Introduction

Sleep is a biological necessity that accounts for approximately one-third of our life. Adequate sleep is important for both mental and metabolic health and helps minimize accidents and injuries caused by fatigue and sleeplessness1. Poor sleep—which is measured by validated methods like polysomnography, simple sleep duration, or sleep time events like snoring—is associated with increased risk of chronic disease, mental health, or even mortality1,2,3,4.

Heat is one of the multidimensional pathways of climate change that leads to increased heat-related illnesses, poor mental health, and poor work capacity5. For instance, higher temperature disrupt insulin activity and may cause an increase in diabetes incidence6 and may negatively affect both blood pressure and cardiovascular health, potentially leading to greater risk of hypertension7. In terms of sleep, a hot sleeping environment can lead to increased core body temperature, which can in turn leads to the disruption of the natural sleep-wake cycle8. This influence of climate change is more pronounced in highly populated urban areas, exacerbating pre-existing urban heat island effect due to urbanization9 and population-weighted impact of global temperature increase is likely to be far higher than the global average5.

Numerous studies on climate change and sleep have pointed out that the increase in ambient temperature leads to sleep disruption. However, many studies were based on small population ranging from a few dozen to a few hundred participants; these may not effectively estimate the impact of the temperature rise due to climate on the general population10,11,12. Few of the large-scale studies, which may represent nationwide or worldwide population, used limited in outcome measurements. Some used subjective sleep reports of participants, which did not use a validated measurement tool13. Others employed mean that can measure limited aspects of sleep quality, such as sleep duration or apnea-hypopnea index (AHI), which may not portray the full extent of climate change’s impact14,15,16.

The Pittsburg Sleep Quality Index (PSQI) is a self-report based assessment tool that evaluates sleep quality over one-month period17. It is comprised of seven components including sleep duration and sleep disturbance that can portray broader, multidimensional impact of climate change on sleep quality. This study explored the impact of the temperature rise due to climate change by using PSQI on nationwide population of Korea.

Methods

We investigated to determine the independent effect of the temperature difference between the yearly average temperature and climate normal on the prevalence of poor sleep quality in nationally representative population of Korea using the cross-sectional data of Korea Community Health Survey-2018 (KCHS-18). KCHS is collects data annually from national representative population of adult Koreans living in 256 administrative districts of Korea18. KCHS changes part of the survey questions periodically, and included Pittsburgh Sleep Quality Index (PSQI) questionnaire for 2018 survey, making KCHS-18 the only nationwide data with validated instrument to measure sleep quality of Korean population.

Of 228,343 KCHS-18 participants, 221,539 Koreans livening in continental Korea mainland were eligible for this study. First, 10,405 participants with missing value for PSQI or potential confounders were removed. A total of 2,581 participants did not respond to PSQI components. Most of the nonresponse or missing values were regarding household income (n = 7,734). In total, 211,134 participants were used for analysis. The response rate was 95.3% (Fig. 1).

Fig. 1
figure 1

Flow chart. KCHS-18 participants who did not respond to or had missing value on Pittsburgh sleep quality index, or potential confounders did not enter this study. Some ineligible participants did not respond to more than one category.

Full size image

Temperature and associated data used for analysis—yearly average temperature, average temperature difference, effective daily sunshine, and normalized difference vegetation index (NDVI)—were measured and analyzed at district level. ERA5 Daily Aggregates from Google Earth Engine were used to estimate annual average temperature for each district. Daily temperatures at the district level were used to calculate the yearly average temperature in 2018. Meteorological data from Korea Meteorological Administration’s 95 Automated Synoptic Observing System (ASOS) was used to extract climate normal temperature, and effective daily sunshine. Yearly average temperature from 1961 to 1990 was used for climate normal temperature. For metropolitan cities, the meteorological data was collected from each ASOS in center. For rural districts, the data was drawn from the closest ASOS. Average temperature difference was calculated by subtracting the climate normal from the yearly average temperature. Visualized 2018 temperature at district level is provided in the supplementary material (Supplementary figure S1, S2) Greenness, measured through NVDI, is known to moderate the impact of temperature on factors such as physical activity, depression, and mortality19,20,21. NDVI measures the vegetation or greenness by calculating the difference between the reflection of infrared and visible light by vegetation20. NDVI ranges between − 1 and 1, with high positive value indicating healthy plant growth in the area, near zero value indicating barren soil, and negative value indicating standing water like lake, ocean, or snow19,20. One of the key reasons of urban heat island effect is the increase in the impervious surface area, which does not allow water to soak into ground, relative to pervious surface area, which allows soaking22. According to a systematic review on urban heat island, NDVI is the most cited indicator in studies on urban heat island effect and land surface temperature23and could be used as a reliable indicator that shows significant negative correlation with urban heat island’s impact24. For this study we used the mean NDVI level at district level between June 2018 and August 2018. NDVI was calculated based on MODIS NDVI data that measures temporal and spatial variations in vegetation activity at 16-day interval with 500 m spatial resolution25. The NDVI zonal mean for each district was calculated using ArcGIS.

Age, sex, socioeconomic status (education level and household income), and health related factors (physical activity, hypertension, diabetes, and depression) were included as potential confounder. Education level was divided into three categories: ≥college, high school, and middle school≥. Household income was divided into three groups at 2 million Korean Won interval. The cutoff for physical activity was at least 30 min of walking for at least five day per week. Hypertension and diabetes status was determined based on the participants’ self-report of doctor’s diagnosis. Patient Health Questionnaire-9 (PHQ-9), a screening test for major depressive disorder, was used to measure depression status26. PHQ-9 score of 10 or more, which represents moderate to severe depressive symptom that require clinical assistance like antidepressant or hospitalization, was used as cutoff27. For regional factors, average yearly temperature, average temperature difference and effective daily sunshine rate were included.

The main outcome of this study is overall sleep quality measured by PSQI collected from each individual who participated in KCHS 2018. PSQI is a self-reported questionnaire that evaluates the sleep quality during the previous month28. The validated Korean version of PSQI was used for this study. The Cronbach’s α was 0.84, showing high internal consistency, and all seven component scores of PSQI showed significant differences between insomniac patients and the healthy control29. The PSQI score of 5 or higher, which indicates poor sleep relative to clinical and laboratory measures was used as cutoff28,30. Sleep disturbance components of PSQI (trouble falling asleep, sleep fragmentation, nocturia, trouble breathing, cough or snore, too cold during sleep, too hot during sleep, bad dreams, and pain) and sleep duration were also analyzed as secondary outcomes. PSQI measures the frequency of sleep disturbance events in four intervals: not during past month, < 1/week, 1–2/week, and ≥ 3/week. To estimate the association between the temperature increase due to climate change and regular occurrence of sleep disturbance event, the frequency of once or twice a week or higher was defined as sleep disturbance event for analysis purposes. Human adults require 7–8 h of sleep2,4. Sleep duration of less than 7 h was defined as insufficient sleep duration for this study.

We used chi-square test and student’s t-test to explore the differences in individual and regional level variables according to district type. Weighted values provided in the KCHS-18 were used to calculate the percentages of each category for categorical variables. To estimate the effects of regional-level and individual-level variables on the prevalence rate of reporting sleep-related problems, we used multilevel analysis with random intercept. Random slopes for temperature variables were tested and found to vary by region, but the final model used random intercept for consistency. Only regional level variables (meteorological variables, and district type) were adjusted for crude models. Adjusted model was adjusted for regional variables and individual variables. The participants were also stratified by district type for analysis. The partition of variance by individual and regional level variables was also calculated with intraclass coefficient based on covariance parameter estimates (Supplementary table S1).

For sensitivity analysis, temperature related variables of 2018 summer (temperature average and temperature difference of July and August 2018) and 2017 (temperature average and temperature difference of 2017) were used to measure potential effect of lagged temperature. Less rigorous frequency cutoff—less than once a week—was used to further test the association between temperature rise due to climate change and the likelihood of sleep disturbance events. Different cutoffs for sleep duration were also use to test the consistency of association. To account for potential interaction between household income and temperature difference, we performed an additional analysis with household income*temperature difference as interaction term.

SAS 9.4 was used (SAS Institute., Cary, NC, USA). The statistical significance level was P-value < 0.05.

Results

Among 211,159 participants, 120,166 (81.5%) lived in urban districts. On average, urban dwellers were younger, attained higher degree of education, have higher household income, and were more likely to participate in physical activity than rural dwellers. The prevalence rate of overall sleep disturbance measured by PSQI were similar between the two district types (Table 1). As for sleep disturbance related events, urban dwellers were more likely to report sleep fragmentation, but were less likely to report in all other sleep disturbance events (Supplementary table S2).

Table 1 The general characteristics of 211,134 participants according to district type.
Full size table

When stratified by district type, urban participants were more likely to report overall poor sleep quality per 1 °C increase in temperature difference. (aOR 1.16 95% CI 1.05–1.29). Increase in temperature average was not associated with overall poor sleep quality for all participants regardless of district type. (Fig. 2, Supplementary table S3). The above associations were consistent when 2018 summer temperature data or 2017 temperature data was used (Supplementary table S4, S5).

Fig. 2
figure 2

The odd ratios of reporting overall poor sleep quality according to district type. *Overall poor sleep quality is defined as pittsburgh sleep quality index score of 5 or higher. Temp avg: per 1 °C increase in 2018 yearly average temperature. Temp diff: per 1 °C increase in 2018 yearly average temperature compared to the climate normal temperature of 1961–1990. Adjusted by age, sex, education, household income, physical activity, hypertension diabetes, sunlight exposure rate, NDVI (normalized difference vegetation index), and district type.

Full size image

Regarding sleep disturbance event, urban participants were more likely to report sleep fragmentation (aOR 1.17 95% CI 1.03–1.32), cough or snore (aOR 1.21 95% CI 1.01–1.44), too cold during sleep (aOR 1.21 95%CI 1.01–1.44), bad dreams (aOR 1.22 95%CI 1.06–1.39), and pain (aOR 1.25 95% CI 1.08–1.45) as temperature difference increased. The increase in temperature average was only associated with being too hot during sleep (aOR 1.15 95% CI 1.03–1.29). (Fig. 3, Supplementary table S6) In contrast, no significant associations were found between the temperature increase and the most of the sleep disturbance events—for instance in sleep fragmentation (aOR 0.98 95% CI 0.92–1.03) or pain (aOR 1.03 95% CI 0.96–1.10)—except for too hot during sleep (aOR 1.15 95% CI 1.03–1.29). The above association between sleep disturbance events and increase in temperature difference among urban dwellers was mostly consistent when more generous cutoff (sleep disturbance event frequency of less than once a week or higher) was used (Supplementary table S7).

Fig. 3
figure 3

The odds of reporting sleep disturbance related events among urban participants due to climate change’s impact on temperature rise. *The cut-off frequency for sleep disturbance is at least once a week. (A) the odds of reporting sleep disturbance related events among urban participants per 1 °C increase in temp diff. (B) the odds of reporting sleep disturbance related events among urban participants per 1 °C increase in yearly average temperature. Adjusted by age, sex, education, household income, physical activity, hypertension, diabetes, sunlight exposure, and NDVI (normalized difference vegetation index).

Full size image

When the interaction term between the household income and temperature difference was included in the analysis, a gradient association between the impact of temperature difference on poor sleep quality and household income was found. By point estimate, the urban participants with high household income were 7% more likely to report poor sleep quality per 1 °C temperature difference, and those with middle or low household income were 17% and 23% more likely to report poor sleep quality (high household income: aOR 1.07 95% CI 0.96–1.20; middle household income: aOR 1.17 95% CI 1.05–1.30; low household income: aOR 1.23 95% CI 1.10–1.37). There was no association between the temperature difference’s impact and poor sleep quality among rural participants, regardless of household income level (Supplementary figure S3).

Discussion

This study showed that temperature rise due to climate change is associated with both poor sleep quality as shown by overall PSQI score, and poor sleep quantity as shown through lower sleep duration. A 1 °C rise from climate normal was associated with a 16% increase the likelihood of reporting an overall sleep-related problem among urban dwellers, and a 1 °C rise from climate normal was associated with 10% increase in the odds of reporting a sleep duration of less than 7 h. The population attributable risk of 1 °C rise from climate normal was 5.9% among urban dwelling Koreans; or in other words, 2.49 million urban dwelling Koreans suffered sleep-related problem per 1 °C rise from climate normal. Considering that the global mean temperature have already reached 1.45 °C above the preindustrial baseline 202331, the number of affected urban dwellers will become higher as the climate change worsens over time. The outcome of this study is consistent with general outcomes of previous studies8.

A multinational study on sleep duration revealed that the residents dwelling in hotter region tend to experience more sleep loss than those living in colder region, suggesting people’s limited adaptation to high ambient temperature32. Our study showed that higher average temperature was associated with sleep disturbance due to too high temperature during sleep, but most other sleep disturbance events were associated with increase in temperature from climate normal rather than higher average temperature. Sleep disturbance due to too high temperature during sleep may insinuate the impact of high baseline temperature on sleep while sleep disturbance due to other causes demonstrates the broader impact of temperature rise from climate normal on sleep. Human sleep cycle is deeply connected with thermoregulation, which is controlled by circadian rhythm and sleep regulation. Heat load suppress the decrease in core body temperature, disrupting thermoregulatory system and leading to delayed sleep onset33. Stress due to delayed sleep onset can also increase cortisol level through the activation of hypothalamic-pituitary-adrenal (HPA) axis which further sleep disruption34.

Urban dwellers have been reported to be more vulnerable to the impacts of climate change than rural dwellers5. Temperature in urban areas are reported to be rising more rapidly than the global average5and this rise in mean temperature over climate normal may contributed to higher mortality in urban populations35. Over 80% of the Korean population represented by KCHS lives in urban environments, far higher than the global average of 55%, and this proportion is expected to increase due to urban expansion and rural-urban migration5. Focusing on city level intervention could be an effective way to control climate change-related impact on sleep quality.

It is important to note that the significant association between temperature rises due to climate change and poor sleep quality persisted after controlling for depression. Poor sleep quality and depression have a bi-directional association36,37. It was particularly interesting find that the independent effect of temperature difference from climate normal significantly increased the likelihood of reporting sleep-related problems and shorter sleep duration, while higher average temperature did not have a significant independent effect on the increased odds of reporting poorer sleep quality.

Climate normal of a region likely represents temperature acclimatized by the region’s denizens. A previous study on temperature and sleep demonstrated that the increase in temperature during night caused delayed sleep onset and advanced offset which led to shorter sleep duration38. Ambient temperature higher than acclimatized climate normal, may hamper thermoregulatory mechanisms associated with sleep. Faster body heat dissipation leads to better subjective sleep latency onset39and it is possible that subjective sleep quality is associated with thermoregulatory functions and body temperature rhythms, which are in turn important for both sleep onset and quality40.

Integrating green-blue infrastructures into urban planning may help mitigating the climate change’s impact on the sleep health of the urban dwellers. Green infrastructures like street trees and green roofs can reduce urban air temperature by cooling and improving microclimate41. Permeable pavements, which allows the underlying soil to absorb precipitation and provide evaporative cooling, can also mitigate the heat burden begot by climate change42. All these integrating solutions should also take account for vulnerable population with fewer resources, like air conditioning for social integration of the measures against climate change42.

Limitation

This study has few limitations, coming from its study design and data source. KCHS-18 collects data from the community dwelling population who are available for computer-assisted personal interviews. Koreans with cognitive impairment or who are hospitalized due to illness are not represented in this study. We could not assess the availability of air conditioning or noise pollution status, which may moderate the association between the temperature rise due to climate change and sleep quality. Information on pre-existing sleep-related disorders, like insomnia or obstructive sleep apnea was not available. As the exact time of the survey was not available, we had to use yearly average temperature, rather than daily temperature, to estimate potential impacts of climate change on sleep within previous month of the KCHS. This study used NDVI to substitute urban heat island effect in its analysis. Future studies may consider using long term urban heat island effect to estimate the impact of the temperature rise due to climate change. ASOSs are concentrated in historically populated areas, and the paucity of ASOSs in rural areas, may have limited the accuracy of this study. Still, ASOSs are the best available meteorological data source that provided data as early as 1960s, and KCHS is the only nationwide survey that extracts survey data from every district of Korea collected by trained interviewers. The authors have used the most accurate and valid data available to represent the independent effect of climate change on the entire Korean population.

Strengths

This study utilized a wide range of potential confounders and nationwide population from KCHS database, and this study’s outcome represents the entire population living in contiguous Korea. To the authors’ knowledge, this is one of the first study that explored the potential long-term influence of temperature difference due to climate change on nationwide population, with a validated tool to measure sleep quality. This study also demonstrated that even after adjusting for depressive symptoms, the independent effect of the temperature rise on poor sleep quality remain significant.

Conclusion

This study showed that temperature rise due to climate change deteriorates sleep quality of urban dwellers through sleep disturbance events and a reduction in sleep time, using the best available regional and individual data that represent every district of Korea. Further studies regarding the daily, short-term effect and the yearly, long-term effect of climate change on the various dimensions of sleep is needed to mitigate the climate change’s impact on sleep health.