in 6 months, or based on degree of concern, consider referral for further neuropsychological assessment.

Study aims and hypotheses

The MyCog trial will test the effectiveness and fidelity of the MyCog paradigm to improve early detection and management of CI/ADRD in primary care settings serving health disparate patient populations. We will partner with a national PCP (Oak Street Health) and conduct a two-arm, clinic-randomised, ‘real-world’ pragmatic trial comparing MyCog to usual care, focusing on populations experiencing CI/ADRD disparities (black, H/L and low socioeconomic status (SES) adults). Our project is guided by the Reach-Effectiveness-Adoption-Implementation-Maintenance (RE-AIM) framework.64 The study aims and hypotheses are listed in table 1.

Table 1

The MyCog trial study aims and hypotheses

Methods and analysis

Study design

Oak Street Health will serve as study performance sites. Oak Street Health is a national network of primary care centres servicing a diverse population of adults 65 and older across the USA, and predominantly Medicare enrollees. Participating Oak Street Health practices (N=24 practices; 45 257 active patients at the proposed practices) will be randomised using parallel assignment to receive either the MyCog paradigm or usual care. Stratified randomisation will be used to achieve balance between the study arms with regard to practice size; specifically, the number of patients seen per site (<1500, 1500–2500, 2500+ eligible patients). Within these three strata, clinics will be randomised at a 1:1 allocation rate using random blocks of 2 or 4 (depending on strata size) to guarantee balance by practice size. Practices randomised to the intervention (n=12) will incorporate the MyCog paradigm into existing workflows as a new standard of care for a 3-year implementation period for all cognitive testing needs.Patients and care partners from each intervention site will be interviewed to guide implementation, and clinical staff will be trained on operating protocols and procedures prior to trial launch. Planned study start date is 1 September 2023 and planned study end date is 7 July 2023.

Sample

Inclusion criteria will be patients 45 years and older who have (1) been seen by an Oak Street health clinician affiliated with one of the 24 enrolled practices, (2) had at least one clinic visit (routine or AWV) during the 3-year study period and (3) not been diagnosed previously with cognitive deficits, impairments or dementias. Children, adolescents and adults younger than 45 will be excluded as CI in these populations is often due to differences other than age-related changes. We will target four geographically diverse states where Oak Street Health operates: Illinois, Michigan, North Carolina and Pennsylvania. Data from the Oak Street Health EHR data warehouse will be retrieved from all eligible patients at participating clinics, regardless of study arm.

Intervention arm

The MyCog paradigm, described previously, will be implemented in all practices randomised to the intervention arm. The MyCog paradigm will be used both during Medicare AWVs, and if a cognitive concern is suspected.

Usual care arm

Standard Oak Street Health practices surrounding cognitive testing will continue in clinics randomised to usual care. Standard procedures typically consist of the use of the standardised Mini-Cog assessment, both for routine testing during Medicare AWVs and if a concern arises. Oak Street practices vary by clinician in terms of making referrals, how results are documented (eg, if diagnosis made (or wait for referral findings), what diagnosis code (International Classification of Diseases-9/10 [ICD-9/10]), placement in problem list or visit diagnosis), and any follow-up plans (eg, increase future visit times, involve family, additional referrals, deprescribing, etc). While we will not make any explicit recommendations to usual care practices with regard to their use of a cognitive assessment, we will ensure that (1) any chosen test is linked to a queriable data field in the EHR, which will allow the clinician to record the results of the test as discrete data and that (2) providers receive a compiled list of local medical and non-medical referrals for any detected cases of CI. The Alzheimer’s Association recommendations for early detection efforts among primary care practices will also be provided to each clinic’s medical leadership.8

Primary outcome

Rates of detected and/or diagnosed cognitive deficits, impairments or dementias including ADRD will be our primary study outcome compared between arms. This will be operationalised as either results of any administered cognitive test suggesting impairment (‘detected’) or having any relevant ICD-9/10 classification code recorded in a patient’s record after the trial launch date and throughout the follow-up observation period (‘diagnosis’). See table 2 for the list of pertinent codes.

Table 2

Pertinent ICD codes for cognitive deficits and impairment

Secondary outcomes

Diagnosis severity, reflected as either mild (eg, MCI) or later stage, moderate/severe impairments or ADRD ICD-10 codes, will also be examined. As it is possible that results from a cognitive test indicating a deficit or impairment may not necessarily translate to an ICD-10 diagnosis, we will examine cases where cognitive assessments were administered, results indicated a deficit or impairment, but no diagnosis was made. We will also conduct text searches leveraging Natural Language Processing (NLP) algorithms to extract and investigate unstructured information from clinical notes pertaining to administration of a cognitive assessment and its results.65 For detected but not diagnosed cases where clinical notes may qualify the extent of cognitive deficit, impairment, test scores will aid this classification via chart review.

In addition, the initiation of cognitive-related referrals (medical (eg, behavioural neurology, neuropsychiatry, neuropsychology, geriatrics), allied health (eg, Occupational Therapy [OT], Physical Therapy [PT], Speech Language Therapy [SLP], social work, clinical psychology) and non-medical services (eg, Alzheimer’s Association, dementia care coordination, home health aide, respite)) will also be investigated. Recommended family member or ‘caregiver’ involvement in care planning, management of health and chronic conditions and at future visits will also be investigated and determined via text search/NLP methods applied in a review of clinic notes among the charts of patients who, after receiving cognitive testing, had a detected and/or diagnosed CI or dementia (notes dating after the assessment).

Exploratory outcome

Rates of cognitive testing will be captured for each year of the trial’s implementation via billing and EHR queries of associated EHR data and flowsheets that document discrete data from structured cognitive assessments (eg, cognitive, functional abilities tab), among other ordered tests. We will also be able to determine whether the cognitive assessment was administered as part of the AWV or not (eg, possible patient or relative-reported concern or clinician concern).

Analysis plan

Analyses, per aim, are described below.

Aim 1: test the effectiveness of the MyCog paradigm to improve early detection of CI and dementias among low SES, black and H/L older adults

Using patients seen at the participating clinics during the 3-year observation period, we will use generalised linear mixed models (GLMM) with a logit link to model the probability of diagnosis of CI, ADRD or MCI specifically. Models will include a fixed effect of strata, and random effect of practice site to account for the intracluster correlation (ICC) within site. In primary models, we will compare the main effect of intervention versus usual care. In secondary models, we will adjust for subject (age, sex and gender, race, ethnicity, comorbidities, language spoken) and site (location, practice size) or visit (routine, problem-specific, AWV) specific characteristics. Per updated Consolidated Standards of Reporting Trials (CONSORT) guidelines,66 we will also present estimates of ICCs.

Aim 2: investigate the presence of disparities in early detection of CI, its diagnosis and rate of referrals by race and ethnicity

We will test rates of diagnoses and referrals between MyCog versus usual care within black race and H/L ethnicity and estimate the effect sizes of any disparities seen between these groups and non-H/L whites. We will again employ GLMMs to account for the clustered study design. Initial models to compare MyCog with usual care will only be fit with the racial or ethnic groups of interest. As the samples for race and ethnicity may overlap, we will fit two separate additional sets of models; first comparing blacks to non-H/L whites, and then comparing H/L to non-H/L whites. To examine racial and ethnic disparities, we will graph the least-square mean rates from these models to visually examine if the difference in rates seen for clinics employing MyCog is less than the difference in rates seen for clinics employing usual care. We will also present estimates of the ICCs overall and within racial and ethnic groups.

Aim 3: determine the fidelity and reliability of MyCog and identify any patient, caregiver, clinician and/or health system barriers to its optimal, sustained implementation

Any instances where the MyCog app does not properly function, does not link and populate test results within the EHR (both in screening tab, flow sheet), or the patient does not fully complete the assessment will be tracked on the app and EHR and documented by practice site. From the EHR, we will compare rates of cognitive testing (all visits, during AWVs, at routine or problem-based visits) between study arms. As in aims 1–2, the outcome is binary at the patient level. We will use GLMMs with a logit link accounting for clinic-level randomisation and an indicator for treatment group. Other aspects of intervention fidelity available for study include: (1) average time for patient completion of the MyCog assessment; (2) frequency, variability of ICD codes added to the problem list or in the visit summary for detected CI; (3) time between a detected CI and follow-up visit for further assessment, per Medicare and patterns of follow-up afterwards and (4) patient, family adherence to medical or non-medical referrals and follow-up recommendations. For process outcomes 2–4, we will compare between study arms. Appropriate bivariate analyses will be performed to examine if any patient factors (eg, age, sex, race, ethnicity, comorbidity) are associated with receipt of a cognitive test, a diagnosis if CI detected, referral and/or follow-up visit. Intervention fidelity differences will also be assessed across practices and providers.

At each intervention site (n=12), we will recruit clinicians (n=4), medical assistants (n=2) and a practice administrator (n=1 per site; N=84 participants from all sites) to 1-hour, postimplementation, debriefing discussion sessions either in-person or via Zoom video conference.67 These will be audiorecorded and transcribed for review and analyses. Discussions will be held separately per profession. To aid discussions, a summary of the quantitative implementation fidelity data, per above, will be shared and tailored to site. Suggestions for additional modifications to the MyCog protocol will be solicited. Mixed methods will be employed using a convergent parallel design to evaluate intervention fidelity and effectiveness. Both quantitative and qualitative analyses will be conducted separately and integrated for side-by-side data comparisons. This will be guided by the normalisation process theory,68 which examines the implementation of innovations into practice. Codebooks will be created to analyse complementary qualitative data from clinician and staff interviews with codes drawn from interview guides and from participants’ words. NVivo will be used to organise data. We will use an iterative approach to identify and group emergent themes that address patient/family, provider or clinic/health system. The use of matrices will help compare and contrast themes across these groups.69 70 Participant quotes will be identified to highlight themes. Attention will be paid to facilitating and impeding factors. An additional layer within impeding factors would be if factors are modifiable, and under facilitating factors, if they are replicable.

Aim 4: assess the cost-effectiveness of the MyCog paradigm from a primary care perspective

Costs of implementing MyCog will be based primarily on time logs and using wage rates to approximate costs from a clinic staffing perspective. Hours spent linking the iPad with the EHR, training clinical staff on when to administer the test, and assisting patients in logging in to MyCog will be recorded weekly along with other technical support time needed to solve patient related issues or concerns. There will also be personnel time involved in programming the EHR to provide clinical guidance to initiate appropriate referrals, creating tailored protocols, training staff on using MyCog results to improve care, establishing care plans and involving caregivers. Actual wages of the personnel involved in the workflow protocol, as well as US average wages for similar occupational titles from the US Bureau of Labor Statistics (BLS) will be considered. We will also have patient time spent in clinic for which we can include a modified societal perspective analysis that estimates cost of their time based on average US wages. Sensitivity analyses with relevant other occupational titles of staff that could potentially be asked to manage the MyCog paradigm (as staffing may vary across clinics) will also be conducted and there will be sensitivity analyses using the 25th and 75th quartiles of wages from the BLS. Other than personnel, patient clinic time, programming, and the costs of maintaining MyCog (eg, charging, sanitising, iOS updates, iPad replacements)—the requisite technology involves the purchase of iPads for which we will have direct market-based cost estimates. The additional per patient costs around using MyCog will be combined with differences in average per person effects to assess the incremental cost-effectiveness of MyCog relative to usual care. Specifically, the ratio of (Cmycog) and (Emycog-Esoc) where Cmycog represents costs of running MyCog, which will be calculated both with and without costs of patient time, and Emycog the effects and Esoc for usual care, measured in aim 1, will be determined and variance in those estimates examined in sensitivity analyses.

Power considerations

For aim 1, unpublished estimates of any diagnosis of CD or CI within the Oak Street Health network are 15%, and estimates of MCI are 3%. Additionally, ICC rates are expected to be 0.03 and 0.02, respectively. Assuming an average number of patients undergoing cognitive tests at each practice is 1000, a sample of 24 centres randomised 1:1 provides 80% power to detect increases in these rates to 22.8% and 6.5%, respectively, at a type I error rate of 5%. These effect sizes denote a meaningful increase in detection; while the overall rates of detection remain below the rate of what we believe to be true CI (ie, our increased yield will likely not be simply individuals with false positives on these tests). For aim 2, using similar diagnosis estimates and assuming that black and H/L populations are approximately 38%–40% of the patient populations at the targeted Oak Street Health practices enrolled in the study (and 20% being non-H/L white), testing ~400 patients at each of 24 centres provides 80% power to detect increases to 26.5% and 6.7%, respectively, within intervention clinics, compared with usual care rates of 15% and 3% for detecting and CI and MCI, respectively. As comparisons of the differences in detection between race and ethnicity will be made using data both within the same clusters and across clusters, and the ICC may differ by race or ethnicity, and potentially within clinics, power calculations to test disparities would be speculative. Additionally, there is less interest in determining if there is a ‘statistically significant’ disparity than in estimating the actual size of the disparities which are observed. All power calculations were made using PASS2021 (Kaysville, UT).

Patient and public involvement

Patients and the public were not directly involved in the design or conduct of this research study; however, stakeholder input will be used to guide intervention implementation.

Ethics and dissemination

The study was approved by the Advarra Institutional Review Board (Pro00064339) and preregistered on ClinicalTrials.gov (NCT05607732). A data safety and monitoring board including a biostatistician and up to five researchers with relevant expertise will be formed to review and approve study methods and analyses and monitor ongoing study activities at least twice annually. Study principal investigators will be responsible for ensuring participants’ safety on a daily basis, and any adverse event reporting to Advarra. Aims 1, 2 and 4 compare interventions implemented at a population (ie, practice clinics) level as standards of care for the study duration; therefore, patients will not be prospectively recruited into the study and individual informed consent will not be obtained. Written informed consent will be obtained by trained personnel for all focus group participants prior to aim 3 activities. We do not anticipate any participants withdrawing from the study, as this is a practice-level, pragmatic trial. Patient-level data will be analysed from the EHR, but will be examined in aggregate form, whereas no member of the study team will have any identifiable patient data; the risk of individuals’ loss of privacy or confidentiality does not apply. All data will be stored on mainframe servers, and terminals will be password protected. Results will be published in peer-reviewed journals and summaries will be provided to the funders of the study. Study investigators will have access to the final trial dataset, and the study dataset will be made available on reasonable request.

Ethics statements

Patient consent for publication

References

Footnotes

  • Twitter @moonk321

  • Contributors MW, CN and RG conceptualised the study. MW obtained funding and AA-C, SB and LY wrote the statistical analysis plan. MBonham, ZH, JYB, MK, RL, MBass, MVD, RG, CN and MW contributed to the final selection of measurements. MBonham, MM and AS manage the project and SRY, GJB, JH, DRB, JYB, CN and MW supervise the project administration. RL wrote the original draft and revised the paper. All authors reviewed and edited initial drafts of the manuscript, and approved the final version.

  • Funding Research reported in this publication is supported by the National Institute of Neurological Disorders and Stroke, (U01NS105562).

  • Disclaimer The funding agencies played no role in the study design, collection of data, analysis or interpretation of data.

  • Competing interests MW reports grants from the NIH, Gordon and Betty Moore Foundation, and Eli Lilly, and personal fees from Pfizer, Sanofi, Luto UK, University of Westminster, Lundbeck and GlaxoSmithKline outside the submitted work. All other authors report no conflicts of interest.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; peer reviewed for ethical and funding approval prior to submission.