The untreated interval

There is a measurement in psychiatry with a deliberately unglamorous name. It is called the Duration of Untreated Psychosis — DUP for short — and it is the number of weeks or months between the first emergence of psychotic symptoms in a young person and the day they begin antipsychotic treatment. It is not a score. It is not a test. It is just a length of time. For much of the twentieth century no one measured it, because no one imagined it mattered. A patient arrived when they arrived. The clock began when the clinician turned it on.

Some time in the late 1980s, a small number of researchers — most of them working in early-intervention services in the United Kingdom, Norway, and Australia — began to notice that the patients with the shortest DUP tended to do better. Not dramatically. Not in every study. But consistently enough that it became difficult to argue the effect away. By 2021, when Oliver Howes and colleagues at King's College London pulled the literature together into an umbrella review in World Psychiatry, there was no longer much doubt about it. Longer DUP was associated with worse symptoms, lower remission rates, and lower global functioning, across studies, across countries, across definitions of "long" and "short" (World Psychiatry, 2021). A 2024 meta-analysis in Schizophrenia Bulletin found that long DUP — roughly a year on average — was associated with lower odds of remission and twice the odds of a suicide attempt at baseline, compared to short DUP of around ten weeks (Schizophrenia Bulletin, 2024). None of this is a claim that shortening DUP cures schizophrenia. It is a claim that the interval before treatment is itself doing clinical work. The patient is not static during it. The disease is not static during it. And the usual assumption — that the clock only starts ticking when treatment begins — is wrong.

Every word of that last sentence is worth sitting with, because it is not a statement about schizophrenia. It is a statement about how medicine keeps its accounts.

Almost every specialty that has gone looking has found its own version. In oncology, the metric is called Time to Treatment Initiation, or TTI, and the most-cited estimate of what it costs was published by Timothy Hanna and colleagues in the BMJ in November 2020, in the middle of the first pandemic winter. Pooling thirty-four studies covering more than 1.27 million patients across seven common cancers, they reported that each additional four weeks of delay between diagnosis and first treatment was associated with a 6–8 per cent increase in the hazard of death for most surgical indications, with larger and more variable estimates for some systemic and radiotherapy indications (BMJ, 2020). The association was not uniform, and the Hanna team were careful about the limits of observational data, including a methodological issue known as the waiting-time paradox, in which the patients with milder disease wait longer because sicker patients are triaged to the front of the queue. Even after the adjustments that could be made, the interval-effect survived in most of the indications that had high-validity studies to draw on. Four weeks of waiting, across much of oncology, is not clinically neutral.

In depression, the same question has been asked in its own vocabulary, and the evidence base is thinner and more contested. A small 2015 observational study by Lucio Ghio and colleagues, following 121 outpatients with major depressive disorder, reported that patients with a Duration of Untreated Depression under six months had about 2.8 times the odds of response to antidepressants at twelve weeks and 4.1 times the odds of remission at twenty-four weeks, compared to patients untreated for longer (Journal of Affective Disorders, 2015). An earlier analysis in the same line of work pointed in a similar direction (Altamura et al., J Affect Disord, 2013). These are small cohorts and naturalistic designs, not definitive evidence, and they should be read as hints rather than proofs. What they hint at is the same direction the psychosis literature points in: that treating earlier seems to work better, and that part of what "earlier" means is earlier in the patient's own untreated interval, not earlier in the disease.

Cardiac rehabilitation has its own version: the weeks between a heart attack and the first session of a cardiac rehab programme, during which a patient's adherence to medication and lifestyle change drifts in ways that later rehab can only partially recover. Stroke rehabilitation has its own, in the narrow window during which neurological recovery is most responsive to structured input. Early-intervention psychiatry has its own, which is how DUP came to be measured in the first place. There is a health-services literature on "time to treatment" and on "treatment delay" that touches most of these specialties, and there is a smaller literature on pre-rehabilitation and pre-habilitation that asks what can usefully be done inside the wait. What there is not — or at least, what this author has not been able to find — is a body of work that treats the pre-treatment interval as a single construct across specialties, with shared definitions, shared outcome measures, and a shared vocabulary for what is happening inside it.

The working hypothesis of this essay is that there is one to give. Call it the untreated interval: the period during which a patient is known to have a diagnosed condition, is known to be eligible for a particular treatment, and is waiting for that treatment to begin, while the underlying disorder continues to do its work. The interval is not a disease. It is not a treatment. It is not a pathway in the sense that healthcare systems use the word. It is a category of time — a specific, measurable, clinically active period — that has fallen outside of every existing accounting framework because it does not correspond to anything existing accounting frameworks were designed to measure.

Consider what the major accounting systems in medicine actually measure. Diagnosis-related groups, the basis of most inpatient reimbursement in the world, measure episodes of care: admission to discharge. They cannot see the interval because it happens before admission. Cost-effectiveness analyses — the QALY-based calculations that underpin much of German health technology assessment, including the work IQWiG does for the Gemeinsamer Bundesausschuss on drugs and procedures — measure the effect of an intervention against a counterfactual. They cannot see the interval because it is, by definition, the period during which there is nothing to measure against. (The DiGA pathway, it is worth noting, runs on a different framework entirely: BfArM assesses digital applications under §139e SGB V and the DiGAV regulation, with its own evidence requirements.) Waiting-time statistics — the headline numbers that ministers quote and patient advocacy groups use to press them — measure the length of the interval, but not what happens inside it. They are a ruler laid against the surface of something they never open up.

The result is that a clinician can describe the untreated interval in great detail, from clinical experience, while the health system that employs that clinician has no mechanism for recording what they describe. There is no form, no code, no line item. A patient who deteriorates during the interval generates costs — in A&E visits, in crisis admissions, in statutory sick pay, in the work of the family members who bring them in — but those costs land in ledgers that do not communicate with each other. From the perspective of any single ledger, the original cause is invisible, and the deterioration is recorded as whatever the nearest available diagnostic code happens to be.

This essay is not an argument that we should be spending more money on every specialty's waiting list. It is narrower, and it is more specific. The argument is that the untreated interval is a distinct object worth its own study, its own measurement, and eventually its own place on the books. Four things follow from that argument. Each of them is worth stating plainly.

The first is that the interval is modifiable. This is the least intuitive claim, because the obvious way to modify the interval is to make it shorter — to add more beds, more therapists, more oncology capacity, more cardiac rehab slots — and the obvious way is not the point. Capacity is expensive and slow. What the interval-as-a-category framing reveals is that there is a second axis of modification: not the length, but the content. The question "what happens during the wait" is separable from the question "how long is the wait". Every one of the specialties mentioned above has quietly invented, somewhere, small programmes that try to fill some of the interval with something. Pre-rehab for orthopaedic patients awaiting joint replacement. Pre-habilitation for cancer patients awaiting surgery. Psychoeducation for patients on psychotherapy waiting lists. Nurse-led check-in calls for patients awaiting bariatric surgery. These are all forms of care, and some of them are treatment in the formal sense, but none of them are the treatment the patient is formally waiting for. They are attempts to make the interval do something other than decay. Collectively, they are a natural experiment the field has not yet noticed it has been running, and the results are encouraging enough that a coordinated literature on them is overdue.

The second is that the interval is where early intervention happens, if early intervention happens at all. The phrase "early intervention" is used loosely in medicine, usually to mean "acting while the disease is less severe". But the early-intervention-in-psychosis literature — the discipline that invented DUP and has spent thirty years trying to shorten it — means something more precise. It means acting during the window in which a person has been identified as needing help but has not yet been given any. That window is the untreated interval, by another name. The reason early intervention works, when it works, is not that it treats a different disease; it is that it treats the same disease earlier in the curve, and at a different point in the patient's own experience. Early in the curve is where small inputs have large effects. Late in the curve is where large inputs have small ones. The untreated interval is, for most conditions, the point at which the curve is still early.

The third is that the patients most exposed to harm during the interval are often the patients least visible to the system. Cancer patients waiting for treatment generate test results and follow-up appointments; they are visible. Patients awaiting hip replacement are booked into pre-op clinics; they are visible. Patients with first-episode psychosis or with substance use disorders, by contrast, often drop off the radar of whoever referred them, because the referral itself is the last recorded interaction until admission. There is no intervening touchpoint. There is no pre-op clinic. There is, in the substance use case specifically, not even a reliable confirmation that the patient still intends to come.

The fourth is that the interval is hardest to see precisely where it matters most. A condition for which society already has a loud vocabulary — cancer, heart disease — generates enough political pressure that its waiting times are published, debated, and occasionally acted on. A condition for which society has a quieter vocabulary — depression, anxiety, substance use, first-episode psychosis — generates waiting times that almost nobody measures with the same rigour, because there is no political payoff to measuring them. In Germany, the Deutsche Rentenversicherung reported an average wait of roughly eleven weeks between the approval of a medical rehabilitation course and the day of admission in 2023, as picked up in consumer-finance coverage of the DRV's figures (Finanztip citing DRV, 2024). That number covers all indications — cardiac, orthopaedic, psychosomatic, addiction — and an addiction-specific figure is not readily available in public aggregate reports. There is also no comparable figure, to this author's knowledge, for the wait between the moment a GP makes a first diagnosis of alcohol use disorder and the moment the patient begins any structured programme aimed at that disorder. It is not that nobody has chosen to count it. It is that the counting has not yet been asked for.

One objection to naming the category at all is that it is too heterogeneous — that the interval in pancreatic cancer and the interval in first-episode psychosis and the interval in substance use disorder are so different in their mechanics that calling them a single thing is a rhetorical move more than an analytical one. The objection has real force, and it deserves an honest answer. The answer is that the interval is heterogeneous in its mechanics and homogeneous in its structural position. Every version of it sits at the same place in the patient's journey: after the diagnosis, before the treatment, inside a waiting relationship between the patient and the system that has agreed to help them. The mechanisms by which deterioration happens vary — a tumour doubles, a psychosis consolidates, a motivation erodes, a clot matures — but the position in the journey does not. A construct that holds a single structural position together across very different mechanisms is not a weaker construct; it is, if anything, a more useful one, because it draws a line around a category of problem that cuts across specialties and therefore cannot be solved inside any one of them.

A second objection is that some of the fields in which the untreated-interval effect has been best documented — DUP in psychosis, DUD in depression — are plagued by a methodological problem known as the waiting-time paradox. The paradox is this: in some conditions, the patients who wait longest are, on average, the ones with milder symptoms, because the patients with severe symptoms are triaged to the front of the queue. When you compare long-wait and short-wait patients' outcomes, you may be comparing not only the effect of the wait itself but also the effect of who gets seen first. The same problem has been raised, in the oncology literature, in a letter to the BMJ commenting on the Hanna et al. meta-analysis (BMJ rapid response, 2021). Every careful researcher in these fields has had to reckon with it, and the most rigorous studies — including the Hanna meta-analysis itself, which restricted its primary analysis to high-validity studies controlling for major prognostic factors — find the interval effect even after adjustment. The paradox is real and it is a reason for caution, but it is not a reason for dismissal.

A third objection is economic. Money spent on the interval is money not spent on the treatment that follows it. If the interval is real but modest in its effects, perhaps the right thing to do is to put the resources into shortening the wait itself — into capacity — rather than into attempting to furnish the wait with content. This is a legitimate question, and it has the same answer in almost every specialty: both, wherever possible, but content is usually the cheaper axis. Shortening a wait by six weeks requires hiring people, expanding buildings, and waiting out the consequences of those decisions for years. Adding structured content to a wait requires designing the content once and delivering it at marginal cost. The two interventions are not substitutes; they are complements, and the reason the content axis has been underused is not that anyone has argued against it but that no one has been responsible for it.

Everything above is a description of a missing category. What a proper research programme on the untreated interval would look like is a different question, and the honest answer is that most of it has not yet been worked out. But the outline is visible. A useful programme would need, first, a common language: a definition of "the interval" that is precise enough to be measured across specialties, and a small set of outcome measures that can be applied inside it. Second, it would need data infrastructure: a way of linking the fragments of a patient's journey through an interval back together, across the ledgers that currently do not speak to each other. Third, it would need a small group of specialties willing to act as the testbed. The best candidates are the ones in which the interval is longest, the patient population is most vulnerable, and the political visibility of the waiting list is lowest — because those are the cases in which the status quo is least defensible and the room for improvement is largest. Substance use disorders, first-episode psychosis, and the long tail of the mental-health waiting list are the natural places to start.

The last piece of the argument is a point about naming. Things that do not have names do not get measured. Things that do not get measured do not get funded. Things that do not get funded do not become specialties, which means they do not acquire the researchers, the clinicians, the journals, and the small bureaucratic constituencies that turn a category of problem into a category of work. Medicine has a long history of discovering that some invisible thing has been doing clinical work all along — bacteria in the mouth causing heart disease, sleep apnoea driving metabolic syndrome, loneliness shortening lifespan — and the naming of that thing is almost always the moment at which measurement becomes possible. Before the name, there is nothing to count. After the name, the counting can begin.

The untreated interval is, today, roughly where sleep apnoea was in 1980: a thing that clinicians in several specialties have noticed independently, that has been measured in piecemeal fashion, and that does not yet have a place on the map. Its day is probably coming. The question is only how long the interval will last between now and then.

Meanwhile, in a GP's surgery in Lower Saxony on a Tuesday morning in March, a 43-year-old machinist is told he has alcohol use disorder. He is given a referral and an appointment for the end of August. The five months between now and then are not a gap in his treatment. They are his treatment, because they are the period during which what happens to him will determine what treatment is capable of doing when it eventually arrives. He does not know this. His GP does not know this. The clinic that will eventually admit him does not know this in any way the clinic can act on. The system does not know this because the system does not yet have a category for it.

That is the thing this essay is asking medicine to notice. The five months are not waiting. The five months are the illness doing its work. The untreated interval is where the disease continues to do its work while the system organises its response, and until we name it, we cannot measure it, and until we measure it, we cannot improve it, and until we improve it, it will go on being one of the most consequential invisible periods in modern clinical medicine. It is a long essay to make a simple point. The point is that something that has no name has no advocate, and something that has no advocate does not, eventually, get built.