Neurotrack

AI is claimed and no learned component is described. The platform page calls Smart Logic an adaptive intelligence layer that chooses which assessments to give based on how a patient is responding, combines the results with EHR context to surface a DSM-5 aligned diagnostic pathway, and generates a patient specific care plan. Neither the home page nor the platform page says any part of this is learned from data, and the assessments themselves (Symbol Match, Path Points, Word Recall) are scored digital tests, which the device regulation Neurotrack is registered under describes as a battery of cognitive tasks producing a score.

That may be rules under an intelligence label, or a model the company has not described; nothing published says which, and adaptive test selection and care plan drafting can be done either way. The product is a validated digital test battery with workflow around it, and its AI rests on the company's claim alone.

The platform page makes a universal claim about the output: 'every output supports confident clinical action, and every result is clinically defensible.' The first half is a claim of reliability: a buyer who believes every output supports confident action has no reason to build the check that would catch a false impairment or a missed one, whatever the review structure. The second half is a claim about basis: in the risk adjustment language Neurotrack sells in, defensible means supported well enough to survive an audit, which holds only where each result shows what it rests on. Nothing here does: nothing published says how Smart Logic chooses assessments or what moves a patient onto the diagnostic pathway.

The structure itself is described in part. The page says Smart Logic supports clinical decision making and never directs it, and that the clinician leads the diagnosis and care plan. But Smart Logic also decides on its own which assessments a patient receives, and nothing published says what happens when the clinician disagrees.

The instruments are now named. The platform page lists Symbol Match (processing speed and visual attention), Path Points (a digitally adapted trail making task for executive function and working memory) and Word Recall (verbal memory), and gives sensitivity and specificity for the Symbol Match and Path Points battery with the journal that published it. Because in cognitive assessment the instrument is the evidence, naming the battery lets a buyer find its literature and known limits. The company began with eye tracking based assessment; the current products use these patient led tests instead.

What is not explained is Smart Logic, the layer that chooses which tests to give, proposes a diagnostic pathway and drafts care plans: its rules or model, the norms scores are compared against, and how and when either changes.

The chain is described by category in the privacy policy (updated 26 March 2025), along with the terms that bind it: secure hosting providers that store and process data, electronic communications providers, data analytics providers, and video conferencing services used to record assessments. These subprocessors may not use the information except to provide the service, may not keep it afterward, and the current list is available on request. The product describes no third party model, and nothing on the platform page suggests one, so the chain as published is short.

That is substantial partial disclosure. No party is named publicly, the list is available only on request, and Smart Logic's mechanism is not described, so nothing shows whether any model provider sits in the care plan or diagnostic pathway step.

The headline claim is that the assessments detect cognitive impairment at the MCI stage, earlier and more accurately than paper tools or other digital tests. The published evidence the platform page cites is a Frontiers in Aging Neuroscience study reporting 94 percent sensitivity and 90 percent specificity for the Symbol Match and Path Points battery, comparing healthy people with people diagnosed with Alzheimer's disease. That is peer reviewed and it answers a different question: separating established Alzheimer's from health is easier than finding mild impairment, and Word Recall is not in the figure. The comparison chart against MoCA, Mini-Cog and MMSE does not say where the comparator figures come from.

The deployment figures are vendor reported. An unnamed national Medicare Advantage plan's three month home visit pilot found 85 percent higher MCI identification and 53 percent higher dementia identification than pen and paper assessment (p below 0.001); Desert Oasis Healthcare reported a 35 percent net increase in new dementia diagnoses, presented at a pharmacists' conference in 2025; internal comparisons against Mini-Cog report five times more impairment identified. More diagnoses is not the same as correct diagnoses, and none of these reports how many flagged patients were confirmed.

The privacy policy (updated 26 March 2025) covers the cognitive assessments and protected health information and publishes categorical commitments: Neurotrack does not sell account information or health data; its subprocessors may not use the data except to provide the service or keep it afterward; and information shared with third parties for research or development excludes personally identifiable information and health data unless the person consents. Deletion requests are handled within 30 days, up to 90 days for backups.

The rest of the lifecycle is looser. The policy says Neurotrack itself uses the information it collects to improve the services, perform data analysis and testing, conduct research and develop new features, which is a use for development rather than a commitment against it; retention is no longer than necessary, with no schedule; and aggregated or de identified data may be disclosed publicly and to partners, including as community benchmarking. The data concerns cognitive decline in older adults, which reaches into driving, finances and legal capacity.

HIPAA compliance is stated on the security page and the platform page, and the privacy policy (updated 26 March 2025) covers the cognitive assessments and protected health information: when a health plan, provider or insurer asks Neurotrack to run an assessment, the policy describes receiving health information on its behalf and sharing results with it and its business associates. The security page describes the administrative, physical and technical safeguards behind the claim.

No business associate agreement or summary of its terms is published, and the policy does not say in plain words that Neurotrack is the business associate when it runs assessments for a covered entity. The treatment of assessment recordings is not described, nor that of de identified data, which the privacy policy allows to be shared for benchmarking.

The security page states that Neurotrack is SOC 2 and HIPAA compliant, and the platform page repeats SOC 2 across all care settings and deployment models. That is a recognized attestation presented as the company's own. The security page also describes administrative, physical and technical safeguards, and says the company conforms to ISO 13485 and FDA quality system regulations; those are device quality standards, which bear on regulatory status rather than security.

The SOC 2 type is not stated, and none of the supporting material is published: no report, trust center or audit period. A Type I report covers the design of controls at a point in time, a Type II their operation over a period, and the difference carries the weight.

The stated position and the marketing contradict each other. The platform page describes the product as an 'FDA-Registered Class II Medical Device' that 'meets regulatory standard for clinical diagnostic tools', and the home page says it equips providers to 'assess, diagnose, and care for people living with MCI and dementia', with Smart Logic surfacing a DSM-5 aligned diagnostic pathway. Registration is not clearance: no 510(k) or De Novo is named. The classification a registered cognitive test battery falls under, the computerized cognitive assessment aid at 21 CFR 882.1470, describes a device that 'does not identify the presence or absence of clinical diagnoses' and 'is not intended as a stand-alone or adjunctive diagnostic device', and its exemption from premarket review does not apply when the device is intended for diagnostic assessment of specific diseases and relies on visual cues or use of the hand, as Symbol Match and Path Points do.

A device registered as an assessment aid and marketed as a diagnostic tool is the D shape. The platform page also says Neurotrack conforms to ISO 13485 and FDA quality system regulations.

Neurotrack asserts even performance across groups without publishing results for them. The platform page says the assessments are validated across diverse patient populations, that Word Recall is a sensitive indicator across diverse linguistic backgrounds, and that assessments can be given in clinic, remotely or at home, in multiple languages, with the same clinical accuracy across every setting. The tests are also described as using culturally agnostic symbols and numbers by design. No results by language, education, age band or setting are published, and the one validation figure on the page compares healthy people with people diagnosed with Alzheimer's disease.

That gap matters because cognitive test scores are known to shift with education, language and familiarity with screens, and a false impairment signal can follow a person into a dementia diagnosis, risk adjustment coding and, where the screen is used in long term care insurance underwriting, a coverage decision.

The mechanism is review by the clinician: the platform page says Smart Logic supports clinical decision making and never directs it, and results, DSM-5 aligned documentation and care plans are written into the record for the provider to act on. Nothing stands behind the output. The page publishes 94 percent sensitivity and 90 percent specificity from a Frontiers in Aging Neuroscience study, but that comparison is of healthy people against people already diagnosed with Alzheimer's disease, not the early stage impairment the product is sold to find, and no warranty, remedy or correction route for the person tested is published.

The person who bears a wrong result is often not the clinician. A false positive can become a coded dementia diagnosis in a risk adjustment record, and the three minute screen is also used in long term care insurance underwriting, where an applicant who scores poorly has no published way to challenge it with Neurotrack.

Integration is claimed and nothing is named. The platform page says Neurotrack integrates directly into existing EHR environments, connects to major EHR platforms, and generates structured documentation that flows automatically into the patient record, including assessment results, DSM-5 aligned documentation and care plans. No record system, marketplace listing, interface standard or production site is named on the home, platform or health plan pages, so there is nothing a buyer can verify before contracting.

The direction matters here because the product writes: diagnosis supporting documentation that lands in the record feeds coding and risk adjustment.

The service is hosted by third party secure hosting providers, according to the privacy policy, and the location is not committed: account information and health data may be stored and processed in the United States or any other country where Neurotrack, its affiliates, subsidiaries, agents or contractors maintain facilities. No hosting provider, region or tenant isolation is named.

The platform page describes administration in clinic, remotely and at home and integration into existing EHR environments, which are ways of using the product rather than places the data rests. The subprocessors include a video conferencing service for recording assessments, so recordings of patients taking the test are among the data held.

No price or pricing basis is published. How the product is sold is discoverable: to health systems, primary care networks and providers, and to health plans through in-home health risk assessments, complex care management programs and provider partnerships. The platform page says cognitive screening and care planning are reimbursable under Medicare Annual Wellness Visits and CPT 99483, which tells a provider how the service can be billed but not what Neurotrack charges.

The health plan case is built on risk adjusted revenue under the V28 model, where dementia severity drives the risk score. It is not stated whether pricing is per assessment, per member or an enterprise platform fee, whether the care planning and Smart Logic features are priced separately from the assessments, or whether any fee is tied to diagnoses or risk score changes.

Coverage is named across settings and buyers: primary care networks, health systems and in-home care, with assessments given in clinic, remotely or at home, and health plans using them in Medicare Advantage in-home health risk assessments, complex care management and provider partnerships. The value based care tie is explicit: the health plan page sells the assessments as documentation for V28 HCC risk adjustment, where dementia severity drives the risk score, and cites a national Medicare Advantage plan that made the assessment standard in its home based Annual Wellness Visit program. Named users include Desert Oasis Healthcare and Sutter Health, and Nationwide reports using the three minute screen in long term care insurance underwriting, a non clinical use with its own stakes.

Validation sits behind part of this: Path Points is described as validated for remote and in-home administration, and the published accuracy figure comes from one comparison of healthy people with people diagnosed with Alzheimer's disease. The claim of the same accuracy in every setting and language is not backed by results.