Acceptable standards for clinic‐based digital cognitive assessments: Recommendations from the Global CEO Initiative on Alzheimer's Disease

Abstract

The rising prevalence of mild cognitive impairment (MCI) and dementia, combined with persistent underdiagnosis, is driving an increased need for scalable cognitive assessment tools. Digital cognitive assessments (DCAs) offer a promising solution by addressing longstanding barriers to routine cognitive testing and diagnosis. However, variations in performance and intended use have created confusion about their clinical applications and utility. The Global CEO Initiative on Alzheimer's Disease convened a DCA Workgroup to define the preferred characteristics of DCAs to meet the needs of patients, health care providers, and regulators for three clinical contexts: (1) initial detection of cognitive impairment, (2) diagnostic support for MCI and dementia, and (3) characterization of cognitive profiles to support identifying etiology. In the near term, ensuring that DCAs meet or exceed the performance of non‐digital tools is a priority. DCAs must be validated in the intended use population with well‐characterized study samples and inclusive designs.

1. INTRODUCTION

The prevalence of dementia and mild cognitive impairment (MCI) is rising (see Prefix Table: Key terms and definitions). Currently, ≈ 10% of older adults are affected by Alzheimer's disease and related dementias (ADRD), while over 20% are living with MCI. ¹ , ² Projections indicate that by 2060, one in five older adults will be living with ADRD, exacerbating the already substantial individual and societal burdens associated with these diseases. ³ , ⁴

Prefix: Key terms and definitions

Term	Definition
Dementia	Clinical syndrome characterized by progressive cognitive decline, affecting memory, executive functions, language, or other thinking abilities to a degree that interferes with a person's daily functioning.
Digital cognitive assessment (DCA)	A test administered on a tablet, computer, or smartphone for supervised use in clinical settings. It may be a novel tool or derived from standard neuropsychological tests and is designed to assess established cognitive domains such as memory, language, and others.
Health care professional (HCP)	A licensed and/or certified individual who provides health care services (e.g., physicians, physician assistants, nurse practitioners, nurses, and allied health professionals).
Mild cognitive impairment (MCI)	Objective cognitive impairment and preserved independence in functional abilities (National Institute on Aging–Alzheimer's Association [NIA‐AA]).
Negative predictive value (NPV)	The likelihood that an individual with a negative result does not have the condition or characteristic being tested, according to a reference standard.
Positive predictive value (PPV)	The likelihood that an individual with a positive result has the condition or characteristic being tested, according to a reference standard.
Target Product Profile (TPP)	A strategic planning tool that defines the minimum and/or preferred characteristics of a product to meet the needs of patients, health care providers, and regulators. TPPs are designed to align research, industry, regulatory, and consumer priorities; guide product development decisions; and support clinicians in evaluating emerging therapies.
U.S. Medicare Annual Wellness Visit (AWV)	An annual clinical visit covered by Medicare Part B that focuses on preventive care, including a health risk assessment and the development or update of a personalized prevention plan.

Despite their increasing prevalence, ADRD remain widely underdiagnosed or diagnosed late in the disease course. ⁵ , ⁶ In the United States, more than 60% of individuals with ADRD may go undiagnosed, and detection rates drop as low as 10% in low‐ and middle‐income countries. ⁶ Detection rates for MCI are even lower, estimated at less than 10% in the United States and other countries. ⁷ , ⁸ , ⁹ Many older adults experience subjective cognitive concerns but may hesitate to raise them with health care professionals (HCPs) due to stigma, whereas others who seek evaluation often face significant delays in assessment, diagnosis, and care. ¹⁰ , ¹¹ , ¹²

Until recently, specific guidance on how and when to conduct cognitive screening have been limited. ¹³ , ¹⁴ New guidelines and workflows focused on early detection of cognitive impairment could provide opportunities to identify and address treatable causes and modifiable risk factors, offer disease education, and develop treatment plans with patients and their care partners. ¹⁵ , ¹⁶ In addition, timely diagnosis facilitates access to approved therapies, clinical trials, and other care, such as the new U.S. Medicare GUIDE services in the United States. ¹⁷ Despite these benefits, only about a third of older adults 65 years of age or older report having been asked about memory concerns or had cognitive screening, whereas less than half of primary care providers say it is their standard protocol to conduct cognitive testing for patients age 65 or older. ¹⁸ , ¹⁹

Several structural and systemic barriers limit early and accurate cognitive assessment. In primary care, clinicians face time constraints and competing priorities, leaving limited opportunities for cognitive testing, interpretation, and disclosure of results. ²⁰ The absence of standardized guidelines and clinical workflows further complicates the use of brief cognitive assessments. In addition, many providers lack adequate training in cognitive evaluation and dementia care and often encounter challenges related to insurance reimbursement for these services. ¹⁵ , ¹⁹ Specialty care is also constrained, largely due to the limited availability of dementia specialists (e.g., neurologists, geriatricians, neuropsychologists, and geriatric psychiatrists), as well as competing health care system priorities and incentive structures. ²¹ These and numerous other challenges lead to delayed referrals and missed opportunities for early intervention.

Brief, conventional, paper‐based cognitive assessments, such as the Mini‐Mental State Examination (MMSE), Saint Louis University Mental Status Exam (SLUMS), Montreal Cognitive Assessment (MoCA), and Mini‐Cog, represent a spectrum of efforts to improve the detection of cognitive impairment and dementia in non‐specialist settings. ²² , ²³ , ²⁴ , ²⁵ However, all are susceptible to examiner bias and errors in response recording, and most have limited sensitivity to detect MCI and limited validity in diverse or sensory‐impaired populations. ²⁶ , ²⁷ , ²⁸ In addition, conventional cognitive assessments are not widely integrated or tracked in electronic health record (EHR) systems, increasing administrative burdens on clinicians and creating another obstacle to testing. ²²

RESEARCH IN CONTEXT

1. Systematic review: A multidisciplinary expert workgroup convened to define use cases, test characteristics, and minimum acceptable performance standards for digital cognitive assessments (DCAs) intended for clinical use. Representatives from health care, industry, academia, policy, and patient advocacy sectors were included in the drafting of the consensus recommendations.

2. Interpretation: The acceptable standards established by the workgroup are described for three DCA profiles (Detection DCA, Diagnostic Aid DCA, and Profile Characterization Aid DCA) based on clinical context of use and need. These expert recommendations serve as general guidelines for DCA developers, as well as health care professionals and health system decision‐makers evaluating DCAs for clinical adoption

3. Future directions: Integrating DCAs into routine clinical practice may improve the feasibility of early detection of cognitive impairment, improve diagnostic accuracy, and facilitate timely intervention. Multi‐stakeholder engagement is required to advance a scientifically based, patient‐focused environment that enables the effective and efficient use of DCAs across health care settings.

Evolving insights into neurodegenerative pathophysiology and advances in treatment have shifted clinical and research priorities toward earlier detection of MCI and driven a growing need for tools sensitive to these stages. ²⁹ , ³⁰ Comprehensive, multi‐domain cognitive evaluations conducted by neuropsychologists, who are experts in test selection, interpretation, and clinical synthesis, have better sensitivity for detecting MCI, but come with additional challenges. Many regions have a shortage of neuropsychologists, referral wait times are often lengthy, and associated costs can be high. ³¹

Digital cognitive assessments (DCAs) may offer an alternative approach to overcome some of the challenges associated with conventional tests, especially those related to administration burden and biases in scoring variability and interpretation. DCAs may feature more immediate, consistent, and automated scoring, reducing training requirements and ensuring greater standardization. ³² , ³³ , ³⁴ In addition, DCAs have the potential to include controlled variations to minimize practice effects for repeat testing, capture more nuanced cognitive performance data (e.g., latency and process), provide real‐time normative comparisons, and integrate with EHR systems for easier documentation and tracking. ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ With careful development and validation, such features could potentially simplify cognitive testing and improve diagnostic accuracy and scalability in primary care. ³⁴ , ⁴⁰ , ⁴¹ Despite their potential advantages, it is important to acknowledge that the integration of DCAs into clinical practice will pose new and different challenges, including factors such as software requirements and cost, digital literacy, test validity and reliability, and data privacy considerations. ³² , ³³ , ⁴² , ⁴³

Many DCAs are currently available or in development for research and increasingly for clinical use. ⁴⁰ , ⁴⁴ Given their rapid proliferation, there is a need for expert consensus on the clinical utility of DCAs prior to widespread clinical implementation. In this Consensus Statement, we present recommendations from the Global CEO Initiative on Alzheimer's Disease (CEOi) DCA Workgroup on the acceptable standards for supervised, in‐clinic DCAs in the United States, as well as the rationale for these recommendations for patients, providers, insurance companies, and test developers. These recommendations build on the prior work of CEOi's Blood‐Based Biomarker Workgroup, which supported the integration of blood tests for Alzheimer's disease (AD) into clinical practice. ⁴⁵

2. METHODS

CEOi is a multistakeholder partnership of experts working collaboratively to address major challenges in ADRD. In 2024, CEOi launched the DCA Workgroup to prepare stakeholders for the widespread adoption of DCAs. The DCA Workgroup consists of academics who are experts in DCA development and validation, clinicians with experience implementing DCAs in practice, diagnostics companies that are developing DCA tests, pharmaceutical companies that expect DCA tests to be useful in treatment pathways, non‐profit organizations supporting the implementation of DCAs in clinical care, and patient advocacy groups that hope to improve care for people with MCI and ADRD.

Within the DCA Workgroup, a workstream was formed to define use cases, test characteristics, and minimum acceptable performance of DCAs for clinical use. The workstream aimed to establish recommendations for any DCA rather than evaluating individual tests. The workstream was co‐led by a behavioral neurologist and a neuropsychologist with expertise in novel assessment and diagnostic approaches to characterizing ADRD in research and practice (Christopher R. Butler and Louisa I. Thompson). A core team was assembled, consisting of physicians, neuropsychologists, and neuroscientists who routinely incorporate cognitive assessments in the clinical diagnosis of MCI and ADRD or conduct research on DCAs in diverse populations, and who also have expertise in cognitive detection and care (A. M. Barrett, Barak Gaster, Dustin B. Hammers, and Sol Fittipaldi). The core team represented three countries (the United Kingdom, Chile, and the United States) and multiple U.S. regions. Both the co‐leaders and core team served voluntarily and without financial compensation.

The DCA Workgroup first defined use cases to support the diagnostic process; recommended test characteristics for DCAs; and gathered feedback from stakeholders in academia, medicine, industry, private foundations, and patient advocacy. The scope of work was limited to consideration of clinically supervised DCAs, excluding self‐administered and other remote, unsupervised procedures. The core team completed a review of existing literature and practices related to the clinical use of DCAs, comparative performance of non‐digital cognitive assessments, as well as validation and reference standards for DCAs across diverse patient populations. The performance of paper‐based cognitive assessments and emerging DCAs was used to inform the minimum acceptable performance of DCAs for clinical use. The workgroup determined that an appropriate benchmark for DCA performance was an established and thoroughly validated conventional neuropsychological test battery for older adults with sensitivity to MCI. A few examples of primarily open‐access measures from clinical research include the Alzheimer's Disease Neuroimaging Initiative (ADNI), National Alzheimer's Coordinating Center Uniform Data Set (UDS), or Preclinical Alzheimer's Cognitive Composite (PACC) batteries (see Table S1 for details). ⁴⁶ , ⁴⁷ , ⁴⁸ Additional licensed clinical measures (e.g., Weschler Memory Scale) exist and may also be appropriate to consider.

The DCA Workgroup developed recommendations for the use of DCAs in three clinical contexts within primary and specialty care and refined them iteratively before achieving final unanimous consensus. Although the recommendations are primarily based on the U.S. health care system, the international composition of the workgroup helped ensure adaptability for national variations in health care organizations, where possible. The core team used the recommendations to draft target characteristics for DCAs, following a format used for other diagnostic tools and previously published recommendations from the CEOi Blood‐Based Biomarker Workgroup. ⁴⁵ The co‐leaders presented the draft of target characteristics to stakeholders and incorporated their feedback into the final recommendations.

3. CLINICAL CONTEXTS OF USE FOR DCAS

The three clinical contexts of use for DCAs are to: (1) detect any possible cognitive impairment as an initial first step to gauge whether further evaluation is needed (Detection DCA), (2) aid in the diagnosis of a cognitive syndrome (i.e., MCI or dementia) by providing an assessment tool that characterizes the degree of impairment (Diagnostic Aid DCA), and (3) aid in differential considerations of underlying disease etiology by providing information on the cognitive profile (Profile Characterization Aid DCA). Tools in each of these contexts serve distinct functions in the clinical pathway to address key questions: (1) Is there any indication of impairment? (Detection DCA), (2) What is the degree of impairment? (Diagnostic Aid DCA), and (3) What is the pattern of impairment? (Profile Characterization Aid DCA). It is important to note that all contexts require consideration of additional workup (e.g., assessment of daily function and laboratory results) and medical history review. Together, DCAs in these contexts provide an integrated approach to cognitive assessment, supporting clinicians in making informed decisions about patient management.

4. TARGET CHARACTERISTICS

The target characteristics for DCAs in each clinical context of use should serve as general guidelines for manufacturers developing DCAs and for HCPs and health system decision‐makers evaluating DCAs for clinical practice.

4.1. DCAs for detecting cognitive impairment

DCAs used for the detection of cognitive impairment are intended to identify individuals with any possible cognitive impairment for further evaluation in the absence of a recognized concern (Table 1). A comparable non‐digital tool would be the Mini‐Cog, Mini Mental Status Exam (MMSE), or a similar test (i.e., very brief, single cutoff for impairment) commonly used in primary care, which DCAs should aim to improve upon. ⁴⁹ , ⁵⁰ This DCA targets adults 65 years or older without a recognized cognitive concern (e.g., given routinely as part of a Medicare Annual Wellness Visit) or younger adults with a family history of ADRD and/or known genetic risk factors. As advances in ADRD biomarkers and treatments move toward earlier intervention, the target population for this DCA may, in some contexts, expand to younger individuals to facilitate early detection.

TABLE 1.

Target characteristics for DCAs in clinical care.

Target characteristics
	DCAs for detecting cognitive impairment (Detection DCA)	DCAs to aid in the diagnosis of MCI or dementia (Diagnostic Aid DCA)	DCAs for cognitive profile characterization to aid etiological workup (Profile Characterization Aid DCA)
What is the test measuring? *	Single or multidomain cognitive function to discriminate cognitive impairment from cognitively unimpaired.	Multi‐domain cognitive function to characterize the severity of impairment, including distinguishing MCI from dementia.	Cognitive function in all cognitive domains to characterize the profile of affected domains.
Target population	Older adult (65 years of age or older) without a recognized cognitive concern. May be initiated in younger individuals with a family history of dementia and/or known genetic risk factors.	Individual who tested positive on a Detection DCA (or similar tool). Individual with a recognized cognitive concern. †	Individual with a diagnosis of MCI or dementia, or whose prior evaluation (including a Diagnostic Aid DCA or similar cognitive assessment) yielded inconclusive results.
Target use settings	Most often used in primary care as part of an evaluation by a PCP or trained HCP. ‡ Infrequently used in specialty care.	Oftentimes used in primary care as part of an evaluation by a PCP. Oftentimes used in specialty care as part of an evaluation by a specialist.	Infrequently used in primary care. § Most often used in specialty care.
Administration time	Short (≈3–5 min)	Medium (≈10–20 min)	≈40 min or more to cover multiple cognitive domains.
Intended use	Abnormal test result indicates concern for cognitive impairment, evaluate further for confirmation. Normal test result indicates lower probability of cognitive impairment, continue to follow and focus on brain health interventions.	Abnormal result ¶ prompts a treatment plan including treatable causes and other interventions to promote brain health; primary care should consider referring to specialty care. Normal result ¶ indicates low likelihood of cognitive impairment, continue to follow and focus on brain health interventions.	Test results will characterize the profile of affected domains to aid in determination of underlying etiology (e.g., MCI due to AD) and assist with selecting appropriate treatment and support plans. Results may also help clarify inconclusive results from prior work up.
Benchmark	Neuropsychological test battery. **	Neuropsychological test battery. **	Neuropsychological test battery. **
Comparative non‐digital tool	Mini‐Cog or similar brief test.	MoCA or similar test.	Testing from specialist neuropsychological evaluation.
Acceptable performance for accuracy	Sensitivity: 80% to detect cognitive impairment. Specificity: 85%	Sensitivity: 85% to detect MCI. Specificity: 90%	Sensitivity: 85% to detect MCI due to AD. # Specificity: 90%

Abbreviations: AD, Alzheimer's disease; DCA, digital cognitive assessment; CT, computed tomography; HCP, health care professional; MCI, mild cognitive impairment; MoCA, Montreal Cognitive Assessment; MRI, magnetic resonance imaging; PCP, primary care physician; TSH, thyroid‐stimulating hormone; B12, Vitamin B12.

^*Specific cognitive domains are not recommended to maximize the flexibility of the target product profiles in guiding the development and use of DCAs across diverse populations and conditions.

^† Concerns reported from individuals, care partners, clinicians, or another informant.

^‡ Routine brain health checks or annual wellness visits for older adults, such as the U.S. Medicare Annual Wellness Visit.

^§ Primary care clinics with infrastructure for advanced cognitive testing, such as access to a neuropsychologist referral for administration and interpretation.

^¶ DCA test result to be interpreted in conjunction with a comprehensive workup. A comprehensive workup should follow best practice guidelines and may include a detailed history, functional and behavioral history, evaluation for treatable causes (e.g., psychological/psychiatric, medication, or medical conditions), labs (e.g., TSH, B12, and syphilis screening), and brain MRI/CT scan as clinically indicated. When determining etiology, biomarker testing may be initiated if clinically indicated.

^# MCI due to AD as an example.

^**A previously validated neuropsychological test battery for older adults (see Table S1 for details).

This DCA will typically be administered and supervised by a primary care HCP. Given the time constraints of primary care visits, this DCA should be a brief (≈3–5 min) assessment measuring single‐ or multi‐domain cognitive function and yielding an estimate of the likelihood of cognitive impairment. An abnormal result warrants further evaluation, including follow‐up with a more in‐depth cognitive assessment, whereas a normal result (low probability of cognitive impairment) may support ongoing monitoring, education, and brain health interventions to mitigate future risk.

DCAs for the detection of cognitive impairment can be integrated into routine preventive care for older adults, alongside brain health conversations, to help normalize cognitive assessment and discussions about cognitive concerns and establish a cognitive assessment baseline for longitudinal monitoring. However, preventive care services remain underutilized, particularly for at‐risk and underserved populations in both the United States and globally. ⁵¹ , ⁵² Routine use of DCAs in this context may also be hindered by limited scientific evidence for the benefits of cognitive screening. In 2020, the U.S. Preventive Services Task Force recommendations concluded that there was insufficient evidence to evaluate the potential clinical benefits or harms of screening for cognitive impairment in older adults. ⁵³ Longitudinal implementation studies with DCAs developed for preventive care settings are needed to track care pathways and patient outcomes following cognitive screening.

Using a neuropsychological test battery as the benchmark (see Table S1 for details), the workgroup recommends a target of 80% sensitivity and 85% specificity for a Detection DCA to identify any cognitive impairment (Table 1). The Detection DCA should outperform most brief conventional assessments (e.g., Mini‐Cog or MMSE) and aim for sensitivity and specificity to detect MCI. ⁵⁴ Although achieving high accuracy may prove challenging with brief assessments, innovative solutions to improve accuracy could be considered. These may include the algorithmic integration of demographic or other risk information (e.g., positive family history of dementia), the use of information from unsupervised elements (e.g., a questionnaire completed during pre‐visit clinical workflows), or the use of artificial intelligence to evaluate other aspects of the patient's cognitive test performance (e.g., patterns of errors or analysis of hand or eye movements) to potentially further refine test accuracy. ³⁹ , ⁵⁵

Although sensitivity and specificity are useful for research, clinical utility is better reflected by positive predictive value (PPV) and negative predictive value (NPV). PPV and NPV provide clinically relevant metrics to evaluate the effectiveness of DCAs to identify true positives and minimize false negatives in real‐world populations. PPV and NPV depend on disease prevalence in the tested population, which may vary widely according to patient characteristics (e.g., age) and HCP characteristics (e.g., specialization). Prevalence studies for the target population should inform the interpretation of DCA results for this context of use.

It is worth noting that in settings with relatively low ADRD or MCI prevalence, minimizing false positives becomes challenging even with DCAs that have high specificity. Particularly when health care systems lack the capacity to manage a high volume of false positives, cognitive screening models that focus on high‐risk individuals and those with cognitive concerns may be more pragmatic. Table 2 presents PPV and NPV calculations for a Detection DCA with 80% sensitivity and 85% specificity at prevalence levels of 10% and 20%. Prevalence estimates are given for a hypothetical U.S. primary care clinic sample of older adults ages 65–95 and are based on the literature suggesting that the prevalence of ADRD is increasing, with ≈10% of older adults currently affected, and over 20% living with MCI. ² , ⁹ The decline in NPV with increasing prevalence underscores the importance of conducting a more in‐depth assessment (e.g., a Diagnostic Aid DCA) and further workup when cognitive concerns are present.

TABLE 2.

Clinical performance of DCAs for each context of use.

	Minimum acceptable performance	Predictive value according to prevalence
		Prevalence of cognitive impairment	Predictive values
Detection DCA (to detect cognitive impairment)	80% sensitivity 85% specificity	10%	PPV 37% NPV 97%
		20%	PPV 57% NPV 94%
		Prevalence of MCI	Predictive values
Diagnostic Aid DCA (to aid in the diagnosis of MCI)	85% sensitivity 90% specificity	30%	PPV 78% NPV 93%
		50%	PPV 89% NPV 86%
		70%	PPV 95% NPV 72%
		Prevalence of MCI due to AD as an example	Predictive values
Profile Characterization Aid DCA (To aid in the diagnosis of a specific etiology)	85% sensitivity 90% specificity	20%	PPV 68% NPV 96%
		50%	PPV 89% NPV 86%
		80%	PPV 97% NPV 60%

Abbreviations: AD, Alzheimer's disease; DCA, digital cognitive assessment; MCI, mild cognitive impairment.

4.2. DCAs to aid in the diagnosis of MCI or dementia

The Diagnostic Aid DCA is intended to support the diagnostic process for all forms of MCI and dementia syndromes (Table 1). It is conceptually similar to non‐digital tools such as the MoCA, which assess multiple cognitive domains and may assist with determining the level of impairment. This type of DCA would be initiated in primary or specialty care following an abnormal result on a Detection DCA (or similar tool) or in response to a cognitive concern raised by an individual, care partner, or clinician. These concerns may include memory or other cognitive changes, confusion, sensory changes, or difficulty with daily activities. Care partners or other informants may observe repetitive questioning, mood and behavioral changes, signs of disorientation, or an increased reliance on assistance, whereas clinicians may note repetition or confusion during visits.

This testing will generally be a 10–20 minute assessment examining multi‐domain cognitive functions, with scoring on a scale indicating impairment severity to assist with differentiating between MCI and dementia and guiding clinical decisions. Of note, the Diagnostic Aid DCA is not a stand‐alone diagnostic tool. This DCA would be implemented as part of a broader diagnostic process that would include a thorough medical history, laboratory testing, structural neuroimaging as clinically indicated, and an assessment of functional status. In addition, input from family or other close observers of the patient's cognition and daily functioning is important to obtain, when possible, to further aid in staging of cognitive impairment.

The evaluation for MCI or dementia should include assessment for treatable causes (e.g., medication side effects, vitamin deficiencies, depression, and so forth) and the initiation of appropriate brain health interventions. ¹⁵ In conjunction with the results of a comprehensive evaluation, an abnormal DCA test result would prompt a treatment plan. A normal result would suggest a lower likelihood of cognitive impairment, although continued monitoring and implementation of brain health interventions remain important. Furthermore, this tool may have an important role in reassuring individuals who have concerns about cognition but no objective impairment.

Given the clinical need for scalable, low‐burden tools to support HCPs in diagnosing MCI, particularly in primary care where patients with MCI typically first present, the development of DCAs with high sensitivity for MCI is a priority. Unlike dementia, which is characterized by both cognitive impairment and functional decline in activities of daily living, MCI requires impairment in only one cognitive domain and functional independence. This distinction makes MCI detection particularly challenging. A well‐designed Diagnostic Aid DCA could help close the significant gap between the prevalence and identification of MCI, ⁷ , ⁸ , ⁹ enabling clinicians to intervene at earlier stages, when treatments and other interventions have the greatest potential benefit. ¹⁵ , ⁵⁶ , ⁵⁷

The workgroup recommends a minimum acceptable performance of 85% sensitivity to aid in the diagnosis of MCI and 90% specificity to maintain acceptable predictive values across a range of expected prevalence levels (30%, 50%, and 70%; Table 2). The sensitivity target of 85% is on par with published ranges for the MoCA and SLUMS for the detection of MCI, whereas the target of 90% is slightly above published ranges for specificity in the same studies to emphasize the importance of avoiding false positives. ⁵⁴ Prevalence estimates of 30% and 50% are for a hypothetical U.S. primary care clinic sample of older adults ages 65–95 with cognitive concerns or a positive Detection DCA (or similar test). Prevalence estimates of 50% and 70% are more representative of specialty care settings.

4.3. DCAs for cognitive profile characterization to aid etiological workup

The Profile Characterization Aid DCA is intended to (1) confirm or rule out the presence of MCI or dementia in individuals with a prior workup, as described in Section 4.2, and (2) characterize the profile of affected cognitive domains to assist in determining the underlying etiology of the cognitive impairment (Table 1). A comparable non‐digital assessment would be a comprehensive cognitive evaluation done by a neuropsychologist. The target population includes individuals with a diagnosis of MCI or dementia or individuals for whom a prior workup, which includes a Diagnostic Aid DCA or similar cognitive assessment, yielded inconclusive results (e.g., when the assessment failed to clearly capture mild impairment in an individual with strong cognitive reserve and persisting cognitive concerns). Notably, patients with moderate‐to‐severe global cognitive or functional impairments may not be suitable candidates for this assessment.

As a more comprehensive assessment tool, the Profile Characterization Aid DCA may typically require ≈40 min or more to evaluate multiple cognitive domains and reliably differentiate cognitive profiles (Table 1). It would be used most commonly in specialty care, following a referral from primary care or another HCP. Ideally, the DCA would be ordered by a dementia specialist, with interpretation by a neuropsychologist or another trained specialist where possible. In some cases, it may be ordered in primary care with consultation from neuropsychology. This DCA should reduce the frequency of need for a full traditional neuropsychological evaluation, thereby helping to alleviate the growing bottleneck of neuropsychology referrals for older adults. In‐depth neuropsychological evaluations should still be utilized for complex cases (e.g., cases involving multiple contributing comorbidities or significant neuropsychiatric symptoms) or for those requiring targeted cognitive phenotyping for diagnosis (e.g., language assessment for suspected primary progressive aphasia).

Although the Profile Characterization Aid DCA provides critical insights into an individual's cognitive profile, it does not independently determine the biological etiology of the disease. Instead, it serves as a complementary tool alongside other diagnostic measures (including biomarker testing when clinically indicated), helping clinicians make more informed decisions about the cause, profile, and degree of cognitive impairment; refine prognosis; and develop tailored care plans. This DCA should also streamline the pathway to receiving a diagnosis by enabling more flexible and efficient point‐of‐care access to cognitive testing, with reduced wait time and fewer referrals.

To illustrate the intended use of this type of DCA, MCI due to AD is described as an example, given that AD is the most common neurodegenerative disorder and has a well‐defined diagnostic framework. ⁵⁸ Here, this type of DCA would be used in conjunction with a comprehensive workup to confirm MCI (level of impairment severity) and provide information on whether the cognitive profile is suggestive of MCI due to AD. For example, the results may show primary deficits in memory as well as a few milder difficulties in executive function and language domains. The absence of other patterns of impairment (e.g., isolated deficits in processing speed or executive dysfunction), in combination with other clinical evidence, would simultaneously help to indicate which other etiologies (e.g., cerebral small vessel disease) are less likely. An abnormal evaluation would confirm the presence of MCI and potentially suggest AD as the primary etiology, prompting AD biomarker testing (if not conducted concurrently) and possible use of AD‐specific interventions, treatments, and care planning. Conversely, a normal evaluation would rule out MCI and guide further management.

Continuing with the example of MCI due to AD, the workgroup recommends a performance threshold of 85% sensitivity and 90% specificity when distinguishing MCI due to AD from normal cognition. Predictive values at three prevalence levels (20%, 50%, and 80%) are provided in Table 2, corresponding to low, intermediate, and high suspicion of MCI due to AD based on prior clinical evaluation. ⁴⁵ Prevalence estimates are representative of specialty care settings with a sample of adults 65–95 years of age with cognitive concerns and some prior testing suggesting possible AD.

5. IDEAL CHARACTERISTICS

As DCAs continue to evolve, developers and health care systems must ensure that these tools are fit for clinical use by prioritizing rigorous validation, scalability, and effective results communication (Table 3).

TABLE 3.

Priorities for DCA design and development.

Validation
Normative data	Normative data should come from large samples broadly representative of the general population. Neuropsychological test batteries are recommended as the reference standard for DCAs and should be validated in demographically representative, longitudinal cohorts that reflect the DCA's intended use.
Diverse adaptation	Validation should encompass diverse cultures, geographic regions, languages, socioeconomic backgrounds, education levels, risk factors, and varying degrees of technological familiarity, using culturally appropriate presentation and stimuli.
Reliability	Test‐retest reliability in the expected use settings should be established for all DCAs to ensure results are not confounded by environmental factors or random error.
Evaluation	Implementation and dissemination studies represent best practices for assessing real‐world performance.
Scalability
Device platform/technology being used	The priority is for the DCA to meet performance criteria, regardless of platform. Web browser compatibility is recommended to ensure the test can be administered on any device available to health care professionals. DCAs are actively being developed for smartphone use. As more data emerges, smartphones are expected to meet performance standards in the near future, with some potentially already suitable for detecting cognitive impairment.
Cost	Developers should clearly define the intended user—whether an individual clinician, health system, academic/research institution, or home user—and outline the DCA's objectives and cost to ensure users can assess its suitability and affordability. Billing codes and insurance reimbursement eligibility should be considered.
Accessibility	Tests should accommodate sensory and motor impairments by offering alternative administration methods. A minimum accessibility standard that assesses for hearing, functional vision, language, and hand coordination should be established.
Usability	Co‐design methodologies, such as focus groups that include healthcare professionals and patients involved in the use of DCAs, should be implemented to involve end users in test development.
Results reporting
Detection DCA	Automated report for provider should clearly indicate whether cognitive impairment is a concern or unlikely, with recommendations for continued monitoring following a normal result. EHR integration/templating of results is encouraged. Sharing patient results through EHR integration is a desirable capability, to be implemented at the discretion of the healthcare system or provider.
Diagnostic Aid DCA and Profile Characterization Aid DCA	Automated report for provider should include information on severity of impairment (Diagnostic Aid DCA) or the pattern and severity of impairment in different cognitive domains (Profile Characterization Aid DCA). EHR integration/templating of results is encouraged. Test results should be contextualized relative to the normative benchmarks applied. Providers should interpret and communicate results directly to patients. Shared decision‐making and discussions throughout the diagnostic process should be promoted, recognizing variations in clinical approaches.

Abbreviations: DCA, digital cognitive assessment; EHR, electronic health record.

5.1. Validation

Historically, normative data have lacked demographic diversity because most cognitive assessments were validated in homogeneous populations (e.g., non‐Hispanic White individuals with high educational attainment). Building on current efforts in clinical neuropsychology, the DCA Workgroup strongly advocates for validation processes that account for the diversity of real‐world clinical populations. ³¹ To achieve this, DCAs and reference standards must be tested in heterogeneous cohorts to ensure appropriate interpretations and use across demographic groups.

Consistency and dependability are critical features for DCAs, particularly those developed for flexible use across different care settings and administered by a range of HCPs. A test should yield similar results when repeated later in the same individual, demonstrating that the outcome is not overly sensitive to environmental factors or random error. Establishing test–retest reliability for all DCAs is therefore an important validation step, regardless of whether the test is designed for single‐timepoint use or cognitive monitoring over time.

Establishing predictive validity using longitudinal data will be especially valuable if early cognitive screening becomes routine practice. This includes evaluating not only the sensitivity of DCAs across the disease course, but also their prognostic accuracy in predicting clinical outcomes. To encourage consistency and rigor, the workgroup recommends using neuropsychological test batteries that assess relevant cognitive domains as the reference standard across all types of DCAs (see Table S1 for details). These reference standards should be validated against clinical and biomarker data in demographically representative, longitudinal cohorts that match the DCA's intended‐use population as much as possible. Regionally specific tests should be considered if validated for the intendeduse population. For example, the African Neuropsychological Battery may be an appropriate reference standard for sub‐Saharan African populations. ⁵⁹

For brief DCAs an additional validation step is recommended: demonstrating superiority (or non‐inferiority at a minimum) over comparative non‐digital cognitive screening measures. This effort will be essential for the successful transition from conventional cognitive screening tools to DCAs in clinical practice.

To ensure transparency and appropriate interpretation, normative data and information on how threshold cutoffs were established must be accessible. Moreover, demographic characteristics, such as age, sex, education, race, ethnicity, geographic region, socioeconomic status, and digital literacy are necessary for interpreting validation data. DCAs should be supported by peer‐reviewed data from well‐powered studies in the intended‐use population before they are presented as clinically validated. For example, the Diagnostic Aid DCA should be validated primarily in clinical samples of older adults with recognized cognitive concerns, whereas the Detection DCA would require testing in broader populations that match the demographics of community dwelling older adults and those accessing primary care within the country or region of intended use.

Beyond initial validation, ongoing monitoring is essential to ensure that DCAs maintain reliability and accuracy in the populations in which they are deployed. A structured evaluation framework should be in place, incorporating implementation and dissemination studies to assess real‐world performance.

5.2. Scalability

Digital device use continues to grow among older adults, improving the feasibility and scalability of widespread DCA adoption. However, significant disparities in digital literacy and access persist among older adults in the United States and globally. ⁶⁰ , ⁶¹ Usability studies—particularly those employing co‐design methodologies with HCPs and patients—are important in supporting large‐scale clinical adoption and appropriate accommodation for low digital literacy among the population of intended use. ⁶² Collaboration between industry, HCPs, and patients, particularly those from underrepresented and underserved groups, is essential to ensure that digital tools are inclusive of diverse needs.

To maximize accessibility, DCAs should be web‐browser compatible to enable administration on any device available to HCPs. Access to DCAs via smartphone holds significant promise for scalability, given the widespread use of these devices globally. In light of accumulating validation research, smartphone‐compatible assessments are expected to meet the recommendations outlined here for minimum acceptable performance in the near future, with several already demonstrating good accuracy and reliability for detecting cognitive impairment. ³⁶ However, given the limitations of small screens, particularly for older adults with sensory changes, assessment results obtained from smartphones will require careful interpretation and confirmation with additional modalities, and may therefore be best suited for initial impairment detection or monitoring uses.

Developers are encouraged to prioritize user‐friendly and intuitive design to facilitate the integration of DCAs into clinical practice. Of note, strong technical performance alone is insufficient—real‐world clinical utility also depends on integration into workflows, clinician uptake, and use in decision‐making. Although a comprehensive discussion of these implementation factors is beyond the scope of this workstream, they are a key focus of ongoing work by the DCA Workgroup.

A clear understanding of the tool's intended user base and clinical applications is also necessary to support successful adoption across health care settings. Tests should be designed to accommodate sensory and motor impairments by offering alternative methods of administration to promote broader usability and minimize bias in cognitive assessment outcomes. Because an individual's hearing, functional vision, language, and hand coordination can influence test performance, the workgroup recommends establishing a minimum accessibility standard that includes evaluating these factors prior to administering a DCA.

5.3. Results communication

Scalable implementation of DCAs also requires careful attention to procedures for results communication. Table 3 outlines recommendations for reporting results across the three clinical contexts of use. Directly communicating test results to patients is recommended, especially for DCAs that are used to support a diagnosis, and shared decision‐making should be promoted throughout the diagnostic process. EHR integration is recommended across all contexts to streamline workflows, reduce provider burden, and enhance clinical efficiency. Careful consideration should be given to how results are released electronically through the EHR. HCPs are encouraged to discuss results with patients and care partners beforehand to support interpretation.

6. REMOTE (UNSUPERVISED) DCAS

These recommendations focus specifically on clinically supervised DCAs, recognizing that remote, unsupervised tools will also be critical to address. Prioritizing standards for supervised use is a pragmatic first step, as the absence of such guidance would likely make integration of remote testing into clinical care unwieldy and pose challenges for clinicians in interpreting or confirming results generated outside the clinic. Given the rapid progress of remote, unsupervised tools, these emerging assessments warrant attention and will be a focus of future work by the DCA Workgroup. To support that effort, this section outlines key considerations related to their clinical application.

Remote DCAs offer potential advantages such as improved accessibility, ease of repeat testing, and reduced clinician burden. However, several challenges must be addressed before widespread implementation can be recommended. Key concerns include the lack of standardized referral pathways following an abnormal result, the impact of variable digital literacy without supervised support, the use of adaptive devices for low vision or hearing impairment, and the impact of uncontrolled testing environments on reliability and validity. Cognitive disorders themselves may also introduce confounding variables, as conditions like apathy, motor impairments, or language deficits may not be recognized in unsupervised testing and could interfere with test completion and accuracy.

To enhance validity, future efforts should focus on standardizing environmental and technological requirements, developing adaptive assessments that account for external factors, and integrating patient‐reported testing conditions. Expert consensus on guidelines for remote DCAs, informed by existing teleneuropsychology standards, ⁶³ will be essential to ensuring their appropriate clinical use.

7. CONCLUSIONS

The increasing prevalence of MCI and ADRD, coupled with the challenges in early diagnosis and detection, underscores the urgent need for improved cognitive assessment tools to detect cognitive changes earlier, during key windows for prevention and early intervention. DCAs offer a promising solution, but the variability in test performance and intended use has led to uncertainty about their clinical utility. The DCA Workgroup recommendations presented here aim to guide the adoption of DCAs into clinical practice by outlining target characteristics for three clinical contexts of use. These guidelines are intended to assist test developers, as well as HCPs and health system decision‐makers evaluating these tools for clinical use. Implementation considerations for the clinical use of DCAs are equally critical for successful widespread adoption of DCAs in primary and secondary care. Guidelines focused specifically on DCA implementation issues—including those related to establishing clinical workflows, navigating billing and reimbursement, and coordinating care and referrals—will be the focus of future work from an expert health care implementation workgroup organized by CEOi.

CONFLICT OF INTEREST STATEMENT

Louisa I. Thompson, Barak Gaster, Dustin B. Hammers, A. M. Barrett, and Christopher R. Butler report no conflicts of interest. Louisa I. Thompson is supported by funding from the National Institute on Aging/National Institutes of Health (NIA/NIH; K23AG080159). Sol Fittipaldi is an Atlantic Fellow for Equity in Brain Health at the Global Brain Health Institute (GBHI) and is supported with funding from GBHI, Alzheimer's Association, and Alzheimer's Society (GBHI ALZ UK‐24‐1068607). Benjamin Tiede is the Executive Director of the Global CEO Initiative on Alzheimer's Disease (CEOi). Katherine A. Partrick and Emily Scholler are paid consultants for the CEOi. Any author disclosures are available in the Supporting Information.

Supporting information

Supporting Information

Supporting Information

1. COLLABORATORS

CEOi Digital Cognitive Assessment Workgroup collaborators.

Name	Affiliation
J. Wesson Ashford, PhD	VA Palo Alto Health Care System, Palo Alto, CA, USA
Sasha Bozeat, PhD	F. Hoffman–La Roche AG, Basel, Switzerland
Kay Bhothinard, MBA	Eisai Inc, Nutley, NJ, USA
David Berron, PhD	German Center for Neurodegenerative Diseases (DZNE), Magdeburg, Germany; Clinical Memory Research Unit, Department of Clinical Sciences Malmö, Lund University, Lund, Sweden
Soo Borson, MD	University of Southern California, Keck School of Medicine, Los Angeles, CA, USA

 

Name	Affiliation
Bryan Cobb, PhD	Eisai Inc, Nutley, NJ, USA
Emrah Duzel, MD	Institute for Cognitive Neurology and Dementia Research, University of Magdeburg, Magdeburg, Germany; German Center for Neurodegenerative Diseases (DZNE), Magdeburg, Germany; Institute of Cognitive Neuroscience, University College London, London, United Kingdom
Darren R. Gitelman, MD	Advocate Medical Group, Behavioral Neurology, Advocate Memory Center, Advocate Lutheran General Hospital, Park Ridge, IL USA
Ishtar Govia, PhD	Amagi Health Ltd, London, United Kingdom
Roos J. Jutten, PhD	Alzheimer Center Amsterdam, Department of Neurology, Amsterdam Neuroscience, Amsterdam UMC, location VUmc, Amsterdam, The Netherlands
Eric G. Klein, PharmD	Eli Lilly and Company, Indianapolis, IN, USA
Nicole A. Kochan, PhD	Centre for Healthy Brain Ageing, School of Clinical Medicine, University of New South Wales, Sydney, Australia
Melissa Lee, PhD	Alzheimer's Drug Discovery Foundation, New York, NY, USA
Nicklas Linz, PhD	ki:elements GmbH, Saarbrücken, Germany
Tim MacLeod, PhD	Davos Alzheimer's Collaborative, Wayne, PA, USA
Soeren Mattke, MD	University of Southern California, Los Angeles, CA, USA
Christopher J. Medberry, PhD	Johnson & Johnson, New Brunswick, NJ, USA
Michelle M. Mielke, PhD	Department of Epidemiology and Prevention, Wake Forest University School of Medicine, Winston‐Salem, NC, USA
Jennifer Murphy, PhD	Biogen, Cambridge, MA, USA
Ziad Nasreddine, MD	MoCA Clinic and Institute, Montreal, QC, Canada
Melissa E. Petersen, PhD	Department of Family Medicine, University of North Texas, Fort Worth, TX, USA
Andrew J. Saykin, PsyD	Indiana Alzheimer's Disease Research Center and Department of Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, IN, USA
Katherine J. Selzler, PhD	Davos Alzheimer's Collaborative, Wayne, PA, USA
John Showalter, MD	Linus, Boston, MA, USA
Dirk Smeets, PhD	Icometrix, Leuven, Belgium
Christa M. Studzinski, PhD	Ontario Brain Institute, Toronto, ON Canada
Pierre N. Tariot, MD	Banner Alzheimer's Institute, University of Arizona College of Medicine, Phoenix, AZ, USA
Vivian Vasallo, MA	UsAgainstAlzheimer's, Washington, DC, USA

Thompson LI, Gaster B, Hammers DB, et al. Acceptable standards for clinic‐based digital cognitive assessments: Recommendations from the Global CEO Initiative on Alzheimer's Disease. Alzheimer's Dement. 2025;21:e70966. 10.1002/alz.70966