A survey was conducted using online and self-administered methods in middle-aged subjects of 40-64 years of age in Seoul and Gyeonggi area from May 5 to May 22, 2020. Among a total of 724 questionnaire collected, 670 questionnaire including 417 online questionnaire (180 men, 237 women) and 253 written questionnaire (140 men, 113 women) were used in the study, after those with 54 missing responses were excluded. A survey response rate was 92.5%.

General characteristics of the subjects included age, body weight status, perceived health status, education level, occupation, and monthly household income. Anthropometric data of the subjects were obtained from self-administered questionnaire and the body mass index (BMI, kg/m²) was calculated using height and weight. BMI was divided into underweight (BMI < 18.5), normal weight (18.5 ≤ BMI < 23), overweight (23 ≤ BMI < 25), and obesity (BMI ≥ 25) according to the Asia-Pacific BMI classifications (World Health Organization 2000) [11].

The stress score used PSS (perceived stress scale) developed by Cohen et al. [12] and adapted by Lee et al. [13]. It was made of a total of 10 questions, and each question was evaluated by Likert 5-point scale. The range of total score was 0–40 points and the perceived stress level was severe as the score increased. The Cronbach's alpha was 0.79 in the study.

A total of 10 questions on dietary habits included three meals a day, breakfast, meal regularity and amount, food combination, green-and-yellow vegetable intake, fruit intake, protein intake, milk and dairy product intake, seaweed intake, and water intake. Each question was evaluated by Likert 5-point scale. The higher score meant good dietary habits. In this study, the upper 30% was classified as high level, the next 40% as middle level, and the lower 30% as low level for analysis. The Cronbach’s alpha was 0.83.

The level of emotional dietary behavior used emotional eater questionnaire (EEQ) developed by Garaulet et al. [14]. It was made of a total of 10 questions, and each question was evaluated by Likert 4-point scale. In the study, 0–5 points were classified as non-emotional eater group, 6–10 points as low emotional eater group, and over 11 points as emotional eater group for analysis. Before the study, the validity and reliability of the questionnaire were established. The Cronbach's alpha was 0.78 in the previous study [15] and 0.87 in this study.

The insomnia level used insomnia severity index (ISI), which was developed by Bastien et al. [16] and adapted by Cho et al. [17] as Korean version of ISI. It was made of a total of 7 questions and each question was evaluated by Likert 5-point scale. In the study, 0–7 points were classified as no clinically significant insomnia group, 8–14 points as subthreshold insomnia group, and over 15 points as clinical insomnia group for analysis. In the previous study [17], the Cronbach's alpha was 0.92; in this study, it was 0.86.

Data were analyzed using SPSS Win 25.0. For general characteristics, the frequency and percentage were calculated, and continuous variables were analyzed using independent samples t-test. One-way analysis of variance test was performed to obtain stress score according to dietary habits, emotional dietary behavior, and insomnia level, and post hoc analysis was performed using Duncan's new multiple range test. Pearson's correlation coefficient test was performed to find the association among dietary habits, emotional dietary behavior, insomnia, and stress. All results were determined the statistical significance at P < 0.05.

For written questionnaire, the instruction was presented to the subjects and the consent form was obtained from the subjects who agreed to the instruction after sufficient explanation before the survey. For online questionnaire, the same instruction and agreement form as in the case of written questionnaire were presented and then the survey was proceeded only if the subject checked the ‘I agree’ box for agreement. It was clearly stated in the instruction and agreement form that the withdrawal of the agreement could be made after the participation in the study. Also, it was explained that data would not be used except for this study purpose, and related data would be stored in the Institutional Review Board (IRB) organization for 3 years after the termination of the study, and then all data would be discarded. The study was performed after the deliberation and approval of Myongji University IRB (deliberation No.: MJU-2020-04-002-01).

General characteristics of the study subjects are show in Table 1.

The highest age range was 50-54 years as 27.5% in men and 27.4% in women. For the weight status, ‘obese’ was the highest as 46.3% in men and ‘normal weight’ was the highest as 51.7% in women (P < 0.001). The BMI was 25.17 kg/m² in men and 22.96 kg/m² in women (P < 0.001). For the perceived health status, ‘average’ was the highest as 61.6% in men and 62.6% in women. The ratio of positive perceived health status was higher in men while the ratio of negative perceived health status was higher in women (P < 0.01). For education level, college graduate was 42.2% and over graduate school was 32.2% in men; college graduate was 50.6% and high school graduate was 31.4% in women (P < 0.001). For occupation, office staff was the highest as 32.5% and then in the order of 22.1% of professionals and 17.9% of self-employed in men. On the other hand, in women, housekeeping was the highest as 30.7% and then in the order of 22.2% of professionals and 16.7% of office staff (P < 0.001). For monthly household income, the highest was over 7 million won in 27.5% of men and 20.0% of women. The lowest distribution was under 2 million won as 6.6% in men and 6–6.99 million won as 6.9% in women.

Table 2 shows the stress score of middle-aged men and women. The perceived stress score by sex was not significantly different in middle-aged subjects: The total stress score was 17.75 ± 4.31 in men and 17.89 ± 4.76 in women, and 17.82 ± 4.55 in all subjects.

Table 3 shows the stress score according to dietary habits level of middle-aged men and women. Stress score according to dietary habits was significantly higher in men with ‘low’ dietary habits level compared to those with ‘moderate’ and ‘high’ levels (P < 0.001). In women, it was significantly higher in those with ‘low’ and ‘moderate’ dietary habits levels compared to those with ‘high’ level (P < 0.05). In all subjects, stress was significantly higher in groups with low dietary habits level (P < 0.001).

Table 4 shows the stress score according to emotional dietary behavior level of middle-aged men and women.

Stress score was higher in emotional eaters in men (P < 0.05) and all subjects (P < 0.001) with significant differences. In women, stress was higher in emotional eaters, with a significant difference (P < 0.001).

Table 5 shows the stress score according to insomnia level of middle-aged men and women.

Stress score was high in men, women, and all subjects in the order of group with no clinically significant insomnia < subthreshold insomnia < clinical insomnia, with a significant difference (P < 0.001).

Table 6 shows the correlations among stress and dietary habits, emotional dietary behavior, and insomnia.

The results of correlation analysis of stress and dietary habit, emotional dietary behavior, and sleep showed significantly negative correlations between stress and dietary habits in middle-aged men, women, and all subjects (P < 0.001), and significantly positive correlations between stress and emotional dietary behavior in men (P < 0.01), women (P < 0.001), and all subjects (P < 0.001). Significantly positive correlation was observed between stress and insomnia in men, women, and all subjects (P < 0.001). Therefore, as stress increased in middle-aged subjects, poorer dietary habits and the quality of sleep habit and more emotional dietary behavior was observed.