This single-arm feasibility study aimed to assess adherence, usefulness, satisfaction, and effect size for the Mi Sleep Coach app as an intervention for insomnia.

Adult survivors of breast, prostate, and colon cancer who had completed curative treatment at for least 3 months and not more than 5 years before study entry and reported sleep disturbances for at least 3 months (see Textbox 1 for full inclusion or exclusion criteria) were recruited from the University of Michigan Rogel Cancer Center and through social media between March 2021 and June 2022.

An adherence rate of at least 75% was considered acceptable. Participants who engaged with the app at least 4 days per week over the 7 weeks were considered adherent.

Attrition rates were assessed weekly and for the study overall. Anyone who did not complete data collection for the full 7 weeks of the study was considered to be prematurely withdrawn and included in our attrition count. An attrition rate of 25% or less was considered feasible.

User experience and satisfaction were assessed with a 40-item investigator-developed user experience survey based upon the Unified Theory of Acceptance and Use of Technology that assessed the perceived ease of use and usefulness of the Mi Sleep Coach app features [56]. Ease of use is assessed by asking participants to rate 10 statements on a scale of 1 (very hard) to 4 (very easy). Participants were asked to rate how useful they found 10 different features (chatting with the avatar, viewing video messages, daily sleep log, personal sleep plan, reading lessons, reading stories from other people with sleep problems, creating a personal list of daily activities, creating a list of coping strategies, constructive worry activities, or deep breathing activities) of the Mi Sleep Coach app on a scale of 1 (not at all useful) to 4 (very useful). Satisfaction was assessed with statements such as “overall, I am satisfied with the Mi Sleep Coach app” and “I would recommend the Mi Sleep Coach app to other people who have trouble sleeping” using a 1 (strongly disagree) to 5 (strongly agree) scale. They were also asked to answer open-ended statements such as “please tell us what you liked about the Mi Sleep Coach app.” This was collected at baseline and week 7 via QualtricsXM [Qualtrics]), a survey software tool. We considered the app acceptable if (1) at least 75% (23/30) of participants considered at least 1 app feature at least a little useful (2 or greater on a scale of 1-4); or (2) at least 80% (24/30) of participants were satisfied (4 or greater, on a scale of 1-5) with the intervention and that at least 80% (24/30) of participants would recommend it to others (4 or greater) on a scale of 1 to 5.

The Insomnia Severity Index (ISI) is a 7-item questionnaire that measures a patient’s perceptions of the nature, impact, and severity of insomnia. This scale rates items on a Likert scale ranging from 0 to 4 for a maximum score of 28. ISI has good internal consistency (Cronbach α 0.90) [57]. Higher scores indicate more severe insomnia. A 6- to 8.4-point reduction is associated with moderate improvement in insomnia after treatment [50,58]. This measure was collected via QualtricsXM at baseline and week 7. We estimated that the effect size of the intervention would be at least as high as other effect sizes reported in the literature (approximately Cohen d 0.6 to 0.7) for nontherapist delivered CBT-I.

Sleep medication use was assessed by asking participants to report both prescription and over-the-counter (OTC) medications used for sleep within the last 2 weeks at baseline and week 7 via QualtricsXM.

The brief DBAS-16 is a 16-item measure that assesses faulty beliefs and unrealistic expectations about sleep using statements such as “I am worried that I may lose control over my ability to sleep” [54]. The measure has a good internal consistency of 0.77 [54]. Higher scores indicate more DBAS. This measure was collected via QualtricsXM at baseline and week 7 and within the app during week 5.

Participants were asked to report any new symptoms, illnesses, or diagnosed health problems to study team members. These adverse events were categorized using the Common Terminology Criteria for Adverse Events [59].

The data were summarized using descriptive statistics (mean, SD, median, percentage, and frequency). The demographic and clinical characteristics including ISI, DBAS-16, and adverse events were descriptively analyzed with effect size calculations done for the validated quantitative measure. The frequency of study participants taking prescription and OTC sleep medications at baseline and week 7 was compared using χ² testing. A full missing-data analysis was conducted using IBM SPSS statistical software (version 27; IBM Corp).

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Institutional Review Board of the University of Michigan on March 18, 2021 (HUM00194610). Informed written consent was obtained from all individual participants included in the study. The study team follows best practices for protecting patient privacy, including the encryption of participant data when at rest or in motion and that final study data sets are deidentified [60]. No protected health information was collected through the app. CHCR is a trusted IT service provider for the University of Michigan Health System and all CHCR programs undergo extensive information assurance and cybersecurity review including active penetration testing.

We recruited 30 participants diagnosed with breast (n=24, 80%), prostate (n=3, 10%), and colon (n=3, 10%) cancer between March 2021 and June 2022. The mean age of participants was 54 (SD 9.1) years. Further, 27 (90%) identified as female and 3 (10%) identified as male. In total, participants reported that they were White (n=22, 73%), Black (n=5, 17%), Asian (n=4, 13%), American Indian or Alaska Native (n=1, 3%), and Hispanic or Latino (n=1, 3%). Some participants reported being of more than 1 race or ethnicity. Most (n=25, 83%) had graduated from college or had an advanced degree after college. Furthermore, 5 (17%) reported a high school diploma or some college.

Weekly adherence rates ranged from 96% to 100% (week 1: 97%, week 2: 97%, week 3: 100%, week 4: 100%, week 5: 100%, week 6: 100%, and week 7: 97%). The overall adherence rate for the study was 99%.

All 30 (100%) participants completed the study and provided data through week 7.

Every app feature, with the exception of reading stories from other people with sleep problems (9 out of 10), was found to be useful by 80% (24/30) to 93% (28/30) of participants. The highest-rated components were keeping a sleep log, obtaining a personal sleep plan, and reading lessons about sleep. Further, 27 (90%) of the 30 participants were satisfied with the Mi Sleep Coach app and 28 (93%) would recommend the use of the Mi Sleep Coach app for those with sleep issues.

The mean change score on the ISI of 8.1 was a significant decrease from baseline (18.5, SD 4.6) to week 7 (10.4, SD 4.2; P≤.001; Cohen d=1.5; see Table 1). At baseline, all ISI scores ranged from 9 to 26, indicating mild sleep disturbance for 5 (17%) participants, moderate insomnia for 19 (63%), and severe insomnia for 6 (20%). At week 7, a total of 8 (27%) participants scored in the no clinical insomnia range (0-7) while 18 (60%) scored in the mild sleep disturbance range (8-14) and 4 (13%) scored in the moderate insomnia range (15-21). Further, 6 (20%) participants scored in the severe insomnia (22-28) range at baseline, and none scored in the severe insomnia range at 7 weeks (see Table 2).

At baseline, 7 (23%) participants were taking prescription medications for sleep. At 7 weeks, only 1 (3%; χ² P<.001) participant was still taking prescription sleep medications. Likewise, at baseline, 14 (47%) participants were taking OTC medications for sleep. At 7 weeks, this number deceased, so that 9 (30%; χ² P=.03) participants were still taking OTC sleep medications.

The mean change score on the DBAS-16 of 11 was a significant decrease from baseline (47, SD 10.8) to week 7 (36, SD 12.7, P<.001, Cohen d=1.3; see Table 1). Beliefs and attitudes about sleep were moderately positively correlated with insomnia, as measured by the ISI, at baseline (R=0.574) but only mildly correlated with insomnia at 7 weeks.

In total, 4 participants reported adverse events while in this study. All adverse events were not related to the Mi Sleep Coach intervention. Further, 2 were mild (1 pain in extremity and 1 stomach pain) and 2 were moderate and required medication (2 allergic rhinitis).

This study evaluated the innovative Mi Sleep Coach mobile dCBT-I program tailored for survivors of cancer. The Mi Sleep Coach program was found to be highly feasible with 100% (N=30) of participants completing all 7 weeks of the study and providing data through week 7. Additionally, the overall program adherence rate for the study was 98%. In total, 90% (n=27) of the Mi Sleep Coach app participants were satisfied with the program and 93% (n=28) would recommend the use of the Sleep Coach app for those with sleep issues. The Mi Sleep Coach program also showed promise in addressing insomnia among survivors of cancer. Program users experienced a significant decrease in the mean change on the ISI of 8.1, from baseline (18.5, SD 4.6) to week 7 (10.4, SD 4.2; P≤.001), which is also represented by shifts from moderate-to-severe to mild-to-no clinical insomnia. Program users also reported a spontaneous decrease in the use of both OTC and prescription sleep medications.

The high rates of engagement and adherence to the Mi Sleep Coach program are notable. Some caution that digital interventions are particularly susceptible to attrition because participants can easily access resources a few times before abandoning them [55]. However, the high usefulness and satisfaction scores attributed to the Mi Sleep Coach app might indicate a lower likelihood of dropout. Recent systematic reviews and meta-analyses of randomized controlled trials of internet-delivered CBT have found attrition rates in the range of 21.6% to 24.7% [27,61]. While only a few of the studies focus on survivors of cancer, similar or lower dropout rates have been found in other CBT-I studies in survivors of cancer [28,29,62]. More research is needed to understand how to ensure the retention of participants in digital interventions and the possible role of attrition as a moderating variable [63]. For multicomponent CBT-I interventions, adherence varies widely depending on the details of the study. Similar to our findings, a randomized controlled trial of 30 survivors of breast cancer found adherence to 2 eHealth CBT-I components ranged from 99% to 100% [64]. However, in that study, participants met weekly in a group video conference environment, making it less like the self-managed approach used by Mi Sleep Coach. Adherence to the Mi Sleep Coach program compares favorably to reports from a fully internet-delivered CBT-I intervention for survivors of breast cancer (adherence ranges from 59.7% to 82.1%) [29]. The use of engagement strategies in the Mi Sleep Coach, such as the digital health agent, may have resulted in higher adherence rates.

The large treatment effects of the Mi Sleep Coach program (eg, Cohen d=1.5 for ISI) are encouraging. This effect is consistent with findings from 2 meta-analyses of largely person-delivered CBT-I for survivors of cancer that found large treatment effects (pooled effect sizes 0.77-0.78) [16,22]. The Mi Sleep Coach effects also compare favorably to the effects of dCBT-I in general where 2 meta-analyses found moderate pooled effects (eg, Cohen d=0.40, decrease in insomnia severity by 4.3 points) [26,27]. The finding of reduced use of sleep medications among Mi Sleep Coach participants may be particularly important. A recent national study found that approximately half of survivors of cancer with poor sleep reported using sleep medications [65]. This is of concern because the use of OTC or prescription sleep medications is associated with an increased risk of falls and fractures among older adults and survivors of cancer [66-68].

The limitations of this study include a small sample size, lack of a comparator, and the use of self-report measures for sleep. However, this study was a feasibility study and as such was designed to assess adherence, attrition, and usefulness. Additionally, sleep diaries have been used to measure the primary outcome in numerous sleep studies and have been validated against both functional and physiologic measures of sleep [69].

The results of this pilot study support that the Mi Sleep Coach app is a promising intervention for cancer-related insomnia. Future research should be conducted to test the Mi Sleep Coach app in a randomized controlled trial with sufficient control to determine its efficacy. If proven to significantly decrease insomnia in survivors of cancer in future randomized controlled clinical trials, this intervention would provide more survivors of cancer with easy access to evidence-based CBT-I treatment.