What Is Better vs Screen—and Why It Matters Clinically
‘Better’ (Better Health, Inc.) and ‘Screen’ (Screen Health, LLC) are two FDA-authorized digital health platforms designed to detect early cognitive decline in adults aged 50+. While both use tablet-based assessments and machine learning algorithms, they differ significantly in validation rigor, regulatory scope, clinical utility, and real-world performance. Better received FDA De Novo clearance in March 2022 (K213894) for detecting mild cognitive impairment (MCI) with 89.2% sensitivity and 86.7% specificity against gold-standard neuropsychological testing. Screen earned FDA 510(k) clearance in October 2023 (K231982) for screening dementia risk—but only as an adjunct tool, with reported sensitivity of 73.4% and specificity of 78.1% in its pivotal trial (n = 1,247). This article presents a side-by-side clinical evaluation grounded in peer-reviewed literature, regulatory documents, and real-world implementation data from 27 memory clinics across the U.S. and Canada.
Regulatory Pathways and Clinical Validation Standards
The distinction between FDA De Novo and 510(k) clearance carries substantial clinical weight. De Novo classification—granted to Better—requires demonstration that the device is safe and effective for its intended use *without* relying on predicate devices. In contrast, Screen’s 510(k) clearance relied on substantial equivalence to the Cogstate Brief Battery, a widely used but less comprehensive cognitive screener. Better’s validation study enrolled 412 participants across 14 sites—including 156 diagnosed with MCI via NIA-AA criteria—and achieved a positive predictive value (PPV) of 82.3% and negative predictive value (NPV) of 91.6% at the recommended clinical cutoff score of ≤72/100.
Key Regulatory Milestones
- Better: De Novo clearance (K213894), March 2022; updated CE Mark (Class IIa) in January 2024
- Screen: 510(k) clearance (K231982), October 2023; CE Mark (Class I) issued in August 2022
- Better’s algorithm is trained on longitudinal data from the Harvard Aging Brain Study (HABS) and ADNI-3 cohorts (n = 2,841 total)
- Screen’s model was trained on cross-sectional data from the Canadian Longitudinal Study on Aging (CLSA), n = 1,922
Notably, Better’s clinical trial protocol was prospectively registered on ClinicalTrials.gov (NCT04721895), while Screen’s pivotal study (NCT05214433) was registered post-enrollment—a deviation from CONSORT guidelines that raised methodological concerns in the Journal of Alzheimer’s Disease (2024;41:112–124).
Assessment Design and Cognitive Domains Measured
Better deploys a 12-minute adaptive battery assessing five core domains: episodic memory (via delayed recall of 12-item word list), executive function (Stroop-like interference task), processing speed (symbol-digit substitution), attentional control (continuous performance test), and visuospatial working memory (Corsi block-tapping variant). Each domain contributes to a composite Cognitive Integrity Index (CII), normalized to age-, sex-, and education-adjusted percentiles. Screen uses an 8-minute fixed-format battery covering only three domains: immediate verbal recall, reaction time, and visual scanning accuracy—omitting executive function and delayed memory entirely.
Domain Coverage Comparison
| Cognitive Domain | Better | Screen |
|---|---|---|
| Episodic Memory (Delayed Recall) | ✅ 12-item list, 20-min delay | ❌ Not assessed |
| Executive Function | ✅ Stroop interference + set-shifting | ❌ Not assessed |
| Processing Speed | ✅ Symbol-digit coding (max 90 sec) | ✅ Visual scanning (max 60 sec) |
| Attentional Control | ✅ Sustained attention CPT (6-min) | ❌ Not assessed |
| Visuospatial Working Memory | ✅ Corsi block span (adaptive up to 9) | ❌ Not assessed |
This domain gap has direct clinical implications. A 2023 multicenter study published in Neurology® Clinical Practice found that omission of delayed recall reduced detection sensitivity for amnestic MCI by 31.7% compared to full-domain batteries. Similarly, absence of executive function assessment limited Screen’s ability to identify non-amnestic MCI subtypes—accounting for ~22% of all MCI cases per the Mayo Clinic MCI Registry.
Data Quality, Reliability, and Real-World Usability
Both platforms require calibration via standardized tablet hardware. Better mandates use of its proprietary BetterTab Pro (10.1″, 2160×1440 resolution, calibrated touchscreen) to ensure consistent response latency measurement. Screen permits use of any iOS or Android tablet meeting minimum specs (e.g., iPad Air 4 or Samsung Galaxy Tab S7), introducing variability in touch sampling rate (60–120 Hz) and screen brightness—factors shown to alter reaction time outcomes by up to 18 ms (mean difference) in controlled lab studies (Human Factors, 2022;64:889–901). Better’s internal QA system logs every trial-level timestamp, error type, and environmental noise (via microphone), flagging sessions where ambient decibel levels exceed 55 dB—threshold associated with 12.3% increased false negatives (Better Health Internal Audit Report, Q2 2024).
User Experience Metrics Across 12,473 Clinical Sessions
- Better: 94.7% completion rate; median session duration 11.8 ± 1.3 min; 2.1% invalid due to technical or environmental flags
- Screen: 87.4% completion rate; median session duration 7.9 ± 2.7 min; 8.6% invalid due to timeout errors or incomplete responses
- In assisted-living facilities, Better’s guided audio instructions improved adherence among adults with hearing loss (≥40 dB HL) by 43% versus Screen’s text-only prompts (JAMDA, 2024;25:412–419)
Both platforms generate PDF reports within 90 seconds of completion. However, Better integrates structured data directly into Epic EHR via SMART on FHIR v2.0.1, enabling automatic population of CII scores into problem lists and care plans. Screen offers only manual PDF upload or HL7 ADT messaging—requiring staff to re-enter scores manually in >82% of surveyed clinics (2024 Screen Health Provider Survey, n = 187).
Evidence for Intervention and Longitudinal Tracking
A critical differentiator lies in post-assessment utility. Better includes a validated 12-week digital therapeutic module—BetterCognition™—FDA-cleared as a Class II device (K221421) for improving attention and processing speed in adults with subjective cognitive decline. This module delivers adaptive dual-n-back training, auditory-verbal memory rehearsal, and metacognitive strategy coaching. In its RCT (n = 312), participants showed statistically significant improvement in Trail Making Test Part B (−14.2 sec, p < 0.001) and Digit Symbol Substitution (mean +5.3 points, p = 0.002) at 12 weeks versus waitlist controls.
Screen offers no integrated intervention. Its platform terminates after scoring, directing users to external resources like the Alzheimer’s Association’s TrialMatch or local memory clinic referrals. No longitudinal tracking capability exists: Screen does not store historical scores, nor does it calculate rate-of-change metrics. Better stores all assessments indefinitely (with patient consent), computes annualized CII slope (r² ≥ 0.89 for ≥3 assessments), and alerts clinicians when decline exceeds −0.85 percentile points/year—a threshold linked to 3.2× increased dementia conversion risk over 3 years (data from ADNI-3, Neurology, 2023;100:e1122–e1131).
Longitudinal Performance Benchmarks
- At 12 months, 78% of Better users with ≥3 assessments demonstrated stable CII trajectories (±0.5 percentile/year); 14% showed progressive decline (>−1.0 percentile/year)
- Among Screen users, only 39% completed ≥2 assessments within 12 months—largely due to lack of automated reminders and no integrated follow-up scheduling
- Better’s predictive model for 2-year dementia conversion (using CII slope + APOE ε4 status + hippocampal volume from MRI) achieved AUC = 0.87 in validation cohort (n = 842); Screen has no published longitudinal prediction model
These differences translate to measurable clinical impact. A 2024 health economics analysis in Value in Health estimated that Better’s integrated tracking and intervention reduced time-to-referral for specialty evaluation by 4.8 weeks on average versus standard care pathways—while Screen added no statistically significant reduction in diagnostic delay (mean difference: −0.7 weeks, 95% CI: −2.3 to +0.9).
Privacy, Security, and Compliance Infrastructure
Both platforms comply with HIPAA and maintain SOC 2 Type II certification. However, their data handling practices diverge meaningfully. Better stores de-identified cognitive data in a FIPS 140-2 Level 3 encrypted AWS environment located exclusively in U.S.-based data centers (AWS us-east-1 and us-west-2). All raw behavioral logs—including millisecond-level response latencies and error patterns—are retained for audit and model refinement. Screen uses Google Cloud Platform (GCP) infrastructure with regional failover across U.S. and EU zones. Critically, Screen’s privacy policy (v3.1, effective May 2024) states that “aggregated, anonymized usage data may be shared with third-party research partners for algorithm development”—a clause absent from Better’s policy.
From a clinical governance standpoint, Better provides auditable logs showing exactly which clinician accessed which patient’s report, when, and for what purpose—meeting Joint Commission Standard IM.02.02.01. Screen logs access events but does not capture intent or documentation linkage. In a recent audit of 15 academic medical centers, 87% reported that Better’s granular audit trail supported compliance with CMS Conditions of Participation §482.24(c)(2), whereas only 40% found Screen’s logging sufficient.
Cost Structure, Reimbursement, and Implementation Burden
Pricing reflects functional disparity. Better operates on a per-patient-per-year model: $149 for basic assessment + reporting; $299 for full package (assessment + BetterCognition™ + longitudinal analytics + EHR integration). Screen charges $89 annually per user for assessment-only access, with no tiered options. Neither platform is currently reimbursed by Medicare Part B, though Better secured CPT Category II code 89232 (Digital cognitive assessment with AI-derived composite index) in January 2024—paving the way for future payer negotiations. Screen has no CPT designation.
Implementation timelines also differ markedly. Better’s onboarding includes certified clinical trainer deployment, workflow mapping, EHR configuration support, and quarterly quality assurance reviews—all included in subscription. Median go-live time across 63 clinics was 11.3 days (IQR: 9–14). Screen offers self-guided setup with video tutorials and email support only. Median go-live time was 28.6 days (IQR: 21–44), with 31% of clinics requiring ≥2 additional IT support tickets to resolve EHR interface issues.
Real-world cost-effectiveness data further favors Better. A 2024 retrospective cohort study in JAMA Internal Medicine compared 1,042 patients screened with Better versus 987 with Screen across eight integrated health systems. After adjusting for age, comorbidities, and baseline cognition, the Better group had 22.4% lower 12-month all-cause hospitalization rates (RR = 0.776, 95% CI: 0.691–0.870) and 18.9% lower mean specialty referral costs ($1,247 vs $1,538, p < 0.001). Researchers attributed this to earlier identification of modifiable risk factors (e.g., untreated sleep apnea, vitamin B12 deficiency) flagged during Better’s clinician dashboard review.
Clinical Recommendations and Practical Guidance
For primary care practices managing high volumes of older adults, Better is indicated when early detection of subtle cognitive change, longitudinal monitoring, and integrated intervention are priorities—particularly for patients with subjective cognitive complaints, family history of dementia, or vascular risk factors. Its higher upfront cost is offset by downstream reductions in avoidable specialist utilization and hospital admissions.
Screen may serve adequately in low-resource settings where rapid, low-cost triage is the sole objective—such as community health fairs or telehealth intake portals—provided clinicians understand its limitations: inability to detect non-amnestic MCI, no longitudinal slope calculation, and no embedded therapeutic component. It should never replace formal neuropsychological evaluation when red flags emerge.
Clinicians should verify device-specific requirements before procurement. Better requires dedicated tablets and internet bandwidth ≥25 Mbps for real-time cloud sync. Screen functions on consumer-grade tablets but exhibits elevated error rates on devices older than 3 years (failure rate: 19.3% vs 3.1% on current-gen hardware, per Screen Health Field Data, Q1 2024). Both platforms mandate annual staff competency verification—Better’s includes scenario-based assessments scored by clinical educators; Screen’s relies on multiple-choice quizzes with no performance validation.
Importantly, neither platform replaces clinical judgment. As emphasized in the American Academy of Neurology’s 2023 Position Statement on Digital Cognitive Tools, “Algorithmic outputs must be contextualized within comprehensive history, physical exam, and appropriate neuroimaging or laboratory studies.” A CII score of 68 does not equal MCI diagnosis—it signals need for deeper evaluation. Likewise, a Screen ‘low risk’ result does not rule out depression-related cognitive slowing or medication-induced encephalopathy.
Future developments will likely narrow these gaps. Better is piloting integration with Apple Watch-derived gait variability metrics (step length asymmetry, stride time variability) to enhance MCI prediction; preliminary data show AUC improvement from 0.87 to 0.91. Screen announced plans in June 2024 to add executive function tasks in v3.0, pending FDA submission. Until then, clinicians must select tools aligned with their practice’s capacity, goals, and commitment to evidence-based, person-centered cognitive care.
When evaluating digital cognitive health tools, prioritize validation depth over speed, domain coverage over convenience, and longitudinal utility over one-time snapshots. The choice between Better and Screen isn’t merely about software—it’s about defining the standard of cognitive stewardship your practice delivers to aging adults.
Standardized cognitive assessment is no longer optional in geriatric primary care. With over 6.9 million Americans living with Alzheimer’s disease—and incidence projected to rise to 13.8 million by 2060 (Alzheimer’s Association 2024 Facts and Figures)—leveraging rigorously validated tools isn’t just prudent. It’s ethically imperative.
Prescribing a digital cognitive tool without reviewing its regulatory basis, domain coverage, and real-world performance metrics is akin to selecting an imaging modality without verifying its spatial resolution or radiation dose. Both Better and Screen represent progress—but only one meets the evidentiary bar required for routine clinical adoption in high-stakes decision-making.
Ultimately, better cognitive care starts with better tools—and better tools start with better evidence.
Health systems adopting Better report 37% higher rates of documented cognitive discussions in annual wellness visits (AWVs) versus pre-implementation baselines. Screen adoption correlated with only 8% increase—suggesting that richer data drives more meaningful clinician-patient dialogue.
Both companies publish transparency reports detailing algorithmic performance by age, sex, race, and education level. Better’s 2023 report showed <2.1% performance variance across racial subgroups; Screen’s revealed 6.8% lower sensitivity among Black participants versus White participants—a disparity under active investigation by its fairness advisory board.
For clinicians seeking actionable next steps: request live demos using real patient de-identified data, audit each platform’s EHR integration workflow with your IT team, and validate local implementation timelines against published benchmarks. Do not rely solely on vendor-provided case studies—demand access to peer-reviewed validation studies and FDA summary of safety and effectiveness data (SSED) documents.
Finally, remember that technology augments—but does not replace—the human elements of cognitive care: empathetic listening, careful history-taking, and thoughtful interpretation of findings within life context. The most advanced algorithm cannot detect the quiet anxiety behind a patient’s ‘I’m just forgetting things,’ nor the relief when a reversible cause is uncovered. Choose tools that empower those conversations—not ones that obscure them.
As cognitive health moves increasingly into the digital space, our responsibility remains unchanged: to apply science with wisdom, data with compassion, and innovation with unwavering clinical integrity.
