Effects of a Tailored Brief Behavioral Therapy Application on Insomnia Severity and Social Disabilities Among Workers With Insomnia in Japan

Key Points Question Does a smartphone application for brief behavioral therapy for insomnia improve insomnia-related symptoms and worker productivity? Findings In this randomized clinical trial including 92 workers with insomnia, use of an application that provided tailored brief behavioral therapy for insomnia for 2 weeks significantly decreased insomnia severity and social disabilities and improved work performance after 3 months. Meaning These findings suggest that individually tailored brief behavioral therapy for insomnia delivered via smartphone application is an inexpensive and effective treatment for workers with insomnia.


Introduction
Nearly 20% of the general adult population is reported to have experienced symptoms of insomnia. [1][2][3] Additionally, 10% to 15% of people with insomnia experience chronic insomnia, 1,4,5 which is associated with the development or relapse of depression, as well as resistance to depression treatment. 6,7 Cognitive behavioral therapy (CBT) for insomnia (CBTI) has been recommended as an effective first choice intervention for chronic insomnia disorder, 8,9 and it has been found to be effective for improving insomnia symptoms in 70% to 80% of patients 10 and to have a long-term preventive effect on symptom recurrence. 11 Meta-analyses indicate that CBTI has moderate to large lasting effects on insomnia severity, sleep quality, sleep efficiency, sleep-onset latency, and wake-up time after sleep onset. [12][13][14] In general, a full course of CBTI is conducted face-to-face and carried out during 4 to 8 weekly sessions, while brief behavioral therapy for insomnia (BBTI) can be carried out in less than 4 weeks in a clinical or primary care setting. 15,16 However, it may be difficult to schedule more than a few appointments with a health care practitioner (eg, physician, clinical psychologist, or nurse) who is also responsible for the care of hundreds of other patients. Since there are many individuals with insomnia symptoms, 3 a face-to-face intervention may not be feasible treatment for individuals with insomnia who need CBTI.
In a 2011 proposal by Mack and Rybarczyk, 17 a stepped-care model of psychological management for insomnia was suggested. A stepped-care model can be thought of as a pyramid. The model by Mack and Rybarczyk 17 proposes attempting lower-cost interventions as the first choice, while more expensive and intensive interventions are reserved for individuals who do not respond to less intensive interventions. The least intensive interventions are self-help courses, such as bibliotherapy and internet-or application-delivered treatments. These interventions are less expensive and potentially more readily available to a greater number of people with insomnia. These interventions are also used to alleviate more insomnia symptoms at the bottom of the pyramid, while fewer individuals with insomnia who do not respond to these treatments receive progressively more intensive and individualized treatments toward the top of the pyramid. A meta-analysis that analyzed computerized CBTI reported relative success with a medium to large effect size. 18 Many studies into the efficacy of digitally delivered CBTI have had experimental limitations that may have affected the outcomes. First, the insomnia remission rate for CBTI with psychological support from CBT experts has been reported as higher than that of CBTI without psychological support (61% vs 24%). 19 The low remission rate associated with CBTI without psychological support may be associated with its use in patients with more severe chronic insomnia disorder. These individuals are likely to require more intensive support from higher levels in the pyramid, such as  20,21 However, despite being highly effective, the second limitation of fully automated CBTI studies is the dropout rate, which has been reported as 39% to 45%, higher than that of face-to-face CBTI. 20,21 Therefore, the number of sessions that patients complete could be a crucial factor for the success of fully automated CBTI as an intervention.
According to the stepped-care model, providing CBTI to people with insomnia symptoms is an important preventive measure for first-and second-stage interventions for individuals with less severe insomnia. Vincent and Walsh 22 examined the factors associated with a patient's movement through a stepped-care pathway using 50 adults with chronic insomnia. They found that digitally delivered CBTI was sufficient to improve insomnia symptoms particularly for younger adults, individuals who were employed, and individuals with less severe insomnia.
For the treatment of depression, individually tailored, internet-delivered CBT has been shown to be more effective for many participants than a standardized intervention. 23 Additionally, Forsell et al 24 reported that individually tailored interventions reduced the number of failed treatments, although it is worth noting that patients in that study also had time with a therapist.
However, to our knowledge, no study of digitally delivered CBTI has made a direct comparison between tailored and standardized interventions that takes into consideration how these differ from waiting list controls. To approach this problem, we developed a smartphone application that provides a 2-week tailored BBTI designed to enhance the effect of digitally delivered CBTI and to reduce the dropout rate during the intervention. The aim of this study was to examine the effects of a fully automated and individually tailored intervention on insomnia-related symptoms, social disabilities, and worker productivity among workers with insomnia compared with standard BBTI, self-monitoring, and a waiting list control group.

Methods
This study was approved by the Ethics Committee of the Waseda University. All participants provided written informed consent. The study was conducted and reported in accordance with the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines.
Therefore, all authors were blind to the allocation. Participants were randomized to tailored BBTI, standard BBTI, self-monitoring with sleep diaries, or waiting list control.
For the purpose of informed consent, participants received information on background and purpose of the research, type of research intervention, voluntary participation, duration, risks, benefits, confidentiality, sharing the results, right to refuse or withdraw, alternative to participating, and who to contact. If they agreed to all contents, they were asked to sign an agreement form.

Assessment Measures
Participants were assessed via an internet form at baseline, 2 weeks after the intervention, and at the 1-month and 3-month follow-ups. The primary outcomes were insomnia severity, measured using the total score of the Japanese version of ISI, 25

Sample Size
Sample size was based on a power analysis conducted for the ISI scores. Effect sizes were estimated from 2 weeks of BBTI pilot data acquired by our group prior to this study: Hedges g in the mean (SD) (2-sided), it was calculated that 12 individuals would be required for each group. Since the reported dropout rate for self-help CBTI is highly variable among studies, 30 we aimed to allow for a 40% dropout rate, which required 20 participants to be recruited per group.

Interventions
Participants downloaded the application for sleep improvement developed for this study to their Android or iPhone smartphone. All participants were assessed for sleep-related daily habits (eg, bed/wake time, physical activities, exposure to bright light) before the intervention, 2 weeks after the intervention, and at the 1-month and 3-month follow-ups. Each participant was randomized to 1 of 4 groups: tailored BBTI, standard BBTI, self-monitoring, or waiting list (control group). Full details on interventions are provided in the trial protocol in Supplement 1.

Tailored BBTI
For participants in the tailored BBTI group, we prepared 26 challenge tasks based on the results of their baseline assessments (eTable 1 in Supplement 2 science was delivered each day. After completing the questionnaires at the 3-month follow-up, participants were allowed to enroll in a program of tailored BBTI.

Waiting List
Participants assigned to the waiting list control group were only asked to complete questionnaires at baseline, 2 weeks after the intervention, and at the 1-month and 3-month follow-ups. After completing the questionnaires at the 3-month follow-up, participants were allowed to enroll in a program of tailored BBTI.

Statistical Analysis
Most analyses were based on the intent-to-treat model. To examine the effect of tailored BBTI on insomnia-related symptoms and productivity, a generalized linear model, which compensates for missing data, was used to compare preintervention, postintervention, 1-month follow-up, and 3-month follow-up data in all groups. If data were missing, they were compensated through the multiple imputation method using Markov chain Monte Carlo estimation because it is significant at Little missing completely at random test (χ 2 27 = 41.75; P = .04). The multiple imputation used age, sex, and variables measured at each period for each group by the multivariate imputation by chained equations algorithm. When main or interaction effects in all analyses were shown, we performed Bonferroni-Holm correction for P values, then conducted post hoc analyses. We introduced the Bonferroni-Holm correction for P values, which provided more conservative P value estimations, to avoid the risk to observing significant P values because of repeated post hoc analyses. Descriptive statistics were computed using R statistical software version 3.4.4 (R Project for Statistical Computing).
In addition, we estimated the effect sizes of scales within and between groups by using correcting biases for Hedges g. In general, an absolute g value of 0.2 or more indicates a small effect size; approximately 0.5, moderate; and 0.8 or more, large. 31 The effect sizes of all scales within groups were computed as baseline vs after the intervention, baseline vs 1-month follow-up, and baseline vs 3-month follow-up. Effect sizes between groups were analyzed after the intervention, at the 1-month follow-up, and at the 3-month follow-up. For all scales, except for the WLQ, the more negative the change in the effect size, the larger the therapeutic effect.
To compare dropout rates among groups, we conducted χ 2 tests. Also, we conducted probit regression analysis by the number of participants who dropped out in each period as the dependent variable, and group and ISI at baseline, postintervention period, or 1-month follow-up as independent variables.

Baseline Characteristics
A total of 288 individuals responded to our recruitment materials, and 196 individuals (68%) were excluded because they did not meet inclusion criteria or did not consent to participate.

Insomnia Severity
We found a significant effect of time (F 3,88 = 25.48; P < .001; η G 2 = 0.09) and interaction (F 9,264 = 5.25; P < .001; η G 2 = 0.06) for the total ISI score (Figure 2). In the post hoc analysis, the tailored BBTI group showed significantly more improvement in ISI score than the waiting list group at the 1-month follow-up (g = -0.

Social Disabilities
The results of the analysis of the all subscales of SDISS showed a significant effect of time (social life:  Post hoc analyses showed that SDISS work performance scores were statistically significantly improved in the tailored BBTI group at the 3-month follow-up compared with the waiting list group

Discussion
This randomized clinical trial examined the effects of a smartphone application of tailored BBTI for workers with insomnia on insomnia severity, social disabilities, and work productivity. The mean age of participants was 42.7 years, which corresponds to a relatively young group of employees. A review by Morin et al 10 reported that older patients are less responsive to behavioral treatments than middle-aged or younger adults. In addition, CBTI that uses digital technology has been reported as particularly effective for younger adults and employees. 22 Therefore, providing BBTI for participants is appropriate according to a stepped-care model for insomnia. After the intervention, the dropout rate for tailored BBTI was lower than that for standard BBTI (17% vs 30%), although the difference was not statistically significant. This means that the participants who received standard BBTI dropped out of the intervention 1.8-fold more than those who received tailored BBTI, which is consistent with results from a 2019 randomized clinical trial. 24 In the primary outcomes, tailored BBTI significantly and greatly improved both insomnia severity and social disabilities at 1 and 3 months after the intervention compared with the waiting list control group. In particular, it is noteworthy that the large effect of 2 weeks of tailored BBTI on insomnia severity and daytime dysfunction after the intervention is similar to the findings of other studies on digitally delivered CBTI 18 and face-to-face CBTI 13 for 4 to 8 weeks. These findings suggest that a tailored intervention based on the stepped-care model can maximize the effect of CBTI, that the effect is sustained after only 2 weeks, and that this may be an effective prophylactic treatment for chronic insomnia disorder. The effects of standard BBTI and self-monitoring on primary outcomes were also medium to large at the 3-month follow-up compared with the waiting list. Out of these, the effect sizes within each group showed a similar change between tailored and standard BBTI. These findings are consistent with a 2012 study 23 of internet-delivered CBT for depression. That study found that standard CBT was effective for participants with mild depression but that tailored CBT could also be effective for those with mild to severe depression. Since the participants of our study had mild insomnia, they received a tailored or standard BBTI or self-monitoring interventions that could theoretically improve insomnia symptoms. However, improvements in insomnia severity at the 1-month follow-up were confirmed only in the tailored BBTI group. This suggests that tailored BBTI is a quicker-acting intervention than standard BBTI, which is the first time this result has been reported for tailored digitally delivered CBT, to our knowledge.

JAMA Network Open | Occupational Health
In the secondary outcomes of dysfunctional beliefs and sleep reactivity, the tailored BBTI group had statistically significant medium to large improvements in both outcomes at 3 months after the intervention compared with the self-monitoring and waiting list groups. Remarkably, in the FIRST assessment, while the participants who underwent tailored BBTI showed statistically significant moderate improvements with sleep reactivity, dysfunctional beliefs were significantly reduced in both tailored and standard BBTI groups. It has been shown that dysfunctional beliefs and sleep reactivity serve as risk factors for maintaining insomnia 32,33 and that sleep reactivity works as a marker for increased risk of insomnia and as a risk factor for the onset of depression. 34 Furthermore,  worker productivity significantly and greatly improved in the tailored BBTI group compared with the standard BBTI group at 3 months after the intervention. The effects of tailored BBTI on sleep reactivity and productivity, as well as on other insomnia-related symptoms, suggest that a 2-week program of tailored BBTI for workers with insomnia is effective for improving sleep and enhancing productivity. Therefore, it is advantageous in terms of both primary and secondary prevention to provide an application for individually tailored BBTI to workers with insomnia.

Limitations
This study has some limitations. First, the participants of this trial were all business workers aged approximately 40 years who had been categorized as experiencing insomnia based on the results of a previous study. 22 According to the stepped-care model, digitally delivered CBTI is also effective for young adults and those with better sleep. 22 In future research, it would be informative to examine the effect of tailored BBTI on insomnia-related symptoms among people with insomnia aged 20 to 30 years. In addition, the intensity of work an individual performs (eg, light office work vs heavy manual labor) might influence the individual's quality of sleep, although we excluded shift workers and the individuals who had higher risk of causing serious harm from sleep loss (eg, individuals who operate heavy machinery). Second, there was a high dropout rate after 3 months. In particular, it was higher in both of the BBTI interventions. The reason for this is not clear from our study design. It would be informative to investigate the reasons for high dropout rates during follow-up periods and the behaviors of people who are prone to dropout in future research.