Abstract
Objective:
Sleep disturbance is experienced by nearly 20% of Americans and is highly comorbid with anxiety. Sleep disturbances may predict development of anxiety disorders. Mindfulness training (MT) has shown efficacy for anxiety yet remains limited by in-person based delivery. Digitally delivered MT may target habitual worry processes, yet its effects on sleep have not been studied. This study tested if app-based MT for anxiety could reduce worry and improve sleep and examined the underlying mechanisms.
Methods:
Individuals reporting worry interfering with sleep were randomized to treatment as usual (TAU, n=40) or TAU + app-based MT (n=40). Treatment-related changes in worry-related sleep disturbances (WRSD), worry, non-reactivity, and anxiety were evaluated via self-report questionnaires at 1- and 2-months post treatment initiation. Fitbit devices were used to record total sleep time (TST) and estimate sleep efficiency (SE). At 2 months, TAU received access to app-based MT and both groups were reassessed at 4 months.
Results:
In a modified intent-to-treat analysis, WRSD scores decreased by 27% in TAU+MT (n=36) and 6% in TAU (n=35) at 2 months (median [IQR] change: 11 [4.3] v. 15 [5.0], P = .001). These WRSD reductions were mediated by decreased worry, particularly improved non-reactivity (P’s < .001). At 4 months, TAU reported a significant 29% reduction after beginning app-based MT at 2 months and TAU+MT maintained its gains. No significant between-group differences in average estimated TST or SE were found after 2 months of using the app.
Conclusions:
Few mindfulness-related apps have been evaluated for clinical efficacy and/or mechanism. Results from this study demonstrate a mechanistic link between MT and increased emotional non-reactivity, decreased worry, and reduction in reported sleep disturbances, suggesting that app-based MT may be a viable option to help individuals who report that worry interferes with their sleep.
Keywords: worry, sleep, anxiety, mindfulness, mHealth, digital therapeutics, TAU = Treatment-as-usual, NIH = National Institutes of Health
Introduction
Sleep disorders are a prevalent problem experienced by nearly 20% of Americans (1). Daytime drowsiness resulting from sleep disturbances (restlessness, difficulty falling asleep or staying asleep, sleeping excessively) can lead to dangerous outcomes like motor vehicle accidents and work injuries. Effects of chronic poor sleep and sleep disorders include exacerbation of psychiatric disorders like generalized anxiety, increased risk for cardiovascular diseases, and metabolic dysfunction (1). Links between disrupted sleep and anxiety appear to be bidirectional, i.e., persistent sleep disturbance predicts the development of anxiety disorders (2,3), and worry plays a major role in increasing bedtime arousal and perceived sleep disturbances (4). There is an approximately 70% comorbidity between generalized anxiety disorder (GAD) and insomnia, and 74% of primary care patients with GAD report sleep disturbances (5). Moreover anxiety-specific treatments are less effective when sleep disturbance is not addressed (3).
Anxiety disorders remain a major public health issue, with an average yearly prevalence of over 19% among U.S. adults. Approximately 1 in 3 U.S. adults experience an anxiety disorder at some point in their lives (6). Worry is a core defining component of anxiety disorders that has been specifically linked to sleep. High levels of worry predict more sleep problems such as increased nocturnal awakenings and lower sleep efficiency (7). Stress-related worry may exacerbate sleep reactivity, which can cause consistent sleep responses to stress over time regardless of stimuli, thus increasing the risk of developing insomnia (8). Sleep-specific worry predicts both sleep disturbance and sleep-related impairment even after controlling for demographics, chronotype, and overall health status (4).
Mindfulness-based Stress Reduction (MBSR), is a promising intervention to mitigate anxiety (9). Defined as nonjudgmental present-centered awareness, mindfulness is a form of mental training that increases interoceptive awareness to reduce reactivity to stress (10,11). Mental processes like worry and rumination contribute to maladaptive behaviors and additional stress surrounding poor sleep (4,10). Mindfulness training (MT) increases metacognitive awareness, which is defined as the ability to monitor and control one’s own cognition to disrupt cycles of worry (12). For example, in individuals with insomnia, chronic sleep disturbances can make one’s bed become a conditioned stimulus that can, in turn, trigger worry and lead to maladaptive coping behaviors like staying in bed despite not being able to sleep. Developing metacognitive awareness may help individuals to notice body sensations and hyperarousal without exacerbating the inability to sleep due to frustration (11). Improving one’s ability to maintain emotional non-attachment from unfavorable sleep outcomes through mindfulness practice addresses the metacognitive component of sleep-related arousal (11,13,14). However, there is still a lack of research showing clear effect of mindfulness-based interventions (MBIs) on improving sleep quality and duration, as well as the mechanism by which this change in sleep might occur. A small number of MBIs have shown varying degrees of efficacy for treating sleep disturbance, though few used randomized controlled trial designs (15).
Non-pharmaceutical interventions have demonstrated success in improving sleep outcomes, but cost and availability are limiting factors for scalability. For instance, cognitive behavioral treatment of insomnia (CBT-I) is an effective therapy that targets cognitions and maladaptive behaviors associated with coping with poor sleep (16), but lack of trained practitioners’ limits patient access (17). Digital therapeutics have the potential to increase access to mental health treatment by virtue of relatively low cost, ready access, and high treatment fidelity as has been demonstrated through successful digital delivery of CBT-I (find studies). A potential advantage of app-based MT for sleep disturbances related to anxiety and worry is that it can be disseminated widely and may mechanistically target worry, disrupting cognitive processes that lead to emotional distress and heightened physiological arousal that may ultimately drive insomnia. In previous research, app-based MT was shown to significantly reduce anxiety in individuals with GAD (18). Furthermore, the effects of MT on anxiety were mediated by decreased worry.
This study aimed to extend the current body of mindfulness research by evaluating the clinical effects and psychological mechanisms of app-based MT on worry and sleep. We hypothesized that: (1) an app-based MT program for anxiety would help individuals reduce worry with concomitant improvements in sleep; (2) reductions in worry would mediate the relationship between mindfulness training and improved sleep quality through a reduction of worry-related sleep disturbances (WRSD).
Methods
This was a parallel group randomized controlled trial. Individuals were recruited using geolocated Facebook advertisements, local paper advertisements, and a Brown University internal advertisement. Inclusion criteria were: (1) age ≥ 18 years; (2) score > 40 on the Penn State Worry Questionnaire (PSWQ) which is suggestive of GAD (19); (3) reported that they had trouble sleeping due to worry at least 40% of the time in the last week; (4) fluency in English; (5) owner of a smartphone; (6) willing to wear Fitbit to track sleep and to keep a sleep diary. We excluded potential participants who reported: (1) changing dose of psychotropic medication in the past 2 months; (2) as-needed use of a hypnotic sleep aid and/or benzodiazepines; (3) known sleep disorder; (4) psychotic disorder; (5) current diagnosis of post-traumatic stress disorder; (6) score ≥ 3 on Patient Health Questionnaire 2-item suggesting depression (20); (7) current shift work employment; (8) evening caffeine use; (9) body mass index > 35 kg/m2, increasing the risk for obesity-related obstructive sleep apnea (21); (10) current or planned (within 6 months) pregnancy; (11) new parents whose babies were not yet sleeping through the night. Individuals with a known sleep disorder and a potential diagnosis of depression were excluded because of the potential of these disorders to act as confounds, and also to focus on worry-related sleep disturbance as a primary outcome. During recruitment, individuals were not clinically evaluated for insomnia and thus are more representative of a broader non-clinical population who experience sleep disturbances. Eligible participants attended an in-person visit at Brown University where they underwent the informed consent process prior to enrollment. Participants were compensated up to $100 in Amazon gift cards for completing questionnaires. The study was approved by the Brown University Institutional Review Board and study design, statistical analyses, and outcomes were pre-registered at ClinicalTrials.gov in addition to the publication of a protocol paper (22). After enrolling, participants completed an online questionnaire via Qualtrics (23) which included demographic and self-report items.
Randomization and Blinding
Upon the completion of the baseline questionnaire, participants received an opaque, sealed envelope which contained their group assignment to treatment-as-usual (TAU) or TAU + MT (1:1 ratio). The envelopes were prepared, reviewed, and sealed by two individuals independent of the research team. The randomization schema was generated by an independent statistician in variable blocks of four and six. The individual responsible for randomization of participants and conducting the assessments was not involved in data analysis. The Principal Investigator, statistician, and individuals conducting data analysis were blinded until all analyses were complete.
Intervention
The MT was delivered through an app-based platform, Unwinding Anxiety, which includes 30 core modules and 8 “theme weeks” (that reinforce core concepts) of brief didactic and experience-based MT (videos and animations, ~10 min/day), app-initiated check-ins and other ecological momentary assessment tools (EMA), user-initiated guided meditations (5–15 min), and brief on-demand mindfulness exercises (30-sec) to help break worry cycles in the moment (22). The intervention content is based on a framework from previously developed in-person and app-based MT protocols that share a core underlying mechanism (reinforcement learning) that have yielded clinically meaningful outcomes (e.g., cessation of smoking and overeating) (24–27).
Orientation and Engagement
All participants were provided Fitbit Inspire devices and instructed how to sync their sleep data. Reminder texts were sent to participants prior to the wear period to improve device wearing and syncing adherence. All participants were also asked to complete paper sleep diaries (PSD). Both groups were provided with referrals to local mental health care providers and instructed to continue taking any prescribed medications. In addition, the TAU + MT group was provided access to the app, given an overview of its features, and asked to demonstrate an understanding of how to use it. Participants in the TAU+MT group received check-in texts on days 3, 7, 14, and 21 to troubleshoot technical issues.
Study Period
The MT intervention spanned a 2-month period. Participants completed the baseline assessment prior to randomization. Post treatment initiation, participants completed assessments at 1 month or upon completion of 14 app modules (whichever occurred first), 2 months or upon completion of 30 modules, and at 4 months. The 1-month follow-up assessment was administered remotely and included only an online questionnaire, while the 2-month and 4-month assessments were in person and followed the same procedure as at treatment initiation. After completing the 2-month assessment, the TAU group was provided with access to the app and the same procedures were followed as during the TAU + MT group’s orientation.
Outcome measures
The primary outcome was change in worry related sleep disturbance (WRSD), measured using five PROMIS sleep disturbance items that were preselected based on their direct assessment of worry interfering with sleep (28). Secondary outcomes included the PSWQ (29), non-reactivity subscale of the Five Facet Mindfulness Questionnaire (30), Generalized Anxiety Disorder-7 Questionnaire (31), and objective sleep measures from the Fitbit. Additional measures can be found in Appendix 1, Supplemental Digital Content 1.
Patient-Reported Outcomes Measurement Information System (PROMIS)
Five items from the PROMIS Sleep Disturbance item bank that directly evaluate the impact of worry on sleep, validated for independent use, were used to assess WRSD (28,32). Scores range from 5 to 25 and higher scores indicate more sleep disturbances and greater sleep-related impairment. Cronbach α = .72 at baseline and α = .82 at 2 months.
Penn State Worry Questionnaire (PSWQ)
The PSWQ is a 16-item questionnaire validated to assess worry (29). Scores range from 16 to 80 and higher total scores indicate increasing levels of worry. At baseline, Cronbach α = .86 and α = .86 at 2 months.
Non-reactivity subscale – Five Facet Mindfulness Questionnaire (FFMQ)
The non-reactivity subscale of the FFMQ was used to assess emotional reactivity. The non-reactivity subscale is validated for use independently and is composed of 7 questions from the 39-item FFMQ (30). Scores range from 7 to 35 with higher scores indicating greater non-reactivity, the ability to not react to or get caught up in thoughts or emotions. Cronbach α = .78 at baseline and α = .86 at 2 months.
Generalized Anxiety Disorder 7-item (GAD-7)
The GAD-7 is a validated 7-item questionnaire used clinically to screen for GAD and assess symptom severity (sensitivity of 89%, specificity of 82%) (33). Scores range from 0 to 21 and greater scores are indicative of increased anxiety. Total scores of 5, 10, and 15 serve as the cut-off points distinguishing mild, moderate, and severe anxiety, respectively (31). Cronbach α = .85 at baseline at α = .85 at 2 months.
Sleep measures
Participants completed a paper sleep diary (PSD) daily for 1 week immediately after treatment initiation and immediately prior to the 1-month and 2-month follow-up assessments. The PSDs asked participants to record their bedtime and wake time and to report their subjective experiences with sleep disturbances and sleep duration. During each diary week, participants wore a Fitbit Inspire to provide an objective measure of sleep patterns. This consumer device was selected rather than a research-grade actigraph because it is most representative of what individuals can access for personal use. While there is a trade-off in terms of accuracy and precision in sleep measurement, consumer devices provide the benefit of convenience and reduce participant burden (34). Fitbit devices were used to estimate total sleep time (TST) as well as total time awake and bed/wake times using the device’s sleep algorithms.
Sample Size
A prior study of MBSR for individuals with anxiety was used to determine the sample size for this study (9). Using the effect size of 0.52, we determined that n=64 would have 80% and 83% power (two-sided type I error) to detect a statistically significant difference in changes non-reactivity scores and WRSD respectively. To account for 20% attrition, we recruited 80 individuals with worry interfering with sleep.
Statistical analysis
Analyses were conducted in R (v.4.0.2) using an intent-to-treat approach (ITT). Fischer’s exact tests for categorical and Kruskal Wallis tests for continuous variables were used to determine any between-group differences in baseline demographics. Welch’s t-tests were used to examine between-group differences in worry, non-reactivity, and PSD data at 2 months. Outcome measures that violated the assumption of normality (WRSD, anxiety, and Fitbit data), as determined by Shapiro-Wilk tests, were assessed using Mann Whitney U tests. The same procedure was followed to explore within-group change for TAU after they received the intervention and multiple comparisons were adjusted for using the False Discovery Rate (35).
Sleep period time (SPT) was calculated from the PSDs by subtracting self-reported bedtime from wake time. Fitbit recorded bedtime, wake time, number of sleep minutes, and number of awake minutes were used to calculate estimated TST and sleep efficiency (SE). To improve reliability, average estimated TST was calculated for participants with ≥ 3 days recorded during the wear period. Analysis used an ITT and a complete case (CC) approach; participants who did not complete the baseline wear period were excluded from analysis. For the ITT analysis, Kruskal Wallis tests were used to evaluate associations between the missing and observed data. The missing data for both Fitbit and PSDs were found to be missing completely at random and multiple imputation was used to create five imputed datasets to correct for any bias introduced by missing data. Using the imputed datasets, t-tests were run to examine between-group differences for both sleep measures. The same tests were run for the CC analysis.
Generalized Estimating Equations (GEE) were used to analyze changes in primary and secondary outcomes from baseline to treatment completion. GEE’s are used for longitudinal data in order to account for within-subject covariance over time (36). Due to a significant between-group difference at baseline, gender was included as a covariate in all GEE models.
Mediation analyses were conducted using the Baron and Kenny (37) four-step approach in addition to causal mediation models (38). In order allow for a better interpretation of directionality and comparison across measures, the mediators and outcome change scores were standardized to have zero means and unit standard deviation. We fit two mediation models: (1) Model A explored the effects of MT on reduction in worry at 2 months mediated by changes in non-reactivity at 1 and 2 months; (2) Model B explored the effects of MT on reduction in WRSD (PROMIS) at 2 months mediated by changes in worry (PSWQ) at 1 and 2 months. In the causal mediation analysis, we focused on the natural direct and indirect effects and used the bootstrap method with 1,000 bootstrapped resamples to compute the standard errors and 95% confidence interval (39,40).
Risk of Bias
Using the Cochrane Collaboration’s tool for evaluating risk of bias, we assessed potential bias across seven domains: random sequence generation, allocation concealment, blinding of participants and researchers, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias (41).
Results
Participants
Eighty participants were recruited and consented to participate in the study from July 2019 to February 2020. 71 participants were included in the ITT analysis (see CONSORT diagram, Figure 1). Baseline demographic characteristics are reported in Table 1.
Figure 1.
Participant flow diagram
Table 1.
Baseline demographic characteristics
| TAU + MT (N = 36) |
TAU (N=35) |
P value | |
|---|---|---|---|
| Age – Mean (SD) | 42.2 (18.5) | 40.6 (16.8) | .65 |
| Sex | .04 | ||
| Female | 25 (69.4%) | 32 (91.4%) | |
| Male | 9 (25.0%) | 3 (8.6%) | |
| Prefer not to answer | 2 (5.6%) | 0 (0.0%) | |
| Race/Ethnicity | .88 | ||
| White | 21 (58.3%) | 23 (65.7%) | |
| Black/African American | 1 (2.8%) | 0 (0.0%) | |
| Hispanic, Latinx, Spanish | 2 (5.6%) | 3 (8.6%) | |
| Asian | 6 (16.7%) | 5 (14.3%) | |
| Two or more races | 6 (16.7%) | 4 (11.4%) | |
| Marital Status | .13 | ||
| Never married | 17 (47.2%) | 16 (45.7%) | |
| Married | 9 (25.0%) | 16 (45.7%) | |
| Separated/divorced | 7 (19.4%) | 2 (5.7%) | |
| Widowed | 3 (8.3%) | 1 (2.9%) | |
| Highest level of education completed | .47 | ||
| High school or GED | 2 (5.6%) | 0 (0.0%) | |
| Some college | 4 (11.1%) | 8 (22.9%) | |
| Associate’s degree | 2 (5.6%) | 3 (8.6%) | |
| Bachelor’s degree | 11 (30.6%) | 8 (22.9%) | |
| Graduate degree | 11 (30.6%) | 8 (22.9%) | |
| Work status | .39 | ||
| Employed full-time | 19 (52.8%) | 22 (62.9%) | |
| Employed part-time | 12 (33.3%) | 6 (17.1%) | |
| Homemaker | 1 (2.8%) | 0 (0.0%) | |
| Retired | 2 (5.6%) | 3 (8.6%) | |
| Unemployed | 2 (5.6%) | 4 (11.4%) | |
| Income | .18 | ||
| Less than $15,000 - $44,999 | 9 (25.0%) | 6 (17.1%) | |
| $45,000 - $99,999 | 12 (33.3%) | 9 (25.7%) | |
| $100,000 or more | 12 (33.3%) | 10 (28.6%) | |
| Prefer not to answer | 3 (8.3%) | 19 (28.6%) |
Safety
There were no adverse events related to the intervention reported in either group.
Engagement
The average number of modules that TAU + MT participants completed at 1, 2, and 4 months was 13 ± 3, 25 ± 7, and 26 ± 7 respectively. At 2 months, 53% (19/36) of TAU + MT participants completed the program (30 modules) which increased to 73% (22/36) at 4 months. In addition to module completion, this group completed meditation exercises and engaged with EMA style tools. At 1 month, the average number of EMA tool uses was 34.1 ± 11.6 and 13.5 ± 3.7 at 2 months. The average number of meditation exercises completed was 12.6 ± 3.5 at 1 month and 4.5 ± 1.5 at 2 months. Average completion per week is shown in Figure 2. After beginning treatment at 2 months, the average number of modules completed by TAU participants was 19 ± 11 and 34% (12/35) completed the program at 4 months.
Figure 2.
App-based MT Engagement over Time
Changes in Primary and Secondary Outcome Measures
Effects of intervention on worry-related sleep disturbances (PROMIS)
At 2 months, the TAU + MT group demonstrated a median reduction of 27% in WRSD compared to a reduction of 6% in the TAU group; the score dropped from 16.5 to 11.0 and 16.0 to 15.0 respectively (P = .001; see Figure 3A). See Table 2 for medians and interquartile ranges in addition to means and standard deviations. A GEE model found a significant time effect for TAU + MT at 1 (β = 3.2, SE = 0.78, P < .001) and 2 months (β = 4.7, SE = 0.79, P < .001) while no effect was found in TAU. At 2 months, there was a significant GROUP X TIME interaction (β = −3.0, SE = 0.95, P = .002; see Figure 4A). A Wilcoxon signed rank test found a median reduction of 5 (29%) in WRSD score in the TAU group after receiving the intervention in months 2 to 4 of the study (P < .001; Padj < .001), and the original treatment group maintained its improvement.
Figure 3.
(A) Change in PROMIS scores; (B) Change in PSWQ scores; (C) Change in non-reactivity scores; (D) Change in GAD-7 scores. PROMIS: Patient-Reported Outcomes Measurement Information System; PSWQ: Penn State Worry Questionnaire; GAD-7: Generalized Anxiety Disorder 7-item.
Color figure available online only at the Psychosomatic Medicine website.
Table 2.
Descriptive statistics for PROMIS, PSWQ, non-reactivity subscale, and GAD-7.
| TAU + MT (N = 36) | TAU (N = 35) | |||
|---|---|---|---|---|
| Mean [SD] | Median [IQR] | Mean [SD] | Median [IQR] | |
| PROMIS | ||||
| Baseline | 16.7 [4.3] | 16.5 [7.3] | 16.3 [3.4] | 16.0 [4.5] |
| 2-month | 12.0 [4.2] | 11.0 [4.3] | 15.3 [4.2] | 15.0 [5.0] |
| 4-month | 12.3 [4.9] | 11.0 [6.3] | 11.5 [3.6] | 11.0 [4.0] |
| Δ at 2-month (%) | −4.7 (−25%) | −4.0 (−27%) | −1.0 (−4%) | −1.0 (−6%) |
| Δ at 4-month (%) | −4.4 (−25%) | −4.0 (−25%) | −4.8 (−26%) | −5.0 (−29%) |
| PSWQ | ||||
| Baseline | 62.3 [9.4] | 62.0 [13.5] | 62.3 [8.3] | 64.0 [11.0] |
| 2-month | 54.6 [9.4] | 53.5 [13.3] | 61.7 [7.1] | 62.0 [10.5] |
| 4-month | 53.2 [12.0] | 54.0 [15.0] | 53.9 [9.7] | 53.0 [13.5] |
| Δ at 2-month (%) | −7.6 (−12%) | −9 (−13%) | −0.7 (−0.3%) | 0.0 (0%) |
| Δ at 4-month (%) | −9.1 (−15%) | −8.5 (−15%) | −8.4 (−13%) | −7 (−12%) |
| Non-reactivity | ||||
| Baseline | 17.8 [3.9] | 17.0 [5.0] | 17.5 [3.4] | 18.0 [3.5] |
| 2-month | 21.7 [3.5] | 21.5 [5.5] | 18.0 [3.3] | 17.0 [3.5] |
| 4-month | 21.8 [4.6] | 22.0 [6.0] | 21.5 [4.0] | 21.0 [5.0] |
| Δ at 2-month (%) | 4.0 (27%) | 4.5 (23%) | 0.4 (4%) | 0.0 (0%) |
| Δ at 4-month (%) | 4.1 (27%) | 4.5 (26%) | 3.9 (28%) | 5.0 (31%) |
| GAD-7 | ||||
| Baseline | 12.0 [5.6] | 12.0 [10.0] | 10.4 [4.0] | 10.0 [5.5] |
| 2-month | 8.0 [4.7] | 6.5 [4.5] | 11.1 [4.0] | 11.0 [6.0] |
| 4-month | 7.3 [5.3] | 5.5 [6.3] | 6.8 [3.4] | 7.0 [4.0] |
| Δ at 2-month (%) | −4.0 (−25%) | −2.5 (−37%) | 0.7 (18%) | 1.0 (17%) |
| Δ at 4-month (%) | −4.6 (−31%) | −4.5 (−32%) | −3.6 (−28%) | −3.0 (33%) |
PROMIS: Worry-related sleep disturbance items from Patient-Reported Outcomes Measurement Information System 29-item; PSWQ: Penn State Worry Questionnaire; GAD-7: Generalized Anxiety Disorder 7-item; Δ: change between baseline and stated time point. TAU received the intervention at 2 months.
Figure 4.
(A) Trajectory for change in PROMIS scores; (B) Trajectory for change in PSWQ scores; (C) Trajectory for change in non-reactivity scores; (D) Trajectory for change in GAD-7 scores. PROMIS: Patient-Reported Outcomes Measurement Information System; PSWQ: Penn State Worry Questionnaire; GAD-7: Generalized Anxiety Disorder 7-item.
Color figure available online only at the Psychosomatic Medicine website.
Effects of intervention on worry (PSWQ)
At 2 months, the TAU + MT group demonstrated an average reduction of 12% in PSWQ score (worry) relative to a reduction of 0.3% in the TAU group; with a change (Δ) from 62.3 to 54.6 and 62.3 to 61.7 respectively (P < .001; see Figure 3B). A GEE model demonstrated a significant time effect for TAU + MT at 1 (β = −3.6, SE = 1.03, P < .001) and 2 months (β = −7.6, SE = 1.33, P < .001) while no effects were found in TAU. There was also a significant GROUP X TIME effect at 2 months (β = −6.4, SE = 1.89, P < .001; see Figure 4B). After receiving the intervention at 2 months, the TAU group reported a similar 8.4 (13%) average reduction in PSWQ score at 4 months (P < .001; Padj < .001), while the original treatment group maintained its improvements.
Effects of intervention on non-reactivity (FFMQ)
At 2 months, TAU + MT reported an average increase of 27% in non-reactivity score relative to an increase of 4% in the TAU group (TAU + MT Δ - 17.8 to 21.7 vs. TAU Δ - 17.0 to 21.5, P < .001; see Figure 3C). At 2 months, a GEE model found a significant time effect for TAU + MT (β = 4.0, SE = 0.71, P < .001). No effects were found for TAU. At 2 months, there was also a significant GROUP X TIME interaction (β = 3.8, SE = 0.78, P < .001; see Figure 4C). After treatment initiation at 2 months, the TAU group reported a significant 3.9 (28%) average increase in non-reactivity score at 4 months (P < .001; Padj < .001), while the original treatment group maintained its improvements.
Effects of intervention on anxiety (GAD-7)
At 2 months, we found a significant between-group difference in anxiety: the TAU + MT group reported a median reduction of 37% in GAD-7 score compared to an increase of 17% in TAU. In TAU + MT, the score dropped from 12.0 to 6.5 while it increased from 10.0 to 11.0 in TAU (P = .002; see Figure 3D). A GEE model found a significant effect of time for TAU + MT (1 month: β = −1.9, SE = 0.90, P = .03; 2 months: β = −4.0, SE = 0.87, P < .001) while no effects were found in the TAU group. There was also a significant GROUP X TIME interaction at 2 months (β = −3.0, SE = 1.06, P = .004; see Figure 4D). After receiving the app at 2 months, the TAU group reported a significant median reduction of 3 (33%) in GAD-7 score at 4 months (P < .001; Padj < .001), while the original treatment group maintained its improvements.
Sleep
Sleep diaries
All participants completed the baseline PSD; 9.3% of the data were identified as missing for the other periods. After imputing missing data, no significant between-group differences were found in any sleep measures at 2 months post treatment initiation (P = .75). TAU + MT (n=24) reported an average increase in SPT from 452 to 458 minutes compared to TAU (n=30) who reported no change (464 minutes). 49 participants were included in CC analysis and at 2 months, no significant between-group differences in average reported SPT were found (P = .76; see Appendix 2, Supplemental Digital Content 2).
Fitbit
Seven participants were excluded from analysis due to not completing the baseline wear period and 30% of the data was missing for the other wear periods. Average and median estimated sleep durations were > 7 hours at all time points in both groups. After imputing missing data, no significant between-group differences were found in average estimated TST or SE at 2 months post treatment initiation (P = .52; P = .70). There was an average estimated increase in TST from 432 to 437 minutes in TAU + MT (n=33) compared to an increase from 427 to 437 minutes in TAU (n=31). SE remained the same in TAU + MT (96%) while it increased from 95% to 96% in TAU. 46 participants were included in the CC analysis which corroborated the findings from the analysis with imputed data for TST (P = .86) however a significant between-group difference was found in SE with 97% in the TAU + MT group and 96% in TAU (P = .04; see Appendix 3, Supplemental Digital Content 3).
Mediation analyses
Mediation by changes in non-reactivity on the effect of MT on changes in worry
Mediation analyses were conducted in 4 steps:
Step 1 – We assessed whether MT led to a reduction in worry at 2 months post treatment initiation, controlling for gender as a covariate. MT was significantly associated with reduction in worry at 2 months with a total effect of β = −0.86 (SE = 0.23, P < .001).
Step 2 – We assessed whether MT produced statistically significant changes in the mediator, non-reactivity, evaluated at 1-month post treatment initiation to meet temporal precedence. However, MT was not significantly associated with increases in non-reactivity at 1 month (β = 0.11, SE = 0.25, P = .67). Therefore, additional testing was not conducted.
Given that our past work has indicated a potential dose-response and/or sleeper effect, an increase in effect over time (18), we ran the same models as described above with the exception of evaluating the mediator, changes in non-reactivity, at 2 months. Results indicated that at 2 months, MT was significantly related to increases in non-reactivity (β = 0.87, SE = 0.23, P < .001), allowing us to proceed to Step 3 (Fig. 5A, a path).
Figure 5.
(A) Causal mediation model with standardized regression coefficients illustrates how the effect of mindfulness training on reduction in worry (PSWQ) is mediated by increases in non-reactivity at 2 months; (B) Causal mediation model with standardized regression coefficients illustrates how the effect of mindfulness training on reduction in WRSD (PROMIS) is mediated by reductions in worry (PSWQ) at 2 months. WRSD: Worry-related sleep disturbances; PROMIS: Patient-Reported Outcomes Measurement Information System; PSWQ: Penn State Worry Questionnaire.
Color figure available online only at the Psychosomatic Medicine website.
Step 3 – We evaluated if increases in non-reactivity at 2 months led to significant reductions in worry (Fig. 5A, b path). Non-reactivity was found to be significantly related to reductions in worry at 2 months (β = −0.49, SE = 0.11, P < .001).
Step 4 – We evaluated if the intervention was still significantly related to reduction in worry from baseline to 2 months after controlling for increases in non-reactivity. MT was no longer associated with reduction in worry (P = .05). The Sobel test was used to formally test the indirect effect (product of a and b path coefficients: .87 x −.49 = −0.43), which found that from baseline to 2 months, the relationship between MT and reduction in worry was significantly mediated by increases in non-reactivity (z’ = −2.99, P = .003).
Since the Baron and Kenny method of mediation analysis assumes that the mediator is exogeneous and can be arbitrarily manipulated, causal mediation analysis was used to evaluate the natural direct and indirect effects (37,38). Using this approach, the direct effect of MT on reduction in worry at 2 months was estimated to be β = −0.44 (95% CI: −.84 to −.02, P = .04; Fig. 5A, c’ path). The relationship between MT and reduction in worry was partially mediated by increases in non-reactivity at 2 months with an estimated indirect effect of β = −0.43 (95% CI: −.77 to −.18, P < .001).
Taken together, these findings indicate that MT reduced worry and increases in non-reactivity accounted for 49% of this effect, suggesting that one mechanism by which participants’ worry decreased was their ability to not get caught up in thoughts and emotions.
Mediation by changes in worry on the effect of MT on changes in WRSD (PROMIS)
The above approach was used to examine reduction in worry (PSWQ) as a mediator for the effect of MT on reduction in WRSD (PROMIS).
Step 1 – We assessed the total treatment effect of MT on reduction in WRSD at 2 months post treatment initiation, controlling for gender as a covariate. MT was significantly associated with a reduction in WRSD post treatment with a total effect of β = −0.76 (SE = 0.23, P = .002).
Step 2 – Next, we assessed whether MT produced statistically significant changes in the mediator, worry, evaluated at 1 month to meet temporal precedence. However, MT was not associated with a significant reduction in worry, so further analyses were not performed (P = .11).
We ran the same models as described above with reduction in worry at 2 months as the mediator. Assessing change at 2 months, MT was significantly related to reduction in worry (β = −0.86, SE = 0.23, P < .001; Fig. 5B, a path).
Step 3 – We evaluated whether the mediator at 2 months led to significant reductions in WRSD and found that reduction in worry at 2 months exhibited a significant direct effect on WRSD post-intervention (β = 0.46, SE = 0.11, P < .001; Fig. 5B, b path).
Step 4 – We evaluated if MT was still significantly related to reduction in WRSD at 2 months after controlling for reduction in worry. MT was no longer associated with reduction in worry (P = .13). The Sobel test was used to formally test the indirect effect (ab = −0.39), which found that at 2 months, the relationship between MT and reduction in WRSD was significantly mediated by a reduction in worry (z’ = −2.82, P = .005).
Using the causal mediation approach, we found the estimated direct effect of MT on reduction in WRSD was β = −0.36 (95% CI: −.86 to .12, P = .11). Reduction in worry at 2 months fully mediated the relationship between MT and reduction in WRSD with an estimated indirect effect of β = −0.39 (95% CI: −.69 to −.15, P < .001; see Fig. 5B, c’ path).
Taken together, these findings indicate that MT reduced WRSD and reductions in worry accounted for 53% of this effect, suggesting that one mechanism by which participants’ WRSD decreased was their ability to decrease hyperarousal through learned non-reactivity.
Risk of Bias
Using Cochrane Collaboration’s criteria for evaluating bias (41), we found a low risk of bias in all 7 categories.
Discussion
To our knowledge, this is the first RCT to evaluate the efficacy of app-based MT in improving anxiety-related sleep disturbance. We found that MT reduced worry and anxiety and decreased self-reported worries about sleep – an effect that was first observed between groups and then replicated within the study when the control group received the active intervention. From a mechanistic perspective, we found that increases in non-reactivity mediated reductions in worry and reductions in worry mediated reductions in WRSD. These findings suggest the potential utility of app-based MT to help individuals with anxiety and WRSD.
Given the results of this study, mechanistically-based and targeted MBIs may be a helpful addition to the other therapies that exist for sleep disturbances. While CBT-I is a prominent non-pharmacological treatment for insomnia, MT differs in mechanism: CBT-I generally focuses on regularizing sleep cycles through implementation of behavioral changes and identifying, challenging, and replacing negative perceptions that impair sleep (42), while MT may help an individual disrupt cycles of habitual worrying through practicing present-centered awareness of thoughts and emotions. MT is characterized by curiosity and nonjudgmental acceptance of anxious thoughts, as compared to modifying or replacing them. This study provides evidence for this mechanism, as app-based MT effectively increased non-reactivity with concomitant decreases in self-reported worry and worry disrupted sleep. Moreover, our participants did not report reductions in time in bed (sleep period time) as would be expected in patients employing CBT-I, supporting the notion that our MT intervention works to improve sleep through a different mechanism than CBT-I. We anticipate that treatments that target different aspects of sleep disturbance will continue to emerge as different insomnia phenotypes are specified. As has been demonstrated with digital CBT-I, delivering MT through a mobile app presents unique advantages to in-person delivered treatments: digital MT requires less time commitment, no transportation, has 100% treatment fidelity and can be less emotionally intensive (43). Digital health has promising utility in improving convenience, cultural and personal tailoring, cost-efficacy, and geographical reach of psychiatric treatment. Accessibility is an important area of future study: a relatively low-cost mobile application can be widely disseminated and may increase treatment access and utilization especially among underserved populations (44–46).
The immense burden of sleep-related worry on the overall U.S. population constitutes an unmet public health need. As a response, the NIH developed the PROMIS Sleep Disturbance item bank to facilitate research that targets self-reported sleep quality. The widespread experience of sleep disturbance was a reason to recruit participants from the broader population. Since our participant population was not clinically evaluated for insomnia, MT in this study is likely targeting worry to improve perceived inadequacies of sleep in a broader population experiencing sub-clinical sleep difficulties. We found that MT effectively reduced worry and decreased self-reported worries about sleep but showed no significant effect on quantitative measures of sleep duration. This may be due to a combination of Fitbit device limitations and the existence of a ceiling effect: sleep efficiency calculated from Fitbit data was 96% at baseline for both groups, mirroring the relatively high sleep durations that participants in both groups reported on their baseline sleep diaries.
The field of mindfulness research has received criticism for lack of reproducibility. Few MBIs surpass Stage 0 or I of the stage model established by the National Institutes of Health (NIH) for clinical science and less than 2% reach the stage of effectiveness or implementation/dissemination research (47). Our evaluation of an app-based MT program leverages a unique study design for replication of effect sizes within the same study. The effects of MT on percent change in WRSD, worry, and non-reactivity scores were reproduced by the TAU group while maintaining expected directionality. These results provide evidence of efficacy that progresses app-based MT for worry and sleep beyond Stage I studies and toward efficacy and implementation research.
Interestingly, time was not significantly associated with a change in non-reactivity scores for the TAU + MT group at 1 month but it was at 2 months. This suggests that, while there is a trend toward significance at 1 month, sustained MT is necessary to have an effect on non-reactivity. This is also reflected in our mediation models, as increases in non-reactivity scores at 1 month did not mediate the effect of MT on reduction in worry. Our previous study of MT for reducing anxiety in individuals with GAD demonstrated that the mediation effect of worry was greater at 2 months than at 1 month (18). The results of the current study provide additional support for the existence of a dose-response and/or sleeper effect.
Limitations
Using a commercial device to measure sleep comes with limitations in accuracy and precision. The Fitbit Inspire was selected because of its convenience and therefore higher expected adherence. Consumer devices are also most reflective of what our study demographic would use on a personal basis to track sleep duration and quality. A recent study found 125 million new wearable devices were shipped worldwide over a 3-month period in 2020 (48). Indeed, concerns about disturbed sleep stemming from people monitoring their sleep on wearable devices and believing device data over more validated and accurate measures (e.g. believing a wearable over an overnight sleep study conducted by a certified lab and interpreted by a licensed sleep medicine physician) has led to the description of orthosomnia, a sleep disorder based on worry about sleep (49). However, it has been shown that Fitbit devices exhibit better sensitivity than specificity, as they overestimate TST and underestimate restlessness (50,51). It is important to note that a cited limitation of the current body of research on MBIs is the narrow focus on novel and/or clinical populations (47). Because our study did not screen participants for clinically diagnosed insomnia, lower precision presents less of a limitation compared to if we had recruited a clinical population. Given that the PROMIS items assess worry at bedtime and difficulty falling asleep, future studies should assess sleep onset latency, number of awakenings, time awake after sleep onset, etc. It is still unclear as to how MT affects sleep quality among people seeking treatment for chronic insomnia. Future studies should aim to recruit participants with clinically diagnosed insomnia to assess whether severity of sleep disturbance has an impact on efficacy of MT. Finally, MBI studies benefit from comparison to TAU or an intervention with similar cognitive effort, though it is often difficult to find appropriate active controls. Future research should compare the clinical efficacy of digital CBT-I and digital MT as interventions that similarly target sleep-related worry and behaviors through distinct mechanisms.
To conclude, sleep disturbances and worry about sleep affect a broad range of individuals at a severity that may not always warrant clinical diagnosis. Making digital therapeutics like app-based MT available to the general population serves to improve preventive treatment-seeking behavior. Our study demonstrates clinical efficacy and full mediation of the effects of MT on perceived sleep disturbances by worry. The mediation models align with mindfulness theory and are suggestive of a treatment mechanism that can more precisely facilitate the process of improving sleep disturbance.
Supplementary Material
Acknowledgments
Source of Funding
This study was supported by the National Institutes of Health (NIH) Science of Behavior Change Common Fund Program through an award administered by the National Institute on Aging (R21DA046957).
Footnotes
Conflicts of Interest
Dr. Brewer is a paid advisor to Sharecare, the company that owns the mindfulness app used in this study. This financial interest has been disclosed to and is being managed by Brown University, in accordance with its Conflict of Interest and Conflict of Commitment policies, including: (1) being restricted from recruitment, (2) being blinded to study group until after analysis, (3) not having access to data or performing analyses. Ms. Roy is also a paid advisor to Sharecare. This financial interest has been disclosed to and is being managed by Brown University, in accordance with its Conflict of Interest and Conflict of Commitment policies. Dr. Sharkey has received royalties from UptoDate. All other authors report no biomedical financial interests or potential conflicts of interest.
Trial Registration: Targeting Worry to Improve Sleep (NCT03684057). URL - https://clinicaltrials.gov/ct2/show/NCT03684057?term=judson+brewer&draw=2&rank=8
References
- 1.Garbarino S, Lanteri P, Durando P, Magnavita N, Sannita WG. Co-morbidity, mortality, quality of life and the healthcare/welfare/social costs of disordered sleep: A rapid review. Vol. 13, International Journal of Environmental Research and Public Health. MDPI AG; 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Gregory AM, Caspi A, Eley TC, Moffitt TE, O’Connor TG, Poulton R. Prospective longitudinal associations between persistent sleep problems in childhood and anxiety and depression disorders in adulthood. Journal of Abnormal Child Psychology 2005;33:157–63. [DOI] [PubMed] [Google Scholar]
- 3.Cox RC, Olatunji BO. A systematic review of sleep disturbance in anxiety and related disorders. Vol. 37, Journal of Anxiety Disorders. Elsevier Ltd; 2016. p. 104–29. [DOI] [PubMed] [Google Scholar]
- 4.Tutek J, Gunn HE, Lichstein KL. Worry and Rumination Have Distinct Associations with Nighttime versus Daytime Sleep Symptomology. Behavioral Sleep Medicine 2021;19:192–207. [DOI] [PubMed] [Google Scholar]
- 5.Dolsen MR, Asarnow LD, Harvey AG. Insomnia as a Transdiagnostic Process in Psychiatric Disorders. Vol. 16, Current Psychiatry Reports. Current Medicine Group LLC 1; 2014. p. 471. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Merikangas KR, He JP, Burstein M, Swanson SA, Avenevoli S, Cui L, Benjet C, Georgiades K, Swendsen J. Lifetime prevalence of mental disorders in U.S. adolescents: Results from the national comorbidity survey replication-adolescent supplement (NCS-A). Journal of the American Academy of Child and Adolescent Psychiatry 2010;49:980–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Weise S, Ong J, Tesler NA, Kim S, Roth WT. Worried sleep: 24-h monitoring in high and low worriers. Biological Psychology 2013;94:61–70. [DOI] [PubMed] [Google Scholar]
- 8.Kalmbach DA, Anderson JR, Drake CL. The impact of stress on sleep: Pathogenic sleep reactivity as a vulnerability to insomnia and circadian disorders. Vol. 27, Journal of Sleep Research. Blackwell Publishing Ltd; 2018. p. e12710. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hoge E, Bui E, Marques L, Metcalf C, Morris L, Robinaugh D, Worthington J, Pollack M, Simon N. Randomized Controlled Trial of Mindfulness Meditation for Generalized Anxiety Disorder: Effects on Anxiety and stress Reactivity. J Clin Psychiatry [Internet]. 2013;74:786–92. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3772979/ [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bishop SR, Lau M, Shapiro S, Carlson L, Anderson ND, Carmody J, Segal ZV., Abbey S, Speca M, Velting D, Devins G. Mindfulness: A Proposed Operational Definition. Clinical Psychology: Science and Practice 2006;11:230–41. [Google Scholar]
- 11.Ong JC, Ulmer CS, Manber R, Ong JC. Improving sleep with mindfulness and acceptance: A metacognitive model of insomnia. Behaviour Research and Therapy 2012;50:651–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Gu J, Strauss C, Bond R, Cavanagh K. How do mindfulness-based cognitive therapy and mindfulness-based stress reduction improve mental health and wellbeing? A systematic review and meta-analysis of mediation studies. Clinical Psychology Review [Internet]. 2015;37:1–12. Available from: 10.1016/j.cpr.2015.01.006 [DOI] [PubMed] [Google Scholar]
- 13.Huberty JL, Green J, Puzia ME, Larkey L, Laird B, Vranceanu A-M, Vlisides-Henry R, Irwin MR. Testing a mindfulness meditation mobile app for the treatment of sleep-related symptoms in adults with sleep disturbance: A randomized controlled trial. Plos One [Internet]. 2021;16:e0244717. Available from: 10.1371/journal.pone.0244717 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Shapiro SL, Bootzin RR, Figueredo AJ, Lopez AM, Schwartz GE. The efficacy of mindfulness-based stress reduction in the treatment of sleep disturbance in women with breast cancer. An exploratory study. Journal of Psychosomatic Research 2003;54:85–91. [DOI] [PubMed] [Google Scholar]
- 15.Rash JA, Kavanagh VAJ, Garland SN. A Meta-Analysis of Mindfulness-Based Therapies for Insomnia and Sleep Disturbance: Moving Towards Processes of Change. Vol. 14, Sleep Medicine Clinics. W.B. Saunders; 2019. p. 209–33. [DOI] [PubMed] [Google Scholar]
- 16.Williams J, Roth A, Vatthauer K, McCrae CS. Cognitive behavioral treatment of Insomnia. Chest 2013;143:554–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Manber R, Simpson NS, Bootzin RR. A step towards stepped care: Delivery of CBT-I with reduced clinician time. Sleep Medicine Reviews 2015;19:3–5. [DOI] [PubMed] [Google Scholar]
- 18.Roy A, Hoge EA, Abrante P, Druker S, Liu T, Brewer JA. Clinical efficacy and psychological mechanisms of an app-based digital therapeutic for generalized anxiety disorder: a randomized controlled trial (Preprint). Journal of Medical Internet Research [Internet]. 2021; Available from: http://preprints.jmir.org/preprint/26987/accepted [DOI] [PMC free article] [PubMed]
- 19.Behar E, Alcaine O, Zuellig AR, Borkovec TD. Screening for generalized anxiety disorder using the Penn State Worry Questionnaire: A receiver operating characteristic analysis. Journal of Behavior Therapy and Experimental Psychiatry 2003;34:25–43. [DOI] [PubMed] [Google Scholar]
- 20.Löwe B, Kroenke K, Gräfe K. Detecting and monitoring depression with a two-item questionnaire (PHQ-2). Journal of Psychosomatic Research 2005;58:163–71. [DOI] [PubMed] [Google Scholar]
- 21.Romero-Corral A, Caples SM, Lopez-Jimenez F, Somers VK. Interactions between obesity and obstructive sleep apnea: Implications for treatment. Chest [Internet]. 2010;137:711–19. Available from: 10.1378/chest.09-0360 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Brewer JA, Roy A, Deluty A, Liu T, Hoge EA. Can mindfulness mechanistically target worry to improve sleep disturbances? Theory and study protocol for app-based anxiety program. Health Psychology 2020;39:776–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Qualtrics 2021. Available from: https://www.qualtrics.com/
- 24.Brewer JA, Mallik S, Babuscio TA, Nich C, Johnson HE, Deleone CM, Minnix-Cotton CA, Byrne SA, Kober H, Weinstein AJ, Carroll KM, Rounsaville BJ. Mindfulness training for smoking cessation: Results from a randomized controlled trial. Drug and Alcohol Dependence [Internet]. 2011;119:72–80. Available from: 10.1016/j.drugalcdep.2011.05.027 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Brewer JA, Ruf A, Beccia AL, Essien GI, Finn LM, van Lutterveld R, Mason AE. Can mindfulness address maladaptive eating behaviors? Why traditional diet plans fail and how new mechanistic insights may lead to novel interventions. Frontiers in Psychology 2018;9:1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Garrison KA, Pal P, O’Malley SS, Pittman BP, Gueorguieva R, Rojiani R, Scheinost D, Dallery J, Brewer JA. Craving to Quit: A Randomized Controlled Trial of Smartphone App-Based Mindfulness Training for Smoking Cessation. Nicotine & tobacco research : official journal of the Society for Research on Nicotine and Tobacco 2020;22:324–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Mason AE, Jhaveri K, Cohn M, Brewer JA. Testing a mobile mindful eating intervention targeting craving-related eating: feasibility and proof of concept. Journal of Behavioral Medicine 2018;41:160–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Yu L, Buysse DJ, Germain A, Moul DE, Stover A, Dodds NE, Johnston KL, Pilkonis PA. Development of Short Forms From the PROMISTM Sleep Disturbance and Sleep-Related Impairment Item Banks. Behavioral Sleep Medicine 2011;10:6–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Meyer TJ, Miller ML, Metzger RL, Borkovec TD. Development and validation of the penn state worry questionnaire. Behaviour Research and Therapy 1990;28:487–95. [DOI] [PubMed] [Google Scholar]
- 30.Baer RA, Smith GT, Lykins E, Button D, Krietemeyer J, Sauer S, Walsh E, Duggan D, Williams JMG. Construct validity of the five facet mindfulness questionnaire in meditating and nonmeditating samples. Assessment 2008;15:329–42. [DOI] [PubMed] [Google Scholar]
- 31.Spitzer RL, Kroenke K, Williams JBW, Löwe B. A brief measure for assessing generalized anxiety disorder: The GAD-7. Archives of Internal Medicine 2006;166:1092–97. [DOI] [PubMed] [Google Scholar]
- 32.Buysse DJ, Yu L, Moul DE, Germain A, Stover A, Dodds NE, Johnston KL, Shablesky-Cade MA, Pilkonis PA. Development and validation of patient-reported outcome measures for sleep disturbance and sleep-related impairments. Sleep 2010;33:781–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Kroenke K, Spitzer RL, Williams JBW, Monahan PO, Löwe B. Anxiety Disorders in Primary Care: Prevalence, Impairment, Comorbidity, and Detection. Annals of Internal Medicine 2007;146:317. [DOI] [PubMed] [Google Scholar]
- 34.Izmailova ES, Wagner JA, Perakslis ED. Wearable Devices in Clinical Trials: Hype and Hypothesis. Clinical Pharmacology and Therapeutics 2018;104:42–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodological) [Internet]. 1995;57:289–300. Available from: 10.1111/j.2517-6161.1995.tb02031.x [DOI] [Google Scholar]
- 36.Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13–22. [Google Scholar]
- 37.Baron RM, Kenny DA. The Moderator-Mediator Variable Distinction in Social Psychological Research. Conceptual, Strategic, and Statistical Considerations. Journal of Personality and Social Psychology 1986;51:1173–82. [DOI] [PubMed] [Google Scholar]
- 38.Van DerWeele TJ, Vansteelandt S. Mediation analysis with multiple mediators. Epidemiologic Methods 2014;2:95–115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Pearl J. Interpretation and identification of causal mediation. Psychological Methods 2014;19:459–81. [DOI] [PubMed] [Google Scholar]
- 40.Imai K, Keele L, Tingley D, Yamamoto T. Causal Mediation Analysis Using R 2010;129–54.
- 41.Higgins JPT, Altman DG, Gøtzsche PC, Jüni P, Moher D, Oxman AD, Savović J, Schulz KF, Weeks L, Sterne JAC. The Cochrane Collaboration’s tool for assessing risk of bias in randomised trials. BMJ (Online) 2011;343:1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Koffel E, Bramoweth AD, Ulmer CS. Increasing access to and utilization of cognitive behavioral therapy for insomnia (CBT-I): a narrative review. Journal of General Internal Medicine 2018;33:955–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Ong JC, Shapiro SL, Manber R. Combining Mindfulness Meditation with Cognitive-Behavior Therapy for Insomnia: A Treatment-Development Study. Behavior Therapy 2008;39:171–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Bender DS, Skodol AE, Dyck IR, Markowitz JC, Shea MT, Yen S, Sanislow CA, Pinto A, Zanarini MC, McGlashan TH, Gunderson JG, Daversa MT, Grilo CM. Ethnicity and Mental Health Treatment Utilization by Patients With Personality Disorders. Journal of Consulting and Clinical Psychology 2007;75:992–99. [DOI] [PubMed] [Google Scholar]
- 45.Goldstein RB, Olfson M, Wickramaratne PJ, Wolk SI. Use of Outpatient Mental Health Services by Depressed and Anxious Children as They Grow Up. Psychiatric Services 2006;57:966–75. [DOI] [PubMed] [Google Scholar]
- 46.Grella CE, Greenwell L, Mays VM, Cochran SD. Influence of gender, sexual orientation, and need on treatment utilization for substance use and mental disorders: Findings from the California Quality of Life Survey. BMC Psychiatry 2009;9:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Dimidjian S, Segal ZV. Prospects for a clinical science of mindfulness-based intervention. American Psychologist 2015;70:593–620. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.IDC. Shipments of Wearable Devices Leap to 125 Million Units, Up 35.1% in the Third Quarter, According to IDC 2020. Available from: https://www.idc.com/getdoc.jsp?containerId=prUS47067820
- 49.Baron KG, Abbott S, Jao N, Manalo N, Mullen R. Orthosomnia: Are some patients taking the quantified self too far? Journal of Clinical Sleep Medicine 2017;13:351–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Moreno-Pino F, Porras-Segovia A, López-Esteban P, Artés A, Baca-García E. Validation of fitbit charge 2 and fitbit alta hr against polysomnography for assessing sleep in adults with obstructive sleep apnea. Journal of Clinical Sleep Medicine 2019;15:1645–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Haghayegh S, Khoshnevis S, Smolensky MH, Diller KR, Castriotta RJ. Accuracy of wristband fitbit models in assessing sleep: Systematic review and meta-analysis. Journal of Medical Internet Research 2019;21:1–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.