An automated internet CBT-I programme held its gains in older adults at 12 months

Cognitive behavioural therapy for insomnia is the guideline-endorsed first-line treatment, strongly recommended by the American Academy of Sleep Medicine, and it is preferred in older adults because sedative-hypnotics in that group have limited benefits and higher rates of adverse events — guidelines caution against their use [s2]. The problem is supply: there are few appropriate specialists and reimbursement is restricted [s2].

That gap is the argument for delivering the therapy through software. A trial published this month reports what happened when a fully automated internet CBT-I programme was tested in people aged 55 and older — the first such trial in this population, according to its authors [s1][s2].

The design

SHUTi OASIS — Sleep Healthy Using the Internet for Older Adults Suffering with Insomnia and Sleeplessness — is a tailored internet CBT-I intervention [s1]. The trial randomised participants to three arms: SHUTi OASIS alone (n=105), SHUTi OASIS plus a stepped support protocol (n=102), and online patient education as the control (n=104) [s1].

Recruitment ran from May 2018 to April 2019, drawing 1,487 completed interest forms, of which 311 were randomised [s2]. Participants came from 38 US states and the District of Columbia [s2]. Mean age was 66.3 (SD 7.2), with a range of 55 to 95 and 31% (n=95) over 70 [s1][s2]. Most were female (68.5%), white (91.0%), and college-educated (74.9%) [s2]. Median duration of sleep problems was 10 years, and 82.0% described themselves as very comfortable using the internet [s2].

That last figure, and the education profile, are the population caveat. This is a sample selected for internet comfort by the recruitment method itself.

The primary result

The primary outcome was insomnia severity on the Insomnia Severity Index. Both SHUTi OASIS arms produced large within-group improvements from baseline sustained across all three follow-ups — Cohen's d of −1.04, −1.24 and −1.29 at post-treatment, six months and 12 months for SHUTi OASIS alone, and −1.35, −1.43 and −1.41 for SHUTi OASIS plus stepped support [s2]. The patient education arm improved too, by small to medium amounts: d of −0.28, −0.50 and −0.67 [s2].

The three-group by four-timepoint interaction was significant (F(6,773)=15.70, p<0.001), and both intervention arms had significantly lower scores than patient education at every timepoint (all p<0.001) [s2].

Categorical outcomes are more legible than effect sizes. Treatment response was defined as a reduction of more than seven points on the index; remission as a score below eight [s2]. At 12 months, response rates were 34/85 (40%) for SHUTi OASIS alone, 39/79 (49%) with stepped support, and 15/83 (18%) for patient education [s2]. Remission rates were 34/85 (40%), 40/79 (51%) and 13/83 (16%) respectively [s2].

What else moved

Sleep-diary secondary outcomes favoured the intervention on wake after sleep onset, sleep efficiency, number of awakenings, sleep quality and fatigue, at every comparison [s2]. Sleep onset latency improved in the intervention arms at post-treatment and six months, though the 12-month comparison was weaker (p=0.01) than the threshold the other comparisons cleared [s2].

By 12 months, the pooled intervention group's mean wake after sleep onset was 47.65 minutes against 70.80 in patient education, and sleep efficiency 85.28% against 79.14% [s2].

Two outcomes did not separate. Total sleep time increased over time in everyone — a significant main effect of time (F(3,772)=18.95, p<0.001) — but with no significant difference between conditions (F(3,772)=0.82, p=0.48) [s2]. Self-reported attention and concentration showed no interaction effect either (F(3,767)=0.70, p=0.55) [s2].

The total sleep time result is the one most likely to surprise a reader. CBT-I works substantially through sleep restriction, which consolidates sleep rather than lengthening it; sleep efficiency rose while total time did not differentiate. People slept better, not longer.

Sleep medication use did shift: there was a significant group-by-time interaction for the percentage of nights on any pharmacological sleep aid (F(3,766)=4.08, p=0.01), with a greater decline in the intervention group [s2].

Engagement, and whether support mattered

During the nine-week intervention period, SHUTi OASIS participants logged in a median of 48 times and submitted a median of 52 sleep diaries [s2]. Sixty-three per cent (n=131) completed all six programme cores within that window; another 41 finished afterwards, bringing eventual completion to 83% (n=172) [s2]. Patient education participants logged in a median of once [s2].

Dose mattered at first and then stopped mattering. Among those completing at least four of six cores within nine weeks (82%, n=169), the estimated marginal mean insomnia score at post-treatment was 9.62 (95% CI 8.83 to 10.40) against 12.48 (95% CI 10.44 to 14.50) for those completing fewer (p=0.01) [s2]. By six and 12 months that gap had closed to non-significance (p>0.29 for both) [s2].

Adding stepped human support produced consistently higher response and remission rates than the automated programme alone, but both arms beat the control decisively, and the trial was not powered to establish the support arm as superior to the unsupported one [s1][s2].

What this does not show

The comparator was online patient education, not another active insomnia treatment and not no treatment. Patient education improved insomnia severity by a small-to-medium margin on its own [s2], so the contrast measures what CBT-I content adds over information — a fair test, but not a test against therapist-delivered CBT-I, which remains the standard this is meant to substitute for.

The sample was 91% white and heavily college-educated [s2], recruited online during 2018–19, and outcomes were self-reported through diaries and questionnaires rather than objective sleep measurement. Participants could not be blinded to which programme they received.

And there is a definitional point worth keeping. Remission was defined as an index score below eight — "no clinically significant insomnia" [s2]. Half the supported arm reached that at a year. Half did not.

Why it matters

The authors' framing is that digital CBT-I offers strong potential to expand access to insomnia treatment for an underserved population [s1]. That claim rests on the durability of the result more than its size: the effects held at 12 months, in an age group where insomnia typically runs for a decade or more [s2], from a programme that requires no clinician time to deliver.

This article is informational and is not medical advice.

Sources

• [s1] A randomized controlled trial of a digital cognitive behavioral therapy for insomnia for older adults — npj Digital Medicine, published online 19 July 2025. https://doi.org/10.1038/s41746-025-01847-0
• [s2] Same paper, open-access full text and tables — Europe PMC, PMC12274496. https://europepmc.org/article/PMC/PMC12274496