Medical Health Cluster

1 junio, 2023

Multidimensional Sleep Health Prior to SARS-CoV-2 Infection and Risk of Post–COVID-19 Condition

Key Points

Question  Is healthy sleep, both before and during the COVID-19 pandemic, prior to SARS-CoV-2 infection, protective against post–COVID-19 condition (PCC)?

Findings  In this cohort study of 1979 women who reported testing positive for SARS-CoV-2, adherence to healthy sleep before infection was inversely associated with the risk of PCC. Compared with women who had a prepandemic sleep score of 0 or 1 (least healthy), those who scored 5 (most healthy) had a 30% lower risk of PCC.

Meaning  These findings suggest that preinfection healthy sleep may be associated with a substantially decreased risk of PCC, indicating a need for further research on sleep health to prevent or improve symptoms of PCC.


Importance  The association of multiple healthy sleep dimensions with post–COVID-19 condition (PCC), also known as long COVID, has not been investigated.

Objective  To examine whether multidimensional sleep health before and during the COVID-19 pandemic, prior to SARS-CoV-2 infection, was associated with the risk of PCC.

Design, Setting, and Participants  This prospective cohort study (2015-2021) included Nurses’ Health Study II participants who reported testing positive (n = 2303) for SARS-CoV-2 infection in a substudy series of COVID-19–related surveys (n = 32 249) between April 2020 and November 2021. After exclusion for incomplete information about sleep health and nonresponse to a question about PCC, 1979 women were included in the analysis.

Exposures  Sleep health was measured both before (June 1, 2015, to May 31, 2017) and early (April 1 to August 31, 2020) in the COVID-19 pandemic. Prepandemic sleep score was defined according to 5 dimensions: morning chronotype (assessed in 2015), 7 to 8 hours of sleep per day, low insomnia symptoms, no snoring, and no frequent daytime dysfunction (all assessed in 2017). On the first COVID-19 substudy survey (returned between April and August 2020), average daily sleep duration and sleep quality for the past 7 days were queried.

Main Outcomes and Measures  SARS-CoV-2 infection and PCC (≥4 weeks of symptoms) were self-reported during 1 year of follow-up. Comparisons were examined between June 8, 2022, and January 9, 2023, using Poisson regression models.

Results  Of the 1979 participants reporting SARS-CoV-2 infection (mean [SD] age, 64.7 [4.6] years; 1979 [100%] female; and 1924 [97.2%] White vs 55 [2.8%] other races and ethnicities), 845 (42.7%) were frontline health care workers, and 870 (44.0%) developed PCC. Compared with women who had a prepandemic sleep score of 0 or 1 (least healthy), those who scored 5 (most healthy) had a 30% lower risk of developing PCC (multivariable-adjusted relative risk, 0.70; 95% CI, 0.52-0.94; P for trend <.001). Associations did not differ by health care worker status. No or little daytime dysfunction prepandemic and good sleep quality during the pandemic were independently associated with a lower risk of PCC (relative risk, 0.83 [95% CI, 0.71-0.98] and 0.82 [95% CI, 0.69-0.99], respectively). Results were similar when PCC was defined as having 8 or more weeks of symptoms or as having ongoing symptoms at the time of PCC assessment.

Conclusions and Relevance  The findings indicate that healthy sleep measured prior to SARS-CoV-2 infection, both before and during the COVID-19 pandemic, may be protective against PCC. Future research should investigate whether interventions on sleep health may prevent PCC or improve PCC symptoms.


Post–COVID-19 condition (PCC), also known as long COVID, defined as having persistent COVID-19–related symptoms more than 4 weeks after infection,1 affects 20% to 70% of persons infected with SARS-CoV-2.2-8 Post–COVID-19 condition encompasses a wide range of symptoms that may impair daily functioning, including persistent cough, depression, fatigue, anosmia, and brain fog.9,10 Notwithstanding the serious public health burden caused by PCC, the etiology, prevention, and treatment of PCC are largely unknown.

Sleep problems are prevalent yet commonly overlooked health risk factors, affecting approximately 1 in 3 people worldwide.11,12 Sleep deprivation, poor sleep quality, and sleep disorders may lead to adverse physical, cognitive, and psychiatric consequences.12 The COVID-19 pandemic compounded both the severity of preexisting sleep disorders and the high prevalence of sleep disturbances.13,14 Unhealthy sleep dimensions individually and in combination (eg, late chronotype, short or long duration of sleep, snoring and sleep apnea, and daytime sleepiness) are prospectively associated with COVID-19 susceptibility, severity, and mortality.15-18 Mendelian randomization analyses have further identified insomnia, extreme sleep duration, and excessive daytime sleepiness as causal risk factors for the severity of COVID-19 and other respiratory infections.16,17,19 These unhealthy sleep dimensions have been linked to chronic low-grade inflammation and immune abberations,12,20,21 which have been implicated in the pathogenesis of PCC and other postinfection syndromes, such as postviral fatigue syndrome.22-24 Despite experimental evidence supporting the role of sleep deprivation and disturbances in infectious disease prognosis via immune dysfunction,20,25 the association between sleep health prior to SARS-CoV-2 infection and risk of PCC has not been investigated.

Leveraging data from a large longitudinal study, the Nurses’ Health Study II (NHS II), we prospectively evaluated the association of multidimensional sleep health prior to infection, measured before and early in the pandemic, with the risk of developing PCC among participants subsequently infected with SARS-CoV-2. We hypothesized that healthier sleep is associated with a lower risk of PCC.

Study Design and Population

In this cohort study, participants were drawn from the NHS II, which was established in 1989 when 116 429 US female nurses (aged 25-42 years) were enrolled.26 Health-related information was updated biennially via questionnaire. Among active NHS II participants, a web-based COVID-19 substudy was launched in April 2020 (henceforth termed baseline). Nurses’ Health Study II participants for whom we had an email address, who had returned the most recent main cohort questionnaire, and who were not currently participating in another substudy were invited (55 925 invited, 39 137 responded [71%]). The follow-up was monthly and quarterly. The final substudy follow-up questionnaire (henceforth termed final questionnaire) was administered 12 months after baseline. The end of follow-up for our analyses was November 3, 2021 (eMethods in Supplement 1).

Among 32 249 substudy participants who returned both the baseline and final questionnaires, 2303 reported a positive SARS-CoV-2 test during the 19 months of follow-up. Of these participants, we excluded 91 who did not have complete information about sleep health and 233 who did not respond to a question about PCC, leaving 1979 in the analysis (eFigure 1 in Supplement 1). The study was approved by the Brigham and Women’s Hospital institutional review board. Return of questionnaires constituted implied consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Assessment of Sleep Dimensions

We measured sleep health before the pandemic (5 dimensions; June 1, 2015, to May 31, 2017, from which we created a healthy sleep score) and during the pandemic (2 dimensions; April 1 to August 31, 2020). Five prepandemic sleep dimensions were queried in 2015 (chronotype) and 2017 (sleep duration, insomnia, snoring, and daytime dysfunction). To assess chronotype, participants were asked to classify themselves as morning or evening people (“One hears about morning and evening types of people. Which one of these types do you consider yourself to be: definite morning, more of a morning, more of an evening, definite evening, or neither?”). This measure has been validated against the criterion standard human circadian marker, dim light melatonin onset.27,28 Daily sleep duration was the average number of sleep hours over a 24-hour period. Spearman correlation between self-reported sleep duration and a sleep diary was 0.79 (P < .001).29 Insomnia was assessed using a validated insomnia rating scale. Participants were asked how frequently they (1) have trouble falling asleep, (2) wake up several times at night, (3) wake up earlier than planned, and (4) have trouble getting back to sleep after waking too early.30 For each question, response options ranged from no (score 0) to 5 or more times per week (score 4). Items were summed to form an insomnia score (range, 0-16), where a score of less than 9 was considered low insomnia symptoms.31 For snoring, participants were asked, “How often do you snore?” Self-report of snoring had high sensitivity (90%) in detecting sleep apnea.32 Regarding daytime dysfunction, participants were asked to what extent daily function (eg, fatigue, mood, ability to work, concentration, memory) was affected by not being well-rested, with options from not at all to very much. This question has been used as a proxy for sleep quality.33

We calculated a prepandemic (2015-2017) sleep score: morning chronotype (definite morning or more of a morning person), adequate sleep (7-8 hours per day), low insomnia symptoms (0-8), no self-reported snoring, and no or little daytime dysfunction. Participants received a score of 1 if they met the criterion and 0 otherwise. For each participant, we calculated the total score by summing healthy sleep dimensions (0-5; higher score indicates healthier sleep). Because only 33 participants scored 0, we combined 0 and 1 for the analysis. A similar sleep score has been associated with health benefits, including lower risk of heart disease and longevity.21,34-36

At COVID-19 substudy baseline, participants were asked about their average daily sleep duration (in hours) and sleep quality (from very bad to very good) in the past 7 days. To jointly capture sleep health both before (2015-2017) and early in the pandemic (April to August 2020), we created a variable in 6 levels. Prepandemic sleep was categorized as either unhealthy (score of 0-3) or healthy (score of 4-5),21 and during-pandemic sleep was a count of healthy sleep duration (7-8 hours per day vs other) and healthy sleep quality (fairly good or very good vs other), which ranged from a score of 0 to 2.

SARS-CoV-2 Infection, Hospitalization, and PCC

Participants reported SARS-CoV-2 infection (confirmed by polymerase chain reaction, antigen, or antibody test), date of test, and hospitalization due to COVID-19 since return of the last questionnaire or, at baseline, since March 1, 2020, on each substudy questionnaire. During the study period, there were 5 major peaks in the US, with dominant variants being Alpha and Delta.37

Ascertainment of PCC in this cohort has been described.38 Briefly, on the final COVID-19 substudy questionnaires, participants were asked whether they had COVID-19–related symptoms lasting for more than 4 weeks; anyone who endorsed this question was considered to have PCC. Of those with PCC, we asked (1) which symptoms they had experienced (fatigue; shortness of breath or difficulty breathing; persistent cough; muscle, joint, or chest pain; smell or taste problems; confusion, disorientation, or brain fog; memory issues; depression, anxiety, or changes in mood; headache; intermittent fever; heart palpitations; rash, blisters, or welts; mouth or tongue ulcers; and other), (2) whether symptoms were ongoing, and (3) duration of symptoms.


Covariates were selected a priori as potential risk factors for PCC. Date of birth, race and ethnicity (American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Pacific Islander, or White), and height were self-reported at cohort enrollment. We included race and ethnicity as a study variable because of the known racial and ethnic disparity in sleep health and PCC.39,40 Diet, including alcohol consumption, was measured in 2015 using a validated semiquantitative food frequency questionnaire.41,42 We adapted the Alternate Healthy Eating Index-2010 score to describe diet quality. Weight, physical activity, smoking history, subjective complaints in memory and cognition,43,44 and lifetime history of physician-diagnosed diseases (cancer, diabetes, hypertension, high cholesterol, and asthma) were assessed in 2017. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Frontline health care worker status (defined as physically working at a site delivering care or not) was self-reported at baseline. Depression and anxiety symptoms at baseline were measured using the Patient Health Questionnaire-2 and the Generalized Anxiety Disorder 2-item, respectively.45 Date of COVID-19 vaccination was self-reported on quarterly follow-up surveys.

Statistical Analysis

We first compared the prevalence of demographic and lifestyle factors according to missingness of exposure and outcome. We then calculated the correlation (ϕ-coefficient) between dimensions of sleep health. Missingness for each covariate was less than 3%. Indicator variables were used for any missing covariate for categorical variables; the median was imputed for continuous variables.46

Poisson regression models were used to estimate the relative risks (RRs) and 95% CIs for the associations of individual sleep dimensions (both as categorical and dichotomized variables) and sleep score (0-5) with risk of PCC, with adjustment for age (continuous, years), race and ethnicity (White vs American Indian or Alaska Native, Asian, Black or African American, or Native Hawaiian or Pacific Islander), frontline health care worker status (yes, no), lifestyle factors (BMI [continuous], physical activity [continuous, metabolic equivalent task-hours per day], smoking [never, past, or current], diet [quintiles, Alternate Healthy Index-2010], and alcohol consumption [0, 0.1-4.9, 5.0-14.9, 15.0-29.9, or ≥30.0 g per day]), and history of comorbidities (cancer, diabetes, hypertension, high cholesterol, and asthma [yes, no]). Because sleep dimensions are correlated with each other,21 we included a model mutually adjusted for all prepandemic healthy sleep dimensions to explore their independent associations with PCC. The linear trend test was performed by treating indicator levels as a continuous variable. Test for trend for sleep duration was not performed because of the well-established U-shaped relationship between sleep duration and health (with 7-9 hours per day being the healthiest level).47

Because frontline health care workers have a higher risk of infection and potentially different sleep behavior due to their job characteristics,48 we investigated whether associations differed by frontline health care worker status by adding a cross-product term in the model. The significance of multiplicative interaction was tested using the Wald test.

We also examined the joint association of sleep health before (2015-2017) and early (April to August 2020) in the pandemic with subsequent risk of PCC. To reduce risk of reverse causation, we excluded 174 participants who reported a positive SARS-CoV-2 test on the baseline questionnaire when pandemic sleep dimensions were assessed.

We conducted several sets of sensitivity analyses. First, alternative definitions of PCC were used. Post–COVID-19 condition was defined as having symptoms longer than 8 weeks and, to reduce risk of recall bias, as having ongoing symptoms at the time of PCC assessment. Because symptoms of sleep deprivation and PCC may overlap, we excluded from PCC cases 155 participants who reported only fatigue, headache, brain fog, memory issues, and depression as their symptoms. To account for the possibility of lack of testing, 912 participants with presumed SARS-CoV-2 infection (without a confirmatory test) were included in the analytic sample. Because vaccination may reduce the risk of PCC,5 we excluded 129 participants who had been vaccinated before infection. To account for preexisting PCC-like symptoms, we additionally adjusted for memory and cognitive issues (the summed score of 6 subjective memory and cognitive complaints) assessed in 2017.43 To explore whether the associations were similar among participants who had less severe acute phase illness, we restricted analysis to 1877 persons who were not hospitalized due to COVID-19. Because sleep disorders have been linked with an increased risk of mental health conditions and mental health has been associated with the risk of PCC,38 we additionally excluded 336 persons with probable anxiety and depression, as measured by the Patient Health Questionnaire-2 and Generalized Anxiety Disorder 2-item, at baseline.45 Finally, multiple imputation was used for missing data.49 All analyses were conducted between June 8, 2022, and January 9, 2023, using SAS, version 9.4 software (SAS Institute Inc). A 2-sided P < .05 was considered statistically significant.


The analytic sample comprised 1979 women (American Indian or Alaska Native, 9 [0.5%]; Asian, 22 [1.1%]; Black or African American, 22 [1.1%]; Native Hawaiian or Pacific Islander, 2 [0.1%]; and White, 1924 [97.2%]), of whom 845 (42.7%) were frontline health care workers. The mean (SD) age of the cohort was 64.7 (4.6) years (range, 55–75 years). Participants with complete exposure and outcome information (n = 1979) vs without (n = 324) were more likely to be White and frontline health care workers (eTable 1 in Supplement 1). At baseline, participants who had a healthier prepandemic (2015-2017) sleep score were more likely to be older and White, have a lower BMI, and adhere to a healthier lifestyle (high diet quality score, high physical activity level, and never smoking) and less likely to have type 2 diabetes, hypertension, and asthma (Table 1). Compared with non–frontline health care workers, frontline health care workers were younger and less likely to have comorbidities (eTable 2 in Supplement 1). Sleep dimensions were weakly to moderately correlated with each other (ϕ-coefficient range, −0.02 to 0.32) (eFigure 2 in Supplement 1).

A total of 870 participants (44.0%) reported having PCC symptoms. Healthier prepandemic sleep score showed a dose-response association with lower risk of PCC (Figure 1). Compared with participants who had a sleep score of 0 or 1, those who scored 5 had a 30% lower risk of developing PCC (RR, 0.70; 95% CI, 0.52-0.94; P for trend <.001). The association between sleep score and PCC did not differ by health care worker status (eTable 3 in Supplement 1).

Of the 5 sleep dimensions assessed prior to the pandemic (2015-2017), chronotype, sleep duration, insomnia, and daytime dysfunction were associated with risk of PCC after adjusting for age and race and ethnicity (Table 2, model 1). Sleep quality early in the pandemic was associated with risk of PCC (Table 2, model 1). Sleep durations of 6, 7, 8, 9, and 10 or more hours were not significantly associated with a lower risk of PCC vs 5 hours or less. The strongest associations were found for daytime dysfunction before the pandemic and sleep quality during the pandemic. Associations were attenuated by 0% to 13% after further adjustment for health risk factors (Table 2, model 2). Sleep duration, insomnia, and daytime dysfunction prior to the pandemic and sleep quality early in the pandemic remained significantly associated with risk of PCC. Similar associations were observed when each sleep dimension was evaluated as a binary variable (eTable 4 in Supplement 1; Figure 2). When we included all prepandemic sleep dimensions in the same model, only daytime dysfunction was independently associated with a lower risk of PCC (RR, 0.83; 95% CI, 0.71-0.98) (Figure 2). Similarly, sleep quality during the pandemic was associated with lower risk of PCC (RR, 0.82; 95% CI, 0.69-0.99) In analyses jointly examining sleep dimensions prior to and early in the pandemic, participants who had healthy sleep at both time points were at lowest risk of PCC (RR, 0.64; 95% CI, 0.50-0.82) vs those who had a poor to intermediate sleep score prior to the pandemic and had neither optimal sleep duration (ie, 7 hours per day) nor fairly good or good sleep quality early in the pandemic (Table 3).

Results were comparable in sensitivity analyses defining PCC as symptoms lasting 8 or more weeks; defining PCC as ongoing symptoms at the time of PCC assessment; excluding participants whose only PCC symptoms were fatigue, headache, brain fog, memory issues, and depression; including presumed COVID-19 cases without a positive test; additionally adjusting for prepandemic memory issues; excluding participants who were hospitalized due to COVID-19, vaccinated, or with probable depression and anxiety at baseline; or using multiple imputation for missing exposure and outcome data (eTables 5 and 6 in Supplement 1).


In this cohort study of female health care workers participating in the NHS II, we found that multiple sleep dimensions measured both before (2015-2017) and early (April to August 2020) during the pandemic were associated with a lower risk of developing PCC. A combined prepandemic sleep score comprising early chronotype, sleep duration of 7 to 8 hours per day, low insomnia symptoms, no self-reported snoring, and no frequent daytime dysfunction was inversely associated with risk of PCC. Compared with women with a 0 or 1 healthy sleep dimension, those who scored 5 had a 30% lower risk of PCC. This association was not explained by the severity of acute phase disease or depression and anxiety. In mutually adjusted models, these associations were primarily driven by daytime dysfunction. Higher self-perceived sleep quality early in the COVID-19 pandemic was also associated with a lower risk of PCC. Women with consistently healthy sleep at both times had the lowest risk of PCC compared with women with consistently unhealthy sleep.

Poorer sleep has been linked to worse COVID-19 outcomes, including hospitalization and mortality.15 Observations from a prospective study using data from the UK Biobank (n = 46 535) indicated that an unhealthy sleep score, including duration, daytime sleepiness, insomnia, and chronotype, measured 10 to 14 years before infection was dose-dependently associated with a higher risk of COVID-19 hospitalization and mortality15; persons with the least healthy sleep pattern had almost twice the mortality risk of those with the most healthy pattern. Although the association between multidimensional sleep health and PCC has not been studied, a single, large, medical record–based study of 400 000 nonhospitalized individuals with confirmed SARS-CoV-2 infection found that obstructive sleep apnea was associated with a borderline increased risk of PCC (adjusted RR, 1.07; 95% CI, 0.99-1.17).50 These results may underestimate this association, however, because clinically undiagnosed obstructive sleep apnea tends to be highly prevalent in the general population and is not accurately captured by medical records. The prevalence of obstructive sleep apnea was only 1.4% in this cohort50 compared with an estimated 10% to 15% prevalence in North America.51,52 Findings from a second study suggested that poor sleep quality at COVID-19 symptom onset is associated with slow recovery from COVID-19 lymphopenia.53

Several pathophysiologic pathways may underlie our findings. All of the dimensions of sleep we examined have been linked to systemic inflammation and immune aberrations.54-57 The proinflammatory state may predispose those with poor sleep health to the occurrence of cytokine storms, which have been proposed to be involved in the multiorgan manifestations of PCC.22,24 Furthermore, short sleep duration and insomnia have been implicated in the development of autoimmune antibodies, which have also been found in higher concentrations among patients with persistent COVID-19–related symptoms.58,59 Dysregulated gut microbiome, possibly through immune function, may also be a shared pathologic process between sleep disorders and PCC.24,60 Coagulopathy, a downstream consequence of sleep disorders, has also been observed in some patients with PCC.61,62

Strengths and Limitations

A strength of this study is the longitudinal design, which allowed us to collect information on multiple sleep dimensions both before and during the pandemic. We also incorporated a multidimensional sleep score to account for the complexities of interconnected sleep behaviors and reduce measurement error. In addition, SARS-CoV-2 infection, hospitalization due to COVID-19, and vaccination against COVID-19 were prospectively collected, with frequent follow-up spanning 19 months during an active phase of the pandemic.

Our study also has several limitations. First, participants were women at middle-age or older and were relatively homogeneous in race (97% White), limiting generalizability of the findings. In addition, nearly one-half of the participants were frontline health care workers, although we did not find differences in the observed associations in frontline health care workers vs those who were not. Second, as the incidence of PCC may differ by SARS-CoV-2 strain,6 our findings may not apply to currently prevalent strains. Third, exposure and outcome were not missing at random, which may have introduced bias. Nevertheless, results were similar in the analysis using multiple imputation. Fourth, because SARS-CoV-2 infection and PCC were self-reported, some misclassification is likely. However, the validity of self-reported health has been high in this cohort.63,64 Fifth, the relatively small sample size limits our ability to detect some modest associations (eg, individuals who are long sleepers or habitual snorers).


The findings from this cohort study suggest that healthy sleep before and during the COVID-19 pandemic, prior to infection, was associated with a reduced risk of PCC. Future research should investigate whether improving sleep health may prevent or alleviate PCC symptoms.


Deja un comentario

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *