The economics of psychedelic therapy is not just the drug price

Fifteen years ago, when I studied psychology, psychedelics were completely absent from the curriculum. They were not treated as a serious part of mental-health care, research, or policy. That has changed. I have now worked on Blossom for seven years, and in that time the field has moved from a small research world into a much larger conversation about evidence, regulation, delivery, and access.

Psychology is still the lens I know best. It helps me think about symptoms, therapeutic relationships, preparation, integration, set and setting, and what it might mean for a person to get better. But health economics asks a different kind of question. It is like moving from the therapy room to the reimbursement table: the subject is still care, but the language shifts to budgets, trade-offs, health gains, opportunity costs, and who pays for what.

That is part of the reason for writing this in public. The economics of psychedelic therapy is where access stops being an abstract hope. It asks whether a therapy that looks promising can also be paid for, staffed, delivered, repeated when needed, and evaluated in ordinary healthcare systems.

In that sense, this is where the rubber meets the road. A therapy can look promising in the final clinical studies needed before approval and still fail to become available. That can happen if delivery is too expensive, if the comparison does not match routine care, if the benefit fades quickly, or if nobody has worked out how to pay for the parts of care that are not the drug.

I spoke about some of these access questions in my ICPR 2026 reimbursement and access talk. This series is a way to go deeper into one part of that puzzle: the economic case for psychedelic therapies, and the evidence needed to make that case credible.

I am also writing this from the position of someone learning in public. The goal is not to pretend that every answer is already settled. It is to build a shared map for readers who care about psychedelic access, but who do not necessarily speak the language of QALYs, ICERs, Markov models, formal coverage reviews, and payer evidence packages. Underneath all of that is one simple question: what would have to be true for these therapies to be worth paying for, and possible to deliver?

Several health-economic papers suggest that psilocybin or MDMA-assisted therapy could be cost-effective, or even cost-saving, under certain assumptions. That is encouraging. But it is also easy to read those headlines too quickly.

Cost-effective does not mean cheap. It means the added health gain is judged worth the added cost, compared with the care people would otherwise receive. Cost-saving is a stronger claim: it means the therapy could reduce total costs over the chosen time horizon.

Most of these claims come from models. They are not simple observations from routine care. A model asks what might happen if a particular population receives a particular treatment, at a particular cost, compared with a particular alternative, over a particular time horizon.

That does not make the models unhelpful. It makes them maps. They show which ingredients matter, where the assumptions sit, and which pieces of evidence are still missing. This first article is about learning to read that map.

Three questions will keep coming back throughout the series. First, what is the full care episode, not just the medicine? Second, how long does the benefit last? Third, what happens when someone does not respond, partly responds, or responds and then relapses? Costs, models, comparators, and payer evidence all look different once those questions are made explicit.

This matters because the field often talks about cost in a compressed way. One headline focuses on the drug price. Another focuses on the number of therapist hours. A third says the treatment may be cost-saving over decades. Each of those claims may help, but none is complete on its own. The economic question depends on the full pathway: who enters care, what they receive, what happens afterwards, what care is displaced, and which costs and benefits count.

It also matters because different readers need different parts of the answer. A patient may care about whether insurance coverage will make treatment affordable. A clinic may care about whether the payment covers staff and room time. A payer may care about budget impact this year. A health economist may care about the ICER, the ratio that compares extra cost with extra health gain, over a longer horizon. A policymaker may care about whether the model includes productivity and informal care. The same therapy sits inside all of those questions at once. So the first move is to pull the questions apart before trying to answer them.

The most obvious cost is the medicine. But for psychedelic therapies, the medicine is only one part of the intervention. The rest is a care pathway: screening, preparation, supervised dosing, monitoring, integration, follow-up, governance, and sometimes retreatment.

This is why a single headline price can mislead. A cheap drug can still be expensive to deliver if it needs many hours of trained staff and dedicated room time. An expensive drug can still be economically attractive if it produces durable benefit and reduces later care. The question is not the price of one component. The question is the full episode of care.

For psilocybin, LSD, and MDMA-assisted therapy, that episode can include clinical assessment, exclusion of higher-risk patients, preparation sessions, one or more long dosing sessions, integration, medical oversight, product handling, emergency protocols, documentation, and follow-up. Some of those costs are visible in a trial protocol. Others become visible only when a service has to run week after week with real staff, real rooms, cancellations, no-shows, training requirements, and payer paperwork.

This is also why the payment definition matters. Is the payer paying only for the medicine, while the rest has to fit into existing psychotherapy codes? Is it paying for the dosing day? Is it paying for a bundled episode that includes preparation and integration? Or is it paying in stages, with follow-up and retreatment treated separately? Those are not just administrative choices. They change who carries financial risk and whether providers can actually deliver the care.

Economic models do not only count costs. They also try to count benefits. In health economics, the best-known measure is the QALY, or quality-adjusted life year. A year in full health counts as 1 QALY. A year with serious symptoms counts for less. This lets economists compare treatments that affect people in different ways.

For psychedelic therapy, the benefit side may include symptom reduction, response, remission, better functioning, fewer crisis episodes, less healthcare use, and sometimes work or caregiving effects. But different decision-makers count different parts of that benefit.

The viewpoint matters. A healthcare payer may count medicine, clinician time, hospital visits, and other covered medical costs. A societal analysis may also count work loss, informal care, disability, and the time burden carried by patients and families. A provider may care less about QALYs and more about whether the reimbursed pathway covers the time and infrastructure needed to deliver care safely.

The same clinical improvement can therefore create different economic stories. If a patient returns to work, society may see a large gain, while a payer may not capture that gain directly. If a patient avoids hospitalisation, both the payer and society may benefit. If a treatment requires many hours in a specialist room, the provider may face a bottleneck even when the model looks attractive on paper. If we blur those lenses, one person's benefit can quietly become another person's invisible cost.

A therapy is not cost-effective in the abstract. It is cost-effective compared with something else. That comparator is one of the choices that most changes the analysis.

In depression, the relevant alternative might be standard care, antidepressant augmentation, psychotherapy, ECT, rTMS, esketamine, ketamine services, or physician's choice at a later treatment line. In PTSD, it might be trauma-focused therapy, medication, specialist care, or usual care for people who have already tried several options.

Ketamine and esketamine matter here as comparators and precedents. They are not the centre of this series, but they are already closer to real-world reimbursement. That makes them concrete examples of how evidence, service burden, repeated treatment, monitoring, and payer restrictions show up in practice.

The practical question is simple: what would happen to these patients if the psychedelic therapy were not available? If the answer is low-intensity usual care, the incremental benefit may look large. If the answer is a specialist pathway with active medication management, psychotherapy, ECT, rTMS, esketamine, or ketamine, the cost and outcome comparison may look different.

Comparator choice also affects credibility. Payers may accept a placebo-controlled trial as proof that a treatment can work, while still asking whether the economic model compares the new therapy with the care that would actually be displaced. For a blog series about access, that distinction is central: approval evidence and reimbursement evidence overlap, but they are not the same thing. Once the comparator is chosen, the next problem is time.

The core problem is time. Trials can tell us what happened over a few weeks or months. Payers often need to decide what to believe over years.

Some longer-term evidence is emerging from places where psychedelic therapies are being used or studied over extended periods, including Switzerland and Australia. But for most questions payers rely on, we still do not have routine-care data that directly answers how often people relapse, what care they use later, how often they need retreatment, and what happens to quality of life over several years.

That is where models come in. A model takes the data we have, adds assumptions where evidence is missing, and projects costs and outcomes over a chosen horizon. The output helps only if we can see the assumptions clearly.

A simple way to think about this is that the trial gives us an anchor, while the model builds the bridge. The anchor might be a response or remission rate at six weeks, twelve weeks, or one year. The bridge asks what that means over a longer period: whether benefit wanes, whether people relapse, whether they receive another course, whether other healthcare use changes, and whether quality of life remains higher.

This is not a problem unique to psychedelics. Many health technologies require modelling because the decision horizon is longer than the trial horizon. What makes psychedelic therapies unusual is the shape of the cost. The spending is front-loaded: screening, preparation, dosing-day support, integration, and monitoring happen early. The hoped-for benefits may accrue over months or years. That makes durability one of the load-bearing assumptions.

The honest version is therefore not that the model proves the therapy will save money. It is that the model estimates value under a stated set of assumptions about durability, pathway cost, retreatment, comparator care, and the time horizon. That sentence is less neat than a headline, but it tells us where the model can break.

Many early economic models focus on one treatment episode. That is a reasonable starting point, especially when the clinical evidence is still young. But real-world care may not be that simple.

I was reminded of this again at ICPR 2026. In conversations around the conference, several people who had received psychedelic treatment, including some who had participated in more than one trial, were still struggling with depression, or struggling again. That is not a criticism of the trials or of the people involved. It is a reminder that relapse, partial response, and repeat treatment are not abstract modelling details. They are live questions for people trying to get well.

Some people may not respond. Some may partially respond. Some may respond and then relapse months later. If a payer is asked to fund another course after 12 weeks, after one year, or after relapse, that needs to be part of the model.

Retreatment is not a footnote. It changes the expected benefit, the total cost, the staffing burden, and the budget risk. It also creates a data problem: we need to know who receives repeat treatment, when, how often, at what cost, and with what outcome.

This matters especially for therapies that are described as potentially durable. If the economic case depends on one expensive episode producing long-lasting benefit, then the relapse and retreatment assumptions carry a lot of weight. A second course after non-response, partial response, or later relapse may be clinically reasonable, but it changes the value proposition.

In a model, retreatment needs rules. Who becomes eligible? How long must benefit last before another course is considered? Does the second course have the same response rate as the first? Are there limits on repeat exposure? Does the repeat course require the full preparation and integration package, or a shorter version? Those choices are not details to hide in an appendix. They define the intervention being paid for.

The current evidence base already helps. It tells us enough to ask better questions. It does not yet answer every question that a payer, provider, or health system will need before broad implementation.

I wrote about a version of this question in 2022 for Lucid News, asking whether the high cost of psychedelic therapy could be worth it. The questions then were already recognisable: who will pay, how much of the cost is therapy time rather than the medicine, how durable are the benefits, what can ketamine and esketamine teach us, and whether insurers would treat a medicine-plus-therapy pathway as something they know how to cover.

Some things have moved since then. There are more economic models, more implementation resources, more discussion of formal coverage-review methods, and more real-world access examples from countries experimenting with legal or authorised use. The field has also become more specific. We can now ask better questions about therapist hours, payment bundles, Dutch versus US payer perspectives, group delivery, and the difference between cost-effectiveness and budget impact.

But many questions from that earlier article remain. The long-term durability of benefit is still doing a lot of work in positive economic models. Retreatment is still not well described. Real-world staffing and room use are still undermeasured. The comparator problem is still difficult. And the practical payer question, especially for a therapy that is not simply dispensed like a conventional medicine, is still not fully solved.

The strongest part of the evidence base is usually the clinical and protocol evidence: what happened in trials, how many sessions were planned, who was included, who was excluded, and what short-term outcomes were measured. That is the starting point for a model.

The weaker parts are often the parts that matter most for reimbursement. We need better evidence on durability, retreatment, routine-care resource use, staff mix, provider costs, local tariffs, patient affordability, and what care is actually displaced. Some of this can be estimated in early models, but if it drives the conclusion, it should be presented as uncertainty rather than hidden as a confident input.

That gives this series a practical agenda. Some questions we can work on now by making the existing models and assumptions clear. Some require better public data, such as local tariffs, provider costs, or real-world follow-up. Some need future registries and managed-access programmes. And some need people in the field to say more plainly what they are actually doing in clinics, payment negotiations, and implementation pilots. The roadmap below is my first attempt to turn that agenda into posts.

This first article is the map. The rest of the series will go into the ingredients one by one. I expect the list to change as I learn more, and I would like readers to help shape it.

Some posts will be more explanatory, such as QALYs and ICERs. Some will be more practical, such as the cost of a therapy episode or the payer evidence checklist. Some will probably be more speculative, because the data are still thin and we need to say what would have to be true for different models to hold.