For this study, we used data from the 2018 Korea Community Health Survey (KCHS), conducted by the Korea Centers for Disease Control and Prevention. The KCHS is a nationwide, cross-sectional survey with a study population from a multi-stage cluster sampling, stratified by primary sample units of civilian, non-institutionalized Korean households by geographic area, age, and gender groups. The survey is conducted annually, and it collects information through face-to-face and a paper-assisted personal interview between interviewers and respondents. The sample size for the KCHS is 900 subjects in each of 253 community units, including 16 metropolitan cities and provinces. Because the population sample is selected using a probability proportional to the sampling methods and, weighted by statisticians who performed post-stratification and considered the non-response rates, samples are representative of the Korean population.

This study included individuals aged 19 years or older. Of the 228340 total survey participants, respondents who did not provide data on their income, employment status, smoking status, alcohol consumption, meal activities, exercise behavior, and scale for sleep quality, or scale for depressive symptoms were excluded. Finally, a total of 89470 eligible participants were included in this study (Fig. 1).

Sleep quality was measured using the Korean version of the Pittsburg Sleep Quality Index (PSQI-K) which included several sleep-related behaviors; in 2018, PSQI-K was included in the KCHS for the first time. The PSQI is a 19-items self-reported questionnaire that measures the quality and patterns of sleep over a 1-month duration [12]. This study consisted of 7 sleep components: C1, subjective sleep quality; C2, sleep latency; C3, total sleep duration; C4, sleep efficiency; C5, sleep disturbance; C6, use of sleep medication; and C7, daytime dysfunction. The PSQI-K has been validated and shown to have adequate reliability [13]. Total PSQI score has been widely used before to measure sleep quality. In this paper, the lower the sleep quality, the higher the total PSQI score, and the higher the sleep quality, the lower the total PSQI score, the higher total PSQI score was regarded as poor sleep quality.

Depressive symptoms were measured using the Patient Health Questionnaire-9 (PHQ-9) that has been widely used in population-based studies because it’s self-reporting and simple. The Korean version of PHQ-9 has also been validated and shown high reliability [14].

The emotional regulation activities measured in this study were as follows: regular breakfast, regular exercises, smoking, and alcohol drinking. The number of days that people had eaten breakfast in the previous week among people who reported eating their meals regularly was used to define regular breakfast. Among the participants who answered to exercise behavior, the most recent one-month regular exercise was defined as regular exercise. We defined alcohol drinking-related behavior with a subscale of The Korean version of Alcohol Use Disorders Identification Test to measure how much alcohol was consumed in the most recent month. The variable of smoking was classified as snoking every day, occasionally smoking, and not smoking.

Since the data from KCHS were collected using stratified, clustered, and systematic sampling, complex survey dataset analysis methods were used while following the guidelines of KCHS analysis, which were published by the Korea Centers for Disease Control and Prevention [15]. Descriptive statistics were used to identify the participants’ baseline characteristics. Logistic regression analysis was done, and the receiver operating characteristic (ROC) curve was drawn to evaluate the feasibility of sleep quality in predicting depressive symptoms. To explore the effects of emotion regulation activities that mediate the link between sleep quality and depressive symptoms, correlation analyses were done. We used statistical methods in a way that created structural equation methods and validated this model. Since there are many missing values, the maximum likelihood with the missing value method is used for estimation. We used a different stepping algorithm in nonconcave regions for maximization. Optimization was done by 300 maximum iterations. Covariance structure used the following: sex, age, household income level, employment status, and current medical comorbidity which could affect sleep quality (diabetes mellitus, hypertension, cerebrovascular disease, asthma, allergic rhinitis, atopic dermatitis). Differences were considered statistically significant at ap-value less than 0.05. Statistical analyses were performed using STATA version 16 (STATA Corp., College Station, TX, USA).

This study was approved by the Institutional Review Board of St. Vincent’s Hospital in Korea (VC21ZASI0245).

The demographics and baseline characteristics of participants are presented in Table 1. The mean age of the participants was 54.52 ± 17.58 years, and 44.78% of the participants were male. The average monthly household income of the participants was 46.88 ± 35.84 million won, and 61.58% of participants were currently employed. 32.68% of the participants had medical comorbidities. 16.01% said they smoke every day, 1.32% said they smoke occasionally and 82.66% said they don’t smoke. 32.55% of the patients said they didn’t drink alcohol, 16.73% said they drink less than once a month, 8.91% said they drink twice or four times a month, 19.86% said they drink two to three times a week, and 8.14% said they drink more than four times a week. Among those who said they eat regularly, 11.33% said they had never eaten breakfast in the past week, 11.33% said they had eaten once, 3.51% said they had eaten twice, 4.44% said they had eaten twice, 2.50%, 5.65%, and 1.46% said they had eaten six times. Of the participants who responded to exercise behavior, 39.20% reported exercising regularly.

When the logistic regression analysis was performed to see how well the total score of PSQI predicted the degree of depressive symptoms measured by PHQ-9, the odds ratio was 1.40 (95% confidential interval [CI] 1.40–1.41, p < 0.001). Based on these results the ROC curve was drawn, and the area under the ROC curve was measured at 0.6690, which is shown in Fig. 2.

Prior to drawing the Structural Equation Model, Pearson correlation analysis was performed to evaluate the association between each variable, and the results are shown in Table 2. Emotion regulation activities were significantly correlated with PSQI total scores but were not significantly correlated with PHQ-9 total scores.

The structural equation model was drawn to evaluate how emotional regulation activities mediated between sleep quality and depressive symptoms, and the results are shown in Fig. 3 and Table 3. The direct effects of the total PSQI scores on each emotional control behavior are as follows: the correlation between the total score of PSQI and eating breakfast was 0.11 (95% CI, 0.10–0.11), the effect for regular exercise was 0.05 (95% CI, 0.05–0.05), the effect for smoking was 0.06 (95% CI, 0.06–0.06). Between PSQI and PHQ-9, the mediating effect of eating breakfast was -0.53 (95% CI, -0.56–-0.51) and the mediating effect of regular exercise was -0.09 (95% CI, -0.12–-0.05). The mediating effects between PSQI and PHQ-9 of smoking and alcohol drinking were -0.21 (95% CI, -0.26–-0.17) and -0.16 (95% CI, -0.17–-0.15), respectively. All the results were significant with a p-value < 0.0001.