{"paper_id":"1fe320b8-01b4-4da2-a760-d76d4b2c5e96","body_text":"Association of Objectively Measured Sleep Patterns Using a Smartphone Application \nwith Work Productivity Loss in Japanese Employees \n \nJaehoon Seol, PhD1,2,3, Masao Iwagami, MD, MPH, MSc, PhD1,4, Masashi Yanagisawa, MD, \nPhD1,5,6,* \n \n1International Institute for Integrative Sleep Medicine (WPI- IIIS), Tsukuba Institute for \nAdvanced Research (TIAR), University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki \n305-8575, Japan  \n2Institute of Health and Sport Sciences, University of Tsukuba, Tsukuba, Ibaraki 305-8574, \nJapan \n3Department of Frailty Research, Center for Gerontology and Social Science, National \nCenter for Geriatrics and Gerontology, Obu, Aichi 474-8511, Japan \n4Department of Digital Health, Institute of Medicine, University of Tsukuba, Tsukuba, Ibaraki \n305-8574, Japan \n5R&D Center for Frontiers of Mirai in Policy and Technology (F-MIRAI), University of \nTsukuba, Tsukuba, Ibaraki 305-8575, Japan \n6Department of Molecular Genetics, University of Texas Southwestern Medical Center, \nDallas, TX 75390 \n \n*Corresponding author: \nProf. Masashi Yanagisawa, MD, PhD  \nInternational Institute for Integrative Sleep Medicine (WPI- IIIS), Tsukuba Institute for \nAdvanced Research (TIAR), University of Tsukuba,  \n1-1-1 Tennodai, Tsukuba, Ibaraki 305-8575, Japan  \nTel.: +81-029-853-5857; Fax: +81-029-853-5857 \nEmail: yanagisawa.masa.fu@u.tsukuba.ac.jp \nORCID: 0000-0002-7358-4022\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\nAbstract \nSleep disturbances are a major yet underrecognized contributor to reduced workplace \nproductivity (“presenteeism”). Previous studies have largely relied on self-reported sleep \ndata, limiting their scalability and objectivity. We examined the association between \nobjectively measured sleep characteristics and presenteeism among Japanese workers, \nusing real-world data from a smartphone sleep application. A total of 79,048 working adults \n(mean age: 42.1 years [range: 18–66 years]; women: 47.8%) provided informed consent \nand at least seven nights of valid sleep data across a 28-day period. Over 2.1 million nights \nof sleep data were analyzed. Sleep variables included total sleep time (TST), sleep latency, \nwake after sleep onset (%WASO), chronotype (MSFsc), and social jetlag. Generalized \nadditive models revealed that both short and long TST were associated with increased \npresenteeism, forming a U-shaped relationship. Greater sleep latency, higher %WASO, \ndelayed chronotype, and greater social jetlag were also independently linked to higher \npresenteeism scores. Unsupervised clustering using UMAP and the Leiden algorithm \nidentified five sleep phenotypes: “Healthy Sleepers,” “Long Sleepers,” “Fragmented \nSleepers,” “Poor Sleepers,” and “Social Jetlaggers.” The latter two groups exhibited the \nhighest levels of insomnia symptoms, excessive daytime sleepiness, and presenteeism. \nThese findings suggest that not only sleep duration but also timing, quality, and regularity \nare critical factors influencing occupational functioning. Smartphone-based sleep tracking \noffers a scalable approach to identify at-risk individuals and may help inform personalized \ninterventions to improve employee health and productivity.\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nIntroduction \nSleep is essential for maintaining physical health, cognitive function, and daily \nperformance.1 Epidemiological studies have consistently demonstrated that inadequate or \npoor-quality sleep is associated with numerous adverse health outcomes, including \nhypertension, cardiometabolic disorders, immune dysfunction, and mood disturbances. 2– 5 \nInadequate sleep also compromises cognitive abilities such as attention and alertness, \nthereby reducing daytime functioning and occupational productivity. 6,7 Hence, sufficient and \nhigh-quality sleep is considered fundamental for sustaining both individual health and work \nperformance.\n6,7 \nDespite its importance, chronic sleep deficiency remains prevalent among working \nadults.8 The United States Centers for Disease Control and Prevention (CDC) has classified \ninsufficient sleep as a public health issue, affecting more than one-third of adults in the \nUnited States.\n9 Similar trends have been reported globally, including in Japan, where 24-h \nsocietal demands and lifestyle factors contribute to inadequate sleep duration. 10,11 This \nwidespread sleep loss has substantial economic implications, with pr oductivity-related \nlosses estimated to reach hundreds of billions of dollars annually, amounting to as much as \n2–3% of gross domestic product (GDP) in some countries.\n12 Notably, Japan stands out as \none of the countries with the shortest average sleep duration globally. According to \ninternational comparisons, Japanese adults consistently report less than 7 h of sleep per \nnight on average—markedly lower than the recommended amount and below the \nOrganization for Economic Co-operation and Development (OECD) average.\n13 This chronic \nsleep restriction may reflect a culture of long working hours, limited rest opportunities, and \npersistent social demands. \nPresenteeism—defined as reduced work productivity despite being physically present at \nwork, often due to health problems—has become a growing concern in occupational health \nresearch.\n14 Sleep disturbances are increasingly recognized as major contributors to \npresenteeism.15,16 Cross-sectional studies have shown that individuals with short sleep \nduration (e.g., <6 h) or poor sleep quality report significantly greater productivity losses \ncompared with well-rested peers.\n8 Recently, interest has expanded beyond sleep quantity to \ninclude timing and regularity. 17 Extended social jetlag, for instance, is associated with \ndepressive symptoms, while delayed sleep–wake phase patterns have been linked to \nimpaired job performance.\n16,18 These findings suggest that sleep timing and consistency, in \naddition to duration, are critical determinants of occupational productivity. 18 However, most \nexisting studies rely on subjective self-reported sleep data and single-timepoint \nassessments,\n16,18,20–22 which are susceptible to recall bias and do not reflect habitual sleep \nbehavior in naturalistic settings. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nTo overcome these limitations, we conducted a large-scale study investigating the \nrelationship between sleep parameters or patterns and presenteeism using objectively \nmeasured, real-world sleep data. In contrast to prior research that has typically focused on \nisolated metrics such as total sleep time or perceived sleep quality, our study employed a \nmultidimensional framework encompassing sleep duration, quality, timing, and regularity.\n21 \nThese metrics were linked to a validated, single-item indicator of work productivity, enabling \nan integrated analysis of habitual sleep behavior and occupational functioning across a \nlarge working population. This approach may help identify modifiable sleep characteristics \nthat contribute to productivity loss and inform targeted workplace interventions. \n \nMethods \nStudy Design and Participants \nThis retrospective cross-sectional study analyzed data from 99,746 individuals living in \nJapan who provided informed consent between February 18 and May 19, 2025 (Figure 1). \nTo match the recall period of the presenteeism measure, which refers to work performance \nover the preceding 28 days, we extracted objective sleep data from up to 28 consecutive \ndays immediately prior to each participant’s questionnaire response. Participants were \nincluded if they had at least 7 days of valid sleep recordings during this period, yielding a \ntotal of over 2.1 million nights of observation. Sleep data were collected under naturalistic \nconditions using a widely available smartphone application. The study protocol was \napproved by the Institutional Review Board of Sapporo Yurinokai Hospital, Japan (approval \nnumber: 036). The reporting of this study followed the Strengthening the Reporting of \nObservational Studies in Epidemiology (STROBE) guidelines. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\n \nFigure 1. Flow of participant inclusion and analysis \nAmong 99,746 individuals with smartphone-recorded sleep data between February and May \n2025, 79,048 were included in the primary analysis after excluding those who identified as \nstudents or “other” (n = 16,745) and those with outlier values in key variables (n = 3,953). \nFor the clustering analysis, an additional 23,338 participants with irregular or undefined \nworkday/free-day patterns were excluded, resulting in a final analytic sample of 55,710. \n \nMeasures \nSleep Variables \nObjective sleep variables—including total sleep time (TST), sleep latency, and percentage \nof wake after sleep onset (%WASO)—were estimated using the Cole–Kripke algorithm, as \nimplemented in the Pokémon Sleep application.\n23 Midpoint of sleep on free days corrected \nfor sleep debt (MSFsc) and social jetlag 17,24 were computed using bedtime and wake time \ndata from the same application, combined with self-reported information on workdays and \nfree days obtained via questionnaire. The definitions and calculation procedures for MSFsc \nand social jetlag followed those described in previous studies.\n17,24 \n \nPresenteeism \nPresenteeism, defined as reduced work productivity despite being physically present at the \nAssessed for eligibility in May 2025\n(n =  99,746; over 2.6 m illion nights of objective sleep data)\nExclusion (n = 20,698)\n R espondents who selected 'student or other‘ (n =  16,745)\n P articipants with outliers in basic demographic or sleep-related \nvariables (n =  3,953) Analysis 2\nAssociation Between Sleep P attern Clusters and P resenteeism\n(n =  55,710)\nAnalysis 1\nAssociation between each sleep variable and presenteeism (GAM)\n(n =  79,048; over 2.1 million nights of data) \nExclusion\n P articipants with non-fixed holidays (n =  23,338)\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nworkplace, is often associated with physical or psychological health issues. In this study, \npresenteeism was assessed using the Single-Item Presenteeism Question (SPQ), a brief \nand validated instrument.\n25 Participants rated their overall work performance during the past \n28 days on a scale from 0% (no performance) to 100% (performance under ideal conditions). \nThe SPQ score was calculated as 100 minus the self-rated performance score, with higher \nvalues indicating greater productivity loss while at work.\n25 \n \nPotential Confounders \nBased on previous studies, 26,27 the following covariates were included in the analyses as \npotential confounders: age (continuous), sex (male, female, or no response), body mass \nindex (BMI; continuous), shift work status (yes or no), alcohol consumption (none, less t han \nonce per week, 1–2 times per week, more than 3 times per week, or daily), smoking status \n(none, <5 cigarettes/day, 5–9/day, 10–19/day, or \n≥ 20/day), caffeine intake (none or rarely, \n<1 cup/day, 2–3 cups/day, 4–6 cups/day, or ≥ 7 cups/day), and history of sleep disorders \n(never, past, or current). Insomnia symptoms and daytime sleepiness were assessed using \nthe Athens Insomnia Scale (AIS) and the Epworth Sleepiness Scale (ESS), respectively.\n28,29 \n \nData Preprocessing \nDaily sleep variables (bedtime, wake time, TST, sleep latency, and %WASO) were screened \nfor outliers using the interquartile range (IQR) method, 30 and values outside the IQR were \nexcluded. The remaining daily values were then averaged for each participant. Participants \nwith extreme values for height, weight, BMI, or age—defined as falling outside the \nIQR—were also excluded.\n30 Moreover, only those who identified their occupation as “worker” \nwere included in the analysis. Students and others were excluded because their daily \nschedules, sleep behaviors, and occupati onal demands differ substantially from those of \nworking adults, potentially introducing heterogeneity and bias into the analysis.  \n \nStatistical Analysis \nAnalysis 1: Generalized Additive Models and Interaction Visualization \nGeneralized additive models (GAMs)\n31 were used to examine the associations between \nsleep variables—including TST, sleep latency, %WASO, MSFsc, and social jetlag—and \nSPQ scores, allowing for potential non-linear relationships. \nTST was used as the primary axis in heatmap visualizations due to its central role in \nsleep-health associations and its relevance to occupational functioning. For each pair of \nsleep variables, bivariate heatmaps were generated by binning values along both axes and \ncalculating the mean SPQ score within each bin. Bins with fewer than 10 observations were \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nmasked and displayed in light gray to maintain interpretability. These visualizations were \nused to explore potential interaction patterns not captured by the GAMs. \n \nAnalysis 2: Sleep Phenotype Clustering and Symptom Associations \nTo provide an overview of the interrelationships among variables, we conducted Pearson \ncorrelation analyses between sleep variables and covariates including age and BMI. The \ncorrelation matrix is presented in Supplementary Figure S1a and S1b. Then, unsupervised \nclustering was conducted to identify latent sleep phenotypes based on the five sleep \nvariables listed above. To adjust for age effects, each variable was first adjusted for age \nusing linear regression. The age-adjusted values were standardized and embedded into a \ntwo-dimensional space using Uniform Manifold Approximation and Projection (UMAP). A \nk-nearest neighbor graph was constructed from the UMAP embeddings, and Leiden \ncommunity detection was applied to define clusters.\n32,33 The resulting clusters were \nvisualized in the UMAP space, with individual sleep variables overlaid to depict spatial \ngradients. We named each cluster according to its characteristics. \nAssociations between sleep phenotype clusters and sleep-related symptoms were \nevaluated using logistic regression, with cluster membership as the independent variable. \nDependent variables were defined as insomnia symptoms (AIS score \n≥  6) and excessive \ndaytime sleepiness (ESS score ≥  11). The “Healthy Sleepers” cluster was used as the \nreference group. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were \ncalculated for each comparison. \nDifferences in SPQ scores across clusters were assessed using multiple linear regression. \nCluster membership was included as a set of dummy-coded predictors, with “Healthy \nSleepers” as the reference. Regression coefficients represent the mean difference in SPQ \nscores compared with the reference group. All models (logistic and linear) were adjusted for \nage, sex, BMI, shift work status, alcohol consumption, smoking status, caffeine intake, and \nhistory of sleep disorders. In supplementary analyses, models were stratified by sex to \nexamine potential sex-specific associations. \n \nResults \nA total of 99,746 participants who consented between February and May 2 025 were initially \nincluded, contributing over 2.6 mi llion nights of objective sleep data. After excluding 16,745 \nindividuals who reported being students or selected “other” for occupation, and 3,953 \nindividuals with missing or outlier values in key variables, 79,048 participants remained for \nthe primary analysis using GAMs, encompassing over 2.1 million nights of data (Figure 1). \nParticipant-level average values of sleep parameters across valid nights were used in all \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nanalyses. For the clustering analysis, an additional 23,338 participants were excluded due \nto irregular or undefined workday/free-day patterns, yielding a final sample of 55,710 \nparticipants (Figure 1).  \nDemographic and sleep-related characteristics of the analytic sample are summarized in \nTable 1. The mean age was 37.5 years (standard deviation [SD]: 10.2), 54.5% were female, \nmean TST was 407.1 min (SD: 91.0), and the mean SPQ score was 20.2 points (SD: 17.5). \nAll variables were available for the full sample, except for MSFsc and social jetlag, which \nwere computed only among participants with regular work/free day schedules (n = 55,710) \n(Table 1). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nTable 1. Demographic and sleep characteristics of the full sample (Analysis 1) \n All \n(n = 79,048) \nMale  \n(n = 34,047) \nFemale \n(n = 43,102) \nNo answer  \n(n = 1,899) \nAge, years 37.5 ± 10.2 36.5 ± 9.7 38.4 ± 10.6 36.4 ± 9.7 \nBMI, kg/m2 22.3 ± 3.3 22.9 ± 3.2 21.9 ± 3.3 21.9 ± 3.3 \nNo. of days of Pokémon sleep recording, day 26.7 ± 3.0 27.0 ± 2.4 26.4 ± 3.3 26.6 ± 3.1 \nPresenteeism score, pts 20.2 ± 17.5 19.0 ± 17.1 21.0 ± 17.5 25.0 ± 21.0 \nShift work status, n (%), (yes) 4,601 (5.8) 2,684 (7.9) 1,827 (4.2) 90 (4.7) \nAlcohol consumption, n (%)     \nNone 35,982 (45.5) 12,718 (37.4) 22,258 (51.6) 1,006 (53.0) \n<1/w 22,221 (28.1) 10,028 (29.5) 11,693 (27.1) 500 (26.3) \n1-2/w 10,061 (12.7) 5,249 (15.4) 4,614 (10.7) 198 (10.4) \n≥ 3/w 4,820 (6.1) 2,744 (8.1) 1,998 (4.6) 78 (4.1) \nDaily 5,964 (7.5) 3,308 (9.7) 2,539 (5.9) 117 (6.2) \nSmoking status, n (%) 71,251 (90.1) 29,166 (85.7) 40,343 (93.6) 1,742 (91.7) \nNone     \n<5 cigarettes/d 1,100 (1.4) 663 (2.0) 409 (1.0) 28 (1.5) \n5–9 cigarettes/d 2,066 (2.6) 1,223 (3.6) 801 (1.9) 42 (2.2) \n10–19 cigarettes/d 3,506 (4.4) 2,214 (6.5) 1,231 (2.9) 61 (3.2) \n≥ 20 cigarettes/d 1,125 (1.4) 781 (2.3) 318 (0.7) 26 (1.4) \nCaffeine consumption, n (%)     \nNone or rarely 14,252 (18.0) 6,037 (17.7) 7,844 (18.2) 371 (19.5) \n<1/d 24,884 (31.5) 10,214 (30.0) 14,123 (32.8) 547 (28.8) \n2–3/d 26,759 (33.9) 11,383 (33.4) 14,723 (34.2) 653 (34.4) \n4–6/d 7,503 (9.5) 3,519 (10.3) 3,790 (8.8) 194 (10.2) \n≥ 7/d 5,650 (7.2) 2,894 (8.5) 2,622 (6.1) 134 (7.1) \nHistory of sleep disorders, n (%)     \nNever 68,705 (86.9) 30,115 (88.5) 37,027 (85.9) 1,563 (82.3) \nPast 5,413 (6.9) 2,084 (6.1) 3,178 (7.4) 151 (8.0) \nCurrent 4,930 (6.2) 1,848 (5.4) 2,897 (6.7) 185 (9.7) \nAIS, pts 5.3 ± 3.4 5.1 ± 3.4 5.5 ± 3.4 6.4 ± 4.1 \nESS, pts 7.6 ± 4.0 7.4 ± 4.0 7.8 ± 4.0 7.8 ± 4.4 \nHabitual bedtime, h 23.5 ± 1.6 23.2 ± 1.7 23.7 ± 1.6 23.5 ± 1.8 \nHabitual waketime, h 7.5 ± 1.6 7.4 ± 1.7 7.5 ± 1.6 7.7 ± 1.8 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nTST, min 398.4 ± 92.1 406.3 ± 94.6 391.7 ± 89.2 407.0 ± 96.6 \nSleep latency, min 17.5 ± 15.8 16.1 ± 15.1 18.6 ± 16.1 18.3 ± 18.0 \nWASO, % 10.0 ± 9.1 11.6 ± 10.0 8.8 ± 8.1 9.3 ± 8.6 \nMSFsc*, h 3.6 ± 1.9 3.6 ± 2.1 3.7 ± 1.6 3.8 ± 2.1 \nSocial jetlag*, h 0.5 ± 0.8 0.5 ± 0.9 0.4 ± 0.7 0.5 ± 0.9 \nNote: BMI, body mass index; TST, total sleep time; WASO, wake after sleep onset; MSFsc, midpoint of sleep on free days corrected for sleep \ndebt; AIS, Athens Insomnia Scale; ESS, Epworth Sleepiness Scale; SPQ, Single-Item Presenteeism Question. *Values for MSFsc and social \njetlag are based on participants with fixed holidays only (n = 55,710; Male: 25,695, Female: 28,745, No answer: 1,270). \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nSleep Variables and Presenteeism (GAMs) \nGAMs revealed significant associations between sleep characteristics and SPQ scores \n(Figure 2a). Monotonic increases in SPQ scores were observed with later MSFsc, longer \nsleep latency, higher %WASO, and greater social jetlag. In contrast, the relationship \nbetween TST and SPQ scores followed a U-shaped curve, indicating elevated presenteeism \namong individuals with both short and long sleep durations. Bivariate heatmaps combining \nTST with each of the other sleep variables (Figure 2B) showed that optimal SPQ scores \nwere observed among individuals with approximately 6–9 h of sleep and relatively lower \nvalues for sleep latency, %WASO, MSFsc, and social jetlag. These associations were \nconsistent in analyses stratified by sex (Supplementary Figure S2 and S3). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\n \n \nFigure 2. Associations between sleep variables and presenteeism score (SPQ) \na) Generalized additive models showing nonlinear associations between individual sleep \nvariables and presenteeism (SPQ score). Blue lines indicate fitted values and red dashed \nlines show 95% CIs. Gray histograms indicate the distribution of each sleep metric. \nb) Bivariate heatmaps showing mean SPQ scores by total sleep time and each of the other \nsleep variables (MSFsc, sleep latency, %WASO, and social jetlag). Gray cells represent \nbins with <10 observations. Lower SPQ indicates better productivity. \nCI, confidence interval; MSFsc, midpoint of sleep on free days corrected for sleep debt; \nSPQ, Single-Item Presenteeism Question; WASO, wake after sleep onset. \n \na\nb\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nUnsupervised Clustering and Sleep Phenotypes \nUnsupervised clustering using UMAP and the Leiden algorithm identified five distinct sleep \nphenotype clusters based on objectively measured sleep variables (Figure 3A and 3B). \nCluster 0, named “Healthy Sleepers,” exhibited a balanced and favorable sleep profile. \nCluster 1, named “Long Sleepers,” had the longest TST. Cluster 2, named “Fragmented \nSleepers,” had markedly elevated %WASO with otherwise average parameters. Cluster 3, \nnamed “Poor Sleepers,” was characterized by prolonged sleep latency and \nelevated %WASO. Cluster 4, named “Social Jetlaggers,” exhibited delayed sleep timing \n(high MSFsc) and substantial social jetlag (Figure 3C and 3D). The detailed sleep \ncharacteristics of each cluster are summarized in Table 2. \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\n \nFigure 3. Identification and characterization of sleep phenotype clusters based on \nobjective sleep variables \na) UMAP projections showing the distribution of five objective sleep variables (MSFsc, sleep \nlatency, total sleep time, %WASO, and social jetlag). Red indicates higher values and blue \nindicates lower values. b) Five clusters were identified using the Leiden algorithm based on \nsleep variables that were adjusted for age effects and standardized. c) Mean (± SD) values \na\nbc\nde\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nof sleep metrics by cluster. d)  Radar plot showing standardized Z-score profiles of sleep \ncharacteristics for each cluster. e)  Odds ratios and 95% confidence intervals for insomnia \nsymptoms (AIS ≥  6) and excessive daytime sleepiness (ESS ≥  11) across clusters. Healthy \nSleepers served as the reference group. \nMSFsc, midpoint of sleep on free days corrected for sleep debt; SD, standard deviation; \nWASO, wake after sleep onset. \n \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nTable 2. Sleep characteristics by cluster: results from the clustering analysis (Analysis 2) \n \nHealthy \nSleepers \n(n = 12,837) \nLong Sleepers \n(n = 10,249) \nFragmented \nSleepers \n(n = 4,416) \nPoor Sleepers \n(n = 15,459) \nSocial \nJetlaggers \n(n = 12,749) \nAge, years 39.9 ± 10.8 37.8 ± 9.9 36.6 ± 9.3 37.4 ± 10.1 36.6 ± 9.9 \nSex, n (%)      \nMale 4,968 (38.7) 5,163 (50.4) 2,868 (64.9) 7,713 (49.9) 4,983 (39.1) \nFemale 7,609 (59.3) 4,804 (46.9) 1,444 (32.7) 7,423 (48.0) 7,465 (58.6) \nNo answer 260 (2.0) 382 (2.8) 104 (2.4) 323 (2.1) 301 (2.4) \nBMI, kg/m2 22.1 ± 3.1 22.1 ± 3.2 22.5 ± 3.1 22.8 ± 3.3 22.2 ± 3.3 \nNo. of days of Pokémon sleep recording, \nday 27.0 ± 2.3 27.6 ± 1.3 27.7 ± 1.1 26.7 ± 2.8 26.1 ± 3.6 \nPresenteeism score, pts 17.4 ± 15.5 19.0 ± 17.2 17.6 ± 16.2 19.8 ± 16.8 20.3 ± 16.4 \nShift work status, n (%), (yes) 161 (1.3) 267 (2.6) 128 (2.9) 266 (1.7) 393 (3.1) \nAlcohol consumption, n (%)      \nNone 5,622 (43.8) 5,023 (49.0) 1,902 (43.1) 6,504 (42.1) 5,451 (42.8) \n<1/w 3,455 (26.9) 2,744 (26.8) 1,294 (29.3) 4,423 (28.6) 3,937 (30.9) \n1–2/w 1,752 (13.6) 1,179 (11.5) 606 (13.7) 2,208 (14.3) 1,793 (14.1) \n≥ 3/w 861 (6.7) 589 (5.7) 297 (6.7) 1,061 (6.9) 749 (5.9) \nDaily 1,147 (8.9) 714 (7.0) 317 (7.2) 1,263 (8.2) 819 (6.4) \nSmoking status, n (%)      \nNone 11,905 (92.7) 9,326 (91.0) 4,039 (91.5) 13,768 (89.1) 11,423 (89.6) \n<5 cigarettes/d 145 (1.1) 127 (1.2) 58 (1.3) 230 (1.5) 184 (1.4) \n5–9 cigarettes/d 239 (1.9) 224 (2.2) 86 (1.9) 480 (3.1) 349 (2.7) \n10–19 cigarettes/d 415 (3.2) 432 (4.2) 181 (4.1) 765 (4.9) 588 (4.6) \n≥ 20 cigarettes/d 133 (1.0) 140 (1.4) 52 (1.2) 216 (1.4) 205 (1.6) \nCaffeine consumption, n (%)      \nNone or rarely 1,959 (15.3) 2,005 (19.6) 850 (19.2) 2,631 (17.0) 2,159 (16.9) \n<1/d 4,093 (31.9) 3,156 (30.8) 1,344 (30.4) 4,865 (31.5) 4,106 (32.2) \n2–3/d 4,699 (36.6) 3,358 (32.8) 1,457 (33.0) 5,277 (34.1) 4,307 (33.8) \n4–6/d 1,184 (9.2) 972 (9.5) 432 (9.8) 1,519 (9.8) 1,267 (9.9) \n≥ 7/d 902 (7.0) 758 (7.4) 333 (7.5) 1,167 (7.5) 910 (7.1) \nHistory of sleep disorders, n (%)      \nNever 11,469 (89.3) 8,831 (86.2) 3,935 (89.1) 13,351 (86.4) 11,315 (88.8) \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nPast 723 (5.6) 725 (7.1) 286 (6.5) 1,107 (7.2) 823 (6.5) \nCurrent 645 (5.0) 693 (6.8) 195 (4.4) 1,001 (6.5) 611 (4.8) \nAIS, pts 4.8 ± 3.2 5.0 ± 3.5 5.0 ± 3.5 5.6 ± 3.5 5.6 ± 3.3 \nESS, pts 7.2 ± 3.9 7.3 ± 4.0 7.4 ± 4.1 7.7 ± 4.0 8.2 ± 4.0 \nHabitual bedtime, h 23.1 ± 1.0 22.6 ± 1.5 22.2 ± 1.0 23.5 ± 1.3 24.5 ± 1.8 \nHabitual waketime, h 6.6 ± 0.9 8.1 ± 1.5 7.4 ± 1.0 7.2 ± 1.3 7.7 ± 1.9 \nTST, min 405.1 ± 45.0 521.9 ± 51.7 465.1 ± 33.1 337.9 ± 66.8 350.6 ± 67.8 \nSleep latency, min 12.9 ± 6.5 9.8 ± 6.3 9.0 ± 5.3 30.9 ± 21.2 16.1 ± 8.8 \nWASO, % 4.2 ± 3.0 5.0 ± 3.5 14.7 ± 3.5 18.2 ± 10.1 6.4 ± 5.5 \nMSFsc, h 2.8 ± 0.8 3.4 ± 1.5 2.8 ± 1.0 3.4 ± 1.2 5.2 ± 2.8 \nSocial jetlag, h 0.2 ± 0.2 0.2 ± 0.3 0.2 ± 0.2 0.3 ± 0.3 1.1 ± 1.4 \nNote: BMI, body mass index; TST, total sleep time; WASO, wake after sleep onset; MSFsc, midpoint of sleep on free days corrected for sleep \ndebt; AIS, Athens Insomnia Scale; ESS, Epworth Sleepiness Scale; SPQ, Single-Item Presenteeism Question \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nSleep Symptoms Across Clusters \nCompared with Healthy Sleepers, all other clusters had significantly higher ORs of reporting \ninsomnia symptoms (AIS ≥  6). The highest ORs were observed in Social Jetlaggers (OR: \n1.56; 95% CI: 1.48–1.64), followed by Poor Sleepers (OR: 1.47; 95% CI: 1.40–1.55), Long \nSleepers (OR: 1.14; 95% CI: 1.08–1.20), and Fragmented Sleepers (OR: 1.13; 95% CI: \n1.06–1.21) (Figure 3E). For excessive daytime sleepiness (ESS \n≥  11), significantly elevated \nORs were found in all clusters except Long Sleepers. The highest ORs were again observed \nin Social Jetlaggers (OR: 1.59; 95% CI: 1.50–1.69), followed by Poor Sleepers (OR: 1.22; \n95% CI: 1.15–1.29) and Fragmented Sleepers (OR: 1.10; 95% CI: 1.01–1.20) (Figure 3E). \nThese associations remained largely consistent in analyses stratified by sex \n(Supplementary Figure S4). \n \nPresenteeism Across Clusters \nSPQ scores also differed significantly across clusters (Figure 4). Compared with Healthy \nSleepers, mean SPQ scores were significantly higher in Social Jetlaggers (+2.96 points, \n95% CI: 2.56–3.36), Poor Sleepers (+2.45 points, 95% CI: 2.07–2.84), and Long Sleepers \n(+1.67 points, 95% CI: 1.26–2.10). In contrast, the difference for Fragmented Sleepers was \nnot statistically significant (+0.25 points, 95% CI: −0.31 to 0.82). These trends were similarly \nobserved in sex-stratified analyses (Supplementary Figure S5). \n \nFigure 4. Differences in presenteeism scores across sleep phenotype clusters \nAdjusted differences in presenteeism scores (SPQ) across five sleep phenotype clusters. \nHealthy Sleepers were used as the reference. Dots indicate estimated means and horizontal \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nlines represent 95% confidence intervals. Numeric labels show adjusted mean SPQ scores. \nSPQ, Single-Item Presenteeism Question.  \n \nDiscussion \nIn this large-scale, real-world study of over 79,000 working adults and more than 2.1 million \nnights of objectively recorded sleep data, we found that multiple aspects of habitual sleep \nbehavior—spanning duration, timing, quality, and regularity—were significantly associated \nwith presenteeism. Using GAMs, we observed that longer sleep latency, greater social jetlag, \nand a later chronotype were each associated with higher presenteeism scores. Notably, TST \nshowed a U-shaped association, with both short and long durations linked to greater \nproductivity loss. Multivariate heatmaps further revealed that optimal productivity was \nassociated with approximately 6–9 h of sleep, shorter sleep latency, lower %WASO, smaller \nsocial jetlag, and an earlier chronotype. Our unsupervised clustering analysis identified five \ndistinct sleep phenotypes with differing sleep architectures and associations with \nsleep-related symptoms and productivity. Workers with delayed and irregular sleep timing \n(“Social Jetlaggers”) and those experiencing poor or fragmented sleep quality had the \nhighest ORs of insomnia symptoms, excessive daytime sleepiness, and reduced work \nperformance. Importantly, these associations were consistently observed in both men and \nwomen, indicating no significant sex differences in the relationship between sleep \ncharacteristics and presenteeism. \nWhile the study using Oura Ring data by Viswanath et al. (2024) successfully identified \nsleep clusters based on sleep duration and fragmentation,\n33 our study extends this \napproach by incorporating chronotype and social jetlag in addition to sleep latency and \nwake after sleep onset. This allowed for a more balanced classification across multiple \ndimensions of sleep health, enabling nuanced visualization of sleep phenotypes that better \nreflect both circadian and homeostatic aspects of sleep health. Notably, Japan is one of the \nfew OECD countries where men sleep longer than women—a reversal of the common \npattern observed globally.\n13,34 Japan is also known for having the shortest average sleep \nduration among OECD nations. 13,34 This trend was mirrored in our study: men slept \napproximately 15 min longer than women, and both sexes showed strikingly short average \nsleep durations of 6.7 h—substantially lower than the OECD average of 8.2 h.\n13 Moreover, \nwe observed patterns consistent with prior studies showing that with increasing age, \nindividuals tend to exhibit earlier chronotypes (lower MSFsc values) and reduced social \njetlag, albeit with modest effect sizes (Supplementary Figure S1a).\n17 Furthermore, later \nchronotypes were associated with greater social jetlag, reinforcing the interplay between \nbiological timing and social constraints (Supplementary Figure S1b). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nPrevious studies examining the link between sleep and presenteeism have often focused \non unidimensional indicators such as TST or self-reported sleep quality. Short sleep duration \n(<6 h/night) has consistently been associated with impaired work performance and \nincreased presenteeism,\n8 while some studies also suggest that long sleep duration may be \nlinked to adverse outcomes—forming a U-shaped relationship. 16 Our findings corroborate \nthis non-linear association using objectively measured sleep data in a large, diverse \npopulation. The lowest SPQ scores were observed in individuals averaging 7–8 h of sleep \nper night. While individual sleep needs vary, consistently long sleep durations have b een \nassociated with nonrestorative sleep or underlying conditions such as obstructive sleep \napnea, hypersomnia, and depressive disorders—all of which may impair daytime \nfunctioning and reduce productivity.\n35 Moreover, long sleep duration has been linked to \nincreased risk of morbidity and mortality in epidemiological studies. 36 For example, \nobstructive sleep apnea—a condition that fragments sleep and causes excessive daytime \nsleepiness—is also strongly associated with depressive symptoms and reduced work \ncapacity.\n35,36 Beyond sleep duration, our study reinforces the emerging view that sleep \ntiming and regularity are critical dimensions of sleep health. 16 We demonstrated that later \nchronotype (MSFsc) and greater social jetlag were independently associated with reduced \nworkplace productivity, even after accounting for sleep duration and quality. These findings \nhighlight the need to adopt a multidimensional approach when evaluating the role of sleep in \noccupational health. \nA major strength of our study lies in its use of high-resolution, objectively measured sleep \ndata collected under naturalistic conditions. While most prior studies have relied on \nself-reports, which are subject to recall and reporting biases,\n37 our approach leveraged \nwearable-based sleep tracking via a widely used smartphone application, enabling the \nassessment of habitual sleep behavior at scale.\n38 Using unsupervised clustering (UMAP and \nLeiden algorithms), 32,33 we identified data-driven phenotypes based on objectively \nmeasured variables. These phenotypes—distinguished by variations in sleep timing, latency, \nfragmentation, and duration—were differentially associated with sleep complaints and \npresenteeism outcomes. Notably, the “Social Jetlagger” and “Poor Sleeper” clusters \nconsistently showed the highest levels of insomnia symptoms, excessive daytime \nsleepiness, and productivity loss (Figure 3E and Figure 4). \nThese results illustrate the heterogeneity of sleep patterns in real-world populations and \nunderscore the value of data-driven phenotyping in identifying at-risk groups. Such \nclassifications could inform more targeted and personalized interventions, as opposed to \ngeneric, one-size-fits-all approaches. Moreover, the observed associations between \nobjective sleep indices and subjective complaints raise important questions regarding the \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nunderlying mechanisms. Prior research has noted a frequent mismatch between subjective \nand objective sleep, particularly in individuals with insomnia or mood disorders.39,40 Although \nour study did not directly examine discrepancies between subjective and objective sleep, \nthe simultaneous inclusion of both types of measures may inform future research on sleep \nmisperception and its impact on productivity. \nOur findings highlight the public health relevance of promoting healthy sleep—not just in \nterms of quantity, but also quality, timing, and regularity. The robust associations between \ncircadian misalignment, sleep fragmentation, and presenteeism suggest that improving \nsleep health could serve as an actionable strategy to enhance workplace productivity. \nInterventions targeting specific sleep phenotypes—such as chronotherapy for delayed sleep \ntiming or cognitive-behavioral therapy for insomnia—may be particularly effective for \nhigh-risk groups.\n41 Additionally, promoting flexible work schedules or circadian-aligned shift \nsystems may help mitigate the negative effects of social jetlag. 42,43 Modern society often \nfavors early chronotypes, with work and school schedules typically aligned to \nmorning-oriented routines. Consequently, individuals with later chronotypes may be \nstructurally disadvantaged, potentially experiencing reduced productivity due to a chronic \nmismatch between their biological rhythms and societal expectations. These insights \nsupport the inclusion of sleep health as a key element in occupational wellness programs \nand economic planning. Validated sleep-tracking tools may offer scalable, low-cost options \nfor population-level screening and behavior modification.\n44 \nThe economic burden of sleep-related productivity loss is substantial. RAND Europe \nestimates that insufficient sleep reduces GDP by 2–3% in countries such as Japan and the \nUnited States.\n12 In line with these estimates, our findings showed that workers classified as \n“Social Jetlaggers” had 2.96-point higher presenteeism scores than “Healthy Sleepers,” \nsuggesting a comparable scale of relative productivity loss. As previously noted, Japan’s \ncharacteristically short sleep duration—among the lowest in the OECD—may have \namplified the negative impact of insufficient sleep on work productivity observed in our \nstudy.\n13,34 Interestingly, presenteeism scores were highest in the “Social Jetlagger” group, \neven exceeding those in the “Poor Sleeper” group, which was defined by prolonged sleep \nlatency and fragmentation. This suggests that circadian misalignment, as reflected by \ndelayed sleep timing and large weekday-weekend discrepancies, may exert more \ndetrimental effects on productivity than poor sleep quality alone.\n17,24,45,46 Through \nunsupervised clustering, we captured nuanced behavioral phenotypes that traditional \nsymptom-based categories may miss. These findings emphasize the potential return on \ninvestment for workplace interventions that improve sleep alignment and regularity. \nUltimately, incorporating sleep metrics into organizational health strategies could enhance \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nemployee well-being, reduce economic burdens, and foster more sustainable and \nproductive work environments.47,48  \nThis study had some limitations. First, the cross-sectional design precludes causal \ninferences regarding the directionality of relationships between sleep variables and \npresenteeism. Longitudinal or interventional studies are needed to explore these \nassociations over time. Second, we lacked detailed occupational data such as job type, \nwork schedule, or organizational demands, limiting contextual interpretation. Third, although \nwe included only self-identified working adults and excluded students to focus on \nwork-related productivity, future research should also consider academic performance (e.g., \nPresenteeism Scale for Students)\n49 metrics to assess the impact of sleep among students. \nFourth, while we collected information on sleep disorders, specific diagnoses such as \nobstructive sleep apnea were not confirmed; this may have influenced the observed \nrelationships, particularly given the known association between sleep apnea, unrefreshing \nsleep, and reduced daytime functioning. Broader assessment of medical and psychosocial \nfactors may be necessary to enhance the generalizability and causal validity of our findings. \n \nConclusions \nThis large-scale, real-world study demonstrated that not only sleep duration but also sleep \ntiming and regularity are strongly associated with presenteeism. Among the five identified \nsleep phenotypes, individuals classified as “Social Jetlaggers” exhibited the highest \nproductivity loss—exceeding even those with poor sleep quality. These findings suggest that \ncircadian misalignment may be an important and potentially modifiable factor influencing \nworkplace performance. The integration of objective sleep tracking with data-driven \nphenotyping offers a promising approach to identifying high-risk individuals and tailoring \npersonalized interventions. Promoting sleep regularity and aligning sleep timing with \ncircadian rhythms may serve as effective strategies to improve both employee productivity \nand overall well-being. \n \nAcknowledgments \nWe would like to thank the personnel of The Pokémon Company (Tokyo, Japan) and \nS’UIMIN Inc. (Tokyo, Japan) for their contributions to data preparation. We would also like to \nthank Editage (www.editage.com) for English language editing. \n \nFunding \nThis work was supported by the World Premier International Research Center Initiative \n(WPI) from the Ministry of Education, Culture, Sports, Science and T echnology (MEXT) to \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nAuthor MY; the Japan Agency for Medical Research and Development (AMED; grant \nnumber: JP21zf0127005) to Authors MI and MY; the COI STREAM initiative launched in \n2013 by MEXT and the COI-NEXT initiative launched in 2020 by MEXT (grant number: \nJPMJPF2017) to Author JS; and the JSPS Fund for the Promotion of Joint International \nResearch (grant number: 22K21351) to Author MY . The funders had no role in the study \ndesign, data collection, analysis and interpretation of data, or in writing the manuscript. \n \nConflict of Interest Disclosures \nAuthor MY reported receiving payment from The Pokémon Company (Tokyo, Japan) for \nconsultation related to the development of the Pokémon Sleep app. No other conflicts of \ninterest are reported. \n \nAuthor Contributions \nJS had full access to all the data in the study and takes responsibility for the integrity of the \ndata and the accuracy of the data analysis. Concept and design: JS, MI, and MY; \nAcquisition, analysis, or interpretation of data: JS and MI; Drafting of the manuscript: JS; \nCritical revision of the manuscript for important intellectual content: JS, MI, and MY; \nStatistical analysis: JS; Administrative, technical, or material support: Not applicable; \nSupervision: MI and MY; Funding acquisition: MY . All authors have read and approved the \nfinal version of the manuscript and agree to be accountable for all aspects of the work. \n \nData Sharing Statement \nThe data that support the findings of this study are owned by The Pokémon Company \n(Tokyo, Japan) and are not publicly available. Data may be made available for future \nresearch upon reasonable request and with permission from The Pokémon Company. \nInterested researchers may contact the corresponding author for more information \nregarding data access. \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nReferences \n1. Shen L, Li BY , Gou W, et al. Trajectories of Sleep Duration, Sleep Onset Timing, and \nContinuous Glucose Monitoring in Adults. JAMA Netw Open. 2025;8(3):e250114. \ndoi:10.1001/jamanetworkopen.2025.0114 \n2. Svensson T, Saito E, Svensson AK, Melander O, Orho-Melander M, Mimura M, \nRahman S, Sawada N, Koh WP , Shu XO, Tsuji I, Kanemura S, Park SK, Nagata C, \nTsugane S, Cai H, Yuan JM, Matsuyama S, Sugawara Y , Wada K, Yoo KY , Chia KS, \nBoffetta P , Ahsan H, Zheng W, Kang D, Potter JD, Inoue M. Association of Sleep \nDuration With All- and Major-Cause Mortality Among Adults in Japan, China, Singapore, \nand Korea. JAMA Netw Open. 2021 Sep 1;4(9):e2122837. doi: \n10.1001/jamanetworkopen.2021.22837.  \n3. Li Y , Sahakian BJ, Kang J, Langley C, Zhang W, Xie C, Xiang S, Yu J, Cheng W, Feng J. \nThe brain structure and genetic mechanisms underlying the nonlinear association \nbetween sleep duration, cognition and mental health. Nat Aging. 2022 \nMay;2(5):425-437. doi: 10.1038/s43587-022-00210-2.  \n4. Musiek ES, Holtzman DM. Mechanisms linking circadian clocks, sleep, and \nneurodegeneration. Science. 2016 Nov 25;354(6315):1004-1008. doi: \n10.1126/science.aah4968.  \n5. Gao C, Guo J, Gong TT, Lv JL, Li XY , Liu FH, Zhang M, Shan YT, Zhao YH, Wu QJ. \nSleep Duration/Quality With Health Outcomes: An Umbrella Review of Meta-Analyses \nof Prospective Studies. Front Med (Lausanne). 2022 Jan 20;8:813943. doi: \n10.3389/fmed.2021.813943.  \n6. Goel N, Rao H, Durmer JS, Dinges DF. Neurocognitive consequences of sleep \ndeprivation. Semin Neurol. 2009 Sep;29(4):320-39. doi: 10.1055/s-0029-1237117.  \n7. Pilcher JJ, Morris DM. Sleep and Organizational Behavior: Implications for Workplace \nProductivity and Safety. Front Psychol. 2020 Jan 31;11:45. doi: \n10.3389/fpsyg.2020.00045. \n8. Ishibashi Y , Shimura A. Association between work productivity and sleep health: A \ncross-sectional study in Japan. Sleep Health. 2020 Jun;6(3):270-276. doi: \n10.1016/j.sleh.2020.02.016. \n9. Ramos AR, Wheaton AG, Johnson DA. Sleep Deprivation, Sleep Disorders, and \nChronic Disease. Prev Chronic Dis. 2023 Aug 31;20:E77. doi: 10.5888/pcd20.230197.  \n10. Kaneita Y , Ohida T. Association of current work and sleep situations with excessive \ndaytime sleepiness and medical incidents among Japanese physicians. J Clin Sleep \nMed. 2011 Oct 15;7(5):512-22. doi: 10.5664/JCSM.1322.  \n11. Morita Y , Sasai-Sakuma T, Asaoka S, Inoue Y . Prevalence and Correlates of Insufficient \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nSleep Syndrome in Japanese Young Adults: A Web-Based Cross-Sectional Study. J \nClin Sleep Med. 2015 Oct 15;11(10):1163-9. doi: 10.5664/jcsm.5092.  \n12. Hafner M, Stepanek M, Taylor J, Troxel WM, van Stolk C. Why Sleep Matters-The \nEconomic Costs of Insufficient Sl eep: A Cross-Country Comparative Analysis. Rand \nHealth Q. 2017 Jan 1;6(4):11. \n13. Organisation for Economic Co-operation and Development (OECD). OECD Time Use \nDatabase. 2021. Available from: \nhttps://www.oecd.org/en/data/datasets/time-use-database.html\n (accessed on 20th \nJune 2025)  \n14. Widera E, Chang A, Chen HL. Presenteeism: a public health hazard. J Gen Intern Med. \n2010 Nov;25(11):1244-7. doi: 10.1007/s11606-010-1422-x.  \n15. Takano Y , Iwano S, Aoki S, Nakano N, Sakano Y . A systematic review of the effect of \nsleep interventions on presenteeism. Biopsychosoc Med. 2021 Nov 17;15(1):21. doi: \n10.1186/s13030-021-00224-z.  \n16. Itani O, Kaneita Y , Otsuka Y , Tokiya M, Jike M, Matsumoto Y , Nakagome S, Kinoshita Y . \nA cross-sectional epidemiological study of the relationship between sleep duration, \nquality, and rhythm and presenteeism in workers. Sleep Biol Rhythms. 2021 Jul \n8;20(1):53-63. doi: 10.1007/s41105-021-00339-4.  \n17. Roenneberg T, Allebrandt KV, Merrow M, Vetter C. Social jetlag and obesity. Curr Biol. \n2012 May 22;22(10):939-43. doi: 10.1016/j.cub.2012.03.038. \n18. Islam Z, Hu H, Akter S, Kuwahara K, Kochi T, Eguchi M, Kurotani K, Nanri A, Kabe I, \nMizoue T. Social jetlag is associated with an increased likelihood of having depressive \nsymptoms among the Japanese working population: the Furukawa Nutrition and Health \nStudy. Sleep. 2020 Jan 13;43(1):zsz204. doi: 10.1093/sleep/zsz204.  \n19. Kaufmann CN, Nakhla MZ, Lee EE, Yoon HK, Wing D, Depp CA, Eyler LT. Inaccuracy \nbetween subjective reports and objective measures of sleep duration and clinical \ncorrelates in bipolar disorder. J Affect Disord. 2019 May 1;250:226-230. doi: \n10.1016/j.jad.2019.03.014. \n20. Otsubo T, Kinjo A, Kuwabara Y , Hongja K, Osaki Y . Lifestyle factors associated with \npresenteeism among city government office workers: a cross-sectional study. J Occup \nHealth. 2024 Jan 4;66(1):uiad012. doi: 10.1093/JOCCUH/uiad012.  \n21. Takano Y , Okajima I, Ando T, Iwano S, Inoue Y . Presenteeism and sleep duration on \nworkdays and days off. Occup Med (Lond). 2024 Jun 11;74(4):283-289. doi: \n10.1093/occmed/kqae028.  \n22. Turner M, Laws M, Griffiths M, Turner K, Dempsey L, Laws SM, Cruickshank T. The \nrelationships between multidimensional sleep health and work productivity in \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nindividuals with neurological conditions. J Sleep Res. 2024 Aug;33(4):e14107. doi: \n10.1111/jsr.14107.  \n23. Seol J, Iwagami M, Kayamare MCT, Yanagisawa M. Relationship Among \nMacronutrients, Dietary Components, and Objective Sleep Variables Measured by \nSmartphone Apps: Real-World Cross-Sectional Study. J Med Internet Res. 2025 Jan \n30;27:e64749. doi: 10.2196/64749.  \n24. Wittmann M, Dinich J, Merrow M, Roenneberg T. Social jetlag: misalignment of \nbiological and social time. Chronobiol Int. 2006;23(1-2):497-509. doi: \n10.1080/07420520500545979.  \n25. Muramatsu K, Nakao K, Ide H, Furui Y . Testing the Construct Validity and \nResponsiveness of the Single-Item Presenteeism Question. J Occup Environ Med. \n2021 Apr 1;63(4):e187-e196. doi: 10.1097/JOM.0000000000002158.  \n26. Lee SY , Lee J, Kwon M. Impacts of heavy smoking and alcohol consumption on \nworkplace presenteeism: A cross-sectional study. Medicine (Baltimore). 2021 Nov \n24;100(47):e27751. doi: 10.1097/MD.0000000000027751.  \n27. Buchvold HV, Pallesen S, Øyane NM, Bjorvatn B. Associations between night work and \nBMI, alcohol, smoking, caffeine and exercise--a cross-sectional study. BMC Public \nHealth. 2015 Nov 12;15:1112. doi: 10.1186/s12889-015-2470-2.  \n28. Soldatos CR, Dikeos DG, Paparrigopoulos TJ. Athens Insomnia Scale: validation of an \ninstrument based on ICD-10 criteria. J Psychosom Res. 2000 Jun;48(6):555-60. doi: \n10.1016/s0022-3999(00)00095-7.  \n29. Lapin BR, Bena JF, Walia HK, Moul DE. The Epworth Sleepiness Scale: Validation of \nOne-Dimensional Factor Structure in a Large Clinical Sample. J Clin Sleep Med. 2018 \nAug 15;14(8):1293-1301. doi: 10.5664/jcsm.7258. \n30. Leys C, Ley C, Klein O, Bernard P , Licata L. Detecting outliers: Do not use standard \ndeviation around the mean, use absolute deviation around the median. J Exp Soc \nPsychol. 2013;49(4):764-766. doi: 10.1016/j.jesp.2013.03.013. \n31. Hastie T, Tibshirani R. Generalized additive models. Stat Sci. 1986;1(3):297-310. \nAvailable from: http://www.jstor.org/stable/2245459\n (accessed on 20th June 2025) \n32. Traag VA, Waltman L, van Eck NJ. From Louvain to Leiden: guaranteeing \nwell-connected communities. Sci Rep. 2019 Mar 26;9(1):5233. doi: \n10.1038/s41598-019-41695-z.  \n33. Viswanath VK, Hartogenesis W, Dilchert S, Pandya L, Hecht FM, Mason AE, Wang EJ, \nSmarr BL. Five million nights: temporal dynamics in human sleep phenotypes. NPJ \nDigit Med. 2024 Jun 20;7(1):150. doi: 10.1038/s41746-024-01125-5.  \n34. Li L, Nakamura T, Hayano J, Yamamoto Y . Age and gender differences in objective \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nsleep properties using large-scale body acceleration data in a Japanese population. Sci \nRep. 2021 May 11;11(1):9970. doi: 10.1038/s41598-021-89341-x.  \n35. Dauvilliers Y , Lopez R, Ohayon M, Bayard S. Hypersomnia and depressive symptoms: \nmethodological and clinical aspects. BMC Med. 2013 Mar 21;11:78. doi: \n10.1186/1741-7015-11-78.  \n36. Cappuccio FP , D'Elia L, Strazzullo P , Miller MA. Sleep duration and all-cause mortality: \na systematic review and meta-analysis of prospective studies. Sleep. 2010 \nMay;33(5):585-92. doi: 10.1093/sleep/33.5.585. \n37. Jackson CL, Ward JB, Johnson DA, Sims M, Wilson J, Redline S. Concordance \nbetween self-reported and actigraphy-assessed sleep duration among \nAfrican-American adults: findings from the Jackson Heart Sleep Study. Sleep. 2 020 \nMar 12;43(3):zsz246. doi: 10.1093/sleep/zsz246.  \n38. Akre S, Cohen ZD, Welborn A, Zbozinek TD, Balliu B, Craske MG, Bui AAT. Comparing \nself reported and physiological sleep quality from consumer devices to depression and \nneurocognitive performance. NPJ Digit Med. 2025 Feb 9;8(1):92. doi: \n10.1038/s41746-025-01493-6.  \n39. Bianchi MT, Williams KL, McKinney S, Ellenbogen JM. The subjective-objective \nmismatch in sleep perception among those with insomnia and sleep apnea. J Sleep \nRes. 2013 Oct;22(5):557-68. doi: 10.1111/jsr.12046.  \n40. Masaki M, Tsumoto S, Tani A, Tominaga M, Seol J, Chiba S, Miyanishi K, Nishida K, \nKawana F, Amemiya T, Hiei T, Kanbayashi T, Yanagisawa M. Discrepancies between \nsubjective and objective sleep assessments revealed by in-home \nelectroencephalography during real-world sleep. Proc Natl Acad Sci USA. 2025 Jan \n21;122(3):e2412895121. doi: 10.1073/pnas.2412895121.  \n41. Shaffer KM, Finkelstein EA, Camacho F, Ingersoll KS, Thorndike F, Ritterband LM. \nEffects of an Internet-Based Cognitive Behavioral Therapy for Insomnia Program on \nWork Productivity: A Secondary Analysis. Ann Behav Med. 2021 Jun 2;55(6):592-599. \ndoi: 10.1093/abm/kaaa085.  \n42. Vetter C, Fischer D, Matera JL, Roenneberg T. Aligning work and circadian time in shift \nworkers improves sleep and reduces circadian disruption. Curr Biol. 2015 Mar \n30;25(7):907-11. doi: 10.1016/j.cub.2015.01.064.  \n43. Härmä M, Karhula K, Ropponen A, Puttonen S, Koskinen A, Ojajärvi A, Hakola T, Pentti \nJ, Oksanen T, Vahtera J, Kivimäki M. Association of changes in work shifts and shift \nintensity with change in fatigue and disturbed sleep: a within-subject study. Scand J \nWork Environ Health. 2018 Jul 1;44(4):394-402. doi: 10.5271/sjweh.3730.  \n44. Lee T, Cho Y , Cha KS, Jung J, Cho J, Kim H, Kim D, Hong J, Lee D, Keum M, Kushida \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint \n\nCA, Yoon IY , Kim JW. Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep \nTrackers: Prospective Multicenter Validation Study. JMIR Mhealth Uhealth. 2023 Nov \n2;11:e50983. doi: 10.2196/50983.  \n45. Ko H, Cho SS, Kang MY . Work ability and health-related productivity loss by \nchronotype: Results from population-based panel study. Sleep Health. 2025 Jun \n2:S2352-7218(25)00081-6. doi: 10.1016/j.sleh.2025.04.005.  \n46. Chakradeo PS, Keshavarzian A, Singh S, Dera AE, Esteban JPG, Lee AA, Burgess HJ, \nFogg L, Swanson GR. Chronotype, social jet lag, sleep debt and food timing in \ninflammatory bowel disease. Sleep Med. 2018 Dec;52:188-195. doi: \n10.1016/j.sleep.2018.08.002.  \n47. Robbins R, Jackson CL, Underwood P , Vieira D, Jean-Louis G, Buxton OM. Employee \nSleep and Workplace Health Promotion: A Systematic Review. Am J Health Promot. \n2019 Sep;33(7):1009-1019. doi: 10.1177/0890117119841407.  \n48. Redeker NS, Caruso CC, Hashmi SD, Mullington JM, Grandner M, Morgenthaler TI. \nWorkplace Interventions to Promote Sleep Health and an Alert, Healthy Workforce. J \nClin Sleep Med. 2019 Apr 15;15(4):649-657. doi: 10.5664/jcsm.7734. \n49. 49. Matsushita M, Adachi H, Arakida M, Namura I, Takahashi Y , Miyata M, Kumano-go \nT, Yamamura S, Shigedo Y , Suganuma N, Mikami A, Moriyama T, Sugita Y . \nPresenteeism in college students: reliability and validity of the Presenteeism Scale for \nStudents. Qual Life Res. 2011 Apr;20(3):439-46. doi: 10.1007/s11136-010-9763-9.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted July 3, 2025. ; https://doi.org/10.1101/2025.07.03.25330692doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}