Non-guided, Mobile, CBT-I-based Sleep Training in War-torn Ukraine: A Feasibility Study

Objectives To study whether a mobile, unguided Cognitive Behavior Therapy-based Intervention for Sleep Disturbance, Sleep2 is feasible, acceptable, and reduces mental health/sleep disturbance symptoms among the Ukrainian population during the ongoing war. Methods A single-arm, open-label, uncontrolled pre-post evaluation study was conducted with 487 registered participants: 283 started, out of which 95 completed without an ambulatory heart rate (HR) sensor and 65 with. Assessments were conducted using online questionnaires and continuous objective measurements via HR sensors. Key outcome measures included sleep disturbance, insomnia, fear of sleep, anxiety, depression, PTSD, perceived stress, and somatic symptoms. Results Engagement with the program was robust, achieving an 80.72% compliance rate, alongside high levels of feasibility and acceptance. Participants reported significant pre- post reductions in the severity of sleep disturbance (by 22.60%), insomnia (by 35.08%), fear of sleep (by 32.43%), anxiety (by 27.72%), depression (by 28.67%), PTSD (by 32.41%), somatic symptoms (by 24.52%), and perceived stress (by 17.90%), all with medium to high effect sizes. Objective sleep measurements showed a slight reduction in sleep onset latency. Conclusion The ‘Sleep2Ukraine’ program demonstrated high feasibility and acceptance, with significant improvements in subjective sleep and mental health measures among participants. These findings demonstrate the potential of scalable mobile-based CBT-I interventions in war-torn regions with or without the instrument, based on the heart rate assessment.


Sleep and Mental Health in Times of War
War can be referred to as a specific form of crisis that inevitably has an impact on various aspects of life (1), including mental health (2) in both military personnel and civilians.
Besides typical trauma-related disorders such as anxiety, depression, PTSD, civil victims of war were shown to develop long-term sleeping problems that may last for years even after the conflict termination (3,4).Thus, as for sleeping problems, war has a significant impact on health via (at least) two pathways: through direct causation of trauma related disorders, as well as indirectly by disturbing resilience factors such as healthy and restorative sleep.It is also well known that sleep issues reinforce mental health complaints (5), with unrestorative sleep being the key maintenance factor for anxiety, depression, and PTSD (6).In turn, improvement of sleep quality usually improves mental health symptoms in general (7,8).As of July 2024, Ukraine is in war for more than two years.Respective recent studies showed significant deterioration of mental health in Ukraine: from the very beginning of the war (9,10) and throughout its progression (11,12).A specific problem seems to be the constant and unpredictable air attacks on the entire territory of Ukraine (13,14).Being mostly carried out during nighttime to complicate air defense, these attacks disturb sleep directly through loud raid alarm systems and likely contribute to fear of sleep through a build-up of anxious apprehension and anticipation of further disturbances and the need to rush for shelter quickly or search for a room without windows at home.Therefore, the need for easily accessible and scalable sleep coaching for Ukrainians is evident and of a high need.

Mobile-Based Sleep Intervention
One of the most efficient sleep interventions is classical, face-to-face, cognitive behavioral therapy for insomnia (CBT-I) (15).European (16) and US (17) clinical guidelines recommend CBT-I as the first line of treatment.It is a non-pharmacological, well-established intervention, teaching relaxation techniques, sleep restriction, stimulus control, psychoeducation, and cognitive strategies (16).Recent meta-analyses show overall high efficiency of such non-pharmacological, conventional, face-to-face interventions (7,18).
Digital CBT-I (dCBT-I) represents the online implementations of these successful programs, with the advantage of being ubiquitously available and cost effective (19).Also, dCBT-I has been studied and has been shown to generate mostly comparable effects to its face-to-face counterpart.In traumatized populations it showed improvements not only on sleep parameters (20,21), but also on other mental health indices (18,22).As such, dCBT-I has also been successfully used in prevention (23).

Digital CBT for Sleep Disturbances in Ukraine
As indicated above, one of the major advantages of mobile-based CBT interventions in Ukraine is their low threshold.The war conditions make mobility to specialized treatment centers difficult in some areas.Mental health resources are limited and focused on the most acute, physical condition, and the armed forces.Thus, mobile interventions provide access for everyone owning a smartphone with internet connection, thus increasing accessibility, feasibility, and likely acceptance.Yet, due to the poor knowledge of English, treatments offered in Ukrainian language are likely to increase feasibility and acceptance.

Study Design and Procedure
This study was planned as a single-arm, open-label, uncontrolled pre-post evaluation study (see Figure 1).After t0 measures, accommodation and adaptation baseline was followed by t1 measures and the 6 weeks CBT-I program, before t2 measures concluded the study.Also, after training discontinuation at t2, we lost contact and could not conduct follow-up data, neither could we follow-up on dropouts.Recruitment and registration for the study started on October 1 st , 2023 and ended on November 1 st , 2023.During registration, participants could voluntarily choose whether they want to receive the heart rate (HR) sensor free of charge, but with the necessity to return it after 6 weeks at the latest.To maintain anonymity, the sensor was shipped to participant through a third-party service provider 1 .All participants have provided written informed consent at t0, which is documented in the survey. 1 The sensors were shipped to a responsible person in Ukraine, who then contacted participants to gather the necessary details for delivery.The delivery was carried out using the courier service "Nova Poshta" (Eng."New Post").Participants were given the option to either collect the belts at the post office or have them delivered CBT-I Program (6 weeks) Baseline (1 week)

T0 T2
sleep diary HR sensor app program

T1
Self-reported data Self-reported data sleep diary HR sensor

CBT-I-based App Program
The Sleep 2 app content was translated, rewritten to fit Ukrainian language grammar, and culturally adapted (27).The program is organized into 6 consecutive levels, each featuring a set of components, based on core elements of CBT-I.It features various modes of communication, e.g., videos teaching sleep hygiene and psychoeducation, a chat-bot designed to help reframe thoughts, audio exercises for relaxation, tips on keeping good sleep habits, and blog posts discussing important scientific sleep topics (28).For details of the program content, see suppl.Table 1.During the 6-week app phase, participants were instructed to complete the program's six levels, which are designed to contain the most important contents of CBT-I.At t2 they were redirected to post-treatment online questionnaires directly from the app.Participants not reaching level 6 and completing t2 measures were considered dropouts (we did not obtain information on partial completion).

Acceptance, Engagement, and Compliance
Acceptance was measured with an adapted 12-item self-report measure about participants' experiences and impressions of the program (suppl.Table 3).Participants gave information on app usability, program helpfulness, content relevance, helpfulness of the pushnotifications, frequency of the reminders, and program duration using a 10-point Likert-type scale ranging from 1 (strongly disagree) to 10 (strongly agree).Engagement was monitored through the app's log data (frequency and duration of app use).Participants were encouraged to engage with the app via daily push-notifications.Compliance was measured by the adherence to the program's requirements: participants were required to complete at least five audio exercises, one video, and one chatbot interaction per each of the 6 levels.We also measured the sleep diary entries.
directly to their address.As such, we did not know the identities of our participants and could not match their data with their personalities, as study emails were used to log into the app, which did not disclose any personal details, ensuring complete anonymity.

Self-report Measures
Self-reported data was obtained with online questionnaires at t0 and t2 focusing on a time range of one month prior to the date of filling out the questionnaire.

Statistical Analysis
Daily/nightly data were averaged over the baseline (first 7 days of the study between t0 and t1) and over the last 7 days before the end of the treatment (t2, see Figure 1) for statistical analysis, as performed in R, version 2023.03.0 (37).Extreme outliers, that is, data points deviating more than ±2 SD from the mean, were excluded from the respective analysis.
Besides completer analyses we also report intention-to-treat (ITT) analyses (with preassessment values (t1) were carried forward to the post-assessment (t2)).

Engagement: Session Completion and Overall App Use
Overall, the N=160 completers saved 80.72% of the 50 required sleep diaries, representing good compliance.Completers used the app on an average of 39.8 days (SD = 19.12)out of a maximum of 50 days.

Usability and Program Acceptance
Most of the participants were very satisfied with the treatment and the app, specifically content clarity, content relevance, usefulness of the relaxation exercises, relevance of the proposed program levels, and with recommending the program to others (see Table 3).Participants rated the app and the sensor on specified criteria using a 10-point Likert scale (1 = strongly disagree to 10 = strongly agree).The evaluation intervals are as follows: very bad (0-2), bad (3)(4), mediocre (5)(6), good (7)(8), very good (9)(10).Formulations of items is presented in Suppl.Table 3. Results are presented for the sensor users (N = 65).
As for qualitative feedback participants reported that the sensor restricted sleeping positions, was uncomfortable to adjust and wear, and had occasional synchronization and battery issues.While many found the objective sleep analysis helpful (78%), others doubted its accuracy (12%).Many (56%) participants preferred a wrist/arm sensor instead of a chest belt.Positive feedback highlighted ease of use, diverse content, helpful relaxation exercises (especially before sleep and after air raid alarms), pleasant design, simple interface, engaging educational content and chatbot, and the ability to take and reflect on notes within the app.
Negative feedback included limited exercise variety, inability to revisit or list all exercises, lack of video adjustments (speed, narrator, subtitles), no late sleep entries or nap recordings, limited chatbot responses, and a discouraging level structure suggesting continuous sleep improvement; a topic-based division was preferred.

Pre-post Changes on Sleep and Mental Health
Completers reported statistically significant improvements on all symptoms after the sleep training including sleep disturbance, insomnia, fear of sleep, anxiety, depression, PTSD, perceived stress, somatic symptoms, and resilience with medium to high effect sizes (Table 4, see also density distributions and clinical cutoffs in Supl.Figure 1).Similar results were found in the ITT analysis (N = 283; see Supl.Table 4), which speaks against selective dropout effects.

Objective Sleep Measures (Sensor Users Only)
Table 5 shows that the N=65 program completers with sensors showed small improvements on objective SOL (reduction of time to fall asleep from ~29 to 25 min).Discussion To our knowledge, this is the first study to report on a non-pharmacological, digital CBT-I intervention during an ongoing war.The results, obtained from Ukraine, demonstrated high uptake/feasibility and acceptance of the program, with significant improvements across all sleep and mental health measures.We will discuss each of these in this order in the following.

Uptake, Feasibility, Acceptance, Usability
Program uptake was moderate: out of 487 registered users, 283 started.Out of 283 starters 160 completed, thus, a completion rate was 56%, which can be considered very good, given the circumstances and the high demands of the program (6 levels, 50 required diaries).
This means that recruitment for the subsequent randomized controlled trials (RCT) should be very feasible.Interestingly, despite additional logistics via mail for the sensor and the discomfort reported by some, 65 out of the 160 program completers proceeded with the daily use of the HR sensor.
Regarding engagement, participants used the app for a duration of 6 weeks, with 57% completing the full training and showing above-average compliance and engagement.The overall dropout rate of 43.46% is similar to other studies investigating the effects of digital programs for insomnia in outpatients: 40% (38) and 34.4% (39).In fact, most participants completed at least half of the program (~20 days), with no obvious demographic differences between dropped-out participants and program completers.Whereas veridical feedback on objective sleep parameter can be very helpful for insomnia, the delivery of a HR sensor is an obvious hurdle to uptake and to largescale rollout.Given our non-randomised assignment to sensor vs. non-sensor groups we can only hint at these effects.Sensor users were more likely to be male and had less favourable financial situation.The further might reflect some gender-stereotypical technology affinity in Ukraine users, whereas the latter points to the potentially higher subjective value of receiving a costly wearable.Importantly, we cannot conclude that the sensor users were more engaged.Should this be confirmed in further studies with random assignment to sensor vs. no-sensor groups then we could remove this hurdle to large scale application (at the cost of objective sleep data experiencing chronic sleep problems, subthreshold and clinical insomnia (44 and 27%, respectively), figures that are nearly three times higher than those normally observed in Western populations (28,42).We also observed very high levels of severe PTSD among participants, potentially reflecting the ongoing air war on civilians, which highlights the necessity for PTSD-specific content in the program.In addition to this, more than 52 percent of our participants had elevated somatic symptoms, 60 percent mild to moderate anxiety or depression, and more than half reporting elevated levels of stress.
Participants were required to respond to all questions, thus yielding no missing values.If no Ukrainian adapted version of the questionnaire was available, items were translated by the authors.For details on scoring ranges, number of items, and reliability statistics see suppl.Table2.Sociodemographic data were obtained on sex, age, employment status, working conditions, financial situation, and internally displaced persons (IDP) status.andprocessed using the deep learning network as described inTopalidis,   Baron, et al. (2023)andTopalidis, Heib, et al. (2023), resulting in a detailed sleep analysis.
Sleep-Related Questionnaires.Sleep disturbance was measured with the Ukrainian version of the 19-item Pittsburgh Sleep Quality Index (PSQI, Mazur et al., 2021).Insomnia symptoms were measured with the 7-item Insomnia Severity Index (ISI; Bastien, 2001).Fear of Sleep in the past month was measured with the 13-item Fear of Sleep Inventory (FoSI; Pruiksma et al., 2014).Mental Health Symptom Questionnaires.Anxiety was measured with the Ukrainian version of the 7-item General Anxiety Disorder-7 scale (GAD-7; Shyroka & Mykolaychuk, 2020).Depression was measured with the Ukrainian version of Patient Health Questionnaire-9 (PHQ-9; Unifikovanyy Klinichnyy Protokol Pervynnoyi, Vtorinnoyi (Spetsializovanoyi) Ta Tretynnoyi (Vysochospetsializovanoyi) Medychnoyi Dopomohy, 2014).PTSD symptoms were measured with the 20-item Ukrainian version of The PTSD Checklist for DSM-5 (PCL-5; Karachevskii, 2016).Perceived Stress was measured with the 4-item Perceived Stress Scale (PSS-4; Warttig et al., 2013).Somatic Symptoms were assessed with the 8-item Somatic Symptom Scale (SSS-8; Gierk et al., 2014).Objective Sleep MeasurementIn participants receiving and using the HR sensor Polar® H10 (Polar Electro GmbH Deutschland), continuous objective sleep measures were available.HR data were read out in the morning in combination with the subjective morning sleep diary, transferred to the server via https protocol

Table 1
Sociodemographic Information, Participant Flow, Engagement, and Compliance The table presents mean values and standard deviations for all measured symptoms, together with the clinical cutoffs per every measure, number of participants and corresponding percentage.
Note.The table represents average values per every group of participants.Compliance is calculated based on the total number of entries from a single participant of daily audio exercises completed (had to be performed minimum one time per day), psychoeducational videos and chat (had to be performed at least once per level).At t0, completers reported high levels of sleep disturbance, insomnia, fear of sleep, anxiety, depression, PTSD, perceived stress, and somatic symptoms (see Table2), confirming that we attracted our target population.On average, 6.36 (SD = 13.51)out of 40 nights were disturbed by air-raid alarms.

Table 3
Summary of the Program EvaluationThe table presents the average survey results on the acceptance of the proposed program, sorted in descending order.

Table 4
Changes in Mental Health Symptoms after the Sleep Training (N=160 completers)The table presents the results of paired-sample t-test for subjective measures obtained with the questionnaires at T0 and T2 (see Figure1for details).Results are presented for program completers (N = 160).

Table 5
Note.NOA (number of awakenings) here is a count of awakenings of more than 2 continuous minutes.TIBtime in bed, TST -total sleep time, SOL -sleep onset latency, WASO -wake after sleep onset, SE -sleep efficiency.Results are presented for sensor users (N = 65).

Sample Characteristics, Pre-Post Effects on Sleep Parameters and Mental Health Participant Flow and Symptom Severity Regarding
). User satisfaction and potential app improvements.Those who completed the training and moved to t2 questionnaires were highly satisfied, especially with the (culturally adapted) training content, relaxation exercises, assigned levels, and were willing to recommend the program.The lowest scores were for the chatbot and push notifications, indicating these features were less effective.As for qualitative feedback, only some sensor users reported significant discomfort.Objective feedback was considered useful, with positive remarks on relaxation exercises and negative feedback on their limited variation and inability to revisit previous exercises.Participants also disliked the inability to record daytime naps and/or sleep shorter than four hours.However, this is a technical limitation as the algorithm has not been trained on nap data and, therefore, can at that point not be verified.The positive reception of the intervention and the relatively low dropout rate suggest that there is a significant demand and openness to internet-based interventions among people in Ukraine.This contrasts with the situation in Western countries, where internet interventions often face challenges with low uptake (40,41).One possible reason for this difference is the low availability of mobile interventions in the Ukrainian language.This indicates a general reluctance among Ukrainians to use interventions available in English or other languages.the pathology of the sample and whether we addressed the program's target group: our sample demonstrated significant sleep impairments, confirming that we reached the intended population.Before starting the training, participants who completed the program exhibited significantly elevated levels of sleep disturbance (60% of the sample), with 30%

Effects on Mental Health Symptoms and Sleep Parameters
).A main finding of our study shows that all trauma-related disorders improve over the course of a predominantly sleep-focused program.This might be related to the abovementioned indirect pathways: as sleep recovers, also depression and PTSD might do so.Yet, many of the taught CBT-based cognitive techniques are applicable to these non-sleep related symptoms just as well.Further, self-efficacy related to sleep improvements might translate onto other domains as well.Limitations and Future DirectionsOur study has several limitations.First, it used a convenience sample and lacked a control group, being a single-arm, open-label, uncontrolled pre-post evaluation study.More comprehensive follow up studies using randomized group assignment and including control groups are necessary to rule out that improvements are not happening also in the absence of any intervention, simply due to passing time, measurement repetition, symptom awareness or any other factors.However, in a war-torn population, it is ethically challenging to use an inactive control or waitlist group, which, together with the fully anonymous data, also led to the absence of the follow-up.Secondly, the lack of representative random sampling may introduce selection bias, as participants who self-subscribed to the sleep training may have more severe sleep problems and/or mental health issues and a higher affinity for digital interventions than the general population.Scalable mobile-based interventions seem crucial for the Ukrainian population, especially in war-threatened areas with available internet and electricity.Our study demonstrates that such interventions can be effectively delivered.Participants are largely reachable and ready to accept help.CBT-I is simple, safe, and effective, showing promising effects within 6 weeks.Given the high feasibility and acceptability, further research, including randomized controlled trials, is necessary to verify the efficiency of mobile-based CBT-I in Ukraine.

Table 2
Scoring Ranges, Number of Items, and Reliability Statistics for Psychological QuestionnaireNote.This table shows names of the variables, corresponding measures, reliability statistics, scoring range, and the number of items per measure.