A person gets told to ask for CBT-I. The name arrives from a doctor, a screening questionnaire, or the fourth page of search results, with a recommendation attached and no description underneath. How many appointments. What happens in each one. What is asked for at home. Which week is the bad week.
That description is what follows. It is not an argument for the treatment, and it is not a set of instructions for running the schedule alone.
4 to 8
sessions in a standard course
85%
the sleep efficiency figure that decides whether the window reopens
5 hours
the floor below which time in bed is not prescribed
14 to 40%
of patients leave before the middle of treatment
The course has a shape, and it is not the shape most people expect
A standard course runs four to eight sessions, usually weekly, with the first one longest. That range comes from the American Academy of Sleep Medicine's treatment guideline, and the variation inside it is real: brief protocols built for primary care compress the work into four appointments, while a full specialist course runs to eight and ends with relapse planning.
The first surprise is that nothing changes in the first appointment. Nothing is prescribed and no bedtime moves. The session is assessment and instruction in record keeping, and then the patient goes away for a fortnight to write things down.
The second surprise is where the difficulty sits. Almost nobody finds week one hard. Week two is where the protocol collects its debt.
| Week | In the session | Between sessions |
|---|---|---|
| Before week 1 | Assessment, screening for other sleep disorders, the diary explained | One to two weeks of diary, filled in each morning |
| Week 1 | The window is calculated and agreed, the rule for leaving the bed is set, daytime sleepiness and driving are discussed | The new window, held every night |
| Week 2 | First review. Sleepiness is at its worst and the window usually does not move | Same window, same diary |
| Week 3 | Efficiency is often high enough for the first fifteen minutes to be handed back | A slightly wider window |
| Weeks 4 to 6 | Titration in fifteen minute steps, up or down. Work on the thinking about tomorrow starts here | The window keeps moving, the diary is still the only evidence |
| Weeks 6 to 8 | Relapse planning. What to do after one bad night, and after a bad fortnight | The schedule held with nobody reading the diary back |
Week numbering is a convenience. What is fixed is the order: record, prescribe, titrate, then plan for the relapse.
Sleep efficiency, and the arithmetic that sets the window
One number runs the protocol. Sleep efficiency is total sleep time divided by time in bed, as a percentage. Somebody in bed for nine hours and asleep for six is running at about 67 percent, and that gap is not neutral dead time. It is three hours a night of practice at being awake in a bed.
The prescription is built from the diary. Average total sleep time across the recorded fortnight becomes the width of the window, with a floor at five hours so that nobody is sent into real deprivation. The six hour sleeper inside a nine hour opportunity gets a six hour window, and picks a rise time first, because the rise time is the anchor and bedtime is what moves to fit.
The prescription starts from how long a person is actually sleeping, not from how long they would like to sleep.
The instinct of somebody short of sleep is to widen the opportunity. The protocol narrows it, on the argument that a bed which reliably produces sleep is worth more than a bed which is merely available for longer.
Then it reopens, on a rule rather than on how the week felt. Best practice guidance puts the threshold at a mean sleep efficiency of 85 percent over five nights: at or above it, fifteen minutes goes back by moving bedtime earlier, and below it the window returns towards the mean sleep time the diary now shows. Titration stops when widening pushes efficiency under the threshold, which is how an actual ceiling gets found instead of an aspirational one.
Spielman, Saskin and Thorpy published the method in 1987, with 35 patients whose insomnia averaged more than fifteen years. A 2021 meta-analysis of the randomised trials since then found large improvements in insomnia severity and medium to large improvements in how long sleep took to arrive.
Why stimulus control asks for getting out of bed
The other active component has nothing to do with duration. Stimulus control is a conditioning argument, and Bootzin's original instructions amount to four sentences: lie down only when sleepy, keep reading and television out of the bedroom, get up and go to another room if sleep is not arriving, and repeat as needed.
A bed is a cue, and cues are learned by repetition regardless of anybody's intentions. A bed used for sleep produces sleepiness. A bed used for lying awake, calculating how many hours remain and being frustrated produces alertness, automatically, after enough nights. Bootzin and Epstein put the mechanism plainly: leaving stops the bed being reinforced as a place where wakefulness happens.
Which is where the commonest objection lands. Lying still with the eyes closed does confer some rest, and nobody in this literature disputes it. But resting quietly in bed while awake is the exact behaviour that built the association, so the rest gets bought at the cost of the thing being treated. Rest can be had in a chair.
The instruction is also less mechanical than its internet version. A fifteen minute rule circulates widely, and Bootzin left the timing vague on purpose, to stop people watching the clock. The cue is the feeling of being clearly awake rather than a stopwatch, with current best practice suggesting an outside limit of about half an hour for anybody who needs a number.
Sleep hygiene is not CBT-I, and the confusion is expensive
A large number of people believe they have already tried this treatment. What they tried was the list.
Sleep hygiene sits inside CBT-I as its weakest component. A 2018 meta-analysis in Family Practice found hygiene education alone clearly less effective than the full treatment, with the reviewers unable to establish a useful role for it even as a first step, and the 2021 AASM guideline recommended against it as a single component therapy at all. The entry on sleep hygiene traces where the list came from. The point here is narrower. Somebody who did the list and got nowhere has not had a trial of this treatment.
The diary is the instrument, and a wrist tracker is not a substitute
One to two weeks of handwritten record, before anything at all is prescribed. Bedtime, rise time, how long sleep took to arrive, time awake in the night, naps, alcohol, and a line on how the following day went. A standardised version exists, the Consensus Sleep Diary, assembled by an expert panel in 2012 and then rewritten in response to focus groups of actual patients.
Why a wrist device cannot stand in for it has a more interesting answer than the usual one. Consumer trackers are not uniformly bad at this. In a laboratory comparison of seven devices against polysomnography, most estimated sleep onset latency to within five minutes, better than the research grade actigraph alongside them. Where they fell apart was wake after sleep onset: two of the wrist devices underestimated time awake in the night by around fifty minutes, and performance degraded specifically on the disrupted nights, which are the nights an insomnia patient is trying to describe.
The deeper reason is what the treatment aims at. Insomnia is a complaint, and the protocol is titrated against the complaint, so the efficiency figure that decides whether the window opens comes from what the person reports. A device reporting a respectable seven hours to somebody who lived through a broken night has not corrected them. It has measured something else, which is why the AASM's actigraphy guideline places wrist monitoring alongside the log rather than instead of it.
The first fortnight is the worst part, and a lot of people leave
This is the fact most write-ups skip, and skipping it is how people arrive at week two feeling ambushed.
Narrowing time in bed works by raising sleep pressure, and raising sleep pressure means being sleepier. For the first two to three weeks patients are measurably more somnolent, and the improvement in how the days feel arrives afterwards rather than alongside. A systematic review of adherence found that between 14 and 40 percent of patients drop out before the midpoint of individual or group treatment, with internet-delivered courses in the same range. A 2026 trial of sleep restriction lost 19 of its 76 participants.
That trial produced a result worth sitting with. Adherence to the prescribed window, measured night by night, did not predict how much insomnia severity improved. Which complicates the usual framing of dropout, in which the patient who could not hold the schedule is the patient who did not try hard enough. It suggests the clinician's job in weeks two and three is partly to keep somebody in the room.
Staying is worth it because of where the gains end up. The improvement is durable past the end of treatment in a way drug effects are not, and the cost sits inside a known number of weeks that can be planned around. Somebody told at week one that week two will be unpleasant, and asked directly about driving, is in a very different position from somebody who improvised a five hour window on a Sunday night.
When CBT-I is not the first step
Bipolar disorder. Sleep loss is a known trigger for mania, and this is not a theoretical caution. When total sleep deprivation was used deliberately as an antidepressant intervention in 206 patients with bipolar depression, 4.85 percent switched into mania and 5.83 percent into hypomania. Sleep restriction is partial sleep deprivation delivered on purpose for a fortnight. The adapted protocol tested by Harvey and colleagues does not merely proceed with care: it sets the floor on time in bed at 6.5 hours rather than five and attaches a written safety plan. The 2023 European insomnia guideline calls the evidence on insomnia comorbid with bipolar disorder insufficient, and lists conditions that sleep loss can worsen, including any form of epilepsy, as contraindications to the components involving partial deprivation. A bipolar diagnosis means the schedule gets set by somebody who knows about it, with mood monitored while it runs. This is the one line here that should not be crossed on anybody's own initiative.
Untreated sleep apnoea. Insomnia and obstructive sleep apnoea coexist far more often than either gets diagnosed alongside the other, and the combination is worse than either alone. Current evidence favours treating both rather than queueing them, which requires the apnoea to have been identified first. Narrowing the window of somebody whose nights are already fragmented by breathing events, and who may be dangerously sleepy by day, is a different proposition. Snoring with witnessed pauses, or unrefreshing sleep at any duration, is a reason to get that looked at before the schedule is touched.
Severe depression. Not a contraindication, and CBT-I is used successfully in depressed patients. What changes is supervision. A protocol whose first fortnight makes people more tired and less able to concentrate needs a clinician watching it, and somebody who cannot reliably get out of bed at a fixed hour is being set a task the illness will not permit.
Shift work. Standard CBT-I assumes a stable schedule, because the fixed rise time is the anchor everything else hangs from, and rotating shifts remove it. Adapted protocols exist and are being trialled, and the reason they had to be written is that sleep restriction and stimulus control in their standard form require regular sleep and wake times. Somebody working nights who is handed an unmodified course has the wrong document.
Digital CBT-I, and the size of the gap
App and web courses are now an ordinary route to the same protocol, and the evidence behind them is substantial rather than provisional. A meta-analysis of 33 randomised trials with roughly 4,700 participants in each arm found digital CBT-I reducing scores on the Insomnia Severity Index by five points relative to control, with the improvement holding at one year.
- Digital CBT-I vs control: 5points
- Face to face vs digital: 3.07points
Face to face treatment outperformed the digital version by 3.07 points on the same scale, which the reviewers judged to sit inside a four point non-inferiority margin. That framing is defensible and the gap is also real. Three points is not nothing when the whole effect against control is five.
Engagement is probably where the difference lives. In a trial of 1,711 adults randomised to a digital programme or to sleep hygiene education, 413 of the 853 assigned to digital CBT completed all six sessions. Under half. A course that titrates a window over six weeks does not work in two instalments, and the person who would have been argued out of quitting at week two is, in the digital version, alone with the sleepiness.
Finding somebody who actually delivers it
The supply problem is severe. A geographic audit counted 752 behavioural sleep medicine providers worldwide, 88 percent of them in the United States, more than half of those concentrated in twelve states, and four American states with none at all. That shortage is what produced the digital programmes.
It also produced something more useful. In a pragmatic trial across 35 general practices in England, 642 adults were randomised either to four sessions of nurse-delivered sleep restriction therapy plus a hygiene booklet, or to the booklet alone. At six months the nurse-delivered group scored 3.05 points lower on insomnia severity, an effect size of 0.74, at an incremental cost of £2,076 per quality-adjusted life year. Practice nurses, trained in the protocol, four appointments.
So the practical answer is not necessarily a specialist psychologist. It is a general practitioner asked for the treatment by name, because the request is what makes the referral happen and a sleeping tablet is the path of least resistance for everybody in the room. Beyond that, the professional societies keep provider directories, hospital sleep clinics often have somebody trained in it, and a validated digital programme is a legitimate route rather than a consolation prize.
Worth bringing to the appointment: a fortnight of diary, and an account of which part of the night fails. MoodyWonder's Insomnia Check scores getting to sleep, staying asleep and the daytime cost separately, since a high onset score with an intact day points somewhere different from all three raised at once. It is a screening-style read rather than a diagnosis, and its only job here is to say whether this is the conversation to start.
The treatment is unglamorous, mildly unpleasant for about a fortnight, and largely made of arithmetic. That is most of what nobody mentions when they recommend it.
Sources
- Edinger, J. D., Arnedt, J. T., Bertisch, S. M., et al. (2021). Behavioral and psychological treatments for chronic insomnia disorder in adults: an American Academy of Sleep Medicine clinical practice guideline. Journal of Clinical Sleep Medicine, 17(2), 255-262.
- Spielman, A. J., Saskin, P., Thorpy, M. J. (1987). Treatment of chronic insomnia by restriction of time in bed. Sleep, 10(1), 45-56.
- Cuddihy, L. J., Grandner, M. A., Nowakowski, S. (2026). Implementation of stimulus control and sleep restriction therapy for insomnia: standard definitions and best practices. Journal of Behavioral and Cognitive Therapy, 36, 100542.
- Maurer, L. F., Schneider, J., Miller, C. B., Espie, C. A., Kyle, S. D. (2021). The clinical effects of sleep restriction therapy for insomnia: a meta-analysis of randomised controlled trials. Sleep Medicine Reviews, 58, 101493.
- Bootzin, R. R., Epstein, D. R. (2011). Understanding and treating insomnia. Annual Review of Clinical Psychology, 7, 435-458.
- Chung, K. F., Lee, C. T., Yeung, W. F., Chan, M. S., Chung, E. W. Y., Lin, W. L. (2018). Sleep hygiene education as a treatment of insomnia: a systematic review and meta-analysis. Family Practice, 35(4), 365-375.
- Carney, C. E., Buysse, D. J., Ancoli-Israel, S., et al. (2012). The Consensus Sleep Diary: standardizing prospective sleep self-monitoring. Sleep, 35(2), 287-302.
- Smith, M. T., McCrae, C. S., Cheung, J., et al. (2018). Use of actigraphy for the evaluation of sleep disorders and circadian rhythm sleep-wake disorders: an American Academy of Sleep Medicine clinical practice guideline. Journal of Clinical Sleep Medicine, 14(7), 1231-1237.
- Chinoy, E. D., Cuellar, J. A., Huwa, K. E., et al. (2021). Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep, 44(5), zsaa291.
- Matthews, E. E., Arnedt, J. T., McCarthy, M. S., Cuddihy, L. J., Aloia, M. S. (2013). Adherence to cognitive behavioral therapy for insomnia: a systematic review. Sleep Medicine Reviews, 17(6), 453-464.
- Looman, M. I., Klau, M. M., Schoenmakers, T. M., Kamphuis, J. H., Blanken, T. F., Lancee, J. (2026). The adherence conundrum of sleep restriction therapy for insomnia. Journal of Sleep Research, 35(4), e70339.
- Colombo, C., Benedetti, F., Barbini, B., Campori, E., Smeraldi, E. (1999). Rate of switch from depression into mania after therapeutic sleep deprivation in bipolar depression. Psychiatry Research, 86(3), 267-270.
- Harvey, A. G., Soehner, A. M., Kaplan, K. A., et al. (2015). Treating insomnia improves mood state, sleep, and functioning in bipolar disorder: a pilot randomized controlled trial. Journal of Consulting and Clinical Psychology, 83(3), 564-577.
- Riemann, D., Espie, C. A., Altena, E., et al. (2023). The European Insomnia Guideline: an update on the diagnosis and treatment of insomnia 2023. Journal of Sleep Research, 32(6), e14035.
- Sweetman, A., Lack, L., Bastien, C. (2019). Co-morbid insomnia and sleep apnea: prevalence, consequences, methodological considerations, and recent randomized controlled trials. Brain Sciences, 9(12), 371.
- Grünberger, T., Höhn, C., Schabus, M., Laireiter, A. R. (2024). Efficacy study comparing a CBT-I developed for shift workers to standard CBT-I. Trials, 25, 562.
- Soh, H. L., Ho, R. C., Ho, C. S., Tam, W. W. (2020). Efficacy of digital cognitive behavioural therapy for insomnia: a meta-analysis of randomised controlled trials. Sleep Medicine, 75, 315-325.
- Espie, C. A., Emsley, R., Kyle, S. D., et al. (2019). Effect of digital cognitive behavioral therapy for insomnia on health, psychological well-being, and sleep-related quality of life: a randomized clinical trial. JAMA Psychiatry, 76(1), 21-30.
- Thomas, A., Grandner, M., Nowakowski, S., Nesom, G., Corbitt, C., Perlis, M. L. (2016). Where are the behavioral sleep medicine providers and where are they needed? A geographic assessment. Behavioral Sleep Medicine, 14(6), 687-698.
- Kyle, S. D., Siriwardena, A. N., Espie, C. A., et al. (2023). Clinical and cost-effectiveness of nurse-delivered sleep restriction therapy for insomnia in primary care: a pragmatic, superiority, open-label, randomised controlled trial. The Lancet, 402(10406), 975-987.