If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author: Megan E. Petrov, PhD, Edson College of Nursing and Health Innovation, Arizona State University, 500 N. 3rd St., Phoenix, AZ 85004, USA. Tel.: 1(602)496-2297.
Edson College of Nursing and Health Innovation, Arizona State University, Phoenix, Arizona, USACollege of Health Solutions, Arizona State University, Phoenix, Arizona, USA
Edson College of Nursing and Health Innovation, Arizona State University, Phoenix, Arizona, USACollege of Health Solutions, Arizona State University, Phoenix, Arizona, USA
Edson College of Nursing and Health Innovation, Arizona State University, Phoenix, Arizona, USACollege of Health Solutions, Arizona State University, Phoenix, Arizona, USA
To describe changes in sleep patterns during the coronavirus disease 2019 (COVID-19) pandemic, develop profiles according to these patterns, and assess sociodemographic, economic, COVID-19 related, and sleep and mental health factors associated with these profiles.
Design, setting, and participants
A 25-minute online survey was distributed worldwide through social media from 5/21/2020 to 7/1/2020.
Measurements
Participants reported sociodemographic/economic information, the impact of the pandemic on major life domains, insomnia and depressive symptoms, and changes in sleep midpoint, time-in-bed, total sleep time (TST), sleep efficiency (SE), and nightmare and nap frequency from prior to during the pandemic. Sleep pattern changes were subjected to latent profile analysis. The identified profiles were compared to one another on all aforementioned factors using probit regression analyses.
Results
The sample of 991 participants (ages: 18-80 years; 72.5% women; 60.3% residing outside of the United States) reported significantly delayed sleep midpoint, reductions in TST and SE, and increases in nightmares and naps. Over half reported significant insomnia symptoms, and almost two-thirds reported significant depressive symptoms. Latent profile analysis revealed 4 sleep pattern change profiles that were significantly differentiated by pre-pandemic sleep patterns, gender, and various COVID-19-related impacts on daily living such as severity of change in routines, and family stress and discord.
Conclusions
In an international online sample, poor sleep and depressive symptoms were widespread, and negative shifts in sleep patterns from pre-pandemic patterns were common. Differences in sleep pattern response to the COVID-19 crisis suggest potential and early targets for behavioral sleep health interventions.
Beginning in December, 2019, our global community experienced the effects of the novel SARS-CoV-2 virus, which resulted in the coronavirus disease 2019 (COVID-19) pandemic. Lifestyles and livelihoods have been upended due to numerous factors including the effects of the illness, disrupted routines and economic/educational opportunities, social isolation, multitasking of work and family responsibilities, and dynamic and abrupt shifts in government-led policies related to social distancing, quarantine, and other measures. These factors contribute to an unprecedented period of chronic, stressful, and traumatic experiences with likely impact on health-promoting behaviors including sleep.
Prevalence of psychiatric morbidity and psychological adaptation of the nurses in a structured SARS caring unit during outbreak: a prospective and periodic assessment study in Taiwan.
Their impact may be drastic, yet may depend on the intersection of prior sleep history, lifestyle, sociodemographic/socioeconomic influences, and regional/policy-level factors. Sleep problems in response to the COVID-19 pandemic are high in prevalence (17.4%-57.2%), as are psychological symptoms.
The enemy who sealed the world: effects quarantine due to the COVID-19 on sleep quality, anxiety, and psychological distress in the Italian population.
The enemy who sealed the world: effects quarantine due to the COVID-19 on sleep quality, anxiety, and psychological distress in the Italian population.
In one exception, a cross-sectional study of 3 European countries reported that despite mild increases in sleep duration and greater consistency between workdays and free days, sleep quality suffered during COVID-19-related lockdowns.
The present exploratory study extended the prior literature in an international sample by (1) describing multidimensional sleep pattern changes from before to during the COVID-19 pandemic, (2) classifying individuals into profiles based on sleep pattern changes, and (3) describing how these profiles differ based on sociodemographic, socioeconomic, COVID-19 pandemic-related impacts, and sleep behaviors and mental health outcomes.
Participants and methods
Study design
In this cross-sectional, exploratory online study, participants completed a 25-minute survey on how the COVID-19 pandemic, the resulting stay-at-home/quarantine, and/or social distancing may have affected sleep patterns, health behaviors, and well-being. The survey was available from 5/21/2020 to 07/01/2020, a period associated with COVID-19 case rate increases when either stay-at-home/quarantine measures or gradual economic reopening were occurring in many regions including the Americas, Middle East and South Asia, and parts of Africa.
Center for Systems Science and Engineering (CSSE) at Johns Hopkins University. COVID-19 dashboard. 2020. Accessed April 12, 2020. Available at: https://coronavirus.jhu.edu/map.html
The protocol was approved by Arizona State University Institutional Review Board, and participants provided electronic consent. Supplemental Table 1 displays our adherence to reporting guidelines for observational studies.
Participants and recruitment
All adults, ages ≥18, who could read and understand English were eligible to participate. Recruitment was conducted through: (1) primarily, paid, globally distributed advertisements through a Facebook/Instagram business account; (2) local institutional banner advertisements; and (3) word-of-mouth. The advertisements through Facebook/Instagram reached 2,278,399 individuals at least once and generated 86,520 clicks. Link clicks connected participants to the survey cover page, including the study description. Participants were asked to provide their email to enter a raffle to win one of five $25 (US) Amazon e-gift cards.
Measures
Sociodemographic and socioeconomic information
Participants provided age, gender (woman/man), ethnicity (Hispanic/Latinx; yes/no), race (White, Black, Asian, Other/Mixed), education (<college degree, bachelor's degree, advanced degree/masters), reduced work hours (yes, no, n/a), shiftwork status (yes/no); and country of residence. Participants reported whether their locality was “currently imposing stay-at-home/quarantine measures” with “yes,” “no,” or “the authorities stopped and/or relaxed stay-at-home quarantine measures in my region recently.”
Sleep measures
Impact of COVID-19 on Sleep Changes Scale (ICV19S). Participants retrospectively reported their sleep patterns that occurred prior to and during the pandemic (at the time of survey administration) using an adapted ICV19S. The ICV19S was provided by the Pennington Biomedical Research Center at Louisiana State University on the PhenX toolkit website operated by RTI International, a nonprofit research institute (https://www.phenxtoolkit.org/covid19/). Adaptations were investigator-created to be specific to the study aims. Per the original ICV19S, participants reported their typical bedtimes (hh:mm; “Prior to COVID-19 [Currently], what time did [do] you usually go to bed?”), and wake times (hh:mm; “Prior to COVID-19 [Currently], what time did [do] you usually wake up?”). These items are similar to items used in other studies.
The following items were added: sleep-onset latency (SOL, min; “Prior to COVID-19 [Currently], how long in minutes did it [has it] usually take[n] you to fall asleep, after you start trying to fall asleep?”), wake time after sleep onset (WASO, min; “Prior to COVID-19 [Currently], how much time, in minutes, did [do] you usually spend awake in between the time you first fall asleep and the time you wake up and start your day?”), and weekly number of naps and nightmares both prior to and during the pandemic in separate questions. The following variables were derived: time-in-bed (TIB = wake time–bedtime), nocturnal total sleep time (nTST = TIB–SOL–WASO), sleep efficiency (SE% = [nTST/TIB]*100%), and sleep midpoint in clock time ([TIB/2] + Bedtime).
Insomnia Severity Index (ISI). The ISI is a validated, 7-item index that assesses the current severity of sleep disturbances and distress over the past 2 weeks.
Sleep Hygiene and Practices Scale (SHPS). The SHPS is a 30-item questionnaire that assesses how often participants engage in sleep-related behaviors or experiences that impinge on sleep hygiene.
Responses range from “never” to “always” on a 6-point Likert scale (range: 30-180). The SHPS includes 4 subscales that examine behaviors related to homeostatic sleep drive and/or circadian sleep-wake rhythms, arousal-associated, eating/drinking, and the sleep environment. The 4 subscales are summed for a total score with greater values indicating greater frequency of non-sleep promoting behaviors.
Overall impact of COVID-19
Participants completed the Coronavirus Impact Scale (CIS), which focuses on the severity of disruption on a 4-point Likert scale (ie, no change, mild, moderate, severe) that the pandemic has had on 8 aspects of daily life including routines, family income/employment, food access, medical healthcare access, mental health treatment access, access to extended family and non-family social supports, experiences of stress related to the pandemic, and stress and discord in the family. The CIS was submitted to the COVID-19 protocols and research collection on the PhenX toolkit website (https://www.phenxtoolkit.org/toolkit_content/PDF/CIS_Stoddard.pdf), and registered with the National Institutes of Health Office of Behavioral Social Sciences Research suite of common instruments.
Depressive symptoms
Depressive symptoms were measured with the 10-item short-form of the Center for Epidemiological Studies–Depression Scale-10 (CES-D-10), which has been validated in the general, community populations and across low and middle-income countries.
Exploring the psychometric properties of the CES-D-10 and its practicality in detecting depressive symptomatology in 27 low- and middle-income countries.
Participants reported how often they experienced each symptom in the past week. Greater scores indicate greater presence of depressive symptoms with a clinical cutoff score of 10.
Statistical analysis
All measures were examined for normality, to identify invalid scores, and to ensure assumptions for subsequent analyses were met. Changes in the 6 sleep metrics (ie, sleep midpoint, TST, TIB, SE%, and nightmare and napping frequency) experienced from before to during the pandemic were examined with paired samples t-tests and repeated measures ANCOVA adjusted for sociodemographic and socioeconomic factors.
Analytic approach for latent profile analysis (LPA)
To form latent profiles of sleep pattern change, we computed the differences between pre-COVID-19 and during COVID-19 values for the 6 sleep metrics. Models having 2-5 latent profiles were fit with maximum likelihood estimation and robust standard errors. Each model was initially estimated using the default number of random starts (20), which was subsequently doubled to ensure that global maxima were replicated and to avoid erroneous local maxima. To assess model adequacy, we examined the Akaike Information Criterion, Bayesian Information Criterion, and sample-size adjusted Bayesian Information Criterion, with smaller values for each supporting a better-fitting model. We also examined values of entropy, which describes overall classification quality, with values ≥0.80 indicating excellent classification accuracy. In addition, we used the Lo-Mendell-Rubin adjusted likelihood ratio test and the Bootstrap Likelihood Ratio Test, with each testing the difference in fit of the current model to the fit of the model having one fewer profile. A statistically significant test indicates that the model with the larger number of profiles has better fit to the data. In addition, for the final model selected, none of the group sizes should be smaller than 5% of the total sample size, which is regarded as a sign of extracting too many classes or overfitting the data.
Masyn K. Latent class analysis and finite mixture modeling. In: Little ED, ed. The Oxford Handbook of Quantitative Methods in Psychology. Vol. 3. Oxford University Press (OUP); 2013:551-611.
Participants were classified into a given profile based on their most likely latent profile membership. We used Mplus software (Version 8.4) to estimate parameters with maximum likelihood estimation and robust standard errors, procedures that are robust to violations of normality and provide for optimal parameter estimates when response data are incomplete.
To examine associations between the sociodemographic/economic, and COVID-19 pandemic-related impact covariates and latent profile membership, we attempted to estimate a multinomial logistic regression model while simultaneously treating missing data on the covariates using a procedure analogous to the saturated correlates model. However, we encountered estimation difficulties and fit a similar model using Bayesian Markov Chain Monte Carlo estimation, which provides for unbiased parameter estimates for MAR data.
This analysis, implemented using uninformative prior distributions, allowed us to estimate a series of probit regression models comparing a given profile to each of the others. We monitored model convergence with the potential scale reduction factor, with a value less than 1.10 indicating convergence. Unlike traditional analyses, Bayesian estimation produces a distribution of values for each parameter, and we requested 24,000 random draws to build these posterior distributions (after 24,000 burn-iterations). The median of these posterior distributions was used to represent final parameter estimates, (ie, probit regression coefficients). Further, the 0.5th and 99.5th values from the distributions were used to form 99% Bayesian credibility intervals, which, when not containing a value of zero, is comparable to achieving statistical significance in traditional analyses with an alpha level of P < .01. We selected this more stringent confidence level (ie, 99%) to strike a reasonable balance between type 1 error control and statistical power. Note that, analogous to bootstrapping, the use of such intervals does not rely on distributional assumptions or large-sample theory. We also report standardized regression coefficients based on the latent variable or threshold probit model, which views membership in the 2 profiles being compared as a function of an underlying and unobserved continuous variable, for which we can compute the standard deviation.
These effect sizes are analogous to conventional standardized regression coefficients and are computed in the same manner, with the expressions for these calculations shown in a note under Table 4. The probit regression analyses were implemented via Mplus (Version 8.4).
Results
Of the 1493 participants who clicked on the study landing page, 502 participants completed none-to-few demographic variables. The total sample size was 991 participants who provided both demographic and sleep pattern change data. Sufficient data to compare completers vs non-completers were not available. Table 1 displays the sample's characteristics. The average respondent was ~38 years (range: 18-80). The sample was primarily women (72.5%), college educated (70.2%), reported White (41.1%) or Asian (39.9%) race, and resided outside of the United States (60.3%). Supplementary Table 2 provides the sample sizes for each of the 79 countries. The majority were currently experiencing or had recently been released from local quarantine orders (86.9%), had experienced moderate-to-severe changes in daily routines (86.4%), stress related to the COVID-19 pandemic (64.4%), and access to social support (51.0%), and reported clinical levels of depressive symptoms (65.0%) and insomnia symptoms (56.5%).
Table 1Descriptive characteristics of the COVID-19 and sleep survey sample (n = 991)
Participants from 79 countries provided data with the greatest representation in the sample as follows: United States = 393 (40.4%), India = 141 (14.5%), Pakistan = 96 (9.9%), Philippines = 44 (4.5%), Bangladesh = 20 (2.1%), Argentina = 15 (1.5%), Bolivia = 14 (1.4%), El Salvador = 14 (1.4%), Nicaragua = 14 (1.4%), Honduras = 13 (1.3%), Nepal = 11 (1.1%), Iran = 11 (1.1%), Egypt = 10 (1.0%).
USA All other participating countries
393 (39.7) 579 (58.4)
1.9
Currently under stay-at-home/quarantine orders? Yes No Authorities recently relaxed orders
381 (41.3) 121 (13.1) 421 (45.6)
6.9
Socioeconomic
Education < college degree Bachelor degree Advanced degree
295 (29.8) 335 (33.8) 360 (36.4)
0.1
Work hours reduced, yes No N/A (not working prior to the pandemic)
351 (35.6) 411 (41.7) 223 (22.6)
0.6
Current night or rotating shifts, yes
91 (9.2)
0.3
COVID-19 Impact Scale
Routines, no change Mild Moderate Severe
29 (3.1) 100 (10.5) 292 (30.8) 528 (55.6)
4.2
Family income/employment, no change Mild Moderate Severe
321 (33.8) 276 (29.1) 260 (27.4) 92 (9.7)
4.2
Food access, no change Mild Moderate Severe
339 (35.7) 438 (46.2) 145 (15.3) 26 (2.7)
4.3
Medical health care access, no change Mild Moderate Severe
334 (35.3) 265 (28.0) 259 (27.3) 89 (9.4)
4.4
Mental health Tx access, no change Mild Moderate Severe
567 (60.4) 161 (17.2) 120 (12.8) 90 (9.6)
5.3
Access to extended social support, no change Mild Moderate Severe
167 (17.6) 297 (31.3) 349 (36.8) 135 (14.2)
4.3
Experiences of stress, no change Mild Moderate Severe
55 (5.8) 283 (29.8) 389 (41.0) 222 (23.4)
4.2
Stress and discord in the family, no change Mild Moderate Severe
278 (29.3) 437 (46.0) 185 (19.5) 49 (5.2)
4.2
Sleep and mental health outcomes
CES-D-10 score (range: 0-30) CES-D-10 cutoff score of ≥ 10
CES-D-10 = Center for Epidemiologic Studies Depression Scale 10-item short form; ISI = Insomnia Severity Index; SHPS = Sleep Hygiene and Practices Scale.
a Participants from 79 countries provided data with the greatest representation in the sample as follows: United States = 393 (40.4%), India = 141 (14.5%), Pakistan = 96 (9.9%), Philippines = 44 (4.5%), Bangladesh = 20 (2.1%), Argentina = 15 (1.5%), Bolivia = 14 (1.4%), El Salvador = 14 (1.4%), Nicaragua = 14 (1.4%), Honduras = 13 (1.3%), Nepal = 11 (1.1%), Iran = 11 (1.1%), Egypt = 10 (1.0%).
Table 2 displays the sleep patterns prior to and during the COVID-19 pandemic. There were significant changes in all sleep metrics except TIB. The sample went to bed over an hour later, slept less at night by ~43 minutes, experienced a 9% reduction in SE%, and increased weekly nightmares (1.4) and naps (0.9). These significant changes were attenuated but remained after adjusting for sociodemographic/economic factors except for TST.
Table 2Sleep change characteristics of the sample, unadjusted paired samples t-tests, and adjusted repeated measures ANCOVAs comparing sleep patterns before and during the COVID-19 pandemic
Adjusted for age, gender, race, ethnicity, country (United States vs all other countries), quarantine status, education, hours of work reduced status, night shift status.
IQR = interquartile range; Md = mean difference of paired data; SE = Sleep efficiency; TST = Total sleep time; TIB = Time in bed.
a TIB was calculated by taking the difference in hours and minutes between reported wake-up time and reported bedtime.
b TST was calculated by subtracting sleep onset latency and wake after sleep onset from time in bed (TIB).
c SE percentage was calculated as (TST/TIB)*100%.
d Bolded statistical values indicate statistically significant change in a given sleep metric at an alpha level of <0.05.
e Adjusted for age, gender, race, ethnicity, country (United States vs all other countries), quarantine status, education, hours of work reduced status, night shift status.
Supplementary Table 3 presents the statistical indices and tests used to determine the number of profiles underlying responses to the indicators. We discarded profile 5 (below the 5% size threshold). Statistical tests indicated a superior fit for the 4-profile model across the various indices, compared to the 3-profile model. Entropy suggested that each of the models yielded highly discriminating profiles.
Table 3 displays the mean differences and ranges for each of the 6 sleep metrics within each profile. Based upon these values, we named the profiles according to their most salient characteristics (Fig. 1) as follows: Delayed Sleep (1), Dysregulated & Distressed (2), Sleep Opportunist (3), and Sleep Lost & Fragmented (4).
Table 3Four-profile results with mean differences (min, max) in each sleep variable from prior to during the COVID-19 pandemic
N = 987. Means and standard deviations for variables across all profiles: Difference in Sleep Midpoint in minutes M = 66.19 (SD = 116.84), Difference in TST in minutes M = −42.42 (SD = 130.90), Difference in SE% M = −9.15 (SD = 17.15), Difference TIB minutes M = −0.72 (SD = 112.80), Difference in Nightmares per week M = 1.35 (SD = 3.25), Difference in Naps per week M = 0.90 (SD = 3.22).
SE = sleep efficiency; TST = total sleep time; TIB = time in bed.
P < .05.
P < .01.
a N = 987. Means and standard deviations for variables across all profiles: Difference in Sleep Midpoint in minutes M = 66.19 (SD = 116.84), Difference in TST in minutes M = −42.42 (SD = 130.90), Difference in SE% M = −9.15 (SD = 17.15), Difference TIB minutes M = −0.72 (SD = 112.80), Difference in Nightmares per week M = 1.35 (SD = 3.25), Difference in Naps per week M = 0.90 (SD = 3.22).
b Profile size is based on an individual's most likely latent class membership as estimated by the 4-profile model.
Fig. 1Change in sleep pattern means by latent profile for scale scores. Of the 987 participants, Profile 1 comprised 64.1% (n = 633), Profile 2 comprised 5.4% (n = 53), Profile 3 comprised 11.1% (n = 110), and Profile 4 comprised 19.4% (n = 191) of the sample. SE = sleep efficiency %; TIB = time in bed in minutes; TST = total sleep time in minutes.
The Delayed Sleep (1) was the most common profile and was characterized by a later sleep midpoint without substantially compromising their original TIB and TST. Yet, they experienced a mild reduction in SE and increases in nightmares and naps. The Dysregulated & Distressed (2) profile had the smallest sample size and was characterized by substantial TIB, TST, and SE loss, increases in naps, and heightened nightmares. The Sleep Opportunist (3) profile significantly increased their TIB, TST, and SE from a prior state of mild-to-moderate sleep deprivation and fragmentation. The Sleep Lost & Fragmented (4) profile lost TST by delaying and reducing their TIB, and with accompanying reductions in SE, heightened nightmares, yet minimal nap compensation.
Table 4 displays probit regression model coefficients that compare each profile to another on the “baseline” sleep patterns, each sociodemographic/economic factor, COVID-19 pandemic impacts on daily life, sleep outcomes and behaviors, and mental health factors. Positive coefficients correspond to a greater likelihood of belonging to the “second profile” in the comparison (or smaller likelihood of belonging to the first profile), whereas negative coefficients indicate a smaller likelihood of belonging to the second profile (or greater likelihood of belonging to the first profile). Note that excessive multicollinearity (variance inflation factor [VIF] = 23.3) was found with the inclusion of the baseline TIB variable. Given that TIB did not substantially shift, on average, in the sample, we elected to remove prior TIB from the model to resolve the multicollinearity (largest VIF = 3.2 with TIB removal).
Table 4Covariate results for the 4-profile model (N = 970)
Values shown are probit standardized regression coefficients, which are computed as β = (b*sdx)/sdy for numeric predictors and β = b/sdy for binary predictors.
99% Bayesian credibility interval does not include 0.
−0.04
−0.15
CIS = Coronavirus Impact Scale; SE = sleep efficiency %; TST = total sleep time; Female = 1 indicates participant is female and 0 = male; Country = 1 indicates participant does not reside in the United States and 0 = otherwise; Night shift =1 indicates participant works a night shift and 0 = otherwise. Ethnicity = 1 indicates participant reported being Hispanic in origin.
p < 0.01 values meeting significance are bolded.
99% Bayesian credibility interval does not include 0.
a Values shown are probit standardized regression coefficients, which are computed as β = (b*sdx)/sdy for numeric predictors and β = b/sdy for binary predictors.
b Profile 1 (Delayed Sleep, n = 622), Profile 2 (Dysregulated & Distressed, n = 51), Profile 3 (Sleep Opportunist, n = 109), Profile 4 (Sleep Lost & Fragmented, n = 188).
Results indicated that there were 22 significant differences found in all of the profile comparisons with 8 involving comparisons between Dysregulated & Distressed (2) and Sleep Opportunist (3). Specifically, when examining the sleep patterns experienced before the pandemic, findings indicated that members of the Dysregulated & Distressed (2) had significantly later midpoint sleep times and greater TST (and therefore the most to lose) than all other profiles. The Sleep Opportunist (3) profile had the least TST (and therefore the most to gain) compared to all profiles, and had significantly greater SE than Dysregulated & Distressed (2) and Delayed Sleep (1), but not Sleep Lost and Fragmented (4). The Sleep Opportunist (3) members also experienced more pre-pandemic nightmares and naps per week relative to the Dysregulated & Distressed (2). Lastly, the Sleep Lost and Fragmented (4) profile had significantly more naps relative to Delayed Sleep (1).
There were no significant differences between the profiles on the sociodemographic/economic factors except for gender. A greater proportion of women belonged to the Sleep Lost & Fragmented (4) profile relative to the Sleep Opportunist profile (3). There were few significant differences between the profiles on severity of change in aspects of daily life. Members of the Sleep Opportunist (3) were more likely to report greater change in daily routines relative to the Dysregulated & Distressed (2). There were no differences across the profiles in engagement in maladaptive behaviors via SHPS scores. However, the Dysregulated & Distressed (2) had greater ISI scores than all other profiles followed by, in order of severity, Sleep Lost & Fragmented (4), Delayed Sleep (1), and Sleep Opportunist (3). The latter 2 profiles did not differ. The sleep profiles did not differ substantially on the mental health variables. CES-D-10 scores, and severity of change in mental healthcare access and pandemic-related stress did not differ across the profiles. However, members of the Sleep Opportunist (3) were more likely to report greater family stress and discord relative to the Dysregulated & Distressed (2).
Discussion
This study was one of few that has assessed an international sample for sleep patterns prior to the pandemic, and identified and classified specific sleep responses to an unprecedented crisis. Overall, sleep patterns significantly deteriorated from patterns experienced prior to the pandemic. Further, clinically meaningful insomnia and depressive symptoms were widespread.
The results are similar to other COVID-19 studies assessing sleep in the general population within nations. The prevalence of insomnia symptoms and other sleep disturbances were high but within the range reported previously (56.5% vs 17.4˗57.2%).
Immediate psychological responses and associated factors during the initial stage of the 2019 coronavirus disease (COVID-19) epidemic among the general population in China.
However, these studies gathered estimates early in the pandemic when socioeconomic and health-related stressors were not yet prolonged, or the full impact not realized. Further, the definitions of significant depressive symptoms varied. Research on stressful traumas and emerging infectious diseases suggest that sleep and mental health implications can be sustained beyond the acute stressor(s) themselves and their immediate consequences and contribute to long-term sub-clinical to poor health outcomes.
Our findings coupled with prior studies suggest that the sleep and psychological fall-out beyond the worst periods of this pandemic may be substantial, reemphasizing the need for prompt, coordinated policy and program responses.
Four distinct profiles of sleep pattern changes were identified: Delayed Sleep (1), Dysregulated and Distressed (2), Sleep Opportunist (3), and Sleep Lost & Fragmented (4). The major differentiators between the profiles were how well members of each profile were sleeping prior to the pandemic and their insomnia severity during the pandemic. For example, the improved sleep experience of Sleep Opportunist (3) members from a pre-pandemic sleep pattern characterized by sleep restriction and probable associated napping and nightmares, suggest that quarantine/stay-at-home measures may have led to an opportunity to adopt sleep and/or work patterns more closely aligned with their endogenous sleep-wake cycle and needs. The Dysregulated & Distressed (2) and the Sleep Lost & Fragmented (4) both had poor sleep responses to the pandemic characterized by heightened insomnia severity and precipitous drops in SE and TST; however, profile 2 reported substantially better sleep pre-pandemic and therefore experienced the greatest relative deterioration. The Delayed Sleep (1) profile delayed their sleep opportunity and had mild disruptions to their sleep. Beyond the decided differences in sleep patterns prior to the pandemic and sleep outcomes between the profiles, few other factors differentiated the profiles.
Our 4-profile solution is similar to the 3-cluster profile solution identified in a study from Canada.
This study also identified a cluster similar to the Sleep Opportunist (3) profile that they named the “Extended Time in Bed,” of which members of this profile also appeared to respond with improved sleep through TIB extension. This study also identified a “Delayed Sleep” cluster similar to Delayed Sleep (1), and a “Reduced Time in Bed” cluster, which most closely resembled Sleep Lost & Fragmented (4)’s sleep pattern changes pre to during pandemic and sleep quality deterioration. Notably, this “Reduced Time in Bed” cluster also had a larger proportion of women as did Sleep Lost & Fragmented (4), relative to Sleep Opportunist (3). This profile, though not the most severe in sleep deterioration, had substantial and clinically meaningful changes in sleep patterns. This finding is consistent with prior studies in Italy and China that found that women's sleep appeared to be most negatively affected by the pandemic.
The enemy who sealed the world: effects quarantine due to the COVID-19 on sleep quality, anxiety, and psychological distress in the Italian population.
The enemy who sealed the world: effects quarantine due to the COVID-19 on sleep quality, anxiety, and psychological distress in the Italian population.
Collectively, this profile was most characterized by women with relatively normal sleep patterns prior to the pandemic that responded with significant sleep deterioration and distress. These results are consistent with extensive literature on the disproportionate burdens of women both psychologically and economically, which are often paired with multiple roles as worker and primary caretaker of family. This already present gender inequity is likely exacerbated by the pandemic,
and may be reflected in sleep deterioration discovered in this study.
The Dysregulated & Distressed (2) profile was characterized by the greatest sleep deterioration. Given the small sample size, the generalizability of this profile is questionable. However, this profile should not be discounted. The severe acuity of this unique response to global trauma indicates these individuals are in need, likely highly vulnerable, and may represent a snapshot of a larger heterogeneous population varying in the circumstances that make them vulnerable to poor sleep health, but nonetheless have a similar response to those circumstances.
Considering the intimate connection between sleep disturbances and depressive symptoms, it is interesting that insomnia severity differentiated the sleep profiles, but not depressive symptoms. One interpretation is that in the present crisis experiencing depressive symptoms is relatively universal no matter prior sleep or mental health history, and one's “sleep pattern response” to the pandemic has less to do with psychological distress and more so with physical and social changes in daily living. However, one's sleep outcomes have much to do with “sleep pattern response.” That is, a temporal model would suggest that the pandemic was met with psychological distress and shift in daily living, but the variety of shifts in daily living prompted the development of sleep pattern responses and, consequently, differing sleep quality response. If such a model were verified, then interventions supporting healthy responses to crisis-related changes in daily routines may be particularly important.
Due to the novelty of the pandemic, it was critical to identify emerging trends in multidimensional sleep health among an international and racially/ethnically diverse sample to better inform future interventions and research. Although the timeliness of this study can tell us much about sleep response to the COVID-19 pandemic and perhaps to other similar global crises, there are several limitations to note. First, this was an explorative, cross-sectional study assessing self-reported, retrospective sleep health. Recall of sleep patterns prior to the pandemic may be skewed by memory and present emotional biases as acknowledged by other similar studies.
Further, the sleep pattern measure was adapted from another questionnaire. However, the adaptation was necessary to extract meaningful data unique to the global context at the time. Second, the worldwide-distributed, online survey was limited to one social media platform and offered only in English. Thus, this study used non-probability, real-time sampling which is subject to topic self-selection bias, and other sociodemographic/economic biases often found in online surveys such as greater completion by women and younger adults.
Although probability sampling is superior, non-probability samples are commonly employed in response to emergent crises including the COVID-19 pandemic,
and they do offer several benefits. One benefit is the ability to collect preliminary, rapid results that match the urgency of the evolving pandemic landscape and its impact on health behaviors that may inform future health policy. Without such an approach, the potential impact of COVID-19 on sleep behaviors and patterns may have remained unknown. Further, probability sampling also cannot garner results that accurately generalize beyond the time frame of assessment. Lastly, because of our use of a couple online recruitment strategies, we are unable to determine the actual reach of our recruitment to calculate the recruitment rate and/or explore the reasons for our sample's composition. Overall, our findings are preliminary and warrant future longitudinal research to confirm.
Conclusions
Sleep health interventions in response to pandemics and other crises are limited, as is research on such interventions. During the COVID-19 pandemic, relevant societies developed recommendations for how to avoid sleep disturbances and promote sleep health.
Dealing with sleep problems during home confinement due to the COVID-19 outbreak: practical recommendations from a task force of the European CBT-I academy.
Such messaging is yet to be widely disseminated, and its implementation formally evaluated. Certain groups are testing solutions to screening and addressing acute sleep complaints using smartphone applications and web-based approaches.
These delivery methods of current, evidence-based behavioral sleep interventions are particularly relevant during pandemics and are generally accessible. Our results may assist in identifying at-risk groups, and facilitating the development of personalized intervention that target distinct sleep profiles, which may vary in the intensity of therapy required within a stepped care model and depend upon the chronicity of the sleep problems. Based upon our results, the dissemination and implementation of such interventions in future research could consider adapting according to demographic and household/family factors, and sleep history.
Declaration of conflict of interest
The authors do not have any conflicts of interest to disclose.
Prevalence of psychiatric morbidity and psychological adaptation of the nurses in a structured SARS caring unit during outbreak: a prospective and periodic assessment study in Taiwan.
The enemy who sealed the world: effects quarantine due to the COVID-19 on sleep quality, anxiety, and psychological distress in the Italian population.
Center for Systems Science and Engineering (CSSE) at Johns Hopkins University. COVID-19 dashboard. 2020. Accessed April 12, 2020. Available at: https://coronavirus.jhu.edu/map.html
Exploring the psychometric properties of the CES-D-10 and its practicality in detecting depressive symptomatology in 27 low- and middle-income countries.
Masyn K. Latent class analysis and finite mixture modeling. In: Little ED, ed. The Oxford Handbook of Quantitative Methods in Psychology. Vol. 3. Oxford University Press (OUP); 2013:551-611.
Immediate psychological responses and associated factors during the initial stage of the 2019 coronavirus disease (COVID-19) epidemic among the general population in China.
Dealing with sleep problems during home confinement due to the COVID-19 outbreak: practical recommendations from a task force of the European CBT-I academy.
P < .05.
00114-5&id=gr1.jpg){kind=link}