-
PDF
- Split View
-
Views
-
Cite
Cite
Julien Coelho, Florian Pecune, Yannick Levavasseur, Etienne De Sevin, Emmanuel D’incau, Patricia Sagaspe, Maria-Montserrat Sanchez-Ortuño, Jean-Arthur Micoulaud-Franchi, Pierre Philip, From improved sleep regularity to reduced sleep complaints and mental health conditions: a population-based interventional study using a smartphone-based virtual agent, Sleep, Volume 46, Issue 9, September 2023, zsad165, https://doi.org/10.1093/sleep/zsad165
- Share Icon Share
Abstract
To explore the effect of sleep regularity on sleep complaints and mental health conditions (i.e. insomnia, fatigue, anxiety, and depressive symptoms) in a population-based interventional study using a smartphone-based virtual agent.
A populational cohort based on the Kanopée application, which provided interactions with a virtual companion to collect data on sleep and make personalized recommendations to improve sleep over 17 days. A pre-intervention sleep diary and interview were used for cross-sectional analysis (n = 2142), and a post-intervention sleep diary and interview were used for longitudinal analysis (n = 732). The intra-individual mean (IIM) and standard deviation (ISD) of total sleep time (TST) were calculated to measure sleep quantity and sleep regularity.
The mean age at baseline was 49 years, 65% were female, 72% reported insomnia, 58% fatigue, 36% anxiety, and 17% depressive symptoms. Before the intervention, irregular and short sleep was associated with a higher likelihood of insomnia (Relative risk [RR] = 1.26 [1.21–1.30] for irregular TST and RR = 1.19 [1.15–1.23] for short TST), fatigue, anxiety, and depressive symptoms. After the intervention, the IIM of the TST increased while the ISD of the TST and sleep complaints and mental health conditions decreased. More regular TST was associated with reduced insomnia and depressive symptoms (RR = 1.33 [1.10–1.52] and RR = 1.55 [1.13–1.98], respectively).
Our results reveal a longitudinal association between sleep regularity and sleep complaints and mental health conditions. Policymakers, health professionals, and the general population should be aware that, beyond its positive effect on sleep health, regular sleep could promote mental health.
Sleep regularity is critical to treat insomnia and reducing fatigue, but its association with mental well-being remains unclear. In a large sample of users of the KANOPEE application, which is a smartphone-based virtual companion that provides a screening and a behavioral personalized intervention for sleep complaints, we demonstrated for the first time the longitudinal associations between sleep regularity improvement (i.e. sleep time consistency) and reduction of depressive symptoms. Though causality cannot be established in this setting, these results indicated the need to warn the population and health professionals about the benefit of regular sleep, not only regarding sleep health but also mental health. Sleep regularity could therefore constitute a prime target to improve the sleep health of the adult population.
Introduction
Sleep health is a positive, multidimensional construct that focuses on how well an individual sleeps rather than on the absence of sleep disorders or sleep disturbances [1]. In his seminal paper published in 2014, Buysse proposed five dimensions (Satisfaction, Alertness, Timing, Alertness, and Duration). This novel definition of sleep health allowed new research avenues (e.g. linking specific dimensions of sleep health with specific health outcomes) and ultimately the development of new interventions targeting poor sleep health profiles as well as the development of public health initiatives emphasizing the improvement of the different dimensions of sleep health. Substantial evidence demonstrates that these five dimensions of sleep health are related to health outcomes [1].
More recently, a sixth dimension was added to the sleep health construct: sleep regularity [2]. Day-to-day variability in sleep behaviors has emerged in recent years as an important factor for health and safety [3]. The daily optimal times for sleeping are frequently altered for multiple reasons, such as enhanced workload, shift work, travel, or career stress [4], leading to irregularity of routine(s) with disturbed sleep episodes [5]. This has repercussions on the quality of wakening [6, 7]. Sleep regularity is thought to treat insomnia and reduce fatigue in clinical populations by maximizing the synchrony between physiological sleep drive and circadian rhythms [8, 9]. Thus, sleep regularity was included as part of the sleep hygiene recommendations for the general population, as stated in a systematic review published in 2015 [10]. Furthermore, the American Academy of Sleep Medicine includes this concept in their sleep recommendations for children [11]. However, its association with sleep complaints has been poorly studied in the nonclinical population [12–14]. Results are inconsistent, probably due to discrepancies in the methodology and metrics used to measure regularity [15]. Additionally, study designs have mostly been cross-sectional and have not systematically considered other sleep dimensions such as sleep duration or timing.
Beyond its effects on sleep quality, sleep regularity is further suspected of having an impact on mental health. Indeed, circadian clocks help regulate many endogenous processes, including inflammation, immunity, and hormone secretion [16–18], and evidence suggests that abnormalities in circadian rhythms may be causal or pathophysiologically significant in psychiatric illness [16]. This relationship might be explained by the alteration of the rhythmicity of the hypothalamic-pituitary (HPA) axis, leading to an increased and dysregulated cortisol secretion [17]. However, few studies have explored the relationship between mental health and sleep regularity [18–23]. Most of them showed a positive association between irregularity and mental health symptoms [18–22]. Overall, most studies have been cross-sectional and have used different metrics, from objective actigraphy measurements to single-item self-assessments of regularity [16, 24]. As a result, findings have been inconsistent. Since they are strongly associated with sleep regularity [16, 25] and possibly mental health [24, 26], the timing and duration of sleep are suspected confounders but were considered in only two studies with inconsistent results. The first one was set in a sample of 771 adults and found a significant cross-sectional association between intra-individual variability of total sleep time (TST) and depression [18]. The second was conducted in a sample of 1048 adult participants and found no association between sleep regularity and anxiety and depressive symptoms in a model that neglected the mediating effect of sleep complaints between regularity and mental health [23, 27]. Another limitation of this study was the use of a self-reported measure of sleep regularity instead of a valid and reliable regularity metrics based on an objective or subjective sleep diary, such as intra-individual variabilities in the duration of mid-sleep and sleep [15].
In 2020, the coronavirus disease 2019 (COVID-19) outbreak massively disturbed sleep among the European population [28], and completely overloaded medical resources, as was the case in France [29]. Consequently, sleep problems have been treated using a large-scale autonomous model. We developed a research program using virtual agents to target and treat sleep complaints and social stress among the general population. To date, 30 000 individuals have downloaded a free app (KANOPEE) from the Google or Apple Store to benefit from the sleep hygiene and stimulus control recommendations. We demonstrated that these autonomous behavioral interventions significantly improved sleep schedules and reduced the overall severity of insomnia [30–32]. Users were also monitored for their anxiety and depressive symptoms, since they represent prevalent and harmful conditions [33] that are highly associated with sleep behaviors and complaints [34, 35].
Based on the ability to improve sleep in the general population, we explored the cross-sectional and longitudinal relationships between a valid sleep regularity measure (intra-individual variability) and sleep complaints and mental health conditions (i.e. insomnia, fatigue, anxiety, and depressive symptoms) while taking into account sleep timing and duration. The questions we sought to address in this study were the following: (1) Is irregular sleep associated with a higher frequency of mental health conditions?, (2) How are sleep schedules, sleep complaints, and mental health conditions modified during the intervention?, and (3) Is improved sleep regularity during follow-up associated with impaired mental health conditions?
Methods
Study design, setting, and participants
This was an ancillary study of the population cohort engaged in the KANOPEE application, which provides interactions with a virtual companion to collect data on sleep and make personalized behavioral recommendations to improve sleep over 17 days. Communication promoting the use of this application targeted social media (Instagram, TikTok, Facebook, and Twitter), national and regional newspapers and TV, and University and hospital mailing lists to increase the download rate and reach the most diverse population. The application was not advertised, and the participants were not compensated. Eligibility criteria were downloading KANOPEE, completing the baseline interview, and being of legal age (age ≥ 18 years). Recruitment was performed on the Google Play Store or the Apple Store by downloading the application. Informed consent was obtained directly on the app and was required from all participants who downloaded the application before any data were collected. Approval of the scientific committees of the University of Bordeaux was obtained as well as the agreement concerning the General Data Protection Regulations by the French authorities (CNIL). The reporting of the study was handled according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Supplementary Table S1) [36].
This was a single-arm intervention using different steps (Figure 1) and previously described [30–32]. First, on day 1, participants interacted with a virtual companion who provided screening for baseline sleep complaints and assessments of mental health conditions. The assessments were completed by 13 343 participants between December 1, 2020 and July 11, 2022. Second, participants were asked to complete a pre-intervention sleep diary between day 1 and day 7 to evaluate their sleep schedules. A minimum of 7 days was completed in accordance with recommendations on subjective sleep measurements of sleep duration and regularity [15, 37]. Third, on day 7, participants met the virtual companion again who conducted a pre-intervention interview (i.e. pre-intervention assessment of sleep complaints and mental health conditions). In all, 2142 participants completed the interview.

Flowchart of inclusion of participants from the KANOPEE application.
The virtual companion provides personalized behavioral recommendations according to sleep diary data and sleep complaint answers. The intervention was delivered on a single 20-minute interview and focused on reinforcing rise time regularity, adjusting the estimated sleep duration with time spent in bed, and developing appropriate behaviors to reinforce circadian rhythms. For example, participants reporting sleep maintenance insomnia symptoms during interviews or prolonged awakenings during the night in the sleep diary were encouraged to apply the stimulus control guideline. The recommendations were simple sentences (e.g. “If I’m awake for more than 15 minutes at night, I get out of bed and only return to bed when I’m sleepy”). The precise set of instructions are available in Supplementary Table S2. After the intervention was delivered, a new tab “personalized recommendations” appeared at the bottom of the screen so that users could access them at any time. Fourth, participants were encouraged to follow these recommendations for 10 days while completing the sleep diary, and they were informed that they would have a third, finial, interview 10 days later.
Fifth, on day 17, participants were asked to interact with the virtual companion for a post-intervention interview and to complete the third and last sleep complaints and mental health conditions assessment (n = 732).
Measures
Participants were asked to complete sleep diaries twice (between baseline [day 1] and the pre-intervention interview [day 7] and between day 7 and the post-intervention interview [day 17]). Data from the two sleep diaries were used to calculate the sleep schedules, including the intra-individual means (IIM) and the intra-individual standard deviations (ISD) of TST (defined as the time elapsed between when first falling asleep and the last awakening minus the time reported awake) and the IIM and the ISD of the circadian midpoint (CM; defined as the midpoint between first falling asleep and the last waking). For both TST and CM, IIM were calculated by summing daily values and dividing by the total number of days for each participant. The IIM of the TST was used as a proxy of sleep quantity while the IIM of the CM was used as a proxy of sleep timing. Likewise, intra-ISD were calculated across daily values using the R package Varian. They were used as proxies of sleep regularity. These overall metrics met the characteristics of this study better than consecutive metrics (large sample, presence of skipped diary days, total length ≥ 7) [15]. Changes in sleep schedules were defined as the absolute difference between the IIM/ISD of the pre-intervention sleep diary and the IIM/ISD of the post-intervention sleep diary. The IIM of the TST was categorized according to the Centers for Disease Control and Prevention (CDC) recommendations [38, 39] (<7, 7–9, and ≥9 hours). Due to the absence of reference thresholds, other sleep schedules and changes in sleep schedules were categorized using quintiles. The extreme quintiles (Q1 and Q5) were used to define extreme sleep schedules (e.g. TST ISD ≥ 128 min defined as irregular TST) and extreme changes in the sleep schedule (e.g. TST IIM decrease ≥ 22 minutes defines shorter TST). The Q2–Q5 quintiles were used to define an average sleep schedule and an unchanged sleep schedule. Regarding CM, extreme quintiles were close to those in previous studies that defined early CM as before 02:00 am and late CM as after 04:00 am [40, 41].
Sleep complaints (i.e. insomnia and fatigue) and mental health conditions (i.e. anxiety and depressive symptoms) were assessed at baseline (day 1), at the pre-intervention interview (day 7), and the post-intervention interview (day 17). Insomnia symptoms were assessed with the Insomnia Severity Index, a 7-item scale with each item rated on a 5-point Likert scale ranging from 0 to 4 [42]. The Cronbach’s α was satisfactory in the original validation study (≥0.76) and our sample (≥0.80). Participants with a total score ≥10 were considered to have clinically significant insomnia. Fatigue was assessed with the Fatigue Severity Scale, a 9-item scale with each item rated on a 7-point Likert scale scored from 1 to 7 [43]. The Cronbach’s α was good in a transcultural validation study (0.93) [44] and our sample (≥0.91). Participants with a mean score ≥4 were considered to have significant fatigue problems. Anxiety and depressive symptoms were measured using the Patient Health Questionnaire (PHQ-4), a reliable and valid short 4-item tool rated on a 3-point Likert scale that explores the past 2 weeks [45]. The first two items (feeling nervous, anxious, or on edge and not being able to stop or control worrying) were summed to obtain the anxiety score, and the last two (little interest or pleasure in doing things and feeling down, depressed, or hopeless) were summed to obtain the depressive score. The Cronbach’s α values were 0.82 for the anxiety subscale and 0.81 for the depressive subscale in the original study and were satisfactory in our sample (≥0.80 for anxiety and ≥0.78 for depressive symptoms). A score ≥ 3 indicated significant anxiety or depressive symptoms.
Participants were also asked about their sociodemographic characteristics at baseline (day 1), including age (continuous), sex (male, female), educational level (middle school, high school, Bachelor’s degree, Master’s degree), and job categories categorized into higher intellectual profession (e.g. engineers, doctors), intermediate profession (e.g. technicians, teachers), employee and worker (e.g. administrative agent, maintenance), and no professional activity. The presence or absence of government mobility restrictions due to the Covid-19 health crisis (lockdown or curfew) has also been reported. In the period between December 1, 2020, and May 3, 2021, there were restrictions (curfew followed by lockdown) while in the period after May 3, 2021, all restrictions had been lifted.
Bias
To evaluate the potential of a selection bias in our study, included and excluded samples were compared for both cross-sectional (n = 2142 vs. n = 11,201, Supplementary Table S3) and longitudinal analyses (n = 742 vs. n = 12,611, Supplementary Table S4) for variables of interest. Overall, excluded samples were younger than included samples (p < 0.001) and reported more severe mental health conditions (p < 0.001). This selection is likely to modify the associations in the direction of an underestimation by the loss of people who have the most disrupted sleep and mental health.
Statistical analysis
Descriptive statistics were calculated as frequencies (%) for categorical variables, whereas means and standard deviations were computed for continuous variables. Correlations between the IIM of the TST, the ISD of the TST, the IIM of the CM, and the ISD of the CM were calculated using Spearman’s coefficient analysis. Multivariate logistic regression analyses were performed for the cross-sectional associations between the sleep schedule from the pre-intervention sleep diary (explicative variables) and sleep complaints and mental health conditions from the pre-intervention interview (day 7) (dependent variables). Relative risk and their 95% confidence intervals were estimated from adjusted odds ratios according to the method proposed by Zhang and Yu [46, 47]. We selected confounding factors that affected sleep and mental health (age, sex, educational level, and mobility restrictions) according to the review of the literature [48–50]. The IIM and the ISD of the TST and CM were adjusted on each other. Sleep diary duration and the delay between evaluations were also included. The linearity hypothesis of quantitative variables was verified using fractional polynomials. All the sleep schedules taken as continuous variables showed a non-linear association with sleep complaints and mental health conditions. They were further analyzed as quintiles. Analysis of the residuals was used to confirm the suitability of the model. Interactions between the sleep schedules were tested for each outcome. None of the interactions was significant at the 0.20 threshold.
To explore the changes in sleep schedules, sleep complaints, and mental health conditions in the follow-up, Cohen’s d-effect sizes were computed between the two sleep diaries for the sleep schedules, and between the pre- and post-intervention interview for sleep complaints and mental health conditions.
Further analyses were performed with a longitudinal setting, exploring how changes in sleep complaints and mental health conditions were related to changes in the sleep schedule following the virtual companion intervention. The mean score and its bootstrap CI were calculated and stratified according to the change in the sleep schedule for each outcome and at each time. Multivariate logistic regression analyses were performed for the longitudinal associations of improvement in the sleep schedule between the pre-intervention and the post-intervention sleep diary and reducing sleep complaints and mental health conditions between the pre-intervention interview (day 7) and the post-intervention interview (day 17). An additional adjustment was made on sleep complaints and mental health conditions at the pre-intervention interview. A P-value < 0.05 was considered significant for all tests. Data analyses were conducted using R v.4.1.2. (The R Foundation for Statistical Computing, Vienna, Austria).
Results
Cross-sectional analyses
After 7 days of the sleep diary and before the behavioral intervention, 2142 participants (mean age: 49.1 years) completed the pre-intervention interview (day 7) and were included in cross-sectional analyses. Among them, 1398 (65.3%) were female, 1630 (76.1%) had at least a bachelor’s degree, and 1426 (66.6%) were under mobility restrictions (Table 1). The baseline characteristics of the included sample (n = 2142) and excluded sample (n = 11 201) are presented in Supplementary Table S3. The IIM of the TST was 440 minutes (7 hours 20) ± 51 minutes and the ISD of the TST was 89 ± 72 minutes. The IIM of the CM was 03:30 am ± 68 minutes and the ISD of the CM was 42 ± 88 minutes. The correlations between IIMs and ISDs of the TST and CM were low, and the maximum was found between the two ISDs (r = 0.51) (Supplementary Table S5). The typical sleep–wake schedules of the participants with different IIMs and ISDs of the TST and CM are presented in Figure 2. In all, 1535 (71.7%) participants reported insomnia, 1243 (58.0%) reported fatigue, 780 (36.4%) reported anxiety, and 370 (17.3%) reported depressive symptoms.
Descriptive Characteristics at Baseline and Sleep Schedules, Sleep Complaints, and Mental Health Conditions at Baseline and in Follow-up
Variables . | Baseline (day 0) . | Pre-intervention (day 7) . | Post-intervention (ay 17) . |
---|---|---|---|
All | 13 343 | 2142 | 732 |
Age (m ± sd) | 47.6 ± 15.3 | 49.1 ± 13.8 | 51.6 ± 12.9 |
Sex | |||
Male | 4759 (35.7%) | 744 (34.7%) | 258 (35.2%) |
Female | 8584 (64.3%) | 1398 (65.3%) | 474 (64.8%) |
Educational level | |||
Middle school | 2152 (16.1%) | 217 (10.1%) | 70 (9.6%) |
High school | 2291 (17.2%) | 295 (13.8%) | 98 (13.4%) |
Bachelor’s degree | 7008 (52.5%) | 1270 (59.3%) | 442 (60.4%) |
Master’s degree | 1892 (14.2%) | 360 (16.8%) | 122 (16.7%) |
Job categories | |||
Higher intellectual professions | 4155 (31.1%) | 822 (38.4%) | 280 (38.2%) |
Intermediate professions | 2161 (16.2%) | 347 (16.2%) | 123 (16.8%) |
Employees and workers | 3310 (24.8%) | 445 (20.8%) | 127 (17.4%) |
No professional activity | 3717 (27.9%) | 528 (24.7%) | 202 (27.6%) |
Mobility restrictions | |||
Yes | 8506 (63.8%) | 1426 (66.6%) | 498 (68.0%) |
No | 4837 (36.3%) | 716 (33.4%) | 234 (32.0%) |
Insomnia (m ± sd) | 14.5 ± 5.2 | 12.7 ± 5.2 | 11.3 ± 5.1 |
Significant (ISI≥10) | 1967 (74.0%) | 1535 (71.7%) | 533 (72.8%) |
Fatigue (m ± sd) | 4.6 ± 1.4 | 4.2 ± 1.4 | 3.9 ± 1.4 |
Significant (FSS≥4) | 9029 (67.7%) | 1243 (58.0%) | 354 (48.4%) |
Anxiety symptoms (m ± sd) | 3.4 ± 1.9 | 2.9 ± 1.9 | 2.5 ± 1.8 |
Significant (PHQ-4≥3) | 6417 (48.1%) | 780 (36.4%) | 203 (27.7%) |
Depressive symptoms (m ± sd) | 2.3 ± 1.8 | 1.8 ± 1.7 | 1.5 ± 1.6 |
Significant (PHQ-4≥3) | 3378 (25.3%) | 370 (17.3%) | 89 (12.2%) |
Total sleep time – IIM | 7 h 20 ± 51 min | 7 h 29 ± 48 min | |
Less than 420 min (7 h) | 681 (31.8%) | 196 (26.8%) | |
Between 420 and 540 min (7–9 h) | 1384 (64.6%) | 513 (70.1%) | |
540 min (9 h) and more | 77 (3.6%) | 23 (3.1%) | |
Total sleep time – ISD | 68 ± 27 min | 65 ± 25 min | |
Q1: Less than 45 min | 429 (20.0%) | 161 (22.0%) | |
Q2-Q4: Between 45 and 88 min | 1284 (59.9%) | 457 (62.4%) | |
Q5: 88 min and more | 429 (20.0%) | 114 (15.6%) | |
Circadian midpoint – IIM | 3:30 ± 68 min | 3:27 ± 66 min | |
Q1: Earlier than 02:42 am | 429 (20.0%) | 138 (18.9%) | |
Q2-Q4: Between 02:42 and 04:14 am | 1284 (59.9%) | 456 (62.3%) | |
Q5: 04:14 am and later | 429 (20.0%) | 138 (18.9%) | |
Circadian midpoint – ISD | 44 ± 25 min | 41 ± 23 min | |
Q1: Less than 26 min | 429 (20.0%) | 169 (23.1%) | |
Q2-Q4: Between 12 and 57 min | 1284 (59.9%) | 449 (61.3%) | |
Q5: 57 min and more | 429 (20.0%) | 114 (15.5%) |
Variables . | Baseline (day 0) . | Pre-intervention (day 7) . | Post-intervention (ay 17) . |
---|---|---|---|
All | 13 343 | 2142 | 732 |
Age (m ± sd) | 47.6 ± 15.3 | 49.1 ± 13.8 | 51.6 ± 12.9 |
Sex | |||
Male | 4759 (35.7%) | 744 (34.7%) | 258 (35.2%) |
Female | 8584 (64.3%) | 1398 (65.3%) | 474 (64.8%) |
Educational level | |||
Middle school | 2152 (16.1%) | 217 (10.1%) | 70 (9.6%) |
High school | 2291 (17.2%) | 295 (13.8%) | 98 (13.4%) |
Bachelor’s degree | 7008 (52.5%) | 1270 (59.3%) | 442 (60.4%) |
Master’s degree | 1892 (14.2%) | 360 (16.8%) | 122 (16.7%) |
Job categories | |||
Higher intellectual professions | 4155 (31.1%) | 822 (38.4%) | 280 (38.2%) |
Intermediate professions | 2161 (16.2%) | 347 (16.2%) | 123 (16.8%) |
Employees and workers | 3310 (24.8%) | 445 (20.8%) | 127 (17.4%) |
No professional activity | 3717 (27.9%) | 528 (24.7%) | 202 (27.6%) |
Mobility restrictions | |||
Yes | 8506 (63.8%) | 1426 (66.6%) | 498 (68.0%) |
No | 4837 (36.3%) | 716 (33.4%) | 234 (32.0%) |
Insomnia (m ± sd) | 14.5 ± 5.2 | 12.7 ± 5.2 | 11.3 ± 5.1 |
Significant (ISI≥10) | 1967 (74.0%) | 1535 (71.7%) | 533 (72.8%) |
Fatigue (m ± sd) | 4.6 ± 1.4 | 4.2 ± 1.4 | 3.9 ± 1.4 |
Significant (FSS≥4) | 9029 (67.7%) | 1243 (58.0%) | 354 (48.4%) |
Anxiety symptoms (m ± sd) | 3.4 ± 1.9 | 2.9 ± 1.9 | 2.5 ± 1.8 |
Significant (PHQ-4≥3) | 6417 (48.1%) | 780 (36.4%) | 203 (27.7%) |
Depressive symptoms (m ± sd) | 2.3 ± 1.8 | 1.8 ± 1.7 | 1.5 ± 1.6 |
Significant (PHQ-4≥3) | 3378 (25.3%) | 370 (17.3%) | 89 (12.2%) |
Total sleep time – IIM | 7 h 20 ± 51 min | 7 h 29 ± 48 min | |
Less than 420 min (7 h) | 681 (31.8%) | 196 (26.8%) | |
Between 420 and 540 min (7–9 h) | 1384 (64.6%) | 513 (70.1%) | |
540 min (9 h) and more | 77 (3.6%) | 23 (3.1%) | |
Total sleep time – ISD | 68 ± 27 min | 65 ± 25 min | |
Q1: Less than 45 min | 429 (20.0%) | 161 (22.0%) | |
Q2-Q4: Between 45 and 88 min | 1284 (59.9%) | 457 (62.4%) | |
Q5: 88 min and more | 429 (20.0%) | 114 (15.6%) | |
Circadian midpoint – IIM | 3:30 ± 68 min | 3:27 ± 66 min | |
Q1: Earlier than 02:42 am | 429 (20.0%) | 138 (18.9%) | |
Q2-Q4: Between 02:42 and 04:14 am | 1284 (59.9%) | 456 (62.3%) | |
Q5: 04:14 am and later | 429 (20.0%) | 138 (18.9%) | |
Circadian midpoint – ISD | 44 ± 25 min | 41 ± 23 min | |
Q1: Less than 26 min | 429 (20.0%) | 169 (23.1%) | |
Q2-Q4: Between 12 and 57 min | 1284 (59.9%) | 449 (61.3%) | |
Q5: 57 min and more | 429 (20.0%) | 114 (15.5%) |
m ± sd: mean ± standard deviation.
Descriptive Characteristics at Baseline and Sleep Schedules, Sleep Complaints, and Mental Health Conditions at Baseline and in Follow-up
Variables . | Baseline (day 0) . | Pre-intervention (day 7) . | Post-intervention (ay 17) . |
---|---|---|---|
All | 13 343 | 2142 | 732 |
Age (m ± sd) | 47.6 ± 15.3 | 49.1 ± 13.8 | 51.6 ± 12.9 |
Sex | |||
Male | 4759 (35.7%) | 744 (34.7%) | 258 (35.2%) |
Female | 8584 (64.3%) | 1398 (65.3%) | 474 (64.8%) |
Educational level | |||
Middle school | 2152 (16.1%) | 217 (10.1%) | 70 (9.6%) |
High school | 2291 (17.2%) | 295 (13.8%) | 98 (13.4%) |
Bachelor’s degree | 7008 (52.5%) | 1270 (59.3%) | 442 (60.4%) |
Master’s degree | 1892 (14.2%) | 360 (16.8%) | 122 (16.7%) |
Job categories | |||
Higher intellectual professions | 4155 (31.1%) | 822 (38.4%) | 280 (38.2%) |
Intermediate professions | 2161 (16.2%) | 347 (16.2%) | 123 (16.8%) |
Employees and workers | 3310 (24.8%) | 445 (20.8%) | 127 (17.4%) |
No professional activity | 3717 (27.9%) | 528 (24.7%) | 202 (27.6%) |
Mobility restrictions | |||
Yes | 8506 (63.8%) | 1426 (66.6%) | 498 (68.0%) |
No | 4837 (36.3%) | 716 (33.4%) | 234 (32.0%) |
Insomnia (m ± sd) | 14.5 ± 5.2 | 12.7 ± 5.2 | 11.3 ± 5.1 |
Significant (ISI≥10) | 1967 (74.0%) | 1535 (71.7%) | 533 (72.8%) |
Fatigue (m ± sd) | 4.6 ± 1.4 | 4.2 ± 1.4 | 3.9 ± 1.4 |
Significant (FSS≥4) | 9029 (67.7%) | 1243 (58.0%) | 354 (48.4%) |
Anxiety symptoms (m ± sd) | 3.4 ± 1.9 | 2.9 ± 1.9 | 2.5 ± 1.8 |
Significant (PHQ-4≥3) | 6417 (48.1%) | 780 (36.4%) | 203 (27.7%) |
Depressive symptoms (m ± sd) | 2.3 ± 1.8 | 1.8 ± 1.7 | 1.5 ± 1.6 |
Significant (PHQ-4≥3) | 3378 (25.3%) | 370 (17.3%) | 89 (12.2%) |
Total sleep time – IIM | 7 h 20 ± 51 min | 7 h 29 ± 48 min | |
Less than 420 min (7 h) | 681 (31.8%) | 196 (26.8%) | |
Between 420 and 540 min (7–9 h) | 1384 (64.6%) | 513 (70.1%) | |
540 min (9 h) and more | 77 (3.6%) | 23 (3.1%) | |
Total sleep time – ISD | 68 ± 27 min | 65 ± 25 min | |
Q1: Less than 45 min | 429 (20.0%) | 161 (22.0%) | |
Q2-Q4: Between 45 and 88 min | 1284 (59.9%) | 457 (62.4%) | |
Q5: 88 min and more | 429 (20.0%) | 114 (15.6%) | |
Circadian midpoint – IIM | 3:30 ± 68 min | 3:27 ± 66 min | |
Q1: Earlier than 02:42 am | 429 (20.0%) | 138 (18.9%) | |
Q2-Q4: Between 02:42 and 04:14 am | 1284 (59.9%) | 456 (62.3%) | |
Q5: 04:14 am and later | 429 (20.0%) | 138 (18.9%) | |
Circadian midpoint – ISD | 44 ± 25 min | 41 ± 23 min | |
Q1: Less than 26 min | 429 (20.0%) | 169 (23.1%) | |
Q2-Q4: Between 12 and 57 min | 1284 (59.9%) | 449 (61.3%) | |
Q5: 57 min and more | 429 (20.0%) | 114 (15.5%) |
Variables . | Baseline (day 0) . | Pre-intervention (day 7) . | Post-intervention (ay 17) . |
---|---|---|---|
All | 13 343 | 2142 | 732 |
Age (m ± sd) | 47.6 ± 15.3 | 49.1 ± 13.8 | 51.6 ± 12.9 |
Sex | |||
Male | 4759 (35.7%) | 744 (34.7%) | 258 (35.2%) |
Female | 8584 (64.3%) | 1398 (65.3%) | 474 (64.8%) |
Educational level | |||
Middle school | 2152 (16.1%) | 217 (10.1%) | 70 (9.6%) |
High school | 2291 (17.2%) | 295 (13.8%) | 98 (13.4%) |
Bachelor’s degree | 7008 (52.5%) | 1270 (59.3%) | 442 (60.4%) |
Master’s degree | 1892 (14.2%) | 360 (16.8%) | 122 (16.7%) |
Job categories | |||
Higher intellectual professions | 4155 (31.1%) | 822 (38.4%) | 280 (38.2%) |
Intermediate professions | 2161 (16.2%) | 347 (16.2%) | 123 (16.8%) |
Employees and workers | 3310 (24.8%) | 445 (20.8%) | 127 (17.4%) |
No professional activity | 3717 (27.9%) | 528 (24.7%) | 202 (27.6%) |
Mobility restrictions | |||
Yes | 8506 (63.8%) | 1426 (66.6%) | 498 (68.0%) |
No | 4837 (36.3%) | 716 (33.4%) | 234 (32.0%) |
Insomnia (m ± sd) | 14.5 ± 5.2 | 12.7 ± 5.2 | 11.3 ± 5.1 |
Significant (ISI≥10) | 1967 (74.0%) | 1535 (71.7%) | 533 (72.8%) |
Fatigue (m ± sd) | 4.6 ± 1.4 | 4.2 ± 1.4 | 3.9 ± 1.4 |
Significant (FSS≥4) | 9029 (67.7%) | 1243 (58.0%) | 354 (48.4%) |
Anxiety symptoms (m ± sd) | 3.4 ± 1.9 | 2.9 ± 1.9 | 2.5 ± 1.8 |
Significant (PHQ-4≥3) | 6417 (48.1%) | 780 (36.4%) | 203 (27.7%) |
Depressive symptoms (m ± sd) | 2.3 ± 1.8 | 1.8 ± 1.7 | 1.5 ± 1.6 |
Significant (PHQ-4≥3) | 3378 (25.3%) | 370 (17.3%) | 89 (12.2%) |
Total sleep time – IIM | 7 h 20 ± 51 min | 7 h 29 ± 48 min | |
Less than 420 min (7 h) | 681 (31.8%) | 196 (26.8%) | |
Between 420 and 540 min (7–9 h) | 1384 (64.6%) | 513 (70.1%) | |
540 min (9 h) and more | 77 (3.6%) | 23 (3.1%) | |
Total sleep time – ISD | 68 ± 27 min | 65 ± 25 min | |
Q1: Less than 45 min | 429 (20.0%) | 161 (22.0%) | |
Q2-Q4: Between 45 and 88 min | 1284 (59.9%) | 457 (62.4%) | |
Q5: 88 min and more | 429 (20.0%) | 114 (15.6%) | |
Circadian midpoint – IIM | 3:30 ± 68 min | 3:27 ± 66 min | |
Q1: Earlier than 02:42 am | 429 (20.0%) | 138 (18.9%) | |
Q2-Q4: Between 02:42 and 04:14 am | 1284 (59.9%) | 456 (62.3%) | |
Q5: 04:14 am and later | 429 (20.0%) | 138 (18.9%) | |
Circadian midpoint – ISD | 44 ± 25 min | 41 ± 23 min | |
Q1: Less than 26 min | 429 (20.0%) | 169 (23.1%) | |
Q2-Q4: Between 12 and 57 min | 1284 (59.9%) | 449 (61.3%) | |
Q5: 57 min and more | 429 (20.0%) | 114 (15.5%) |
m ± sd: mean ± standard deviation.

Typical sleep–wake schedules according to the different sleep patterns.
After adjusting for the covariates, a high ISD of the TST was associated with a higher likelihood of insomnia (RR = 1.26 [1.21–1.30] for irregular TST compared to regular TST), fatigue (RR = 1.41 [1.26–1.55]), anxiety (RR = 1.55 [1.30–1.82]), and depressive symptoms (RR = 1.90 [1.41–2.51]). A short TST was associated with a higher likelihood of insomnia (RR = 1.19 [1.15–1.23] for short TST compared to mid-TST), fatigue (RR = 1.18 [1.10–1.26]), and anxiety symptoms (RR = 1.24 [1.10–1.38]), and late CM with a higher likelihood of depressive symptoms (RR = 1.42 [1.13–1.75] for late CM compared to neutral CM. A high ISD of the CM was associated with a higher likelihood of insomnia (RR = 1.17 [1.11–1.22] for irregular CM compared to regular CM), fatigue (RR = 1.19 [1.05–1.33]) and anxiety symptoms (RR = 1.26 [1.04–1.50]) (Table 2).
Adjusted Cross-Sectional Associations Between Sleep Schedules and Sleep Complaints and Mental Health Conditions (n = 2142)
. | Insomnia . | Fatigue . | Anxiety . | Depression . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total sleep time – IIM | ||||||||
Less than 420 min (7 h) | 1.19 [1.15 to 1.23] | 1.17 [1.09 to 1.24] | 1.24 [1.10 to 1.38] | 1.11 [0.90 to 1.36] | ||||
Between 420 and 540 min (7–9 h) | Ref | <0.001 | Ref | <0.001 | Ref | 0.002 | Ref | 0.283 |
540 min (9 h) and more | 0.87 [0.70 to 1.02] | 1.13 [0.93 to 1.31] | 0.94 [0.66 to 1.27] | 1.36 [0.87 to 2.02] | ||||
Total sleep time – ISD | ||||||||
Q1:Less than 45 min | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 |
Q2-Q4: Between 45 and 88 min | 1.15 [1.10 to 1.20] | 1.26 [1.13 to 1.37] | 1.35 [1.14 to 1.56] | 1.45 [1.10 to 1.90] | ||||
Q5: 88 min and more | 1.26 [1.21 to 1.30] | 1.41 [1.26 to 1.55] | 1.55 [1.30 to 1.82] | 1.90 [1.41 to 2.51] | ||||
Circadian Midpoint – IIM | ||||||||
Q1: Earlier than 02:42 am | 1.15 [1.09 to 1.21] | <0.001 | 1.05 [0.95 to 1.14] | 0.611 | 1.09 [0.94 to 1.25] | 0.492 | 1.07 [0.83 to 1.37] | 0.013 |
Q2-Q4: Between 02:42 and 04:14 am | Ref | Ref | Ref | Ref | ||||
Q5: 04:14 am and later | 1.03 [0.95 to 1.10] | 1.03 [0.93 to 1.13] | 1.05 [0.90 to 1.21] | 1.42 [1.13 to 1.75] | ||||
Circadian midpoint – ISD | ||||||||
Q1: Less than 26 min | Ref | <0.001 | Ref | 0.027 | Ref | 0.050 | Ref | 0.151 |
Q2-Q4: Between 12 min and 57 min | 1.11 [1.05 to 1.15] | 1.10 [0.99 to 1.21] | 1.17 [1.00 to 1.35] | 1.29 [0.99 to 1.67] | ||||
Q5: 57 min and more | 1.17 [1.11 to 1.22] | 1.19 [1.05 to 1.33] | 1.26 [1.04 to 1.50] | 1.29 [0.93 to 1.76] |
. | Insomnia . | Fatigue . | Anxiety . | Depression . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total sleep time – IIM | ||||||||
Less than 420 min (7 h) | 1.19 [1.15 to 1.23] | 1.17 [1.09 to 1.24] | 1.24 [1.10 to 1.38] | 1.11 [0.90 to 1.36] | ||||
Between 420 and 540 min (7–9 h) | Ref | <0.001 | Ref | <0.001 | Ref | 0.002 | Ref | 0.283 |
540 min (9 h) and more | 0.87 [0.70 to 1.02] | 1.13 [0.93 to 1.31] | 0.94 [0.66 to 1.27] | 1.36 [0.87 to 2.02] | ||||
Total sleep time – ISD | ||||||||
Q1:Less than 45 min | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 |
Q2-Q4: Between 45 and 88 min | 1.15 [1.10 to 1.20] | 1.26 [1.13 to 1.37] | 1.35 [1.14 to 1.56] | 1.45 [1.10 to 1.90] | ||||
Q5: 88 min and more | 1.26 [1.21 to 1.30] | 1.41 [1.26 to 1.55] | 1.55 [1.30 to 1.82] | 1.90 [1.41 to 2.51] | ||||
Circadian Midpoint – IIM | ||||||||
Q1: Earlier than 02:42 am | 1.15 [1.09 to 1.21] | <0.001 | 1.05 [0.95 to 1.14] | 0.611 | 1.09 [0.94 to 1.25] | 0.492 | 1.07 [0.83 to 1.37] | 0.013 |
Q2-Q4: Between 02:42 and 04:14 am | Ref | Ref | Ref | Ref | ||||
Q5: 04:14 am and later | 1.03 [0.95 to 1.10] | 1.03 [0.93 to 1.13] | 1.05 [0.90 to 1.21] | 1.42 [1.13 to 1.75] | ||||
Circadian midpoint – ISD | ||||||||
Q1: Less than 26 min | Ref | <0.001 | Ref | 0.027 | Ref | 0.050 | Ref | 0.151 |
Q2-Q4: Between 12 min and 57 min | 1.11 [1.05 to 1.15] | 1.10 [0.99 to 1.21] | 1.17 [1.00 to 1.35] | 1.29 [0.99 to 1.67] | ||||
Q5: 57 min and more | 1.17 [1.11 to 1.22] | 1.19 [1.05 to 1.33] | 1.26 [1.04 to 1.50] | 1.29 [0.93 to 1.76] |
Adjusted for age, sex, educational level, sleep diary duration, and time between evaluations.
Adjusted Cross-Sectional Associations Between Sleep Schedules and Sleep Complaints and Mental Health Conditions (n = 2142)
. | Insomnia . | Fatigue . | Anxiety . | Depression . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total sleep time – IIM | ||||||||
Less than 420 min (7 h) | 1.19 [1.15 to 1.23] | 1.17 [1.09 to 1.24] | 1.24 [1.10 to 1.38] | 1.11 [0.90 to 1.36] | ||||
Between 420 and 540 min (7–9 h) | Ref | <0.001 | Ref | <0.001 | Ref | 0.002 | Ref | 0.283 |
540 min (9 h) and more | 0.87 [0.70 to 1.02] | 1.13 [0.93 to 1.31] | 0.94 [0.66 to 1.27] | 1.36 [0.87 to 2.02] | ||||
Total sleep time – ISD | ||||||||
Q1:Less than 45 min | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 |
Q2-Q4: Between 45 and 88 min | 1.15 [1.10 to 1.20] | 1.26 [1.13 to 1.37] | 1.35 [1.14 to 1.56] | 1.45 [1.10 to 1.90] | ||||
Q5: 88 min and more | 1.26 [1.21 to 1.30] | 1.41 [1.26 to 1.55] | 1.55 [1.30 to 1.82] | 1.90 [1.41 to 2.51] | ||||
Circadian Midpoint – IIM | ||||||||
Q1: Earlier than 02:42 am | 1.15 [1.09 to 1.21] | <0.001 | 1.05 [0.95 to 1.14] | 0.611 | 1.09 [0.94 to 1.25] | 0.492 | 1.07 [0.83 to 1.37] | 0.013 |
Q2-Q4: Between 02:42 and 04:14 am | Ref | Ref | Ref | Ref | ||||
Q5: 04:14 am and later | 1.03 [0.95 to 1.10] | 1.03 [0.93 to 1.13] | 1.05 [0.90 to 1.21] | 1.42 [1.13 to 1.75] | ||||
Circadian midpoint – ISD | ||||||||
Q1: Less than 26 min | Ref | <0.001 | Ref | 0.027 | Ref | 0.050 | Ref | 0.151 |
Q2-Q4: Between 12 min and 57 min | 1.11 [1.05 to 1.15] | 1.10 [0.99 to 1.21] | 1.17 [1.00 to 1.35] | 1.29 [0.99 to 1.67] | ||||
Q5: 57 min and more | 1.17 [1.11 to 1.22] | 1.19 [1.05 to 1.33] | 1.26 [1.04 to 1.50] | 1.29 [0.93 to 1.76] |
. | Insomnia . | Fatigue . | Anxiety . | Depression . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total sleep time – IIM | ||||||||
Less than 420 min (7 h) | 1.19 [1.15 to 1.23] | 1.17 [1.09 to 1.24] | 1.24 [1.10 to 1.38] | 1.11 [0.90 to 1.36] | ||||
Between 420 and 540 min (7–9 h) | Ref | <0.001 | Ref | <0.001 | Ref | 0.002 | Ref | 0.283 |
540 min (9 h) and more | 0.87 [0.70 to 1.02] | 1.13 [0.93 to 1.31] | 0.94 [0.66 to 1.27] | 1.36 [0.87 to 2.02] | ||||
Total sleep time – ISD | ||||||||
Q1:Less than 45 min | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 | Ref | <0.001 |
Q2-Q4: Between 45 and 88 min | 1.15 [1.10 to 1.20] | 1.26 [1.13 to 1.37] | 1.35 [1.14 to 1.56] | 1.45 [1.10 to 1.90] | ||||
Q5: 88 min and more | 1.26 [1.21 to 1.30] | 1.41 [1.26 to 1.55] | 1.55 [1.30 to 1.82] | 1.90 [1.41 to 2.51] | ||||
Circadian Midpoint – IIM | ||||||||
Q1: Earlier than 02:42 am | 1.15 [1.09 to 1.21] | <0.001 | 1.05 [0.95 to 1.14] | 0.611 | 1.09 [0.94 to 1.25] | 0.492 | 1.07 [0.83 to 1.37] | 0.013 |
Q2-Q4: Between 02:42 and 04:14 am | Ref | Ref | Ref | Ref | ||||
Q5: 04:14 am and later | 1.03 [0.95 to 1.10] | 1.03 [0.93 to 1.13] | 1.05 [0.90 to 1.21] | 1.42 [1.13 to 1.75] | ||||
Circadian midpoint – ISD | ||||||||
Q1: Less than 26 min | Ref | <0.001 | Ref | 0.027 | Ref | 0.050 | Ref | 0.151 |
Q2-Q4: Between 12 min and 57 min | 1.11 [1.05 to 1.15] | 1.10 [0.99 to 1.21] | 1.17 [1.00 to 1.35] | 1.29 [0.99 to 1.67] | ||||
Q5: 57 min and more | 1.17 [1.11 to 1.22] | 1.19 [1.05 to 1.33] | 1.26 [1.04 to 1.50] | 1.29 [0.93 to 1.76] |
Adjusted for age, sex, educational level, sleep diary duration, and time between evaluations.
Longitudinal analyses
After the behavioral intervention, 732 participants completed the post-intervention sleep diary and the final interview (day 17) and were included in the longitudinal analyses. The baseline characteristics of the included sample (n = 732) and excluded samples (n = 12 611) are presented in Supplementary Table S4. The IIM of the TST increased for 59% of the participants and increased for more than 30 minutes for 27% of the participants (d = 0.16, P < 0.001). The ISD of the TST decreased from 68 to 65 minutes (d = 0.11, P = 0.048). Insomnia decreased from 12.9 at the pre-intervention interview (day 7) to 11.3 (d = 0.30, P < 0.001) at the post-intervention interview (day 17); fatigue decreased from 4.2 to 3.9 (d = 0.22, P < 0.001); anxiety decreased from 2.8 to 2.5 (d = 0.18, P < 0.001); and depressive symptoms decreased from 1.7 to 1.5 (d = 0.11, P = 0.029).
The changes in insomnia and depressive symptoms following the behavioral intervention varied according to the change in the sleep schedule (Figure 3). Greater improvements in insomnia were observed in participants with a longer and more regular TST. The improvement in anxiety symptoms was greater in participants with a longer TST and the improvement in depressive symptoms was greater in participants with a more regular TST.

Changes in sleep complaints and mental health conditions according to changes in total sleep time (n = 732).
After adjusting for the covariates, including sleep complaints and mental health conditions at the pre-intervention interview, a more regular TST was associated with a higher likelihood of reduced insomnia (RR = 1.33 [1.10–1.52]) and depressive symptoms (RR = 1.55 [1.13–1.98]) compared to a more irregular TST (Table 3). A longer TST was associated with a higher likelihood of reduced insomnia (RR = 1.37 [1.17–1.52]) and anxiety symptoms (RR = 1.52 [1.09–1.98]) compared to a shorter TST.
Discussion
In this large nonclinical population, irregular sleep duration, as measured by the ISD of the TST, was associated with an increased likelihood of insomnia, fatigue, anxiety, and depressive symptoms at baseline. Estimates were higher for the ISD of the TST (2-fold higher likelihood of depressive symptoms for irregular sleepers than regular sleepers) than those with the IIM of the TST (1.5-fold higher likelihood of depressive symptoms for short sleepers than mid-sleepers). Improved sleep regularity between the post-intervention and the pre-intervention sleep diary was associated with a greater reduction in sleep complaints and mental health symptoms between the post-intervention and the pre-intervention interviews. These findings indicate the importance of sleep rhythms in sleep and mental health. The other sleep variables studied (i.e. improvements in mean sleep quantity and timing) and sociodemographic characteristics (i.e. age, sex, educational level, and mobility restrictions) had lower estimates in association with the mental health condition, confirming the importance of regular sleep as a major determinant of sleep health. These initial results were confirmed after the behavioral interventions and showed that sleep complaints and mental health conditions improved during follow-up.
Our study had several strengths. First, consistent associations were observed between sleep regularity and sleep complaints and mental health conditions in a large nonclinical adult sample, using cross-sectional and longitudinal analyses. We also examined other sleep dimensions (i.e. duration and timing) that are closely related to sleep regularity [51] and that are thought to affect sleep complaints and mental health conditions [24, 26].
Second, this study showed for the first time that improvements in sleep regularity and duration are associated with fewer sleep complaints and an improved mental health condition using a fully automated intervention program. Further research using a controlled design in a clinical population should confirm these population findings.
Third, to the best of our knowledge, this is the first study to provide data obtained from a virtual agent application and showed not only that this application can be used in a research setting but also that it may be useful to facilitate changes in the sleep schedule, which would, in turn, improve sleep health as well as mental health.
Our results are consistent with previous studies that reported an association between sleep regularity and sleep complaints [12–14], and anxiety and depressive symptoms [12, 18–22]. Among them, two studies were conducted during the Covid-19 health crisis [21, 22], while most were conducted before the outbreak [12–14, 18–20]. One study on a sample of 771 adults reported that the ISD of the TST was associated with depressive symptoms after adjusting for the IIM of the TST. However, the ISD of the CM was not associated with anxiety or depressive symptoms [18]. Our study confirms these findings and shows that these associations persisted longitudinally. Improvements in regular sleep duration were associated with fewer anxiety and depressive symptoms in the post-intervention interview but not sleep timing regularity. As these are different physiological processes [52], regularity in sleep timing and sleep duration are not strongly correlated and it is not surprising that they have different effects on mental health. These two dimensions should be systematically considered separately concerning adverse health outcomes in future studies. However, within a prevention approach, we suggest addressing sleep regularity holistically, because both duration regularity and timing regularity are related to sleep complaints.
Sleep duration, a well-studied sleep behavior that may be associated with depression [26], was not associated with depressive symptoms in this study. This could be due to the choice of the threshold (±7 hours), based on international recommendations [38]. However, in supplementary analyses, a TST < 360 minutes (6 hours) was associated with depressive symptoms compared to participants with a TST > 420 minutes (7 hours) (RR = 1.73 [1.24–2.33]). We found no increased risk for a mental health condition in long sleepers, as in previous studies [26]. Sleep duration was also associated with insomnia, fatigue, and anxiety symptoms at the follow-up, but estimates were slightly lower than with sleep duration regularity. A joint analysis of these two variables suggested that sleep regularity should be considered as an independent and equally important dimension in sleep health recommendations [38, 53]. Owing to individual, social, and societal constraints that determine an individual’s opportunity to sleep [54], it would be interesting to explore whether it is better to promote sleep regularity even if this means reducing sleep duration or whether it is better to promote adequate sleep duration, even if it means being irregular. Sleep timing, which has been less studied, is associated with mental health [24]. However, in our study, sleep timing was not associated with sleep complaints or mental health conditions, as in a previous study [41]. Further studies are needed to clarify the importance of sleep timing concerning sleep health and mental health.
Our findings must be considered in light of certain limitations. First, the sleep schedules were self-reported. However, the sleep diary is a worldwide validated and reliable day-to-day tool [55], which may limit misclassifications. Subjective measurements of sleep schedules yielded higher estimates of time-in-bed and TST and lower estimates of sleep latency and wake-after-sleep-onset compared to objective measurements [56]. Moreover, this bias might be differential since patients with chronic insomnia might have a more objective/subjective discrepancy than good sleepers [57]. This might favor the relationship between the IIM of the TST and sleep complaints and mental health outcomes in our study. However, the extent to which the ISDs of the TST and the CM varied across objective/subjective measurements is unknown and should be explored in future studies.
Second, this was an ancillary study of a single-arm intervention based on behavioral recommendations. The lack of a control group or blinding prevents us from drawing conclusions about the effectiveness of the intervention. Although we cannot exclude a placebo or an interventional effect, these results raise the need to integrate regularity in the design of future digital interventions, at least as an early marker of success, and at best as an element on the causal path. An observational study should confirm these results but this is outside the scope of this intervention.
Third, being a population-based study, the follow-up evaluation was set at 10 days to allow participants enough time to implement the recommendations but not too much time to maximize the participation rate. Thus, our study did not benefit from a medium/long-term follow-up, which may explain the slight effect on mental health. Although the impact of the sleep schedule on sleep complaints appeared to be direct and highly reactive [9], a longer term may be required to generate mood changes. The first hypothesis involves the internal desynchronization of the hypothalamic-pituitary-adrenal axis, leading to heightened stress reactivity [58]. Additionally, altered sleep quality may be a critical mediator of emotion regulation [59] and resilience to stress [60]. Further studies should explore the longitudinal relationship between sleep timing regularity and mental health using a longer-term follow-up.
Fourth, a high percentage of participants were lost to follow-up, which may have led to a selection bias. In addition, the excluded sample was younger than the included sample and reported more severe mental health conditions. This selection is likely to have modified the associations in the direction of an underestimation by the loss of the people who have the most disrupted sleep and mental health.
Fifth, no objective measurements of sleep were used. Therefore, the effects of other sleep disorders, such as obstructive sleep apnea, as risk factors for sleep complaints and mental health conditions were not considered. Nevertheless, because we focused on insomnia complaints, which is mainly a subjective sleep complaint, we believe that our results are relevant. Similarly, there was no information on several health outcomes associated with poor sleep (body mass index, smoking status, diabetes, hypertension). However, a sensitivity analysis conducted across a small sample with available data for smoking status showed similar results (Supplementary Table S6). Moreover, work schedules were not assessed despite their known sleep disruption [61]. However, several studies hypothesize that sleep hygiene recommendations could benefit shift workers [62, 63], and night work concerns less than 2% of the French general population [64]. Therefore, it is unlikely that it would have changed the results of this study.
In conclusion, our results show that a more regular sleep schedule is associated with a decreased risk of sleep complaints and mental health conditions at follow-up. The small magnitude of the associations might be explained by the short follow-up period. Our results raise the need to consider regularity as an important target to improve the sleep health of the adult population [38, 53], independently of other sleep dimensions. Policymakers, health professionals, and the general population should be made aware that beyond their positive effects on sleep health, regular schedules could enhance mental health. Further studies with a controlled design are needed to confirm that improving sleep regularity benefits both sleep health and mental health.
Supplementary Material
Supplementary material is available at SLEEP online.
Adjusted Longitudinal Associations Between Changes in Sleep Schedules and Changes in Sleep Complaints and Mental Health Conditions (n = 732)
. | Decreased insomnia . | Decreased fatigue . | Decreased anxiety symptoms . | Decreased depressive symptoms . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total Sleep Time – IIM | ||||||||
Q1: Shorter (decrease ≥ 22 min) | Ref | 0.001 | Ref | 0.166 | Ref | 0.019 | Ref | 0.132 |
Q2-Q4: Unchanged | 1.08 [0.90 to 1.25] | 1.18 [0.99 to 1.36] | 1.46 [1.10 to 1.84] | 1.32 [0.97 to 1.72] | ||||
Q5: Longer (increase > 40 min) | 1.37 [1.17 to 1.52] | 1.16 [0.93 to 1.38] | 1.52 [1.09 to 1.98] | 1.08 [0.71 to 1.54] | ||||
Total sleep time – ISD | ||||||||
Q1: More irregular (increase > 44 min) | Ref | 0.007 | Ref | 0.117 | Ref | 0.359 | Ref | 0.014 |
Q2-Q4: Unchanged | 1.28 [1.09 to 1.45] | 0.89 [0.73 to 1.04] | 1.12 [0.86 to 1.41] | 1.12 [0.82 to 1.48] | ||||
Q5: More regular (decrease ≥ 54 min) | 1.33 [1.10 to 1.52] | 1.03 [0.84 to 1.20] | 0.95 [0.66 to 1.28] | 1.55 [1.13 to 1.98] | ||||
Circadian midpoint – IIM | ||||||||
Q1: Earlier (≥ 21 min) | Ref | 0.033 | Ref | 0.879 | Ref | 0.763 | Ref | 0.194 |
Q2-Q4: Unchanged | 0.98 [0.81 to 1.13] | 1.00 [0.83 to 1.17] | 1.03 [0.78 to 1.31] | 1.14 [0.84 to 1.49] | ||||
Q5: Later (> 22 min) | 1.18 [1.00 to 1.33] | 1.04 [0.83 to 1.24] | 0.94 [0.66 to 1.26] | 0.87 [0.57 to 1.26] | ||||
Circadian midpoint – ISD | ||||||||
Q1: More irregular (increase > 16 min) | Ref | 0.079 | Ref | 0.604 | Ref | 0.601 | Ref | 0.931 |
Q2-Q4: Unchanged | 1.19 [1.01 to 1.34] | 1.07 [0.89 to 1.23] | 0.91 [0.69 to 1.14] | 1.06 [0.77 to 1.39] | ||||
Q5: More regular (decrease ≥ 24 min) | 1.22 [1.00 to 1.40] | 1.11 [0.89 to 1.30] | 0.86 [0.61 to 1.15] | 1.03 [0.70 to 1.44] |
. | Decreased insomnia . | Decreased fatigue . | Decreased anxiety symptoms . | Decreased depressive symptoms . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total Sleep Time – IIM | ||||||||
Q1: Shorter (decrease ≥ 22 min) | Ref | 0.001 | Ref | 0.166 | Ref | 0.019 | Ref | 0.132 |
Q2-Q4: Unchanged | 1.08 [0.90 to 1.25] | 1.18 [0.99 to 1.36] | 1.46 [1.10 to 1.84] | 1.32 [0.97 to 1.72] | ||||
Q5: Longer (increase > 40 min) | 1.37 [1.17 to 1.52] | 1.16 [0.93 to 1.38] | 1.52 [1.09 to 1.98] | 1.08 [0.71 to 1.54] | ||||
Total sleep time – ISD | ||||||||
Q1: More irregular (increase > 44 min) | Ref | 0.007 | Ref | 0.117 | Ref | 0.359 | Ref | 0.014 |
Q2-Q4: Unchanged | 1.28 [1.09 to 1.45] | 0.89 [0.73 to 1.04] | 1.12 [0.86 to 1.41] | 1.12 [0.82 to 1.48] | ||||
Q5: More regular (decrease ≥ 54 min) | 1.33 [1.10 to 1.52] | 1.03 [0.84 to 1.20] | 0.95 [0.66 to 1.28] | 1.55 [1.13 to 1.98] | ||||
Circadian midpoint – IIM | ||||||||
Q1: Earlier (≥ 21 min) | Ref | 0.033 | Ref | 0.879 | Ref | 0.763 | Ref | 0.194 |
Q2-Q4: Unchanged | 0.98 [0.81 to 1.13] | 1.00 [0.83 to 1.17] | 1.03 [0.78 to 1.31] | 1.14 [0.84 to 1.49] | ||||
Q5: Later (> 22 min) | 1.18 [1.00 to 1.33] | 1.04 [0.83 to 1.24] | 0.94 [0.66 to 1.26] | 0.87 [0.57 to 1.26] | ||||
Circadian midpoint – ISD | ||||||||
Q1: More irregular (increase > 16 min) | Ref | 0.079 | Ref | 0.604 | Ref | 0.601 | Ref | 0.931 |
Q2-Q4: Unchanged | 1.19 [1.01 to 1.34] | 1.07 [0.89 to 1.23] | 0.91 [0.69 to 1.14] | 1.06 [0.77 to 1.39] | ||||
Q5: More regular (decrease ≥ 24 min) | 1.22 [1.00 to 1.40] | 1.11 [0.89 to 1.30] | 0.86 [0.61 to 1.15] | 1.03 [0.70 to 1.44] |
Adjusted for age, sex, educational level, sleep diary duration, time between evaluations, and sleep complaints and mental health conditions at the pre-intervention interview.
Adjusted Longitudinal Associations Between Changes in Sleep Schedules and Changes in Sleep Complaints and Mental Health Conditions (n = 732)
. | Decreased insomnia . | Decreased fatigue . | Decreased anxiety symptoms . | Decreased depressive symptoms . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total Sleep Time – IIM | ||||||||
Q1: Shorter (decrease ≥ 22 min) | Ref | 0.001 | Ref | 0.166 | Ref | 0.019 | Ref | 0.132 |
Q2-Q4: Unchanged | 1.08 [0.90 to 1.25] | 1.18 [0.99 to 1.36] | 1.46 [1.10 to 1.84] | 1.32 [0.97 to 1.72] | ||||
Q5: Longer (increase > 40 min) | 1.37 [1.17 to 1.52] | 1.16 [0.93 to 1.38] | 1.52 [1.09 to 1.98] | 1.08 [0.71 to 1.54] | ||||
Total sleep time – ISD | ||||||||
Q1: More irregular (increase > 44 min) | Ref | 0.007 | Ref | 0.117 | Ref | 0.359 | Ref | 0.014 |
Q2-Q4: Unchanged | 1.28 [1.09 to 1.45] | 0.89 [0.73 to 1.04] | 1.12 [0.86 to 1.41] | 1.12 [0.82 to 1.48] | ||||
Q5: More regular (decrease ≥ 54 min) | 1.33 [1.10 to 1.52] | 1.03 [0.84 to 1.20] | 0.95 [0.66 to 1.28] | 1.55 [1.13 to 1.98] | ||||
Circadian midpoint – IIM | ||||||||
Q1: Earlier (≥ 21 min) | Ref | 0.033 | Ref | 0.879 | Ref | 0.763 | Ref | 0.194 |
Q2-Q4: Unchanged | 0.98 [0.81 to 1.13] | 1.00 [0.83 to 1.17] | 1.03 [0.78 to 1.31] | 1.14 [0.84 to 1.49] | ||||
Q5: Later (> 22 min) | 1.18 [1.00 to 1.33] | 1.04 [0.83 to 1.24] | 0.94 [0.66 to 1.26] | 0.87 [0.57 to 1.26] | ||||
Circadian midpoint – ISD | ||||||||
Q1: More irregular (increase > 16 min) | Ref | 0.079 | Ref | 0.604 | Ref | 0.601 | Ref | 0.931 |
Q2-Q4: Unchanged | 1.19 [1.01 to 1.34] | 1.07 [0.89 to 1.23] | 0.91 [0.69 to 1.14] | 1.06 [0.77 to 1.39] | ||||
Q5: More regular (decrease ≥ 24 min) | 1.22 [1.00 to 1.40] | 1.11 [0.89 to 1.30] | 0.86 [0.61 to 1.15] | 1.03 [0.70 to 1.44] |
. | Decreased insomnia . | Decreased fatigue . | Decreased anxiety symptoms . | Decreased depressive symptoms . | ||||
---|---|---|---|---|---|---|---|---|
. | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . | RR [95%CI] . | P-value . |
Total Sleep Time – IIM | ||||||||
Q1: Shorter (decrease ≥ 22 min) | Ref | 0.001 | Ref | 0.166 | Ref | 0.019 | Ref | 0.132 |
Q2-Q4: Unchanged | 1.08 [0.90 to 1.25] | 1.18 [0.99 to 1.36] | 1.46 [1.10 to 1.84] | 1.32 [0.97 to 1.72] | ||||
Q5: Longer (increase > 40 min) | 1.37 [1.17 to 1.52] | 1.16 [0.93 to 1.38] | 1.52 [1.09 to 1.98] | 1.08 [0.71 to 1.54] | ||||
Total sleep time – ISD | ||||||||
Q1: More irregular (increase > 44 min) | Ref | 0.007 | Ref | 0.117 | Ref | 0.359 | Ref | 0.014 |
Q2-Q4: Unchanged | 1.28 [1.09 to 1.45] | 0.89 [0.73 to 1.04] | 1.12 [0.86 to 1.41] | 1.12 [0.82 to 1.48] | ||||
Q5: More regular (decrease ≥ 54 min) | 1.33 [1.10 to 1.52] | 1.03 [0.84 to 1.20] | 0.95 [0.66 to 1.28] | 1.55 [1.13 to 1.98] | ||||
Circadian midpoint – IIM | ||||||||
Q1: Earlier (≥ 21 min) | Ref | 0.033 | Ref | 0.879 | Ref | 0.763 | Ref | 0.194 |
Q2-Q4: Unchanged | 0.98 [0.81 to 1.13] | 1.00 [0.83 to 1.17] | 1.03 [0.78 to 1.31] | 1.14 [0.84 to 1.49] | ||||
Q5: Later (> 22 min) | 1.18 [1.00 to 1.33] | 1.04 [0.83 to 1.24] | 0.94 [0.66 to 1.26] | 0.87 [0.57 to 1.26] | ||||
Circadian midpoint – ISD | ||||||||
Q1: More irregular (increase > 16 min) | Ref | 0.079 | Ref | 0.604 | Ref | 0.601 | Ref | 0.931 |
Q2-Q4: Unchanged | 1.19 [1.01 to 1.34] | 1.07 [0.89 to 1.23] | 0.91 [0.69 to 1.14] | 1.06 [0.77 to 1.39] | ||||
Q5: More regular (decrease ≥ 24 min) | 1.22 [1.00 to 1.40] | 1.11 [0.89 to 1.30] | 0.86 [0.61 to 1.15] | 1.03 [0.70 to 1.44] |
Adjusted for age, sex, educational level, sleep diary duration, time between evaluations, and sleep complaints and mental health conditions at the pre-intervention interview.
Funding
This project was supported by the grants LABEX BRAIN (ANR-10-LABX-43), EQUIPEX PHENOVIRT (ANR-10-EQPX-12-01) and funding from the Region Nouvelle-Aquitaine (IS-OSA project, Contract No: 18000389).
Acknowledgments
We express all our thanks to all the institutions and professionals involved in developing the application (University of Bordeaux and especially Emeric Labbé) and in its dissemination (University of Bordeaux, Mutuelle Nationale des Hospitaliers MNH®, Fédération Hospitalière de France FHF).
Disclosure Statement
None declared.
Data availability
The data used in this research cannot be shared publicly for reasons of privacy. They can be shared on reasonable request to the corresponding author.
Comments