Why completing more of a digital CBT-I program is linked to better sleep
Automated digital CBT-I delivers the same core treatment as therapist-led CBT-I—changing sleep habits and unhelpful beliefs about sleep—but through a self-guided program rather than a clinician. A systematic review of 103 CBT-I studies found that for digital programs, completing more treatment modules was linked to improvements on the Insomnia Severity Index (ISI), a standard questionnaire measure of insomnia severity [4]. That is the clearest adherence–outcome signal in the evidence, but it rests on only five digital studies, so it should be read as a promising pattern rather than a settled fact [4].
The same review complicates the picture for therapist-led CBT-I: neither overall adherence nor adherence to any specific component reliably predicted sleep outcomes [4]. The authors suggest this may be because studies measured adherence in very different ways—sleep diaries, questionnaires, interviews, actigraphy, or module counts—making it hard to compare results across studies [4]. In other words, the weak adherence–outcome link in therapist-led care may partly reflect measurement noise rather than adherence truly being irrelevant.
Real-world uptake can be built into a health system, but ordering is not the same as finishing
One large implementation study shows that real-world uptake is not doomed to stay low. Within the Henry Ford Health System, clinicians ordered digital CBT-I for 2,576 patients during an 18-month period of active implementation support, and orders continued at roughly 159 per month for the 12 months after that support ended—about 4,487 patients in total by December 2025 [2]. The authors describe this as self-sustaining adoption, supported by embedding the order process in the electronic health record and by staff training and reminders [2].
That result is about orders, not about how many patients actually started or completed the program, and the study did not report insomnia outcomes [2]. So it answers a question about clinical uptake—whether doctors will offer and order the treatment—rather than about whether patients stick with it or sleep better. The gap matters because poor completion is a well-documented problem in digital CBT-I even when access is easy [5].
Who drops out, and what might keep them engaged
Patient feedback helps explain why adherence is poor despite high accessibility. In interviews, written feedback, and a focus group with low-socioeconomic-status racial minorities, users generally found the program content engaging and valued doing treatment privately at home, and many said the virtual therapist character increased their engagement [5]. Barriers clustered around eHealth literacy—for example, difficulty filling out the sleep diary—and a wish for more human contact [5].
Notably, non-completers largely agreed that additional support would have prevented dropout, suggesting things like a help button or someone to ask questions, and some felt the program did not account for physical causes of their sleep problems [5]. The authors conclude that adding personalized support and tailored content may improve adherence and completion, though this is a proposal based on patient perspectives rather than a tested intervention [5].
How trial results and real-world use diverge
The evidence does not include a direct head-to-head comparison of automated digital CBT-I trials against real-world use on uptake, adherence, and attrition together. What it does show is that these are different questions with different answers. Trials and meta-analyses can link module completion to insomnia improvement [4] and can measure symptom outcomes with standardized sleep metrics [3], while implementation studies can show that ordering and adoption persist at health-system scale [2] without reporting whether patients completed treatment or slept better.
There is also a hint that automated digital CBT-I may produce weaker symptom outcomes than individual therapist-delivered CBT-I in some populations: among people with comorbid mood and anxiety disorders, response and remission rates were lower for automated digital CBT-I than for individual CBT-I at about 9 weeks and again at about 33 weeks [1]. Because that finding comes from a comorbid population, it may not generalize to unselected insomnia patients, but it is a reminder that the format of delivery can matter for outcomes, not just for adherence.
Sources used in this answer
Digital and telehealth cognitive behavioral therapy for insomnia in comorbid mood/anxiety disorders
In people with comorbid mood and anxiety disorders, automated digital CBT-I produced lower response and remission rates than individual CBT-I at about 9 weeks and about 33 weeks.
0442 Real-World Sustainability of Digital CBT-I: Healthcare System Implementation to Self-Sustaining Adoption
A health-system implementation of digital CBT-I generated 4,487 patient orders over about two and a half years, with monthly ordering continuing and even increasing after active implementation support ended, though insomnia outcomes were not reported.
Sleep Healthy Using the Internet: A Meta-Analysis of Twenty Years of Research on a Digital Cognitive Behavioral Therapy for Insomnia Program
A 20-year meta-analysis of the SHUTi digital CBT-I program reports adherence rates and insomnia outcomes using standardized sleep metrics.
P093 A Systematic Review of Adherence to Cognitive Behavioural Therapy for Insomnia (CBT-I) - Key Findings
Across 103 CBT-I studies, adherence measurement was highly heterogeneous; for digital CBT-I, module completion was linked to ISI improvement in five studies, while adherence did not reliably predict sleep outcomes in therapist-led CBT-I.
0464 Patient perspectives on facilitators and barriers to engagement with digital CBT-I
Patients found digital CBT-I content engaging and valued privacy and the virtual therapist, but lower health literacy and a desire for human contact were barriers to engagement, and non-completers said additional support would have prevented dropout.