DTx Paradox

2026-08-07 — Opened (ben-steer)

Opened as a dedicated child thread of mh-evidence-watch, seeded from the EBP digital/AI-science research crawl. (EBP digital/AI-science crawl, 2026-08-07) ⟨steer 2026-08-07⟩

PHTI’s newest report, “Payment for Clinical AI,” extends the identical argument it made about mental-health DTx in 2025 — fee-for-service and capitation both misalign incentives, “autonomous clinical AI requires new payment models, not incremental modifications” — to all of clinical AI, using hypertension as the case study. The diagnosis has spread past mental health even though no commercial fix has emerged anywhere it’s been applied; PHTI’s own scope just broadened from “the mental-health DTx referee” to “the payment-model referee for all of digital/AI health.” (PHTI) (wave: mh-q07) ⟨daily 2026-08-07⟩

Akili’s OTC pivot (EndeavorOTC) generated real early traction at launch, but total company revenue was still only ~$1.94M trailing-twelve-months as of April 2026 — three years post-launch and two years into private ownership. Click Therapeutics (Rejoyn’s maker) raised a reported $50M Series D around the same month alongside reported workforce cuts — capital-raising-plus-cutting is the same pattern that preceded Pear’s collapse, now visible again roughly two years after Rejoyn’s FDA clearance. (pharmaphorum, Behavioral Health Business) (wave: mh-q07) ⟨daily 2026-08-07⟩

DiGA’s ≥20% outcomes-linked pricing mandate became legally live January 1, 2026 — but the performance data that would actually drive any re-pricing isn’t due to the regulator (BfArM) until April 15, 2027. Mechanically, no DiGA product has been re-priced on performance yet; quarterly real-world usage data collection (AbEM) doesn’t even begin until Q3 2026. (Inside EU Life Sciences) (wave: mh-q07) ⟨daily 2026-08-07⟩

From 2026, Germany’s DiGA digital-therapeutics reimbursement scheme adds mandatory outcomes monitoring and ties at least 20% of a product’s price to performance — a direct regulatory response to the evidence quality problem the 2025 review below identified. Exact effective date within 2026 not specified in the underlying crawl. (PubMed) ⟨steer 2026-08-07⟩

PHTI (Peterson Health Technology Institute, ICER-partnered) — now the field’s most credible independent DTx referee — found in its 2025 mental-health category assessment that blended-care tools show the largest clinical benefit of any category but the worst cost structure: flat access-fee pricing regardless of usage inflates employer/payer spend rather than lowering it. Self-guided tools (Headspace, Talkspace, SilverCloud) cut net payer spend with modest clinical benefit; prescription DTx (DaylightRx, Rejoyn) help mainly combined with other care. PHTI recommends outcomes-based payment instead. (MedCity News) ⟨steer 2026-08-07⟩

CMS’s DMHT HCPCS codes (G0552 device/onboarding, contractor-priced; G0553 $20.06 for the first 20 minutes/month of treatment management; G0554 $19.73 per additional 20 minutes) have been live since the CY2025 Physician Fee Schedule, but CMS’s own rulemaking record acknowledges claims volume remains low. Outside counsel attributes this partly to CMS never setting a national payment rate for the device itself — pricing is deferred to opaque, inconsistent Medicare Administrative Contractor decisions, a bureaucratic replay of the reimbursement failure that sank Pear Therapeutics, now built into Medicare’s own architecture. CMS’s 2026 proposal would extend eligibility to FDA-special-controls ADHD digital therapy devices. (Nixon Law Group, AAPC) (wave: mh-q07) ⟨daily 2026-08-07⟩

A systematic review of all 23 DiGA approval studies as of March 2024 (npj Digital Medicine, published mid-2025; n=56-1245 per study, 1.5-12 month durations) found every one reported significant medium-to-large effects — but risk of bias was consistently high, especially around outcome measurement and missing data. The evidence looks strong in aggregate and shaky study-by-study — the direct predicate for the 2026 reform above. (PubMed) ⟨steer 2026-08-07⟩

Digital mental-health attrition runs 18.6% at posttest, rising to 28.4% at follow-up, per an npj Digital Medicine review — and there is no agreed metric for “engagement” and no clean evidence that more engagement causes better outcomes. Reminders and human support improved retention; gamification did not. Month of publication not specified in the underlying crawl. (Nature) ⟨steer 2026-08-07⟩

The lead author’s own follow-up work explicitly frames mental-health apps as defensible mainly “if no evidence-based first-line intervention is available” — a fallback, not a first choice — in an updated systematic review and meta-analysis (Linardon, Torous, Firth et al., Lancet Digital Health, 2025) building on the 176-RCT base below. Month of publication not specified in the underlying crawl. (The Lancet) ⟨steer 2026-08-07⟩

The two largest meta-analyses of mental-health apps (176 RCTs, ~33,567 participants combined) find small but statistically real effects: depression g=0.28, generalized anxiety g=0.26, NNT approximately 11-12 (Linardon et al., World Psychiatry, 2024). This is the evidence-isn’t-the-bottleneck baseline the rest of this thread argues against: apps clear a real, modest efficacy bar, and still lose on reimbursement and business-model design. (PubMed) ⟨steer 2026-08-07⟩

Pear Therapeutics — $1.6B valued, 40+ studies including multiple RCTs in top journals, one of the strongest RCT evidence bases of any prescription digital therapeutic — went bankrupt anyway. Industry post-mortems attribute the collapse to spending roughly $150M on a broad pipeline before market validation, over-reliance on Medicaid reimbursement, and trying to disrupt both care delivery and payment simultaneously. The founding case for this thread’s core claim: strong evidence did not translate to reimbursement or adoption. (MedCity News) ⟨steer 2026-08-07⟩