A sleep protocol for serious mental illness that community clinic staff deliver in four 20-minute sessions
- Source and design. In Implementation Science (July 2025), Harvey and colleagues at UC Berkeley cluster-randomised ten California counties between two versions of TranS-C, a modular treatment for sleep and circadian problems, and within each county randomised patients of community mental health centres to immediate treatment or to usual care with treatment deferred. Participants: 396 adults with serious mental illness – 74 on the standard version, 124 on the adapted one, 198 in the deferred arm – treated by 93 centre staff. Psychosis was reported by 31.08% and 32.26% of the two treated arms, bipolar disorder by 29.73% and 23.39%. NCT04154631; funded by NIMH.
- Entry and dose. On the sleep side, eligibility rested on one PROMIS item scored 4 or 5, with no insomnia diagnosis asked for. Adapted TranS-C then takes 20 minutes a week for four weeks, standard TranS-C 50 minutes a week for eight.
- Sleep, end of treatment against usual care, both versions together, self-report only. PROMIS Sleep Disturbance, the primary outcome: b = -10.91, p < 0.001, d = -1.52. PROMIS Sleep-Related Impairment: b = -9.52, p < 0.001, d = -1.06. No diary or actigraphy was collected.
- Psychiatric outcomes, same comparison. DSM-5 Cross-Cutting Measure, scored 0–52: b = -6.72, p < 0.001, d = -0.52. Sheehan Disability Scale: b = -5.12, p < 0.001, d = -0.71. The deferred arm had started the protocol before the six-month assessment, which leaves the trial without a between-arm result at six months.
The sleep criterion is met by a single answer: 4, "quite a bit", or 5, "very much", on any of the eight PROMIS Sleep Disturbance items covering the past seven days. No insomnia diagnosis was required, and no minimum duration of the complaint was checked. The remaining conditions are a serious mental illness confirmed by the referring provider or a MINI interview, and a guaranteed place to sleep for at least two months that is not a shelter. Suicidal ideation alone excluded no one; night shifts more than twice a week and pregnancy did.
Sessions and modules
The adapted version of TranS-C is four weekly sessions of 20 minutes. Each carries four cross-cutting elements: case formulation, education about sleep and circadian rhythms, motivational enhancement and goal setting. The core work is split into five short modules – irregular sleep and wake times, winding down, waking up, daytime impairment, and keeping the changes going – with one optional module on worry about sleep. Unhelpful beliefs about sleep, a module of the standard eight-session version, were left out. No session-by-session plan is given beyond a note that core modules 1–3 go to separate sessions. Completers had a mean of 4.99 sessions (SD 1.91) in the adapted arm and 8.95 (SD 8.01) in the standard one.
The staff who delivered it
Half the patients in the standard arm stopped during treatment; in the adapted arm the figure was 25.81%. The versions also differed in how long their providers trained – four hours for the adapted version and 6–8 hours for the standard one – while supervision in both was a weekly drop-in session offered by external facilitators and attended at will. Among the 93 providers, one held a nursing degree and one a medical degree. Marriage and family therapists and social workers were the largest groups, and just over half were licensed. The eligibility rules asked for client-facing work at the centre, interest in the protocol and consent; prior training in sleep treatment was not among them, and each centre decided whom to put forward.
Against usual care
Control patients had usual care for four or eight weeks and then received the same protocol. From the start to the end of treatment, with both versions pooled, sleep disturbance moved from a T-score of 62.76 to 50.83 against 62.80 to 61.65 (d = -1.52); it was the primary outcome and the only patient measure corrected for multiple testing. Psychiatric symptoms on the DSM-5 Cross-Cutting Measure went from 24.3 to 18.88 against 24.18 to 23.35 (d = -0.52), and functional impairment changed with d = -0.71.
Every sleep measure was a questionnaire; the authors judged a diary and actigraphy unrealistic for this sample. Because the control group had also been treated by the six-month follow-up, no comparison between arms exists beyond the end of treatment. Suicidal ideation was compared only between versions (no difference, p > 0.10); neither the paper nor its supplement compares it with usual care.
Four 20-minute sessions, delivered by community centre staff of whom no prior training in sleep treatment was asked.
The control arm was usual care followed by delayed treatment, with no attention-matched comparison, and the comparison with it covers only the start to the end of treatment; by the six-month follow-up the control group had received the protocol, so no between-group result exists there. Every sleep outcome was self-report: PROMIS scales and a composite built from questions about the past week, with no sleep diary, actigraphy or polysomnography. Diagnoses rested on self-report, checked either by the referring clinician or with the MINI, not on a structured interview for every patient, which the authors explain by the need to limit patient burden in a large multi-site study. A sleep diary and actigraphy were left out for a different reason: the authors judged them unrealistic here and cite a view that they are not essential for treatment studies of large community samples in routine practice. They present both choices as aspects that might look like limitations but are essential elements of the study. Assessors knew the allocation at the pre-treatment assessment; the authors tried to keep them blind at post-treatment and follow-up. Dropout in the immediate-treatment arms was 14.14% before the first session and 34.48% between the first session and the end of treatment. Randomisation was by county, ten clusters, and the arms were unequal because recruitment differed between counties. Psychiatric symptoms and functioning were secondary outcomes and were not part of the multiple-testing correction; the authors write that the secondary outcomes are not confirmatory and that the delayed-treatment design may inflate the effect size differences. Results against usual care are reported for the two versions pooled. Beyond a note in Additional File 1 that core modules 1–3 are delivered in different sessions, no session-by-session plan for the adapted version is given. Neither the main text nor the supplement compares TranS-C with usual care on suicidal ideation: the supplement tables that report it contrast only the two versions, and the only supplement table that sets outcomes against usual care (Table 9) is an exploratory moderation analysis of the primary and secondary outcomes. The setting was publicly funded community mental health centres in California.