Blog | Creyos | Cognitive Testing for Better Brain Health

Population-Level Quality Metrics: The Cognitive Data Gap

Written by Andrea Welsh, LPN, CHPC | Nov 17, 2026, 3:43:48 PM

Health systems and accountable care organizations comparing population health management solutions often ask which software solution will help them improve healthcare quality metrics on a population level. These platforms organize and analyze the health data an organization already holds, so their results depend on what clinicians have assessed and documented.

Cognitive impairment is a common gap in that record. In 2015–2019 Medicare data, an estimated 8% of expected mild cognitive impairment (MCI) cases had been diagnosed, although impairment can affect medication adherence, chronic condition management, and follow-through on care plans.

For quality, population health, and clinical leaders, the sections below cover what population health software does well, where undetected cognitive impairment leaves gaps in the data it depends on, and what adding standardized cognitive assessment requires: which population to start with, who owns follow-up, the capacity the work takes, and how to measure success.

Article Highlights

  • Which software solution will help improve healthcare quality metrics on a population level depends partly on the clinical data feeding it, since population health platforms work largely from conditions that clinicians have assessed and documented.
  • Cognitive impairment is a common gap in that data. An estimated 8% of expected MCI cases in 2015–2019 Medicare data had been diagnosed, and 31% of surveyed beneficiaries reported formal cognitive screening during their Annual Wellness Visit.
  • Undetected impairment can appear as missed follow-up, medication nonadherence, or emergency department use before it appears as a documented clinical finding.
  • Standardized cognitive assessment adds a consistent clinical input that population health platforms can use, with results prompting clinical evaluation and, when dementia is confirmed, accurate documentation.
  • Results depend on leadership decisions about which population to prioritize, who owns follow-up, what capacity the work requires, and how success is measured.

How population-level quality measurement works

Population-level quality metrics measure how consistently an organization delivers recommended care across a defined group of patients, such as an attributed population, a health plan’s members, or a chronic disease registry. Measures are commonly grouped into structural, process, and outcome types. Structural measures assess the resources and systems an organization has in place, process measures assess whether recommended care such as screenings and follow-up visits was delivered, and outcome measures assess results such as hospitalization rates or blood pressure control.

Programs such as HEDIS, Medicare Star Ratings, and MIPS each define their own measure sets and reporting rules. In value-based contracts, performance on those measures affects shared savings, quality bonuses, and payment adjustments. Every program also relies on clinical data that is accurate, standardized, and documented where measurement systems can find it, and that reliance shapes what any software platform can deliver.

What population health management solutions do well, and what they depend on

Population health management solutions bring together data from electronic health records, claims, pharmacy records, lab results, and patient engagement systems into a more unified patient record. That aggregation reduces manual reconciliation across disconnected systems and gives quality teams a view of performance across sites, service lines, and contracts.

Population health analytics then group patients by diagnosis history, utilization, medication use, clinical values, or social determinants of health such as transportation access and housing stability. Care teams use those groupings to prioritize outreach and plan targeted interventions for the patients most likely to benefit.

Each of these functions works from conditions already recorded in a structured, usable form. A diagnosis buried in an unstructured note is harder for analytics to surface. Tools such as eRADAR can flag possible undiagnosed dementia from existing EHR data, but standardized assessment with clinical evaluation adds the actionable information a risk signal alone cannot provide. For conditions that depend on specific screening, the completeness of the clinical input sets the limit on what the platform can show.

For organizations comparing software, the quality of clinical data capture belongs in the evaluation alongside analytics capability.

“

If you’re not able to measure it, you’re not able to improve it. You’re not even able to define success.

— Benjamin Todd, SVP of Strategy and Analytics, Claremedica

Undetected cognitive impairment as a gap in the population record

Many chronic disease care gaps, such as an overdue A1c or a missed cancer screening, are tracked systematically. As Anthony Zizza, MD, Chief Medical Officer at Element Care and Clinical Advisor at Creyos, wrote in RISE Health, “They are converted into measurable objects: denominators, registries, EHR prompts, outreach lists, care-management queues, and quality dashboards. Cognition is the conspicuous exception.”

Medicare builds a cognitive-detection requirement into the optional Annual Wellness Visit (AWV), yet practice varies. In a national survey of Medicare beneficiaries, 81% reported having had an AWV, while 31% reported formal cognitive screening and 35% reported being asked about memory concerns.

Cognitive impairment can change how patients understand instructions, manage medications, and follow through on a care plan. A systematic review of medication adherence in older adults with cognitive impairment identified barriers such as understanding new directions and fitting medications into a daily routine. When cognitive status is missing from the record, a missed refill or follow-up visit may be read as disengagement when it reflects a barrier the care team hasn’t yet identified.

Undetected impairment often surfaces through acute care. Among Medicare beneficiaries, a new dementia diagnosis was associated with a 40% increase in the likelihood of an emergency department visit over the 12 months surrounding diagnosis, and in the month before diagnosis, the ED visit rate was 13 per 100 patients compared with 2.95 per 100 matched peers. The study is observational and leaves open whether those visits were preventable, but the pattern suggests impairment can reach the record through utilization ahead of routine assessment.

When cognition is absent from the record, quality measurement and care planning both work from an incomplete picture of the population.

Adding standardized cognitive assessment to the clinical input

Standardized cognitive assessment gives clinicians a consistent way to measure cognition during routine care. A result that suggests impairment identifies a concern that warrants clinical evaluation, and that evaluation determines whether further workup, a diagnosis, or a cognitive care plan follows. That can mean evaluating function and reversible contributors, simplifying medications, involving a caregiver, and naming an owner who confirms follow-through.

For population health, the value lies in consistency. When an organization defines what triggers an assessment and when it occurs, results can be compared across sites and patient groups and entered into the record in a structured form that population health platforms can use. The platforms continue to organize care gaps and outreach, now with patient data that includes cognition.

The same consistency supports accurate documentation. When clinical evaluation confirms dementia, documenting it with its stage matters under the CMS-HCC model (V28) used for non-PACE Medicare Advantage plans in 2026, which assigns mild, moderate, and severe dementia to separate hierarchical condition categories. Accurate documentation keeps risk scores aligned with the complexity a care team is actually managing.

At Claremedica, a full-risk Medicare Advantage primary care organization with more than 35 Florida sites, the share of patients screening positive for possible dementia more than doubled after the shift from an MMSE-based approach to standardized digital assessment with Creyos, across more than 34,000 assessments in 7 months. Most patients who screened positive showed early or mild impairment and had been with the organization for years, suggesting impairment may have been present before it appeared in the data. Claremedica is now tracking how earlier identification relates to utilization and cost across its population.

What to look for in a cognitive assessment solution

Once cognitive status is treated as a clinical data input, evaluating an assessment tool comes down to whether it produces information that is clinically credible and usable at scale. Key criteria include:

  • Clinical validation. Peer-reviewed validation and normative comparisons in the intended population, with clear measurement properties across domains such as memory, attention, reasoning, and executive function. Domain-level results support care planning more than a single pass-fail score.
  • Workflow fit and integration. Simple administration, clear reporting, and integration with the electronic health record so that assessment runs inside routine visits and results arrive where clinicians already work.
  • Accessibility. Language options, accommodations for hearing and vision differences, and digital accessibility, so scores reflect cognition.
  • Consistent use across settings. Population-level data depends on every site administering and scoring assessments the same way. In one large multi-state health network, paper screening tools produced inconsistent results depending on who administered them, and steps such as separate logins or manual scoring can lower completion further.
  • Repeat measurement. Tracking change over time requires evidence that a tool can be repeated reliably without scores rising from familiarity alone. Traditional tools such as the Montreal Cognitive Assessment (MoCA) have a limited number of alternate versions and simple exercises that patients can learn, so practice effects can make repeat results difficult to interpret. Tools that draw from larger item banks are built to support longitudinal cognitive monitoring.
  • Reporting that supports follow-up. Output that gives care management, documentation, and referral decisions what they need, so results translate into actionable insights for the care team, including follow-up for false-positive results.

Implementing cognitive assessment at scale

Population-level assessment depends on identifying cognitive concerns in routine care, with specialist referral reserved for the patients who need it. In a national survey of primary care physicians, 55% said there were not enough dementia care specialists in their area to meet patient demand, rising to 71% in rural areas.

Implementation starts with which population to prioritize. A defined population and clinical use case, such as patients due for an Annual Wellness Visit or enrolled in a chronic condition management program, lets a pilot surface completion barriers and staff workload before expansion. Because the U.S. Preventive Services Task Force has found the evidence insufficient to assess routine screening of older adults without symptoms, a defined clinical use case keeps the work tied to patients whose care the results can change.

Each workflow should name who orders the assessment, who reviews the result, and who owns follow-up. Governance sets documentation standards and defines how aggregate results will feed quality reporting and population analysis.

Administering a digital assessment can fit into existing workflows without specialized staff, and in the full-risk organization described above, screening time fell from about 12 minutes with the MMSE to about 6. Reviewing results, arranging further evaluation, and coordinating follow-up still require clinical and care management capacity, which should be planned and resourced before rollout.

Measures of success are best defined before launch. Assessment completion, completed evaluations, care-plan changes, and time from a positive result to further evaluation show whether the pathway is working, and the key performance indicators the organization already tracks, such as medication adherence and emergency department use, show over time whether care is changing, measured against baseline data, a comparison group where feasible, and total clinical and technology costs.

Built this way, cognitive assessment becomes part of continuous quality improvement, giving population health platforms a clinical input they have largely been missing.

 

Written by Andrea Welsh, LPN, CHPC

Andrea Welsh, LPN, CHPC, is a nurse and medical writer with clinical experience in geriatric, primary care, and behavioral health settings. She has managed quality and compliance programs supporting value-based care in community health centers and holds the Certified in Healthcare Privacy Compliance (CHPC) credential. She speaks at healthcare conferences on data literacy and clinical documentation.

How we keep this article accurate
We review our content regularly to ensure it is up to date.
Share
 
Published
November 17, 2026
Reviewed by
Dr. Anthony Zizza
Chief Medical Officer, Element Care and Clinical Advisor at Creyos