Back

Covered in code, not in dollars: what a payer must do to make chronic insomnia's first-line treatment orderable

Jun 11, 2026 · 11:35 AM

Abstract

Chronic insomnia affects roughly 8% of US adults, yet fewer than 10% of patients are referred to cognitive behavioral therapy for insomnia (CBT-I), the first-line treatment in every major guideline; only 2.6% of US adults used any form of it in the past year. We built a four-stage dossier for payer and formulary committees: a cost-of-inaction model, a head-to-head dossier of the two FDA-cleared digital CBT-I devices and their 2025 Medicare codes, a deprescribing pathway for adults 65+ on long-term hypnotics, and an integrated action plan. The status quo costs an estimated $92.49M per 100,000 adult members per year (sensitivity range $16.7M–$146.1M; $5.7M–$68.1M counting medical spend only). The headline finding: Medicare's January 2025 coverage of digital CBT-I is nominal, because the device code G0552 is contractor-priced and no published fee was findable as of June 2026. We recommend five payer actions, led by publishing an explicit device rate (anchors £45 to $899 per course, 3%–43% of the $2,083 observed first-year savings) and a mailed deprescribing package for members 65+ (26.2% vs 7.5% discontinuation, at printing-and-postage cost).

Background

The burden numbers are not in dispute. RAND prices US insomnia-driven productivity loss at $207.5 billion per year, with affected workers losing 44–54 workdays annually [1]. The drugs prescribed instead of CBT-I harm the patients most likely to receive them long term: in adults 65+, benzodiazepines carry a falls odds ratio of 1.57 [8], z-drugs a fracture odds ratio of 1.63 [7], and both classes are "avoid" medications in the 2023 Beers Criteria [11]. The ACP (2016) and AASM (2021) make CBT-I first-line for all adults [12, 13]. The bottleneck is delivery: a 2016 census found 659 credentialed behavioral sleep medicine providers in the US, with four states having zero [4]. The problem is fixable at low cost. The UK's NICE recommended the digital CBT-I program Sleepio at £45 per person in 2022 [23], and a 2024 randomized trial (YAWNS NB) got 26.2% of older long-term hypnotic users off the drugs with a mailed self-help package, against 7.5% under usual care [27].

Method

An autonomous research loop produced four artifacts in sequence: (M1) a per-100,000-adults cost-of-inaction model with line-by-line arithmetic and a sensitivity range; (M2) a product and reimbursement dossier covering FDA clearances, pivotal trials, pricing, and the CY2025/CY2026 CMS rules; (M3) a clinician-facing deprescribing-and-substitution pathway for adults 65+ assembled from published guidelines and trials; (M4) the integrated payer dossier with a seven-row objections table. Each artifact passed an independent fresh-eyes evaluator that checked the milestone's done-criteria and spot-checked three citations against source text. M1–M3 passed on first attempt; M4 failed once (two completion percentages attributed to a paper that never reported them) and passed after the unsupported claim was removed. The evaluator independently reproduced every line of the M1 model. Artifacts: M1, M2, M3, M4.

Findings

1. Medicare's 2025 digital CBT-I coverage exists in code, not in dollars

The naive reading is that CMS solved this problem on January 1, 2025, when the CY2025 Physician Fee Schedule created three codes for digital mental health treatment devices [19]. It did not. The monthly management codes are nationally priced but small (G0553 ≈ $20.06, G0554 ≈ $19.73 per 20 minutes). The code that pays for the device itself, G0552, is contractor-priced: each Medicare Administrative Contractor sets its own amount, and no MAC-published fee was findable as of June 2026. The older per-course code A9291 still has no national Medicare payment. The CY2026 final rule retained and expanded the codes [20], so the policy direction is favorable, but coverage that exists in code without a published dollar amount does not change prescribing. A payer that publishes its own rate converts nominal coverage into an orderable benefit; none was identified that has done so.

2. The status quo costs about $92M per 100,000 adult members per year

Component Central estimate What it counts
A. Productivity loss, untreated sufferers $76.32M RAND $207.5B prorated per adult, 95% untreated share [1]
B. Modifiable healthcare spend $15.83M 7,600 untreated sufferers × $2,083 demonstrated first-year digital CBT-I savings [18]
C. Hypnotic-attributable hip fractures, 65+ $0.34M 6.77 excess fractures × $50,508 first-year cost
Total $92.49M Sensitivity range $16.7M–$146.1M

A committee that credits only medical spend should use B + C alone: $5.7M–$68.1M per 100,000 adults (central ~$16.2M). The supply-side framing of the same gap: ~7,600 untreated chronic insomnia sufferers per 100,000 adults are served by ~0.26 trained providers, about 29,800 untreated patients per provider [4].

3. The covered category is exactly two products, both RCT-backed, one a bankruptcy survivor

An openFDA query for product code QVO returns exactly two cleared devices [14], so a category-level coverage policy is fully enumerable.

Somryst (Nox Health) SleepioRx (Big Health)
FDA 510(k) K191716, cleared 2020-03-23 K233577, cleared 2024-08-05
Pivotal evidence Ritterband 2017, N=303: 1-year remission 56.6% vs 27.3% [17] Freeman 2017, N=3,755: Sleep Condition Indicator +4.78 points (95% CI 4.29–5.26), d=1.11 [16]; Espie 2012, N=164: sleep-efficiency d=0.95–1.06 [15]
Published price $899 per course (2020, Pear era); current price not found US price not found; NICE UK contract £45 per person [23, 24]
Real-world economics Conference data only Matched claims cohort, N=21,797: −$2,083 per patient (−42%) in year one; pharmacy −$1,745 (−46%) [18]

Market risk is demonstrated, not hypothetical: Somryst's original maker, Pear Therapeutics, filed Chapter 11 in April 2023 [21], and Nox Health bought the product for $3.9M [22]. The clearance survived. Covering the category by regulation number (21 CFR 882.5801), with data-portability and continuity clauses, is the hedge.

4. For members 65+, paper and pharmacists do most of the deprescribing work

The highest-harm group needs almost no clinician time to start exiting the drugs.

Intervention RCT result (discontinuation, intervention vs control) Delivery cost driver
Mailed self-help package (YAWNS NB, N=565) 26.2% vs 7.5% at 6 months [27] Printing and postage; no clinician time
EMPOWER brochure (N=303) 27% vs 5%, NNT 4 [26] Pharmacist hand-off
Pharmacist opinion to prescriber (D-PRESCRIBE) 43.2% vs 9.0% at 6 months [28] Pharmacist professional fee
Taper + therapist CBT-I (Morin 2004; Baillargeon 2003) 85% vs 48%; 77% vs 38% (70% vs 24% at 12 months) [29, 30] 8–10 therapist sessions
Nurse-delivered brief behavioral treatment (N=79) Remission 55% vs 13%, NNT 2.4 [31] ~2.5 hours of nurse time

No published RCT combines an FDA-cleared digital CBT-I product with a structured taper in adults 65+. The recommended program therefore stands on the mailed-package and pharmacist steps, which are RCT-proven in this age group, with digital CBT-I as one of three delivery options for the therapy step.

5. Dropout is priced into the headline evidence, not hidden by it

The standard objection to digital CBT-I is attrition, and it is real: a 2025 meta-analysis found 59.33% average completion in fully automated programs (range 16.67%–85.71%) [33]. But in Freeman 2017, only 18% of the digital arm completed all six sessions, and the intention-to-treat effect was still d=1.11 [35]. The trial effect sizes already include the people who quit. In the older-adult RCT closest to the target program (SHUTi OASIS, N=311), completion was 83% with effects of d −1.04 to −1.43 [32].

Recommendations

# Action Executor Cost Anchor
1 Add G0552/G0553/G0554 to the fee schedule and publish an explicit G0552 device rate (A9291 for commercial/Medicaid lines) Medical policy + network contracting Device rate between £45 and $899 per course, 3%–43% of the $2,083 observed first-year savings [18, 23, 24]
2 Issue a medical-necessity policy covering the category by regulation (21 CFR 882.5801 / QVO), naming K191716 and K233577 Medical policy committee Policy drafting only Highmark precedent [34]; openFDA [14]
3 Claims-triggered deprescribing program for members 65+ on a hypnotic >4 weeks: mailed package plus reimbursed pharmacist opinions Pharmacy team, mailroom, community pharmacists Printing, postage, pharmacist fee YAWNS NB 26.2% vs 7.5% [27]; D-PRESCRIBE 43.2% vs 9.0% [28]
4 Refill-point prompt (not a hard edit) on hypnotic refills past ~4 weeks offering covered CBT-I; no step therapy requiring drug failure first PBM / claims system One claims-system edit Beers 2023 [11]; ACP 2016 [12]
5 Contract for vendor failure (data portability, mid-course continuity) and track G-code volume, completion, and sleep outcomes Contracting + analytics Contract language Pear Chapter 11, $3.9M asset sale [21, 22]

Sequencing: Action 3 can launch immediately at near-zero cost and pays off regardless of the digital coverage question. Actions 1 and 2 together make the digital option orderable and are the precondition for Action 4's prompt to offer something real. Action 5 rides along with the contracting in Action 1.

Limitations

  1. Productivity loss is 83% of the central cost estimate and is the softest input; RAND's figure does not fully isolate insomnia from comorbid conditions, and the comorbidity-adjusted alternative is 3.3 times lower. The medical-only floor ($5.7M per 100,000) survives this objection.
  2. The $2,083 savings figure is observational, vendor-adjacent, and conditioned on prescribed-and-engaged patients; selection bias likely inflates it.
  3. The digital-CBT-I-plus-taper combination in adults 65+ is an extrapolation with no supporting RCT. If it underperforms, program yield falls toward the mailed-package rate (~26% discontinuation), still 3.5 times usual care.
  4. No payer was found that has set a G0552 rate, so Action 1 has no named precedent, and MACs may resolve pricing without payer action; waiting 12 months is defensible, with the cost of delay bounded by the per-100,000 figures.
  5. The harm odds ratios are observational and carry confounding-by-indication and reverse-causation risk; insomnia itself raises fall risk, and prodromal dementia may drive benzodiazepine prescribing rather than the reverse.
  6. Pricing inputs are stale or foreign: the $899 list price is from 2020 under a now-bankrupt maker, and the £45 figure is a UK national-procurement price. Current US contract prices for both products are unpublished.

One label was corrected at publication: the Freeman 2017 effect (+4.78, d=1.11) was carried through M2 and M4 as an "ISI" result; it is the trial's Sleep Condition Indicator co-primary outcome. The values match the source; only the instrument name was wrong.

References

  1. RAND, RRA2166-1: The societal and economic burden of insomnia in adults (2023). https://www.rand.org/news/press/2023/03/17.html
  2. Behavioral Sleep Medicine (2026), national survey of CBT-I awareness and use. https://doi.org/10.1080/15402002.2025.2610674
  3. Koffel et al., J Gen Intern Med (2018). https://pmc.ncbi.nlm.nih.gov/articles/PMC5975165/
  4. Thomas et al., J Clin Sleep Med (2016), behavioral sleep medicine provider census. https://pmc.ncbi.nlm.nih.gov/articles/PMC5070478/
  5. FDA Drug Safety Communication, benzodiazepines (2020). https://www.fda.gov/media/142368/download?attachment=
  6. ClinCalc DrugStats: Zolpidem. https://clincalc.com/DrugStats/Drugs/Zolpidem
  7. Treves et al., Age and Ageing (2018), z-drug fracture meta-analysis. https://academic.oup.com/ageing/article/47/2/201/4564456
  8. Woolcott et al., Arch Intern Med (2009), falls meta-analysis. https://pubmed.ncbi.nlm.nih.gov/19933955/
  9. Billioti de Gage et al., BMJ (2014), benzodiazepines and Alzheimer's. https://doi.org/10.1136/bmj.g5205
  10. Richardson et al., BMC Medicine (2020), z-drugs in dementia. https://doi.org/10.1186/s12916-020-01821-5
  11. American Geriatrics Society, 2023 Beers Criteria. https://agsjournals.onlinelibrary.wiley.com/doi/abs/10.1111/jgs.18372
  12. Qaseem et al., ACP guideline, Ann Intern Med (2016). https://www.acpjournals.org/doi/10.7326/M15-2175
  13. Edinger et al., AASM behavioral treatment guideline, J Clin Sleep Med (2021). https://jcsm.aasm.org/doi/10.5664/jcsm.8986
  14. openFDA 510(k) query, product code QVO. https://api.fda.gov/device/510k.json?search=product_code:QVO&limit=20
  15. Espie et al., SLEEP (2012), Sleepio RCT. https://pmc.ncbi.nlm.nih.gov/articles/PMC3353040/
  16. Freeman et al., Lancet Psychiatry (2017), OASIS trial. https://pubmed.ncbi.nlm.nih.gov/28888927/
  17. Ritterband et al., JAMA Psychiatry (2017), SHUTi RCT. https://pubmed.ncbi.nlm.nih.gov/27902836/
  18. J Health Econ Outcomes Res (2025), SleepioRx matched claims cohort. https://doi.org/10.36469/001c.146434
  19. CMS, CY2025 Physician Fee Schedule final rule summary (MM13887). https://www.cms.gov/files/document/mm13887-medicare-physician-fee-schedule-final-rule-summary-cy-2025.pdf
  20. CMS, CY2026 Physician Fee Schedule final rule fact sheet (CMS-1832-F). https://www.cms.gov/newsroom/fact-sheets/calendar-year-cy-2026-medicare-physician-fee-schedule-final-rule-cms-1832-f
  21. SEC 8-K, Pear Therapeutics Chapter 11 (2023). https://www.sec.gov/Archives/edgar/data/0001835567/000183556723000020/peartherapeuticsfilesforch.htm
  22. GlobeNewswire, Nox Health acquisition of Somryst (2023). https://www.globenewswire.com/news-release/2023/06/20/2691426/0/en/Nox-Health-Closes-Acquisition-of-Somryst-the-Only-FDA-Cleared-Digital-Insomnia-Treatment-From-Pear-Therapeutics.html
  23. NICE, MTG70: Sleepio (2022). https://www.nice.org.uk/guidance/mtg70/chapter/2-The-technology
  24. MedCity News, Somryst US launch price (2020). https://medcitynews.com/2020/11/pear-rolls-out-digital-therapeutic-for-insomnia/
  25. Pottie et al., Can Fam Physician (2018), BZRA deprescribing guideline. https://pmc.ncbi.nlm.nih.gov/articles/PMC5951648/
  26. Tannenbaum et al., JAMA Intern Med (2014), EMPOWER trial. https://pubmed.ncbi.nlm.nih.gov/24733354/
  27. Murphy et al., JAMA Psychiatry (2024), YAWNS NB trial. https://jamanetwork.com/journals/jamapsychiatry/fullarticle/2823668
  28. Martin et al., JAMA (2018), D-PRESCRIBE trial. https://pmc.ncbi.nlm.nih.gov/articles/PMC6248132/
  29. Morin et al. (2004), taper plus CBT-I RCT. https://pubmed.ncbi.nlm.nih.gov/14754783/
  30. Baillargeon et al., CMAJ (2003), taper plus CBT-I RCT. https://pmc.ncbi.nlm.nih.gov/articles/PMC236226/
  31. Buysse et al., JAMA Intern Med (2011), nurse-delivered BBTI. https://pmc.ncbi.nlm.nih.gov/articles/PMC3101289/
  32. npj Digital Medicine (2025), SHUTi OASIS older-adult RCT. https://www.nature.com/articles/s41746-025-01847-0
  33. npj Digital Medicine (2025), completion meta-analysis of fully automated dCBT-I. https://www.nature.com/articles/s41746-025-01514-4
  34. STAT News (2022), Highmark PDT coverage policy. https://www.statnews.com/2022/10/19/highmark-insurers-decision-to-cover-prescription-digital-therapeutics/
  35. Freeman et al. (2017), full text with session-uptake data. https://pmc.ncbi.nlm.nih.gov/articles/PMC5614772/

Provenance

This report was produced by an autonomous research loop: a planner wrote the four-milestone spec, a generator produced each artifact, and an independent evaluator with fresh eyes passed each one against quoted done-criteria, including citation spot-checks (12 sampled, 11 verified exactly; the one failure forced a rewrite before M4 passed). All artifacts and pass/fail verdicts are in artifacts/ and verdicts/ in this directory.