In April 2023, the company that made history with the first FDA-authorised prescription digital therapeutic filed for bankruptcy. Pear Therapeutics had regulatory clearance, clinical trials and a product, reSET, aimed at one of medicine’s hardest problems: substance use disorder. What it did not have was enough payers willing to cover it.
The coverage it did win came mostly from a handful of state Medicaid programmes. Private insurers largely stayed away, and prescriptions never reached the volume the business needed.
The standard story about digital therapeutics is that payers won’t pay. The more uncomfortable story is that nobody in the care pathway was clearly responsible for making them work, and payers noticed.
The German experiment answered the payment question
If lack of reimbursement were the core problem, Germany should have solved it. Since 2020, its DiGA pathway has let certified digital health applications be prescribed and reimbursed by statutory health insurance, with a provisional listing while evidence is gathered.
It worked, in the narrow sense. By the end of 2024, according to the statutory insurers’ association GKV, 68 applications were listed in the directory kept by the federal drug and device regulator BfArM. Insurer spending on them rose 71% between 2023 and 2024, and cumulative spending since September 2020 reached about €234 million. The average price per prescription in 2024 was €541.
It also produced the most honest data set in the field. By mid-2025, eleven applications had been removed from the directory, six of them because they could not demonstrate a positive healthcare effect.
So Germany showed that if you guarantee payment, products will come. It also showed that guaranteed payment does not guarantee evidence of benefit, and that a meaningful share of what arrives will not prove its case.
Payment unlocked supply. It did not create a buyer who was confident about what they were buying.
Why the US route is so hard
In the United States, the problem is fragmentation. A digital therapeutic has to convince hundreds of payers, each with its own evidence thresholds, and fit into a billing system built for pills, procedures and visits.
Medicare took a real step with its 2025 physician fee schedule, which created three new codes for digital mental health treatment devices. G0552 covers supply of a device cleared or authorised by the FDA under a specific classification; G0553 and G0554 cover monthly treatment management time. It is a genuine opening.
Look at what the codes say, though. The device must be used “incident to” behavioural health treatment under a plan of care. The management codes pay clinicians roughly $20 for each 20 minutes of work reviewing device data, per CMS estimates. Payment for the device itself is left to contractor pricing.
In other words, Medicare did not buy an app. It bought a clinician’s time supervising an app. That is a sensible design, and it exposes the real bottleneck: digital therapeutics only work when a clinician is paid, trained and motivated to prescribe them, onboard the patient and act on the data. Most of the business models in the sector assumed the product could substitute for that clinician, not depend on them.
The incentive underneath
Here is the trade-off that rarely gets said aloud. A drug has a manufacturer with an incentive to sell it and a pharmacist who dispenses it. A device has a surgeon who implants it. A digital therapeutic, once prescribed, has no one in the system whose income depends on the patient actually opening it on day ten.
Payers see that gap clearly. They are asked to pay drug-like prices for products whose real-world effect depends on engagement they cannot observe. Oklahoma’s Medicaid programme tried to manage that risk with a value-based arrangement with Pear, paying upfront with the understanding that the company would reimburse the state if savings did not materialise. After the bankruptcy, STAT reported, the state paused its digital therapeutics efforts.
That is what a burned buyer looks like. And the next company pitching the same payer pays the price.
What would actually move reimbursement
The sector’s instinct is to lobby for more codes and dedicated pathways. Those help at the margins. But the evidence from Germany and the US points to a different set of priorities:
- Report engagement and outcomes together. A trial result tells a payer what happens when patients use the product as intended. Payers need to know how many actually will.
- Price for the pathway, not the product. If a clinician must supervise use, the offer should include that clinician’s time and workflow, not assume it will appear for free.
- Accept outcome risk that is real. Value-based deals only build trust if the developer can survive the downside. A risk-sharing agreement with a company that may not exist next year is not risk-sharing.
- Target conditions where the alternative is nothing. Digital therapeutics compete poorly against a cheap generic. They compete well where waiting lists for therapy are months long.
A worked example: the insomnia app and the regional insurer
Consider a scenario that plays out, in some version, in payer meetings everywhere.
A company has a digital cognitive behavioural therapy programme for insomnia. Its randomised trial shows a clinically meaningful improvement in sleep scores compared with a control group. It proposes a price per patient that is a fraction of what the insurer pays for a course of in-person therapy, and argues that the insurer will save money on sleeping pills, GP visits and sick days.
The insurer’s medical director asks three questions.
First: of the patients in your trial, how many completed the programme? The answer is a respectable majority. Then: in your commercial deployments, how many patients who receive a prescription finish at least half of the modules? The answer is less clear, and lower.
Second: who will prescribe it? The company says GPs. The medical director knows GPs in the region have ten-minute appointments and no time to explain an app, onboard a patient or follow up when they drop out.
Third: if patients don’t use it, do we still pay? The company says yes, because the price is per prescription.
At that point, the insurer is not deciding whether the app works. The trial settled that. It is deciding whether it wants to pay a fixed price for an uncertain dose of a treatment delivered through a workflow that doesn’t exist. Put that way, the hesitation looks rational.
Now change one variable. Suppose the company offers to charge only for patients who complete an agreed minimum number of sessions, provides its own onboarding staff by phone, and reports outcomes quarterly by practice. The price per completing patient goes up. The insurer’s risk goes down. The deal becomes much easier to sign.
Nothing about the software changed. The offer did.
The strongest counterargument
The most serious objection to this framing comes from the developers, and it has force.
They argue that asking digital therapeutics to prove real-world engagement and outcomes before being paid is a standard no drug faces. When a payer covers a statin, it doesn’t demand evidence of how many patients took their pills every day. Adherence to many oral drugs is mediocre, yet those drugs are reimbursed on the strength of their trials. Why should software be held to a stricter test?
This is a fair point, and partly correct. Payers are inconsistent. Digital products do face a heavier burden than medicines with comparable evidence, and some of that reflects unfamiliarity rather than principle.
But the analogy only goes so far. A drug’s active ingredient is the same in every tablet. Its dose is controlled by the manufacturer. A digital therapeutic’s “dose” is the patient’s engagement, which depends on design, onboarding, reminders and the clinician’s involvement, all of which vary enormously in practice. Two patients prescribed the same app can receive very different treatments. That is a genuine difference in kind, and it justifies a closer look at real-world use.
The Germans seem to have reached a similar conclusion. Their system started by granting provisional access on the basis of a plausible benefit, and has since removed products that could not prove their effect. The direction of travel is toward more evidence, not less.
Second-order effects of the current model
The way reimbursement works today shapes what gets built, in ways that are not always good for patients.
Products drift toward what is billable, not what is needed. If the only viable route to payment in the United States is to fit an existing code category, companies will design products to fit that category. The Medicare codes that arrived in 2025 are tied to a specific device classification for digital mental health treatment. That will steer investment toward mental health and away from conditions that lack a code.
Clinicians become gatekeepers without incentives. Where products must be prescribed, clinicians decide who gets them. Unless they are paid for the extra work, or the product saves them time, they have little reason to prescribe. Adoption will then track clinician enthusiasm rather than patient need.
Consolidation may be the likeliest winner. Small companies with one product struggle to carry payer risk or support workflows. Larger health technology firms, pharmaceutical companies and electronic record vendors can bundle digital therapeutics into broader offers. That may make products more sustainable. It may also make them more expensive and less independent.
Evidence standards may diverge by country. Germany, France, the UK and the US are building different pathways with different thresholds. A company can now design an evidence strategy around the most lenient market, and export whatever reputation it earns there.
What this means for decision-makers
For payers and ministries, the lesson from Germany is to pair guaranteed access with a strong mechanism for delisting products that don’t show benefit, and to insist on engagement data from real-world use. Paying for the pathway, including clinician time, is more likely to produce benefit than paying for a licence alone.
For hospital and health system leaders, it is to treat a digital therapeutic as a service line, not a product. Before prescribing at scale, ask who will onboard patients, who will review data and who will follow up with those who stop.
For investors and developers, it is to stop presenting reimbursement as a policy problem someone else will solve. The companies most likely to survive are those that price for outcomes and build the human support around the software.
For clinicians, it is to ask the same question you would of any treatment: what proportion of patients like mine actually use this, and what happens to them?
What to watch next
Three signals will say a lot about where the field is going: whether Medicare’s digital mental health codes see meaningful billing volume or languish like earlier integration codes; whether Germany tightens evidence and pricing rules for DiGA; and whether any major private payer in the US signs outcome-based contracts at scale. If those three move, the sector will have found its buyer.
The takeaway
Digital therapeutics were sold as software that behaves like medicine. Payers have treated them, reasonably, as medicine that behaves like software: easy to prescribe, easy to ignore, and hard to verify.
The companies that get paid will be the ones that stop asking for reimbursement as a category and start proving, to a specific buyer, that a specific patient group uses the product and gets better. That is slower. It is also the only argument payers have shown they will accept.
If you’ve prescribed or paid for a digital therapeutic, what made it stick, or not? We’d like to hear from you.
