Longitudinal association of nonrestorative sleep with subjective health and work performance: a cohort study in the Japanese construction industry

Abstract

Previous studies have reported associations between nonrestorative sleep (NRS) and various physical and mental illnesses. However, its long-term relationship with subjective health and work performance remains unclear. This study aimed to investigate the association between NRS and subjective health and work performance among Japanese workers. A longitudinal cohort study was conducted using health examination and medical claims data from fiscal years 2018 to 2023, involving 36,160 workers. NRS was assessed at baseline based on responses to the question, “Do you feel refreshed after a typical night’s sleep?” Longitudinal associations of NRS with subjective physical and mental health and work performance were analyzed using generalized linear models. The prevalence of NRS was 55.4%. NRS was significantly associated with poorer subjective physical health (aOR=1.77, 95% CI: 1.65–1.91) and mental health (aOR=1.66, 95% CI: 1.54–1.79). In addition, workers with NRS exhibited 3.6% lower work performance (95% CI: 3.25–3.88) compared with those with restorative sleep and incurred an estimated annual productivity-related economic loss of approximately JPY 160,000 per worker. These findings indicate that NRS is associated with decreased subjective health and reduced work performance. Early screening and intervention for NRS may help improve workers’ health and productivity.

Introduction

Approximately 40% of workers have experienced working while suffering from some form of illness¹⁾. In this context, it is important to consider not only whether workers have an illness but also how they perceive their own health. Therefore, in occupational health, subjective health is an important indicator alongside medical health status, which refers to the absence of disease. Moreover, work performance is a key concern for organizations in occupational health. Work performance is known to be influenced by factors such as mental health^{2, 3)} and physical activity^{4, 5)}. Interventions targeting these factors are implemented with the goal of improving labor productivity.

In recent years, sleep has received increasing attention as a factor influencing workers’ subjective health and work performance. Previous studies have shown that sleep disorders worsen subjective health⁶⁾ and reduce work performance⁷⁾. The economic impact of sleep disorders on national economies is substantial, estimated at 1.86–2.92% in Japan, 1.56–2.28% in the United States, and 1.36–1.86% in the United Kingdom, with Japan experiencing the greatest effect⁸⁾.

Among sleep-related conditions, nonrestorative sleep (NRS) has attracted increasing attention. NRS is defined as “a subjective experience of unrefreshing sleep”⁹⁾, and is one of the comprehensive concepts of sleep. Because it is a subjective experience, it can be easily screened, and even a single-item question is considered sufficient to assess the primary symptom¹⁰⁾. The prevalence of NRS varies by country and measurement method, ranging from 1.4% to 42%^{9, 11, 12)}. NRS is associated with an increased risk of physical conditions such as diabetes^{13, 14)} and hypertension^{14, 15)}, as well as mental disorders, including depression^{16, 17)} and anxiety disorders¹⁶⁾. This suggests that NRS may negatively affect subjective health. Furthermore, NRS is linked to daytime sleepiness¹⁸⁾, which is believed to impair work performance. Indeed, some studies suggest that NRS affects workers’ performance¹⁹⁾; however, these studies were small and cross-sectional, and long-term associations remain unclear. Moreover, limited research has examined the impact of NRS on subjective health.

Therefore, this study aimed to clarify the long-term effects of NRS on subjective health and work performance among workers using large-scale cohort data from Japanese workers.

Methods

Participants

This retrospective cohort study used health examination data from members of the Health Insurance Association for Architecture and Civil Engineering companies, one of Japan’s major health insurance associations. The baseline period was from April 2018 to March 2019 (Fiscal Year 2018). A total of 46,827 individuals were followed through March 2024 (FY 2018–FY 2023). Data from participants who were successfully followed and had no missing values were included in the analysis.

Outcomes

The outcomes of this study were subjective physical health, subjective mental health, and work performance, as measured by a newly implemented questionnaire administered during health checkups conducted between April 2023 and March 2024.

Subjective health: Respondents were asked to rate their physical and mental health in response to the question, “How has your health been over the past month?” Responses were recorded on a 5-point scale: very poor, poor, fair, good, and excellent. “Excellent”, “good”, and “fair” were categorized as “good health”, while “very poor” and “poor” were categorized as “poor health” and used as outcome variables.

Work performance: The Single-Item Presenteeism Question (SPQ)²⁰⁾ was used. The SPQ asks: “On a scale from 1% to 100%, where 100% is the best job performance you could have at your job if unimpeded by sickness or injury, how would you rate your overall job performance on the days you worked during the past four weeks (28 d)?” Respondents provided a numerical rating from 0% to 100%. This measure has established reliability and validity.

Nonrestorative sleep (NRS)

NRS was assessed using baseline statutory health examination data. Participants responded yes or no to the question, “Do you feel refreshed after a typical night’s sleep?” Those who answered “no” were classified into the NRS group, while those who answered “yes” were classified into the restorative sleep (RS) group. This single-item question is widely used for both practical and research purposes. Although this question has not been validated, previous studies have regarded it as effective in assessing NRS, as it captures the main symptom of NRS^{10, 14)}.

Covariates

Sex, age, medication history, comorbidities, and lifestyle habits in FY2018 were included as covariates. Medication history was categorized as prescribed or not prescribed from the medical claims data based on whether participants were prescribed any type of medication in FY2018. Comorbidities were assessed using the Charlson Comorbidity Index (CCI)²¹⁾, calculated from diagnosis data in insurance claims. Higher scores of CCI indicate a greater number and severity of pre-existing chronic conditions.

Lifestyle habits included exercise, dietary habits, smoking, and alcohol consumption. Participants were classified as having appropriate exercise habits if they met two or more of the following criteria: (A) engaging in moderate exercise (enough to produce a light sweat) for ≥30 min at least twice a week for at least one year, (B) walking or performing an equivalent level of physical activity for ≥1 h per day as part of their daily routine, and (C) walking faster than peers of similar age and sex. Appropriate eating habits were defined as meeting none or only one of the following: (A) eating faster than others, (B) eating within 2 h of bedtime three or more times a week, (C) consuming snacks or sweet beverages between meals daily or occasionally, and (D) skipping breakfast ≥3 times a week. Smoking habits were defined as having smoked ≥100 cigarettes over six months and having smoked within the previous month. Appropriate alcohol consumption was defined as meeting any of the following: (A) abstaining from alcohol, (B) occasional consumption of fewer than three drinks, or (C) daily consumption of fewer than two drinks. One drink was defined as equivalent to 500 ml of beer or 60 ml of whiskey in alcohol content.

Statistical analyses

For nominal variables, frequencies and proportions were calculated, and for continuous variables, means and standard deviations were computed. Differences between the NRS and RS groups were compared using χ² tests for nominal variables and t-tests for continuous variables. Next, univariate generalized linear models were used, with NRS/RS as the explanatory variable and subjective physical health, subjective mental health, and work performance as the respective outcomes. From these models, odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated to estimate the likelihood of poorer subjective health among the NRS group compared with the RS group, as well as the difference in work performance between the two groups, along with its 95% CI. Similar analyses were conducted using multivariate generalized linear models incorporating covariates.

Finally, the following formula was used to estimate the economic loss associated with decreased work performance per individual. The difference between the NRS and RS groups was then compared using multivariate generalized linear models:

Estimated annual productivity-related economic loss associated with NRS

= average monthly salary in the civil engineering industry in Japan (stratified by sex and age)²²⁾

× 12 months

× percentage decrease in work performance

In addition, to show the robustness of the analyses, we performed sensitivity analyses by reclassification of the groups of subjective health and sex-stratified analyses. All statistical analyses were performed using SPSS version 29.0, with a two-tailed significance level set at 5%.

Ethical considerations

The data used in this study were pre-existing anonymized data collected as part of the legally mandated services of the health insurance associations under the National Health Insurance Act and the Act on Assurance of Medical Care for Elderly People in Japan. In accordance with the Ethical Guidelines for Medical and Biological Research Involving Human Subjects issued by the Government of Japan, research using such legally collected, pre-existing, and properly anonymized data is exempt from institutional review board approval for both health insurance associations and research institutions. Therefore, this study was deemed exempt from ethical review.

Results

Demographic profiles

Table 1 presents the demographic characteristics of the study participants. A total of 36,160 individuals were analyzed (80.6% male, 19.4% female), with a mean age of 46.5 yr (SD=10.7). The prevalence of NRS was 55.4%. Participants reporting good subjective physical health accounted for 86.3%, and those reporting good subjective mental health accounted for 87.1%. The mean work performance score was 84.5% (SD=15.3). Compared with the RS group, the NRS group had a higher proportion of women, was younger on average, and had a greater proportion of participants taking prescription medications. Additionally, the NRS group had higher proportions of inappropriate exercise and dietary habits, with differences also observed in smoking and alcohol consumption.

Table 1. Sociodemographic characteristics of participants (N=36,160).

Variablesa		Total	RS (N=16,144)	NRS (N=20,016)	p-valueb
Sex	Male	29,144 (80.6)	13,220 (81.9)	15,924 (79.6)	<0.001
	Female	7,016 (19.4)	2,924 (18.1)	4,092 (20.4)
Age (yr)		46.5 (10.7)	47.4 (10.7)	45.7 (10.7)	<0.001c
CCI		0.23 (0.7)	0.23 (0.8)	0.23 (0.7)	0.645c
Use of prescription medication	Yes	27,260 (75.4)	11,988 (74.3)	15,272 (76.3)	<0.001
Exercise habit	Appropriate	8,974 (24.8)	4,927 (30.5)	4,047 (20.2)	<0.001
Dietary habit	Appropriate	15,580 (43.1)	7,831 (48.5)	7,749 (38.7)	<0.001
Smoking habit	Yes	10,271 (28.4)	4,429 (27.4)	5,842 (29.2)	<0.001
Alcohol consumption	Appropriate	31,302 (86.6)	13,885 (86.0)	17,417 (87.0)	0.005
Subjective physical health (fair=Good)	Good	31,211 (86.3)	14,765 (91.5)	16,446 (82.2)	<0.001
Subjective physical health (fair=Poor)	Good	8,143 (22.5)	4,938 (30.6)	3,205 (16.0)	<0.001
Subjective mental health (fair=Good)	Good	31,499 (87.1)	14,840 (91.9)	16,659 (83.2)	<0.001
Subjective mental health (fair=Poor)	Good	8,776 (24.3)	5,201 (32.2)	3,575 (17.9)	<0.001
Subjective work performance (%)		84.5 (15.3)	86.7 (14.5)	82.7 (15.8)	<0.001c

^aAt baseline, except for outcomes (subjective health and work performance). ^bThe χ² test was used except in the variables where the t-test was performed. ^cStudent’s t-test was used. RS: restorative sleep; NRS: nonrestorative sleep; CCI: Charlson Comorbidity Index.

NRS and subjective health

Table 2 presents the association between NRS and subjective health. In the unadjusted model, NRS was associated with poorer subjective physical health (cOR=2.32, 95% CI: 2.18–2.48) and poorer subjective mental health (cOR=2.29, 95% CI: 2.14–2.45). This association remained significant in the adjusted model, with NRS associated with poorer subjective physical health (aOR=1.77, 95% CI: 1.65–1.91) and poorer subjective mental health (aOR=1.66, 95% CI: 1.54–1.79).

Table 2. Association between NRS and subjective health.

Outcomes	cOR (95% CI)	p-value	aORa (95% CI)	p-value
Subjective physical health (ref. good subjective health)	2.32 (2.18–2.48)	<0.001	1.77 (1.65–1.91)	<0.001
Subjective mental health (ref. good subjective health)	2.29 (2.14–2.45)	<0.001	1.66 (1.54–1.79)	<0.001

^aAdjusted for sex, age, CCI, use of prescription medication, exercise habit, dietary habit, smoking habit, alcohol consumption, and subjective physical health or mental health. NRS: nonrestorative sleep; CCI: Charlson Comorbidity Index; OR: odds ratio.

Furthermore, when “fair” health was reclassified as “poor” for sensitivity analysis, NRS continued to be associated with poorer subjective physical health (cOR=2.31, 95% CI: 2.20–2.43; aOR=1.63, 95% CI: 1.49–1.77) and poorer subjective mental health (cOR=2.19, 95% CI: 2.08–2.30; aOR=1.87, 95% CI: 1.74–2.00). In addition, the sex-stratified analysis showed consistent significant associations (Table 3).

Table 3. Sex-stratified analysis for the association between NRS and subjective health.

Outcomes	Male		Female
	aORa (95% CI)	p-value	aORa (95% CI)	p-value
Subjective physical health (ref. good subjective health)	1.77 (1.63–1.93)	<0.001	1.78 (1.52–2.08)	<0.001
Subjective mental health (ref. good subjective health)	1.68 (1.54–1.83)	<0.001	1.58 (1.34–1.86)	<0.001

^aAdjusted for age, CCI, use of prescription medication, exercise habit, dietary habit, smoking habit, alcohol consumption, and subjective physical health or mental health. NRS: nonrestorative sleep; CCI: Charlson Comorbidity Index; OR: odds ratio.

NRS and work performance, and annual productivity-related economic loss

Table 4 presents the association between NRS and work performance. In the unadjusted model, NRS was significantly associated with 4.04% lower work performance (95% CI: 3.72–4.35) compared with RS. This association remained significant in the adjusted model, with work performance being 3.57% lower (95% CI: 3.25–3.88).

Table 4. NRS and subjective work performance.

		Subjective work performance (95% CI)	Difference (95% CI)	p-value
Unadjusted	RS	86.72 (86.48–86.95)	4.04 (3.72–4.35)	<0.001
	NRS	82.68 (82.47–82.89)
aAdjusted	RS	86.66 (86.30–87.02)	3.57 (3.25–3.88)	<0.001
	NRS	83.09 (82.74–83.44)

^aAdjusted for sex, age, CCI, use of prescription medication, exercise habit, dietary habit, smoking habit, and alcohol consumption. RS: restorative sleep; NRS: nonrestorative sleep; CCI: Charlson Comorbidity Index.

Table 5 presents the annual productivity-related economic loss associated with NRS and RS. In both the unadjusted and adjusted models, NRS was associated with a significantly greater economic loss, exceeding that of RS by JPY 162,664 per year (95% CI: 148,440–176,887) per person. Sensitivity analysis, which is stratified by sex, showed a consistent increase in economic loss in both males and females (Table 6).

Table 5. Annual productivity-related economic loss associated with NRS.

		Economic loss (JPY) (95% CI)	Difference (95% CI)	p-value
Unadjusted	RS	584,644 (574,102–595,186)	168,382 (154,213–182,551)	<0.001
	NRS	753,026 (743,558–762,493)
aAdjusted	RS	530,566 (514,464–546,668)	162,664 (148,440–176,887)	<0.001
	NRS	693,230 (677,403–709,056)

^aAdjusted for sex, age, CCI, use of prescription medication, exercise habit, dietary habit, smoking habit, and alcohol consumption. JPY: Japanese yen; RS: restorative sleep; NRS: nonrestorative sleep; CCI: Charlson Comorbidity Index.

Table 6. Sex-stratified analysis for annual productivity-related economic loss associated with NRS.

		Economic loss (JPY) (95% CI)	Difference (95% CI)	p-value
Male	RS	609,427 (593,826–625,028)	179,449 (162,722–196,176)	<0.001
	NRS	788,876 (773,267–804,485)
Female	RS	467,903 (425,319–510,488)	98,550 (75,328–121,772)	<0.001
	NRS	566,453 (524,585–608,321)

Adjusted for age, CCI, use of prescription medication, exercise habit, dietary habit, smoking habit, and alcohol consumption. JPY: Japanese yen; RS: restorative sleep; NRS: nonrestorative sleep; CCI: Charlson Comorbidity Index.

Discussion

To our knowledge, this study is the first to clarify the longitudinal association between NRS and workers’ subjective health and work performance using large-scale cohort data. This large-scale longitudinal study utilized a worker health examination database to examine the impact of NRS on subjective health and work performance five years later. The results showed that NRS was associated with deterioration in both subjective physical and mental health, as well as a reduction in work performance of approximately 4% compared with the RS group. Furthermore, the economic loss resulting from decreased work performance associated with NRS was estimated to be approximately JPY 160,000 per person per year, exceeding that of the RS group.

Prevalence of NRS

The prevalence of NRS in this study was 55.4%. Although some studies estimate the prevalence of NRS to be less than 20%, most of them are based upon a different definition of NRS, which says it needs to last at least 3 d a week^{17, 23,24,25)}. Given that previous reports using the same single-item question have estimated the prevalence of NRS in Japanese populations to be 19–37%^{11, 26)}, and in the US population to be 26–42%²⁷⁾, the prevalence observed in this study was relatively high. Prior studies have typically focused on the general population, whereas most of the participants in this study were construction workers. It is known that workers primarily engaged in physical labor exhibit different sleep patterns from office workers^{28, 29)}. Although physical activity is a known protective factor against sleep problems³⁰⁾, the adverse effects of a harsh work environment may outweigh its protective benefits. Therefore, the higher prevalence of NRS among construction workers is plausible.

NRS and subjective health

NRS has been shown to increase the risk of various disorders and is thought to contribute to deterioration in subjective health through its effects on medical health. NRS increases the risk of physical conditions such as cardiovascular disease¹⁵⁾ and diabetes^{13, 14)}. Sleep apnea syndrome, a condition that can cause NRS, is a known risk factor for vascular disorders, likely due to elevated levels of factors such as plasminogen activator inhibitor-1³¹⁾, resulting from nocturnal hypoxemia. Sleep deprivation associated with NRS is known to affect glucose metabolism and endocrine function, increasing the risk of diabetes and obesity^{32, 33)}, further demonstrating the link between NRS and physical health conditions.

NRS also increases the risk of mental disorders such as depression^{16, 17)} and anxiety¹⁶⁾. Although the mechanisms by which NRS affects mental health are not fully understood, it has been suggested that changes in amygdala activity due to sleep deprivation may influence mental well-being³⁴⁾. While further research is needed, a close association between NRS and mental disorders is therefore likely.

Actual medical conditions are known to reduce subjective health³⁵⁾. In this context, NRS is thought to worsen objective health, and awareness of this deterioration may further impair subjective health. Moreover, because NRS was self-reported, the perception of “not feeling rested” was already present, suggesting that this perception alone may directly affect subjective physical and mental health. In any case, early intervention targeting NRS, such as sleep education, cognitive behavioral therapy for insomnia (CBT-I), and pharmaceutical treatment, may help improve subjective health.

NRS and work performance

This study revealed that, compared with RS, NRS is associated with an approximately 4% decline in work performance and an increase in economic losses of about JPY 160,000 per person annually. Previous research has also suggested that NRS may reduce work performance¹⁹⁾. Importantly, the association between NRS and work performance remained significant after adjustment for comorbidities, suggesting that NRS is not merely a proxy for disease burden but a subjective sleep-related condition that impairs daytime functioning. NRS and underlying sleep disorders cause daytime dysfunctions such as sleepiness, reduced concentration, and impaired cognitive function, which directly lead to decreased work performance^{36, 37)}. Furthermore, NRS may indirectly reduce work performance by increasing susceptibility to the physical and mental health conditions described above. Workers with any symptoms are known to have lower work performance compared with symptom-free individuals^{38, 39)}. For example, NRS is a risk factor for hypertension^{14, 15)}, diabetes¹³⁾, and obesity¹⁴⁾, which can in turn cause several symptoms such as headaches and joint pain that reduce work performance. In addition, NRS increases the risk of mental disorders, and cognitive impairments associated with depression and anxiety have been shown to contribute to reduced work performance⁴⁰⁾.

Thus, it is plausible that NRS indirectly contributes to reduced work performance by increasing the risk of various health conditions. NRS diminishes work performance both directly and indirectly, leading to increased overtime at the individual level and reduced labor productivity at the corporate level. Furthermore, the economic losses resulting from decreased labor productivity due to NRS are substantial. Given that the prevalence of NRS in the general population is approximately 20% or higher, its impact on national economic losses is likely considerable. Therefore, early screening for NRS may be beneficial not only for improving individual workers’ health and quality of life but also for enhancing corporate productivity and strengthening national economic performance.

Limitations

This study demonstrated the above findings; however, several limitations should be noted. First, the measurement of NRS may not be accurate enough. In this study, NRS was assessed by a single-item questionnaire, which has not been validated. In addition, since a single-item question cannot assess multiple aspects of NRS, such as severity, the assumption of the pathway that NRS leads to poor subjective health and economic loss is limited. However, previous studies have demonstrated that it captures core symptoms and it is useful for practical purposes such as mass screening in health checkups. Therefore, the results shown in this study are still valuable.

Second, the majority of participants were construction workers. As mentioned earlier, construction workers are known to have different sleep patterns compared with other occupational groups^{28, 29)}. Because they experience greater physical strain than white-collar workers⁴¹⁾, a decline in sleep quality and the occurrence of NRS may have a particularly pronounced impact on their work performance. Conversely, white-collar workers, whose tasks rely primarily on cognitive functioning, may be more sensitive even to slight declines in concentration, potentially leading to a greater perceived reduction in work performance. Therefore, the impact of NRS on work performance may differ by occupation. Moreover, although sex was included as a covariate in this study, the majority of participants were male, indicating a participant bias. Although generalizability is limited, focusing on a high-risk industry such as construction may enhance occupational relevance.

Third, all variables used as outcomes and explanatory factors in this study were based on participants’ subjective responses to single-item questions. Objective assessments of health status and sleep—such as physiological markers, imaging tests, and polysomnography—are well established. Moreover, even for subjective assessments, validated standardized questionnaires are available. Therefore, future studies should include more detailed evaluations using objective measures and validated questionnaires. Nonetheless, health checkup screenings need to remain simple, and the associations identified through single-item subjective assessments, as used in this study, are also of practical significance.

Fourth, there may be residual confounding factors beyond those adjusted for in this study. Because this analysis was based on health examination data, it did not account for detailed occupational categories, educational attainment, income level, or other socioeconomic factors. In addition, information on over-the-counter (OTC) medication use and undiagnosed conditions was not included. Although baseline comorbidities were adjusted for using the CCI, conditions not captured by the CCI or diseases that developed during the follow-up period may have influenced the results. Additionally, the changes in sleep status, lifestyle factors, occupational factors, or other health conditions were not taken into account due to the nature of the dataset, which made it difficult to examine the causation, and there might be a reverse causation. Therefore, future studies should aim to collect and incorporate data on a broader range of covariates, including the factors mentioned above.

Conclusions

This large-scale longitudinal study of Japanese workers demonstrated, for the first time, that NRS is associated with poorer subjective health and approximately 4% lower work performance compared with RS. While NRS has been linked to various physical and mental disorders, the present findings suggest that it also adversely affects individuals’ subjective health, independent of actual disease onset. Furthermore, NRS contributes to decreased labor productivity. Accordingly, early intervention, including sleep education and CBT-I, may help improve workers’ health and enhance productivity at both individual and organizational levels.