Seeking Well-Being
April 20, 2026
Digital therapeutics are crossing from clinical curiosity into reimbursed, prescribed medicine — but the patient experience remains fragile. Evidence is strongest in diabetes and mental health; engagement collapses after the first weeks; trust in digital health is at a historic low; and the populations most likely to benefit are least equipped to access or sustain it.
Digital therapeutics — software-delivered interventions designed to prevent, manage, or treat medical conditions — are moving from clinical curiosity to a reimbursed, prescribed category of care. The evidence base for their effectiveness is stronger than it was three years ago, particularly in chronic disease management and mental health. But the patient experience of using these tools remains fragile: engagement drops sharply after the first weeks, trust in digital health overall is at a historic low, and the populations most likely to benefit from digital care are often the least equipped to access or sustain it. The field is bifurcating: a small cohort of clinically validated, FDA-cleared, and now CMS-reimbursed products is gaining legitimacy as prescribed medicine; the vast majority of the 318,000+ available health apps operate without clinical evidence, regulatory oversight, or meaningful integration into care. Patients are left to navigate this landscape largely alone.
The strongest evidence for digital therapeutics is concentrated in a handful of condition categories. A 2025 JMIR systematic review and meta-analysis of randomised controlled trials found digital interventions for diabetes management produced measurable improvements in HbA1c, blood pressure, and weight. IQVIA's analysis of digital health apps across five conditions — diabetes, asthma, cardiac rehabilitation, pulmonary rehabilitation, and depression — estimates $7 billion in potential annual U.S. savings through reduced acute care utilisation if validated apps were deployed at scale.
For mental health, the picture is meaningful but should not be overstated. A meta-analysis of 176 randomised controlled trials (PMC, 2024) found mental health apps produced statistically significant but small effects on depression (effect size g=0.28) and generalised anxiety (g=0.26). Effect sizes were meaningfully larger when apps incorporated cognitive behavioural therapy or chatbot features, and when targeting specific conditions rather than general wellness. A 2025 PLOS One systematic review of evidence-based mental health apps confirmed consistent but modest improvements across depression, anxiety, and suicidal behaviours.
WellDoc's BlueStar — one of the most-studied prescription digital therapeutics — has 571 published clinical studies on app efficacy across its diabetes management platform, making it among the most evidence-rich DTx products in existence. This is exceptional; most DTx products have far thinner evidence portfolios.
The gap between "clinically validated" and "app store available" is enormous. Of the 318,000+ health apps available globally (Deloitte), the overwhelming majority have no clinical validation at all. The DTx Alliance and AMCP have called for evidence requirements as a condition of reimbursement — but until coverage is broadly linked to evidence, patients have no reliable signal for which tools actually work.
Confidence: HIGH for efficacy in diabetes and chronic disease management DTx; MEDIUM for mental health apps (consistent but small effect sizes, significant variation across products); LOW for the broader app market, where evidence is largely absent.
Sources: JMIR (digital therapeutics for diabetes, systematic review and meta-analysis, 2025), IQVIA (The Growing Value of Digital Health — $7B savings estimate), PMC meta-analysis of 176 RCTs (mental health apps, 2024), PLOS One (evidence-based mental health apps, 2025), Osmosis/WellDoc BlueStar (571 published studies), Deloitte (318,000 health apps worldwide)
Medication adherence in chronic disease is already poor — only 50% of chronically ill patients take medications as prescribed. Digital therapeutics face a compounding version of this problem: they must first achieve uptake, then sustain engagement, and only then can they deliver clinical benefit.
A 2025 JMIR systematic review on mobile apps for medication adherence in chronic conditions found that reminder functionality increased adherence from 50% to 67.8% — the single most effective feature tested. But reminder-only design is insufficient for long-term engagement. The same review found that while reminders were used in 92% of apps, behavioural strategies appeared in only 45% and educational strategies in just 9%. The tools are built around the least sophisticated lever.
A Frontiers in Digital Health 2025 paper found that adherence to a digital therapeutic mediates the relationship between momentary self-regulation and health risk reduction — in other words, the patients who most need the tools are often the ones least equipped to use them consistently. A 2025 Nature publication identified the motivational strategies with the strongest evidence: feedback, goal-setting, social interaction, gamification, and motivational messages — but noted these are underrepresented in deployed apps relative to their evidence base.
The dropout problem is acute. A npj Digital Medicine 2026 scoping review on treatment dropout in blended digital therapy found patients reporting low therapeutic alliance — a weak sense of human connection — are 2.58 times more likely to drop out. Financial barriers compound this: patients with even modest copay obligations ($40 at first visit) were twice as likely to disengage. The implication is that dropout is not primarily a design problem within the app — it is a relational and financial problem that surfaces within the app.
Virtual care platforms that have prioritised longitudinal engagement over episode-based interaction are showing better results. Wheel's 2026 data shows 70% of users return for repeat visits when virtual care is designed for continuous management rather than one-off encounters.
Confidence: HIGH for reminder effectiveness on short-term adherence; HIGH for dropout risk from low therapeutic alliance and financial barriers; MEDIUM for long-term sustained engagement — evidence is limited beyond 6-month follow-up windows in most studies.
Sources: JMIR (mobile apps and medication adherence, systematic review 2025), Frontiers in Digital Health (self-regulation and DTx adherence, 2025), Nature (motivational strategies in digital health, 2025), npj Digital Medicine (dropout in blended therapy, 2026), Wheel (70% returning virtual care users, 2026)
Patient trust in digital health tools is not a neutral backdrop — it is an active risk factor. A national STAT News survey of 443,000 U.S. adults (March 2026, newsroom) found physician and hospital trust collapsed from 72% to 40% between 2020 and 2024 — a 30-point drop spanning all sociodemographic groups. Rushed AI deployments are accelerating this decline. Patients who don't trust their clinicians are structurally unlikely to trust clinician-prescribed digital tools.
The AI-specific trust failure is already playing out in consumer health. A Cult of Mac investigation (January 2026, newsroom) documented Washington Post tech columnist Geoffrey Fowler connecting 29 million steps and 6 million heartbeats of Apple Watch data to ChatGPT Health — receiving a cardiac assessment that swung wildly between F and B grades across repeated identical queries. The assessment was described as "baseless." This is not an edge case; it is a preview of what happens when AI health tools are launched into a consumer context without clinical guardrails, and the reputational damage extends to legitimate DTx products that share the same digital health category in patients' minds.
A PatientPoint survey (newsroom) found 88% of patients prefer doctor-delivered medical information over AI, and trust physicians four times more. This is a significant finding for DTx design: tools prescribed by a trusted clinician carry a fundamentally different adoption profile than tools discovered through an app store. The therapeutic pathway matters as much as the therapeutic content.
Perplexity Health's March 2026 launch — integrating Apple Health with 1.7 million provider records, Fitbit, Withings, and Clue — illustrates where consumer-grade and clinical-grade digital health are colliding. The product is powerful, but it sits in ambiguous territory between personal health assistant and medical tool, without the clinical validation or prescriber context that gives patients a reason to trust it.
Confidence: HIGH for the trust collapse in medicine broadly; HIGH for preference for clinician-mediated tools over app-store tools; MEDIUM for the specific impact of AI failures on DTx adoption — directionally strong but not yet quantified rigorously.
Sources: STAT News (30-point trust collapse in medicine, 443,000-patient survey, March 2026), Cult of Mac (ChatGPT Health and Apple Watch failure, January 2026), PatientPoint (88% prefer physician info over AI), 9to5Mac/Perplexity Health launch (March 2026) — all newsroom items
2025 marked a genuine inflection point for DTx reimbursement. CMS introduced three new HCPCS codes covering digital mental health treatments under Medicare, effective January 1, 2025 — the first time a broad digital therapeutic category has received defined Medicare coverage. In November 2025, CMS expanded coverage to include DTx for ADHD. The newsroom's Galen Growth item noted that evidence-based digital therapeutics with CMS reimbursement are delivering 20–40% improvements in utilisation and cost-of-care reduction in mental health. Cigna Healthcare announced commercial coverage of FDA-approved digital therapeutics in September 2025.
CMS's ACCESS Model, launching in 2026, reimburses digital chronic care tools — including CGMs, biosensors, wearables, and apps — for the more than two-thirds of Medicare patients with targeted chronic conditions (diabetes, hypertension, chronic pain, depression). Over 60 companies have pledged collaboration with CMS on this model.
The friction points are real, however. The HCPCS codes require a "buy-and-bill" model — providers pay upfront and seek reimbursement later — which is unfamiliar to most clinicians and particularly burdensome for mental health professionals operating in solo practices. The Access to Prescription Digital Therapeutics Act (H.R.1458), which would create a formal Medicare benefit category for DTx similar to pharmaceuticals, has not yet passed. Without it, DTx coverage remains fragmented, model-dependent, and administratively complex. A 2025 npj Digital Medicine paper documented that only a handful of DTx products have achieved claims-based reimbursement due to insufficient long-term outcome evidence.
Confidence: HIGH for the CMS coverage milestone as a regulatory signal; MEDIUM for real-world implementation — provider adoption of new billing models is historically slow, and commercial coverage from Cigna does not yet represent industry-wide shift.
Sources: Healthcare Brew (CMS Medicare coverage of DTx, April 2025), CMS ACCESS Model documentation (newsroom), Galen Growth (DTx reimbursement and outcomes, newsroom), APA Services (reimbursement pathways for DTx), npj Digital Medicine (DTx reimbursement barriers, 2025), Covington Digital Health (CMS 2026 developments, newsroom)
The populations with the highest chronic disease burden — lower-income, less educated, racially and ethnically diverse communities — are the least likely to be using digital health tools consistently. A PMC 2024 study found significant disparities in remote patient monitoring, medical apps, and wearables use in lower-income and racially diverse populations, even when controlling for smart device ownership and internet access. The barrier is not just connectivity — it is digital literacy, language and cultural fit of the tools themselves, and the structural absence of a prescribing clinician who might bridge the gap.
At least 21 million Americans lack broadband access (Johns Hopkins Bloomberg School of Public Health, 2025). Mid-2025 federal policy changes cancelled Digital Equity Act awards that had funded health-focused broadband and digital literacy projects, removing a significant structural support at a critical moment. Higher education and income ($35,001+) remained strongly associated with medical app and wearables use.
The newsroom's CMS ACCESS Model item is significant in this context: it explicitly targets chronic condition management for Medicare patients — a population that skews older, lower-income, and often less digitally fluent. Whether the model's product mandate translates into actual patient use will depend on implementation choices that are not yet resolved.
The consumer electronics market is moving in the opposite direction. Apple's introduction of the Withings BeamO ($299.95) — a clinical-grade ECG, thermometer, lung scanner, and oximeter — represents the normalisation of medical-grade diagnostics for patients who can afford them. Apple AirPods Pro's FDA clearance as a hearing aid and diagnostic tool follows the same logic. These are genuine advances for the patients who have them. For the patients who don't, they widen the capability gap.
Confidence: HIGH for the equity gap in digital health adoption; MEDIUM for the long-term policy risk from Digital Equity Act cuts — too early to quantify impact; MEDIUM for whether CMS ACCESS Model will reach underserved populations — design intent and implementation reality may diverge.
Sources: PMC (equity in remote patient monitoring, apps, and wearables, 2024), Johns Hopkins Bloomberg School of Public Health (digital divide framework, 2025), Health Affairs (digital inclusion and health equity, 2025), Frontiers in Digital Health (redefining the digital divide, 2025), 9to5Mac/Withings BeamO and AirPods Pro (newsroom), CMS ACCESS Model (newsroom)
Three product categories are doing the most to reshape what patients experience as "digital therapeutics":
Prescription digital therapeutics (PDTs) — FDA-cleared, clinician-prescribed software products — are the most clinically credible but also the most friction-laden to access. The buy-and-bill reimbursement model, limited prescriber familiarity, and front-loaded patient effort create a consistent gap between clinical availability and actual use. Products in this category are now appearing in oncology, psychiatry, substance use, and chronic pain management.
Consumer wearables with clinical-grade sensing are collapsing the boundary between wellness tracking and medical monitoring. Continuous glucose monitors, ECG-capable watches, blood oxygen sensors, and now the Withings BeamO are generating the kind of longitudinal physiological data that supports proactive chronic care management — but only for patients who can afford them, use them consistently, and have a care team prepared to interpret and act on the data.
AI-powered health companions — Perplexity Health, ChatGPT Health, and similar products — are normalising the idea of a personal health AI but are doing so ahead of the evidence base. The ChatGPT Health/Apple Watch failure documented in the newsroom illustrates the specific risk: patients making health decisions based on AI output that is inconsistent and unvalidated. The patient experience of this failure is not "I got a bad AI result" — it is "I can't trust digital health tools," which has downstream effects on legitimate DTx adoption.
Social prescribing platforms represent a quieter thread worth watching. The Johns Hopkins International Arts + Mind Lab research finding that one or more artistic experiences per month can extend life by up to 10 years has accelerated social prescribing across 30 countries — a reminder that the most effective "digital therapeutic" for some populations may be a navigated connection to community, not a software intervention.
Does clinician prescription meaningfully change patient engagement with DTx? The trust data suggests it should — patients trust clinician-recommended tools far more than app-store discoveries. But the mechanistic evidence for how much prescriber endorsement affects sustained engagement (vs. initial uptake) is thin. If the therapeutic alliance is with the prescribing clinician rather than the app itself, what happens when a patient changes providers?
What is the real cost of dropout? Studies track dropout rates but rarely measure the downstream impact on patients who disengage. Does early dropout lead to treatment disengagement more broadly, or is it relatively neutral? For conditions like depression or substance use disorder, this is not a small question.
Can the CMS ACCESS Model close the equity gap, or will it replicate existing disparities? The model's chronic disease targeting is well-aligned with need. But if implementation requires digital fluency, device ownership, and a tech-enabled care team, it may serve connected, suburban Medicare patients while missing rural and low-income populations.
What counts as "evidence" for a digital therapeutic as the category matures? RCTs are the gold standard but are expensive, slow, and often conducted in highly selected populations that don't represent real-world users. Real-world evidence (RWE) approaches are being explored but don't yet have consistent payer acceptance. The evidentiary standards the field agrees on will determine which products survive.
How do patients experience the difference between a wellness app and a digital therapeutic? From a design and regulatory standpoint, these are distinct categories. From the patient's perspective — especially for mental health tools — the line is often invisible. This creates both a trust risk (if a wellness app fails, it colours perception of a clinical tool) and a design opportunity (if clinical tools can feel as intuitive as consumer apps, uptake improves).
1. The prescriber relationship is the product's most important design element. The trust data is unambiguous: patients engage more with tools their clinician recommended and explained. Every DTx product should have a designed prescriber onboarding experience that is as carefully considered as patient onboarding — not because clinicians are the customer, but because their endorsement is the patient's primary engagement signal. Tools that treat the prescriber touchpoint as an afterthought will face chronic dropout.
2. Reminder-only apps are not digital therapeutics — they are alarms. The adherence evidence is clear: reminders help in the short term but don't sustain behaviour change. Apps that want to produce lasting clinical benefit need to invest in behavioural features — goal-setting, feedback loops, social elements, motivational scaffolding — even though these are harder to build and harder to evaluate.
3. Design for dropout, not just for engagement. The 2.58x dropout risk from low therapeutic alliance is a design brief in itself. What does graceful re-engagement look like? What signals early dropout? How does the app maintain a minimal useful presence for patients who are struggling with adherence without creating guilt or abandonment? The post-dropout experience is almost never designed for.
4. Consumer AI failures have clinical consequences. When ChatGPT gives a patient an inconsistent cardiac assessment, it doesn't just create a bad experience — it shifts the patient's prior about all digital health tools. DTx companies should treat the consumer AI reliability crisis as an external threat to their category, and should build explicit trust signals into their products (clinical validation badges, prescriber connection, FDA-cleared labelling) that differentiate them from unvalidated tools.
5. Equity must be a design constraint, not an afterthought. The populations most burdened by chronic disease are the least reached by current digital therapeutics. Designing exclusively for smartphone-fluent, broadband-connected, English-speaking patients is not a neutral choice — it is an active decision to serve a smaller, less-sick population. DTx products that want to demonstrate real-world clinical impact must set explicit equity targets in their design requirements: literacy level of content, language accessibility, low-data mode functionality, and offline capability.
This report was produced using the StillHuman intelligence methodology: parallel newsroom query and web research synthesis, with confidence scoring applied to all major claims. Items marked "evergreen" have no confirmed publication date and should not be cited as recent findings.
Intelligence reports are synthesised from public health data, published research, and industry sources.