Comparison of sleep and health behaviors among diabetic patients and non-diabetics in Phitsanulok, Thailand: a cross-sectional study

Study site

This study utilized data from a previous case-control study on diabetes and pesticide exposure (Juntarawijit & Juntarawijit, 2018). The data on health behaviors were collected from February to May 2016 from diabetic patients (T2DM) and a non-T2DM control group living in the rural community of Bang Rakam, a district with 95,098 people (in the year 2018) in Phitsanulok province, Thailand. The district is located in the lower northern part of Thailand, about 400 km from Bangkok.

Study participants

The diabetic patients were those who had come to receive follow-up services at seven health promoting hospitals, which were randomly selected, using random number tables, from all 21 local hospitals in the target area. All diabetic patients who met the inclusion criteria were approached at their home by village health volunteers to take part in the study. In this study, the T2DM group was limited to those aged 30–85 years and free from other chronic diseases, such as heart disease, allergies, chronic pulmonary disease, and cancer. For each diabetes case, one healthy control (non-T2DM) who was free from diabetes and met the same inclusion criteria as the case was also approached by the same health volunteer based on the convenience sampling method. The control group were neighbors of the diabetic patients matched for gender and age (± five years.).

Study questionnaire

In addition to demographic information, data on sleep duration and other health behaviors were collected using an interviewer-administered questionnaire during a face-to-face interview, which was written in the Thai language (Juntarawijit, 2019b). Before use, the questionnaire was tested for question sequencing and understanding. An interview took place at home of each participant. The participants’ self-reported sleep duration was collected using the question “How many hours do you usually sleep per day?”. Participants were classified as ‘current smoker’ if they had smoked 100 cigarettes or more in their lifetime and they currently smoke cigarettes. Those who drank alcohol 2-4 times a week were classified as ‘alcohol use’. Data on food consumption, including consumption of meat, sausage, vegetable, fruit, sweets, rice and sweet soft drinks, were also collected using ‘yes or no’ questions. In this study, a modified Food Frequency Questionnaire (FFQ) was used (Barrat et al., 2012). Only types of foods found to be related to diabetes and those often found in Thailand were included in the survey. Information on personal lifestyle (whether they are a morning person or an evening person) was collected using the question “What is the lifestyle that best describes you, morning people or evening people: “morning people” refer to those who usually go to bed early and like to work or being active during the day; “evening people” are those who go to bed late, being alert and prefer to work at night?”. Participants were also asked to report how frequently they did certain physical activities (walking, biking, playing sports or sweating excessively from exercise or physical activity but not from hot climate or health problems) and watched television during their leisure time using two categories: absent (never, rarely) and present (sometimes, often, almost always). A modified Baecke Habitual Physical Activity Questionnaire (BHPAQ) was used (Florindo & Latorre, 2003). Body mass index (BMI) was calculated by dividing body weight (in kg) by height (in meters squared). The high BMI group was those with BMI ≥25.00. For waist to hip ratio (WHR), a high WHR referred to men with WHR ≥0.90 or women with WHR ≥85. For waist circumstance (WC), a high WC referred to men with WC ≥90.0 or women with WC ≥80. All of these measurements were assessed by the health volunteers. Data were collected by 50 village health volunteers who were trained on how to use questionnaires and how to interview study participants.

Statistical analysis

Demographic and health behaviors were analyzed using descriptive statistics and Chi-square test for comparison of categorical data. To identify predictive factors of sleep duration, logistic regression was performed, adjusted for gender, age (continuous), waist to hip ratio (WHR) and lifestyle (evening person vs morning person). The predictive factors of physical activity were also analyzed using ordinal regression, with physical activity categorized as never, rarely, sometimes, often and almost always. All analyses were performed using IBM SPSS statistics (version 19). Confidence intervals of 95% were used to determine significant statistics and all p-values are two two-sided. In this study, listwise or case deletion was used to handling of missing data.

Ethical statement

This study was approved by the Ethics Board of Naresuan University (project number 402/59). Written informed consent for an interview and participation in the study was obtained from each of the subjects before the interview process.

Health behaviors among diabetes and non-diabetes groups

It was found that people with diabetes in the rural area of Thailand had healthy behaviors regarding eating, smoking, alcohol consumption, sleep pattern and duration, but not for physical activities and exercise. These results contradicted the common perception that people with diabetes all have unhealthy lifestyles. A large study in Australia reported no positive improvements among the recently diagnosed T2DM in their lifestyle behaviors, such as physical activities and fruit and vegetable consumption (Chong et al., 2017). These findings implied that lifestyle prevention of diabetes (especially type 2 diabetes) included healthy lifestyle based on dietary patterns alone, may not be sufficient and should always include physical activities as an integral part.

Most of the participants in this study had a healthy sleep pattern and lifestyle. Over 80% of the participants usually sleep 7–9 hours per day, which is considered to be a healthy amount of sleep (Ip & Mokhlesi, 2007). However, a closer look revealed that there was a higher percentage of the T2DM group compared to the control group who sleep less than six hours per day (3.2% vs. 2.4%) or more than nine hours per day (9.3%% vs. 7.5%) (Table 1) than the control group. Comparing these results to the literature, these participants had a better sleep pattern. The National Sleep Foundation in the United Sates reported that about 30% of middle-aged men and women slept less than six hours per night (Ip & Mokhlesi, 2007). Another study in the United States reported about half of diabetes group had sleep problems (Shamshirgaran et al., 2017).

Further analysis revealed that sleep ≥8 hours was significantly associated with being women (OR = 1.27, 95% CI 1.03-1.56) and having a high WHR (OR = 1.28, 95% CI 1.04-1.59). A short sleep duration (≤5 hours) was significantly associated with a high WC (OR = 3.14, 95% CI 1.13-8.75) and women with a high WC (OR = 3.47, 95% CI 1.21-9.97) and being an evening person (OR = 5.92, 95% CI 3.46-10.13) and women who are evening people (OR = 7.55, 95% CI 4.17-13.66). These results were supported by a study that reported that sleep disturbance increases metabolic disorders and obesity (Chattu et al., 2019).

Compared to the control group, the T2DM group were more overweight and had a higher BMI, (24.9 ± 4.7 vs. 23.8 ± 4.3), higher WHR (0.91 ± 0.14 vs. 0.90 ± 0.11) and higher WC (36.8 ± 11.8 vs. 35.6 ± 11.9) (Table 1). This was not surprising since obesity is a well-established risk factor of diabetes. In an epidemiological study, a short sleep was associated with BMI and weight gain (Leproult & Van Cauter, 2010). In laboratory studies, sleep deprivation affected sympathovagal balance, evening concentrations of cortisol and ghrelin hormones or hunger hormones, but decreased glucose tolerance, insulin sensitivity and leptin, a hormone controlling body weight (Van Cauter & Knutson, 2008). These changes increase blood glucose (Nedeltcheva & Scheer, 2014) and appetite for carbohydrate-rich food (Ip & Mokhlesi, 2007).

Compared with other studies, the number of cigarette smokers (10.8% of T2DM and 14.3% of non-T2DM) and alcohol users (6.6% of T2DM and 9.7% of non-T2DM) in this study were relatively small. In the United States, the prevalence of cigarette smoking among adults with diabetes was about 23.6% (Ford et al., 2004). A study in California, United States, reported that 50% of participants in the diabetes group consumed alcohol (Ahmed et al., 2006).

Compared to the control group, the prevalence of smoking and alcohol consumption among the T2DM group was significantly lower (p<0.01). A recent study in Australia also reported a higher rate of smoking cessation (OR for quitting smoking = 2.71, 95% CI 1.59-4.63) among recently diagnosed T2DM as compared to a healthy control group (Chong et al., 2017). This behavior change was often claimed to be a result of diabetes care programs and global trends of cigarette and alcohol consumption (Shi et al., 2013).

Concerning eating behaviors and choice of food, the diabetes group trended to be healthier than the control group and were more likely to eat foods that are believed to be good for health. Compared to non-T2DM, there was a significantly higher percentage of the T2DM group who reported eating vegetables (91.5% vs. 89.0%, p=0.03) and chicken (16.8% vs. 13.5%; p=0.03) and the opposite was true for beef (34.7% vs. 40.6%, p<0.01) and eating more than a cup of rice per meal (37.4% vs. 42.6%, p<0.01). A similar study in Australia also found a lifestyle change among newly diagnosed diabetic patients (Chong et al., 2017). These results may be useful for the diabetic prevention program, by advising their patients to eat good quality and healthy foods.

However, it must be noted that there were a large portion of participants in both T2DM and non-T2DM groups who reported eating fruit (63.6% vs. 61.7%), sweets (40.6% vs. 38.5%), and drinking sweet soft drinks (46.4% vs. 50.7%). Eating these foods might affect blood sugar and sleep pattern.

Comparing the frequency of several physical activities performed during their leisure time, participants in the T2DM group were less active than the control group. A significantly higher number of T2DM participants admitted to being less active compared with people of the same age and not doing exercise or playing sports as often as the control group (Table 1). There were also fewer T2DM participants who reported doing walking (41.5% vs. 51.1%) and riding a bicycle (17.4% vs. 26.7%) during their leisure time. These results are supported by other studies. In a study in rural communities of Missouri, Tennessee, and Arkansas, it was reported that 37% of T2DM patients had no leisure-time physical activity (Deshpande et al., 2005). Hays & Clark (1999) also found that over half of T2DM (54.6%) patients, mostly elderly females, had no weekly physical activity. However, a study in Nepal reported that 52% of diabetic patients were moderately active and 28% were highly active (Kadariya & Aro, 2018). This discrepancy in physical activity might be related to the culture and lifestyle of the patients. It was expected that study participants in this study would be more active because they are rural people who mainly work in agriculture.

Further ordinal regression analysis revealed that physical activities were associated with obesity (BMI). Those with a lower BMI trended to have a higher rate of walking, biking and exercising (Table 5). This result is well consistent with literature. A study on outpatients with T2DM and their matched controls found that the total energy expenditure (<300 kcal/day), number of steps (<1500 /day), physical activity duration (<130 min/day) and active energy expenditure/day (<300 kcal) were all lower in the diabetes group (p<0.05) (Fagour et al., 2013; Hamasaki, 2016). This lower physical activity might be partially due to a fear of joint or leg pain (Dutton et al., 2005) or hypoglycemia (Brazeau et al., 2008). Therefore, this problem requires greater attention, and exercise plans specifically designed for those with diabetes need to be developed.

Association between diabetes and sleep

After applying several grouping methods, the association was found to be significant only when sleep duration was grouped to ≤5 hours, 6–7 hours and ≥8 hours. In comparison to the 6–7 hours group, further analysis using logistic regression found that a significant association between diabetes and sleep was only among women with ≥8 hours of sleep (OR = 1.27, 95% CI 1.03-1.56) after being adjusted for age, gender, WHR and lifestyle (Table 3). A similar result was also reported in a study in Finland, which found that sleep of ≥8 hours increased the risk of diabetes among middle aged women (Tuomilehto et al., 2008). The effect of oversleeping on the risk of developing diabetes is well established, but most of the previous studies use 7-8 hours of sleep as a reference and defined oversleep to be 10–12 hours per day (Chattu et al., 2019). It was also noticed that the relative risk of diabetes in this study (OR = 1.27) was rather low compared with those reported in previous studies, which mostly reported sleep to increase diabetes risk by 2–3 times (Heianza et al., 2014; Yaggi et al., 2006). These results might be explained by the fact that sufficient sleep depends on both quantity and quality of sleep (Chattu et al., 2019). Since most participants in this study are rural villagers with a healthier lifestyle, they were more likely to have a good sleep and thus, require a shorter sleep duration. This was supported by the fact that only approximately 7% of the participants (T2DM and non-T2DM) who classified themselves to be evening people that like to stay up late at night (Table 1).

It was found that less than 7% of the study participants are evening people and the association between diabetes and lifestyle was not statistically significant. These results were in contrast to previous studies. In a large study in Korea, there was a large proportion of people who were evening people, and this group was at risk of diabetes (OR = 1.73, 95% CI 1.01-2.95). (Yu et al., 2015). The differences in workload, lifestyle, social activities and technology might affect sleep patterns of the two groups. Most of the participants in this study were rural villagers, while those in the Korean study were urban people with a modern lifestyle. In Thailand, most villagers usually go to bed early after being exhausted from hard, physical work on the farm during the day and they usually wake up early in the morning to have enough time to prepare food for their family members and the Buddhist monks. More study should be conducted to verify this issue.

Study limitations

One of the main limitations of this study was that it used a cross-sectional design and there was a lack of data on diabetes onset. Since the relationship between diabetes and sleep is double-sided, it therefore cannot be determined whether sleep causes diabetes or the disease interferes with the sleep pattern of the patient (Chattu et al., 2019). This bias will cause a positive effect and overestimate the association of diabetes and sleep. Without data on diabetes onset, behavior change and disease duration cannot be analyzed. This is also true of the effect of sleep duration on glucose control. These two issues are often reported in literature (Chong et al., 2017). This study only collected data regarding sleep quantity, while previous studies have suggested that both the duration and quality of sleep could have an effect on diabetes (Lee et al., 2017). Lastly, the information regarding sleep duration and other potential risk factors were self-reported, and the reliability of the questionnaire used was not validated, and therefore, recall bias was likely to occur. However, if the bias do occur, it could equally affect both the study and the comparison group.

Underlying data

Figshare: Sleep and health behaviors among diabetic and non-diabetic groups. 10.6084/m9.figshare.8246780 (Juntarawijit, 2019a)

This project contains the following underlying data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Extended data

Figshare: Questionnaire-sleep and health behavior among diabetes. 10.6084/m9.figshare.8298689 (Juntarawijit, 2019b)

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).