Abstract
Background
Dichotomizing a continuous outcome, sleep quality score, is associated with loss of information, and bias. Here we compared the performance of dichotomizing sleep quality scores according to cut-point as outcome variable in multivariable logistic regression with using that as continuous outcome in an adjusted distributional method.
Methods
In this study, the data from the second phase of the Shahroud eye cohort study (ShECS) on 4710 adults were used. Sleep quality score using Pittsburg index was normalized using item response theory (IRT) method. Sleep quality score was used as a dichotomized variable in a logistic regression model and as a continuous variable in an adjusted distributional method.
Results
The overall prevalence of poor sleep quality was 44.9 % (95 %CI, 43.4–46.2). In the adjusted distributional model poor sleep quality was associated with female gender (OR = 2.08; 95 %CI: 1.9–2.2), old age (OR = 1.2; 95 %CI: 1.0–1.4), low economic status (OR = 1.3; 95 % CI: 1.1–1.4), Illiteracy (OR = 1.4; 95 %CI: 1.2–1.7), diabetes (OR = 1.1; 95 %CI: 1.0–1.2), hypertension (OR = 1.2; 95 %CI: 1.0–1.3) and tobacco smoking (OR = 1.2; 95 %CI: 1.1–1.4). There was no difference between the size and direction of the observed association between two competing models. The confidence interval of the ORs and the marginal differences in proportions of poor sleep quality for the diabetic and non-diabetic people using the distributional method was more precise (narrower confidence interval) than logistic regression.
Conclusion
Using the adjusted distributional method based on linear regression instead of dichotomizing the continuous outcome in logistic regression leads to narrower and more precise CIs for ORs but size and direction of associations between two models are identical. Comparison between two models showed that statistical performance of two models is equals. In addition to increasing age, women have higher odds of poor sleep quality than men. Some other modifiable predictors such as smoking, diabetes, and hypertension can be investigated to improve sleep quality.
Keywords: Distributional method, Dichotomization, Logistic regression, Sleep quality
Background
The American Academy of Sleep Medicine and Sleep Research Society (AASM/SRS) and the National Sleep Foundation (NSF) recommends that people over the age of 18 must get at least 7 h of sleep a day to stay healthy [1,2]. Inadequate sleep (<7 h per day) is associated with mild to severe physical and mental health problems, injury, reduced productivity, and premature death [[3], [4], [5]]. The prevalence of inadequate sleep duration (<7 h per day) among working adults in the United States has increased from 30.9 % in 2010 to 35.6 % in 2018 [6]. Population-based studies have reported the prevalence of poor sleep quality in the general population to be 35.9 % in Germany [7], 26.6 % in China [8], 39.4 % in Hong Kong [9], 17.8 % in Italy [10], 38.2 % in Spain [11] and 32 % in Austria [12]. The results of systematic reviews reveal that approximately half of older adults (over 60 years) [13] and 40 % of older people experience poor sleep quality [14]. The results of a study on the adult population in southeastern Iran in 2013 showed that 34.3 % of people have daily sleepiness and 57.5 % have low sleep quality [15]. Another study in Iran reported the prevalence of poor sleep quality with 66.6 % [16].
Sleep disorders and lack of inadequacy in sleep cause excessive daytime sleepiness, which can affect a person's mood, alertness, memory, security, and daily performance [17]. Research have shown that short or long sleep and poor sleep quality are risk factors for several diseases and conditions, including cancer [18], depression and generalized anxiety disorder [19,20], schizophrenia [21], dementia [22], sarcoidosis [23], high blood pressure, diabetes, stroke, obesity, heart disease and death [[24], [25], [26], [27]]. Various factors can affect the quantity and quality of sleep, including socio-economic status [28], gender [29], aging, physical activities, smoking, lifestyle, diet, and physical environment [30].
One of the widely used tools in measuring sleep quality is the Pittsburgh Sleep Quality Questionnaire, which has 7 main areas and 19 questions, and the sum of scores shows the final state of a person's sleep quality. To investigate the factors related to sleep quality, we have to convert the quantitative score obtained from the questionnaire based on the cut-point determined in the questionnaire (a score higher than 5 according to the questionnaire guide) into a dichotomous variable, such as being good and poor quality [31]. Dichotomous grouping of a continuous variable based on a cut-point is a common approach in medical and epidemiological research, and despite criticism, this method is widely used due to its better interpretation and presentation. The main limitations of converting a continuous outcome to a dichotomous variable are the loss of information leads to reducing the statistical power in detection of association and bias. The strength and degree of association depend on the choice of cut-point. Dichotomising a continuous, sleep quality, measure can affect the results or inference [32,33]. Some authors have investigated the result of dichotomizing the outcome in regression analysis. Breitling and Brenner have proven that sometimes the dichotomization of the outcome variable leads to the creation of spurious interactions between two predictor variables according to the type of their relationship with the outcome [34]. Paige et al. compared different models for body weight changes as a continuous or categorical outcome and concluded that the significance of model variables depends on the type of outcome defined [35]. “Distributional method” is an extended dual approach that analysis continuous outcome using both adjusted means and proportions derived from a comparison of means to replace dichotomisation alone [36,37]. Proportions obtained from this method provides less biased estimates (validity) [38]. In a systematic review on the analysis of birth-weight as continuous and dichotomous variables showed that using distributional method in primary studies restrict selective outcome bias [39].
Obtaining information related to the general population's sleep quality and the role of socio-economic and health-related factors on sleep quality is of particular importance. Considering the quantitative nature of the sleep quality score, dichotomising will be common problem in sleep research field. Furthermore, the effects of this practice in the context of sleep research in contrast to traditional methods of analysis have not been fully understood. Thus, in this study we investigated the relationship between categorized covariates with sleep quality in the population of people aged 45–65 in Shahroud, northeast Iran using the adjusted distributional method based on linear regression. Finally we compared the performance of dichotomizing sleep quality scores according to cut-point as outcome variable in multivariable logistic regression with using that as continuous outcome in an adjusted distributional method.
Material and methods
Sleep quality data were collected cross-sectionally during the second phase of the Shahroud Eye Cohort Study (ShECS), a population-based cohort study in Shahroud, Northeast Iran, evaluating the prevalence, incidence, and related factors of visual impairment and major eye conditions in adults. The present study is a cross-sectional study conducted on 4710 adults participating in the second phase of the ShECS [40]. The study protocol has been reviewed and approved by the Ethics Committee of Shahroud University of Medical Sciences with code IR.SHMU.REC.1393.023.
Study population
Shahroud eye cohort study (ShECS) was conducted in three phases in 2009, 2014, and 2019, and the details of the methodology of this study have been previously published [40]. In this study, using a random stratified cluster sampling, a total of 300 clusters in nine strata were randomly selected in Shahroud city. At least 20 individuals were selected from each cluster. Finally 5190 people from the 40 to 64-year-old population of Shahroud participated in the first phase of the study. All participants of the first phase were invited to return for a follow-up examination after a period of 5 in 2014 and 2019.In the second phase in 2014, 4737 of them were participated (response rate = 91.3 %). The study's second phase collected sleep quality data from 4710 participants, along with clinical, biochemical, genetic and physical evaluation data.
Study variables
In this study, independent socio-demographic and health-related variables were including age, gender, marital status, occupation, economic status, education, body mass index (BMI), smoking, diabetes, hypertension, and risk of cardiovascular disease. The outcome variable was sleep quality score obtained from the Pittsburgh Sleep Quality Index (PSQI) which was standardized using the item response theory (IRT) method as a latent variable [16].
The sleep quality questionnaire was created with the aim of checking the quality of sleep during the last month and includes 19 statements items. The Cronbach's alpha coefficient calculated for the reliability of the Persian version of the components of Pittsburg questionnaire was 0.77 [41] and the validity of the components of this questionnaire was investigated by the method of item response theory (IRT) [16]. The seven main components of the questionnaire include subjective quality of sleep, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, use of sleeping medications, and daily dysfunction. The total score of the questionnaire is from zero to 21 [31]. Since the seven components of this questionnaire have different discrimination values in examining sleep quality [16], therefore, for the continuous variable of sleep quality, first, the sleep quality score obtained from the Pittsburgh questionnaire was converted into a normal continuous variable with a mean of zero using the item response theory (IRT) method [16]. This model assumes: 1) the PSQI measures a single dimension of sleep quality; 2) item responses are independent; and 3) responses to polytomous items are a logistic function of the latent variable, with higher response levels indicating a greater level of the latent variable [16]. In this method, the sleep quality score is displayed as a hidden variable with a z-score ranging from −3 to 3 with a mean of zero and a standard deviation (SD) of 1. The standardized sleep scores greater than mean (that is zero) indicate a person's poor sleep quality. This group includes people who have severe problems in at least 2 domains or moderate problems in more than 3 domains. The details of this method in calculating the sleep quality score are presented in another article [16].
Health-related variables including hypertension and diabetes were diagnosed based on blood pressure and fasting blood sugar or HbA1c measurements. Hypertension in this study was equal to systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg or taking blood pressure medications, and systolic blood pressure 120–139 or diastolic blood pressure 80–89 was defined as prehypertension [42]. In this study, fasting plasma glucose ≥126 mg/dL and/or HbA1c level ≥ 6.5 % and/or the use of hypoglycemic medications were defined as having diabetes [43,44]. Variables such as age, gender, systolic blood pressure, total cholesterol, HDL cholesterol, use of blood pressure medications, smoking, and diabetes were used to calculate the 10-year CVD risk percentage based on the Framingham model [45]. Participants with a 10-year risk of more than 20 % were considered as high risk for heart disease.
The economic status of people was calculated using the principal component analysis (PCA) method based on household assets. The asset-index variable was divided into three quantiles, high, medium, and low in terms of economic status [46]. Educational attainment was categorized into five levels based on years of schooling: illiterate, primary school (1–5 years), middle school (6–8 years), high school (9–12 years), and college (13+ years). The definition of a smoker in this study included the use of cigarettes, water pipe, and pipe on most days of the week for at least 6 months [47].
Data analysis
The primary score of sleep quality was a quantitative variable ranging 0 and 20. The mean score of sleep quality was equal to 6.9 (±4.3). In the univariate analysis the association between health related and socio-demographic variables such as age, sex, marital status, education, job, economic status, BMI, tobacco use, 10-year risk of cardiovascular, diabetes and hypertension with poor sleep quality were evaluated. For comparison of two proportions a z- test and for assessing the linear trend of proportions a Cochrane-Armitage chi-square test was used. All covariates with p-values less than 0.2 in univariate analysis were entered the models. Proportion of poor sleep quality was estimated for total population and sub-population of covariates.
For constructing a standardized and continuous outcome variable, we utilize the degree of difficulty and discrimination of each part of the PSQI and the IRT method. This standardized continuous variable was considered the sleep quality score with zero mean. In the next step, the relationship between predictor variables and the standardized continuous sleep quality score as outcome was investigated using the multivariable logistic regression and the adjusted distributional method based on multivariate linear regression. In the logistic regression method, the quantitative variable of standardized sleep quality was dichotomized into two categories of good and poor sleep quality based on the defined cut point (that is zero) [16]. In the adjusted distributional method, the relationship between the standardized continuous sleep quality score, and covariates was investigated in a multivariable linear regression model, and the regression coefficients of each level of the predictors (they are age, gender, marital status, occupation, education, economic status, hypertension, diabetes, cardiovascular diseases risk, BMI, and smoking) were calculated. The obtained regression coefficients indicate the difference of the adjusted means for the predictor variable level compared to the baseline level. In the next step, according to the fitted multivariate regression model, the post-estimated marginal means of the sleep quality score has been calculated for the levels of each predictor variables. In the last step, the marginal difference in means between two groups using the distributional method and normal distribution parameters, transformed into a comparison of proportions of people that fall above the threshold point (that is zero). The distributional method works using the delta method to obtain the proportions of the population under a threshold value using the parameters of the normal distribution estimated from the data. For example the normal distribution shape of poor sleep quality score (outcome variable) in female and male (predictor variable) had same shape (common standard deviation) but with different means. This situation leads to a shift in the whole distribution and consequently changes the proportion of observations below a certain cut-point for sleep quality score. Finally, these proportions were used for estimation of the odds ratio (OR) [38,48]. Distributional model using linear regression assumes a distributional shift (same shape but different means) between the different levels of exposure, residuals of the linear regression are normally distributed and a linear relationship between the outcome and the predictors [36,48]. Using the results obtained from two methods of multivariable logistic regression and the adjusted distributional method based on multivariable linear regression, the association between predictor variables were investigated with sleep quality.
Comparison of performance of models: Also, to check the precision of the estimation of the two models, the marginal proportion for one of the predictor variables (here diabetes) was calculated and compared for both the logistic regression model and the distributional method[36]. Adjusted marginal proportion of poor sleep quality (adjusted for age, gender, education level, body mass index, economic status, hypertension, and tobacco use) was calculated for diabetics and non-diabetics.
To adjust the effect of cluster sampling, we estimated weighted proportion of poor sleep quality for categories of predictors. The STATA-v14 software and the command reg_distdicho were applied for data analysis. The level of significance in this study was considered 0.05.
Results
In this study, the data of 4710 people out of 4737 participants in the second phase of the Shahroud eye cohort study (ShECS) in 2014 were available. Of these, 2777 (59.0 %) were female and the mean (SD) age of the participants was 55.9 (±6.2) years. In Table 1, mean (standard deviation) and range of sleep quality score (standardized and non-standardized) illustrated in terms of diabetes status. Fig. 1 shows the histogram/density plot of standardized sleep quality scores for diabetic and non-diabetics.
Table 1.
summary of sleep quality scores (standardized and non-standardized) in terms of diabetic status as a predictor variable.
| Continuous variables | Mean (SD) | Range |
|---|---|---|
| Sleep quality score⁎ | ||
| Diabetic | 7.5 (4.1) | 0–20 |
| Non-diabetic | 6.7 (4.0) | 0–20 |
| Total | 6.9 (4.0) | 0–20 |
| Standardized sleep quality score using IRT⁎ | ||
| Diabetic | 0.12 (0.87) | −1.96–2.7 |
| Non-diabetic | −0.04 (0.86) | −1.96–2.6 |
| Total | 0 (0.87) | −1.96 - 2.7 |
Mean differences between diabetic and non-diabetic is significant at 0.05 level
Fig. 1.
Histogram/density plot of standardized sleep quality scores for diabetic and non-diabetics
Based on the defined cut point and the standardized continuous score in the IRT model for the Pittsburgh questionnaire (standardized mean sleep quality score over than zero is defined as poor sleep quality), the prevalence of poor sleep quality in this study was 44.9 % (95 %CI: 43.4–46.2). The frequency of sleep quality categories according to socio-demographic and health related variables is presented in Table 2.
Table 2.
Proportion of poor sleep quality in adult population of Shahroud in terms of Socio-demographic and health related characteristics.
| Variables |
Sleep Quality |
||||
|---|---|---|---|---|---|
|
Good sleep quality (n = 2586) (%proportion) |
Poor sleep quality (n = 2124) (%proportion) |
Total (n = 4710) No (%) |
Test and p- value⁎⁎, ⁎ | ||
| Gender | |||||
| Male | 1315 (68.0) | 618 (32.0) | 1933 (100.0) | Z = −9.5, <0.001 |
|
| Female | 1271 (45.8) | 1506 (54.2) | 2777 (100.0) | ||
| Age group (year) | |||||
| 45–49 | 493 (56.7) | 377 (43.3) | 870 (100.0) | Trend chi-square = 16.1, <0.001 |
|
| 50–54 | 734 (57.8) | 535 (42.2) | 1269 (100.0) | ||
| 55–59 | 651 (55.2) | 528 (44.8) | 1179 (100.0) | ||
| 60–64 | 464 (54.0) | 395 (46.0) | 859 (100.0) | ||
| 65–69 | 244 (45.8) | 289 (54.2) | 533 (100.0) | ||
| Marital status | |||||
| Single | 26 (56.5) | 20 (43.5) | 46 (100.0) | Trend chi-square = 31.2 <0.001 |
|
| Married | 2376 (56.3) | 1841 (43.7) | 4217 (100.0) | ||
| Widowed | 166 (40.9) | 240 (59.1) | 406 (100.0) | ||
| Divorced | 18 (43.9) | 23 (56.1) | 41 (100.0) | ||
| Educational level | |||||
| College | 370 (70.5) | 155 (29.5) | 525 (100.0) | Trend chi-square = 125.7, <0.001 | |
| High school | 893 (61.4) | 562 (38.6) | 1455 (100.0) | ||
| Guidance | 362 (50.8) | 350 (49.2) | 712 (100.0) | ||
| Primary | 737 (49.5) | 752 (50.5) | 1489 (100.0) | ||
| Illiterate | 222 (42.2) | 304 (57.8) | 526 (100.0) | ||
| Occupational status | |||||
| Practitioner | 960 (71.2) | 389 (28.8) | 1349 (100.0) | Trend chi-square = 234.1, <0.001 |
|
| Retired | 455 (59.0) | 316 (41.0) | 771 (100.0) | ||
| Unemployed | 16 (47.1) | 18 (52.9) | 34 (100.0) | ||
| Out of work | 16 (61.5) | 10 (38.5) | 26 (100.0) | ||
| Housekeeper | 1098 (44.5) | 1369 (55.5) | 2467 (100.0) | ||
| Other | 39 (65.0) | 21 (35.0) | 60 (100.0) | ||
| Economic status | |||||
| High | 1008 (64.1) | 564 (35.9) | 1572 (100.0) | Trend chi-square = 96.4, <0.001 | |
| Medium | 846 (53.9) | 724 (46.1) | 1570 (100.0) | ||
| Low | 732 (46.7) | 836 (53.3) | 1568 (100.0) | ||
| BMI (kg/m2) | |||||
| Normal (<25) | 600 (60.9) | 386 (39.1) | 986 (100.0) | Trend chi-square = 44.8, <0.001 |
|
| Overweight (25–29.9) | 1129 (57.7) | 826 (42.3) | 1955 (100.0) | ||
| Obesity (≥30) | 856 (48.6) | 906 (51.4) | 1762 (100.0) | ||
| Blood pressure | |||||
| Normotensive | 551 (61.1) | 351 (38.9) | 902 (100.0) | Chi-square = 33.3, <0.001 |
|
| Pre-hypertensive | 535 (59.2) | 368 (40.8) | 903 (100.0) | ||
| Hypertensive | 1500 (51.6) | 1405 (48.4) | 2905 (100.0) | ||
| Diabetes Mellitus | |||||
| No | 2024 (57.0) | 1529 (43.0) | 3553 (100.0) | z = 3.3, <0.001 |
|
| Yes | 542 (48.5) | 575 (51.5) | 1117 (100.0) | ||
| 10-year Cardiovascular disease risk | |||||
| Low risk | 1805 (54.1) | 1532 (45.9) | 3337 (100.0) | Z = 1.3 0.19 |
|
| High risk | 755 (57.2) | 564 (42.8) | 1319 (100.0) | ||
| Tobacco use | |||||
| No | 2272 (54.1) | 1926 (45.9) | 4198 (100.0) | Z = 2.4 0.01 |
|
| Yes | 314 (61.3) | 198 (38.7) | 512 (100.0) | ||
Weighted estimates of proportions adjusted for survey clustering were calculated.
P value for the comparison between Good sleep quality and Poor sleep quality proportions, using the Z-test for binary predictors and chocrane-armitage trend test for predictors with more than two levels.
According to the results of Table 2, prevalence of poor sleep quality was 54.2 % (N = 1506) in females and 32 % (N = 618) in men. The poor sleep quality was predominant in the age group of 65–69 years with 289 of 533 (54.2 %), and widowed people with 240 of 406 (59.1 %). Housewives with 55.5 % (N = 1369), and people with practitioner jobs with 28.8 % (N = 389) had the highest and lowest poor sleep quality, respectively. The highest percentage of poor sleep quality was related to illiterate people with 57.8 % (N = 304). The prevalence of poor sleep quality for obese people, hypertensive people with hypertension, people with diabetes, and people with a high CVD risk was 51.4 %, 48.4 %, 51.5 %, and 42.8 % respectively. Also, 53.3 % of people with low economic status (N = 836) and 38.7 % of people who used tobacco (N = 198) had poor sleep quality.
Relation between sleep quality and socio-demographic and health-related predictors were illustrated in Table 3 using two comparative multivariate models (distributional and logistic models). Based on the results of the adjusted distributional model analysis in Table 3, a significant association was observed between poor sleep quality and gender, age, education level, economic status, hypertension, diabetes, and smoking. The odds of poor sleep quality in women was 2.08 (95 % CI: 1.9–2.2) times that of men. Only the odds ratio (OR) of poor sleep quality in the age group of 65–69 years was significant compared to the baseline age group (45–49 years). The odds of poor sleep quality were higher in illiterate people than in academic people, and poor sleep quality was inversely related to economic status. Also, the odds of poor sleep quality increase with hypertension and diabetes and were not associated to BMI. The OR in people who used tobacco was 1.29 (95 % CI: 1.1–1.4) compared to that of who did not use it.
Table 3.
Relation between socio-demographic and health related factors with sleep quality among adult population of Shahroud using adjusted distributional and multivariable logistic regression models.
|
Variable |
Multivariable Logistic Regression Model |
Adjusted Distributional Model |
||||
|---|---|---|---|---|---|---|
| OR | 95 % CI | P value | OR | 95 % CI | P value | |
| Gender | ||||||
| Male | 1 | – | – | 1 | – | – |
| Female | 2.37 | 2.05–2.74 | <0.001 | 2.08 | 1.92–2.25 | <0.001 |
| Age group (year) | ||||||
| 45–49 | 1 | – | – | 1 | – | – |
| 50–54 | 0.96 | 0.80–1.15 | 0.673 | 1.00 | 0.87–1.15 | 0.97 |
| 55–59 | 1.02 | 0.84–1.23 | 0.811 | 1.10 | 0.96–1.27 | 0.17 |
| 60–64 | 1.03 | 0.84–1.27 | 0.726 | 1.05 | 0.90–1.22 | 0.51 |
| 65–69 | 1.38 | 1.08–1.75 | 0.009 | 1.25 | 1.05–1.48 | 0.02 |
| Educational level | ||||||
| College | 1 | – | – | 1 | – | – |
| High school | 1.10 | 0.88–1.38 | 0.372 | 1.20 | 1.02–1.41 | 0.03 |
| Guidance | 1.50 | 1.16–1.95 | 0.002 | 1.38 | 1.15–1.65 | 0.001 |
| Primary | 1.36 | 1.07–1.73 | 0.011 | 1.30 | 1.11–1.53 | 0.003 |
| Illiterate | 1.39 | 1.03–1.88 | 0.028 | 1.46 | 1.20–1.77 | 0.001 |
| Economic status | ||||||
| High | 1 | – | – | 1 | – | – |
| Medium | 1.26 | 1.08–1.47 | 0.003 | 1.25 | 1.12–1.40 | <0.001 |
| Low | 1.42 | 1.20–1.68 | <0.001 | 1.33 | 1.18–1.48 | <0.001 |
| BMI (kg/m2) | ||||||
| Normal (<25) | 1 | – | – | 1 | – | – |
| Overweight (25–29.9) | 1.00 | 0.84–1.18 | 0.989 | 1.01 | 0.90–1.15 | 0.77 |
| Obese (≥30) | 1.12 | 0.94–1.34 | 0.181 | 1.07 | 0.95–1.22 | 0.28 |
| Blood pressure | ||||||
| Normotensive | 1 | – | – | 1 | – | – |
| Pre-hypertensive | 1.13 | 0.93–1.38 | 0.213 | 1.05 | 0.91–1.22 | 0.46 |
| Hypertensive | 1.29 | 1.09–1.52 | 0.002 | 1.21 | 1.07–1.37 | 0.002 |
| Diabetes Mellitus | ||||||
| No | 1 | – | – | 1 | – | – |
| Yes | 1.19 | 1.03–1.38 | 0.014 | 1.16 | 1.06–1.27 | 0.007 |
| Tobacco use | ||||||
| No | 1 | – | – | 1 | – | – |
| Yes | 1.39 | 1.12–1.72 | 0.002 | 1.29 | 1.14–1.46 | 0.002 |
| Intercept | 0.22 | 0.16–0.29 | – | – | – | |
Using the multivariable logistic regression model (Table 3), the variables of gender, age group, education level, economic status, hypertension, diabetes, and smoking had a significant association with poor sleep quality. The ORs as measure effect of predictors for two models are shown in Table 3. The comparison of the results showed that there is no difference between the type and direction of the association between predictor variables and the outcome, as well as the number of significant variables of models, but in estimating the effect size with the adjusted distributional method, the confidence interval was narrower.
Considering that in the adjusted distributional method, the marginal proportion of the poor sleep quality is calculated, therefore the marginal proportion of the poor sleep quality in the levels of diabetes (exposure) has been calculated using the logistic regression model in Table 4 and compared with the estimate obtained from the distributional model. The marginal estimated difference in the proportion or prevalence of poor sleep quality in diabetics and non-diabetics by logistic regression was 0.041 (SE = 0.017). The difference in the proportion of poor sleep quality in these two groups with the adjusted distributional method was 0.038 (SE = 0.013). The results of these estimates for a unique predictor show less standard error in the adjusted distributional method and higher precision.
Table 4.
Comparison of adjusted regression-based distributional and logistic regression methods for the quality of sleep of diabetics versus non-diabetics.
| N = 4710 | Estimate (SE)⁎ | 95 % CI | p- value⁎⁎ | |
|---|---|---|---|---|
| Adjusted mean difference of sleep quality score (linear regression)a,b | 0.079 (0.029) | 0.020–0.137 | 0.008 | |
| Adjusted regression-based distributional modela | ||||
| Marginal difference in proportions | 0.038 (0.013) | 0.015–0.060 | 0.005 | |
| Marginal odds ratio | 1.16 (0.063) | 1.06–1.27 | 0.007 | |
| Multivariable Logistic regressiona | ||||
| Marginal difference in proportions | 0.041 (0.017) | 0.008–0.075 | 0.01 | |
| Odds ratio | 1.19 (0.088) | 1.03–1.38 | 0.01 | |
Adjusted for age, gender, education level, body mass index, economic status, hypertension, and tobacco use.
The difference in the mean sleep quality score in two diabetic and non-diabetic groups using multivariate linear regression model.
Standard Error,
A Z-test for differences between two adjusted marginal proportions.
Discussion
With the development of urban life and the industrialization of societies, sleep problems have become more evident. Despite importance of sleep quality, there is still a significant research gap in understanding socio-demographic and health related factors affecting sleep quality [49]. The present study investigated the prevalence and role of various factors on sleep quality in a population-based study using two distinct multivariable regression models.
The prevalence of poor sleep quality in this study was 44.9 % (95 %CI: 43.4–46.2), while the prevalence of poor sleep quality in other countries were less than our estimate. The prevalence of poor sleep quality in Germany [7], China [8], Hong Kong [9], Italy [10], Spain [11] and Austria [12] were 35.9 %, 26.6 %, 39.4 %, 17.8 %, 38.2 % and 32 % respectively. In a review the global prevalence of poor sleep quality was estimated 40 % in older people [14]. One of the reasons for the higher prevalence in current study is older age group (45–69 years), in this study. Also, the lifestyle of people in different countries is not identical, which can be another reason.
Based on the regression-based distributional model, women had worse sleep quality than men. The findings of Spoormaker and colleagues [50] also indicate a higher prevalence of insomnia complaints in women compared to men. Also, the studies conducted in Austria [12], Japan [30], China [8], Italy [10], Spain [11] and Iran [51,52] showed that the prevalence of sleep problems was significantly higher in women than in men. Hormonal changes during the menstrual cycle, menopause, pregnancy, and postpartum in women may affect the circadian rhythm and sleep pattern, causing frequent sleep disturbances and worsening sleep quality.
In the present study, a significant association was observed between poor sleep quality and age which is similar to the findings in Hunan China [8], Austria [12], Spain [11] and Iran [52]. However, in a study conducted in Germany, no significant association was observed between poor sleep quality and age group [7], which may be due to the difference in the age groups of that study (18 to 80 years). Also, in our study, a significant association was observed between education level and sleep quality, and the results showed that illiteracy was associated to poor sleep quality, which is in line with the results of studies conducted in Korea [53], Italy [10] and Germany [7]; but in a study conducted by Jinsong Tang in China, it was observed that a higher level of education was associated with insomnia [8]. One of the reasons could be the increasing job stress among educated people in China [54]. In the current study, a significant association was observed between economic status and sleep quality, which is in line to the findings of studies conducted in Germany [7] and another study [51] in Iran. The association between sleep quality and occupation [7] and economic status indicate the impact of low socioeconomic status on sleep quality, which is related to low quality of life in low socioeconomic classes.
In addition, the results of our study indicate a significant association between diabetes and hypertension with sleep quality, which is consistent with the study by Soon Young Lee in Korea [53]. Various studies have shown that people with chronic diseases have worse sleep quality [55,56] and these chronic diseases lead to poor sleep quality because it is chronic, and a person has to take different medicines in the long term. In addition, the results of some studies showed that poor sleep quality could be associated with an increased risk of chronic diseases [[25], [26], [27]]. According to the type of the present study, this association may be influenced by the reverse causality bias and a causal inference needs more criteria.
Also, in this study, a significant association was observed between tobacco smoking and sleep quality, which is consistent with the studies conducted in Korea [53], Germany [7], China [8], and Iran [51]. Smoking and receiving nicotine can stimulate nicotine receptors in the brain and cause the release of neurotransmitters such as acetylcholine, serotonin and gamma-amino butyric acid, which can affect the sleep and wake cycle [57]. Finally, insomnia increases the desire to smoke [58]. Since the start of smoking in this adult population was many years ago and during their youth (average onset age in our study = 25.5 years), therefore, there is a temporal relationship between exposure and outcome, and smoking can be mentioned as a risk factor for poor sleep quality.
In the present study, no significant association was observed between sleep quality and BMI, which is in contrast to studies conducted in Germany [7] and China [25]. In the study in China and Germany, the lower average age of the population and the lower prevalence of obesity [59] can be the reasons for this difference.
The comparison of multivariable logistic regression method and the regression-based distributional method in identifying and examining the role of related factors showed that two models are equally efficient despite what we expected, the distributional model for examining the sleep quality relationship would have less error. The precision or significance of the statistical test in distributional method, unlike when the data are directly dichotomized as pointed out by Altman et al. [33], does not depend on the choice of threshold because the statistical test is performed based on the comparison of the mean values of the continuous outcome. Using a multivariable regression model for estimating adjusted mean and standard error (SE) of distributions of predictor levels is among strengths of regression-based distributional method in comparison to distributional method. The regression-based distributional method can also be used to estimate the marginal proportions of predictors. This work is more efficient than the comparison of the proportions obtained from the data based on the dichotomous outcome and saves time and money [36,38].
We have also shown that using the adjusted distributional method based on linear regression, enable us to calculate the odds ratio (also proportion or relative risk) with a narrower confidence interval for a continuous variable, which is similar to the results of Sauzet's study [38,48].
Comparing the performance of these two models were evaluated by comparing the difference of the marginal proportion calculated for the variable of diabetes (has/does not have). The results of this comparing in Table 3 showed that the adjusted distributional method lead to the lower SE for marginal difference in proportions, which is similar to the results of Sauzet's study [48]. Simulation results in other studies show that using this method to calculate the confidence interval for effect estimates was efficient even in a small sample size [36]. Although, the results of other studies emphasized on advantages of the adjusted regression-based distributional method in estimation of the odds ratio [38,39], however our results showed no conceptual difference in the number of predictors and direction of the observed association between the predictor variables and sleep quality between the two methods.
Strengths and weaknesses of the study.
Strengths: A population-based study with representative and high sample size, accuracy in measurements and sampling, appropriate quality in completing questionnaires by well trained nurses, using the item response theory (IRT) method to normalize the sleep quality score, using real data to compare the power of two different data analysis methods.
Weaknesses: Considering the cross-sectional nature of the study, the investigation of causal relationships between independent variables and sleep quality can be influenced by reverse causality bias. For example, people with hypertension may experience more sleep problems, but sleep problems may cause behavioral or hormonal changes that lead to high blood pressure. Also, this study was conducted in the age group between 45 and 69 years old, which limit the generalizability of our findings. Potential confounders such as mental health status, alcohol consumption that may influence sleep quality were not considered. Measurement of quality of sleep using Pittsburgh is neither a gold standard nor equivalent to polysomnography (PSG). The generalizability of results was limited by the specific age group (45–70 years), participation in a cohort study, and ongoing follow-up. We recommend for future to conduct a new research among different age groups and remarketing other factors affecting sleep quality such as social variables, psychological stress level and quality of life.
Conclusions
The prevalence of poor sleep quality is high in Iranian adults. Using the adjusted distributional method based on linear regression instead of dichotomizing the continuous outcome in logistic regression leads to narrower and more precise CIs for ORs of the relationship between the variables. There was no conceptual difference in the number and direction of the observed associations between variables, therefore the statistical performance of two models is equals. Large sample size and normalization of outcome variable using IRT model can affect and dilute the differences between two methods. Despite the relative superiority of the adjusted distribution method over the logistic regression model, the limitation of using quantitative continuous variables as independent variables in the model makes its application difficult. Utilization of this method for assessing the relationship between a continuous outcome and some categorical independent variable is suitable.
In addition to increasing age, women have higher odds of sleep quality disorders than men. Some other modifiable covariates such as smoking, diabetes, and hypertension can be investigated to improve sleep quality.
Ethics approval and consent to participate
A written informed consent obtained from all subjects. The study protocol has been reviewed and approved by the Ethics Committee of Shahroud University of Medical Sciences with code IR.SHMU.REC.1393.023.
Consent for publication
Not applicable.
Availability of data and materials
The datasets generated and/or analyzed during the current study are not publicly available due the commitment of the authors to the administrators of the ShECS in the use and publication of data but are available from the corresponding author on reasonable request.
Funding
The ShECS was conducted with the financial support of Noor Ophthalmology Research Center and Shahroud University of Medical Sciences (Grant Number: 9826).
CRediT authorship contribution statement
Hajar Golbabaei Pasandi: Writing – original draft, Project administration, Formal analysis, Data curation. Ahmad Khosravi: Writing – review & editing, Supervision, Methodology, Formal analysis, Conceptualization. Mohammad Hassan Emamian: Writing – review & editing, Supervision, Project administration, Methodology, Conceptualization. Seyed Abbas Mousavi: Writing – review & editing, Methodology, Conceptualization. Hassan Hashemi: Writing – review & editing, Supervision, Project administration, Methodology, Investigation, Conceptualization. Akbar Fotouhi: Writing – review & editing, Supervision, Project administration, Methodology, Investigation, Conceptualization.
Declaration of competing interest
The authors declare that they have no competing interests or financial disclosure about this publication.
Acknowledgements
This article is extracted from the master's thesis of epidemiology with reference code 968 in the Vice-Chancellery for research and technology at Shahroud University of Medical Sciences. The data of this study was taken from the Shahroud Eye Cohort Study (ShECS).
Contributor Information
Hajar Golbabaei Pasandi, Email: h.golbabaei@yahoo.com.
Ahmad Khosravi, Email: khosravi2000us@yahoo.com.
Mohammad Hassan Emamian, Email: emamian@shmu.ac.ir.
Seyed Abbas Mousavi, Email: mmm89099@gmail.com.
Hassan Hashemi, Email: hhashemi@tums.ac.ir.
Akbar Fotouhi, Email: afotouhi@tums.ac.ir.
Refrences
- 1.Watson N.F., Badr M.S., Belenky G., Bliwise D.L., Buxton O.M., Buysse D., et al. Joint Consensus Statement of the American Academy of Sleep Medicine and Sleep Research Society on the Recommended Amount of Sleep for a Healthy Adult: Methodology and Discussion. Sleep. 2015;38:1161–1183. doi: 10.5665/sleep.4886. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hirshkowitz M., Whiton K., Albert S.M., Alessi C., Bruni O., DonCarlos L., et al. National Sleep Foundation’s updated sleep duration recommendations: final report. Sleep Health. 2015;1:233–243. doi: 10.1016/j.sleh.2015.10.004. [DOI] [PubMed] [Google Scholar]
- 3.Barnes C.M., Watson N.F. Why healthy sleep is good for business. Sleep Med Rev. 2019;47:112–118. doi: 10.1016/j.smrv.2019.07.005. [DOI] [PubMed] [Google Scholar]
- 4.Litwiller B., Snyder L.A., Taylor W.D., Steele L.M. The relationship between sleep and work: A meta-analysis. J Appl Psychol. 2017;102:682–699. doi: 10.1037/apl0000169. [DOI] [PubMed] [Google Scholar]
- 5.Kiley J.P., Twery M.J., Gibbons G.H. The National Center on Sleep Disorders Research-progress and promise. Sleep. 2019;42 doi: 10.1093/sleep/zsz105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Khubchandani J., Price J.H. Short Sleep Duration in Working American Adults, 2010–2018. J Community Health. 2020;45:219–227. doi: 10.1007/s10900-019-00731-9. [DOI] [PubMed] [Google Scholar]
- 7.Hinz A., Glaesmer H., Brähler E., Löffler M., Engel C., Enzenbach C., et al. Sleep quality in the general population: psychometric properties of the Pittsburgh Sleep Quality Index, derived from a German community sample of 9284 people. Sleep Med. 2017;30:57–63. doi: 10.1016/j.sleep.2016.03.008. [DOI] [PubMed] [Google Scholar]
- 8.Tang J., Liao Y., Kelly B.C., Xie L., Xiang Y.T., Qi C., et al. Gender and Regional Differences in Sleep Quality and Insomnia: A General Population-based Study in Hunan Province of China. Sci Rep. 2017;7:43690. doi: 10.1038/srep43690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wong W.S., Fielding R. Prevalence of insomnia among Chinese adults in Hong Kong: a population-based study. J Sleep Res. 2011;20:117–126. doi: 10.1111/j.1365-2869.2010.00822.x. [DOI] [PubMed] [Google Scholar]
- 10.Ausserhofer D., Piccoliori G., Engl A., Marino P., Barbieri V., Lombardo S., et al. Sleep Problems and Sleep Quality in the General Adult Population Living in South Tyrol (Italy): A Cross-Sectional Survey Study. Clocks & sleep. 2025:7. doi: 10.3390/clockssleep7020023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Madrid-Valero J.J., Martínez-Selva J.M., Ribeiro do Couto B., Sánchez-Romera J.F., Ordoñana J.R. Age and gender effects on the prevalence of poor sleep quality in the adult population. Gac Sanit. 2017;31:18–22. doi: 10.1016/j.gaceta.2016.05.013. [DOI] [PubMed] [Google Scholar]
- 12.Zeitlhofer J., Schmeiser-Rieder A., Tribl G., Rosenberger A., Bolitschek J., Kapfhammer G., et al. Sleep and quality of life in the Austrian population. Acta Neurol Scand. 2000;102:249–257. doi: 10.1034/j.1600-0404.2000.102004249.x. [DOI] [PubMed] [Google Scholar]
- 13.Parisa K., Ehsan M., Nasim S., Mohamad S., Leyla P., Fatemeh K. Worldwide prevalence of poor sleep quality in older adults: a systematic review and meta-analysis. Iran J Psychiatry. 2025:20. doi: 10.18502/ijps.v20i2.18207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Canever J.B., Zurman G., Vogel F., Sutil D.V., Diz J.B.M., Danielewicz A.L., et al. Worldwide prevalence of sleep problems in community-dwelling older adults: A systematic review and meta-analysis. Sleep Med. 2024;119:118–134. doi: 10.1016/j.sleep.2024.03.040. [DOI] [PubMed] [Google Scholar]
- 15.Tirgari B., Azzizadeh Forouzi M., Iranmanesh S., Khodabandeh Shahraki S. Predictors of sleep quality and sleepiness in the Iranian Adult: A population Based Study. J Community Health. 2013;1:144–152. [Google Scholar]
- 16.Khosravi A., Emamian M.H., Hashemi H., Fotouhi A. Components of Pittsburgh Sleep Quality Index in Iranian adult population: an item response theory model. Sleep Med X. 2021;3 doi: 10.1016/j.sleepx.2021.100038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Stanley N. The physiology of sleep and the impact of aging. Eur Urol Suppl. 2005;3:17–23. [Google Scholar]
- 18.Akman T., Yavuzsen T., Sevgen Z., Ellidokuz H., Yilmaz A.U. Evaluation of sleep disorders in cancer patients based on Pittsburgh Sleep Quality Index. Eur J Cancer Care (Engl) 2015;24:553–559. doi: 10.1111/ecc.12296. [DOI] [PubMed] [Google Scholar]
- 19.Ohayon M.M., Roth T. Place of chronic insomnia in the course of depressive and anxiety disorders. J Psychiatr Res. 2003;37:9–15. doi: 10.1016/s0022-3956(02)00052-3. [DOI] [PubMed] [Google Scholar]
- 20.Li Z., Zhong T., Meng X. A meta-analysis study evaluating the effects of sleep quality on mental health among the adult population. BMC Public Health. 2025;25:2992. doi: 10.1186/s12889-025-23709-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Doi Y., Minowa M., Uchiyama M., Okawa M., Kim K., Shibui K., et al. Psychometric assessment of subjective sleep quality using the Japanese version of the Pittsburgh Sleep Quality Index (PSQI-J) in psychiatric disordered and control subjects. Psychiatry Res. 2000;97:165–172. doi: 10.1016/s0165-1781(00)00232-8. [DOI] [PubMed] [Google Scholar]
- 22.Curcio G., Tempesta D., Scarlata S., Marzano C., Moroni F., Rossini P.M., et al. Validity of the Italian version of the Pittsburgh Sleep Quality Index (PSQI) Neurol Sci. 2013;34:511–519. doi: 10.1007/s10072-012-1085-y. [DOI] [PubMed] [Google Scholar]
- 23.Bosse-Henck A., Wirtz H., Hinz A. Subjective sleep quality in sarcoidosis. Sleep Med. 2015;16:570–576. doi: 10.1016/j.sleep.2014.12.025. [DOI] [PubMed] [Google Scholar]
- 24.Effect of short sleep duration on daily activities–United States, 2005–2008. MMWR Morb Mortal Wkly Rep. 2011;60:239–242. [PubMed] [Google Scholar]
- 25.Hung H.C., Yang Y.C., Ou H.Y., Wu J.S., Lu F.H., Chang C.J. The association between self-reported sleep quality and overweight in a Chinese population. Obesity (Silver Spring) 2013;21:486–492. doi: 10.1002/oby.20259. [DOI] [PubMed] [Google Scholar]
- 26.Qureshi A.I., Giles W.H., Croft J.B., Bliwise D.L. Habitual sleep patterns and risk for stroke and coronary heart disease: a 10-year follow-up from NHANES I. Neurology. 1997;48:904–911. doi: 10.1212/wnl.48.4.904. [DOI] [PubMed] [Google Scholar]
- 27.Hoevenaar-Blom M.P., Spijkerman A.M., Kromhout D., van den Berg J.F., Verschuren W.M. Sleep duration and sleep quality in relation to 12-year cardiovascular disease incidence: the MORGEN study. Sleep. 2011;34:1487–1492. doi: 10.5665/sleep.1382. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wilk P., Stranges S., Maltby A. Geographic variation in short and long sleep duration and poor sleep quality: a multilevel analysis using the 2015–2018 Canadian community health survey. Sleep Health. 2020;6:676–683. doi: 10.1016/j.sleh.2020.02.018. [DOI] [PubMed] [Google Scholar]
- 29.Sa J., Samuel T., Chaput J.P., Chung J., Grigsby-Toussaint D.S., Lee J. Sex and racial/ethnic differences in sleep quality and its relationship with body weight status among US college students. J Am Coll Heal. 2020;68:704–711. doi: 10.1080/07448481.2019.1594829. [DOI] [PubMed] [Google Scholar]
- 30.Asai T., Kaneita Y., Uchiyama M., Takemura S., Asai S., Yokoyama E., et al. Epidemiological study of the relationship between sleep disturbances and somatic and psychological complaints among the Japanese general population. Sleep Biol Rhythms. 2006;4:55–62. [Google Scholar]
- 31.Buysse D.J., Reynolds C.F., 3rd, Monk T.H., Berman S.R., Kupfer D.J. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28:193–213. doi: 10.1016/0165-1781(89)90047-4. [DOI] [PubMed] [Google Scholar]
- 32.Ragland D.R. Dichotomizing continuous outcome variables: dependence of the magnitude of association and statistical power on the cutpoint. Epidemiology. 1992;3:434–440. doi: 10.1097/00001648-199209000-00009. [DOI] [PubMed] [Google Scholar]
- 33.Altman D.G., Royston P. The cost of dichotomising continuous variables. Bmj. 2006;332:1080. doi: 10.1136/bmj.332.7549.1080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Breitling L.P., Brenner H. Odd odds interactions introduced through dichotomisation of continuous outcomes. J Epidemiol Community Health. 2010;64:300–303. doi: 10.1136/jech.2009.089458. [DOI] [PubMed] [Google Scholar]
- 35.Paige E., Korda R.J., Banks E., Rodgers B. How weight change is modelled in population studies can affect research findings: empirical results from a large-scale cohort study. BMJ Open. 2014;4 doi: 10.1136/bmjopen-2014-004860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Peacock J.L., Sauzet O., Ewings S.M., Kerry S.M. Dichotomising continuous data while retaining statistical power using a distributional approach. Stat Med. 2012;31:3089–3103. doi: 10.1002/sim.5354. [DOI] [PubMed] [Google Scholar]
- 37.Casella G., Berger R. Wadsworth & Brooks/Cole; Statistics/Probability Series Pacific Grove, California: 1990. Statistical Inference. [Google Scholar]
- 38.Sauzet O., Peacock J.L. Estimating dichotomised outcomes in two groups with unequal variances: a distributional approach. Stat Med. 2014;33:4547–4559. doi: 10.1002/sim.6255. [DOI] [PubMed] [Google Scholar]
- 39.Ofuya M., Sauzet O., Peacock J.L. Dichotomisation of a continuous outcome and effect on meta-analyses: illustration of the distributional approach using the outcome birthweight. Systematic Reviews. 2014;3:63. doi: 10.1186/2046-4053-3-63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Fotouhi A., Hashemi H., Shariati M., Emamian M.H., Yazdani K., Jafarzadehpur E., et al. Cohort profile: Shahroud Eye Cohort Study. Int J Epidemiol. 2013;42:1300–1308. doi: 10.1093/ije/dys161. [DOI] [PubMed] [Google Scholar]
- 41.Farrahi Moghaddam J., Nakhaee N., Sheibani V., Garrusi B., Amirkafi A. Reliability and validity of the Persian version of the Pittsburgh Sleep Quality Index (PSQI-P) Sleep Breath. 2012;16:79–82. doi: 10.1007/s11325-010-0478-5. [DOI] [PubMed] [Google Scholar]
- 42.Chobanian A.V., Bakris G.L., Black H.R., Cushman W.C., Green L.A., Izzo J.L., Jr., et al. Seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure. Hypertension. 2003;42:1206–1252. doi: 10.1161/01.HYP.0000107251.49515.c2. [DOI] [PubMed] [Google Scholar]
- 43.Ebrahimi H., Emamian M.H., Hashemi H., Fotouhi A. High Incidence of Diabetes Mellitus Among a Middle-Aged Population in Iran: A Longitudinal Study. Can J Diabetes. 2016;40:570–575. doi: 10.1016/j.jcjd.2016.05.012. [DOI] [PubMed] [Google Scholar]
- 44.Diagnosis and classification of diabetes mellitus Diabetes Care. 2014;37(Suppl. 1):S81–S90. doi: 10.2337/dc14-S081. [DOI] [PubMed] [Google Scholar]
- 45.D’Agostino R.B., Sr., Vasan R.S., Pencina M.J., Wolf P.A., Cobain M., Massaro J.M., et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117:743–753. doi: 10.1161/CIRCULATIONAHA.107.699579. [DOI] [PubMed] [Google Scholar]
- 46.Wagstaff A., O’Donnell O., Van Doorslaer E., Lindelow M. World Bank Publications; 2007. Analyzing health equity using household survey data: a guide to techniques and their implementation. [Google Scholar]
- 47.Khosravi A., Emamian M.H., Hashemi H., Fotouhi A. Transition in tobacco use stages and its related factors in a longitudinal study. Environ Health Prev Med. 2018;23:39. doi: 10.1186/s12199-018-0728-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Sauzet O., Breckenkamp J., Borde T., Brenne S., David M., Razum O., et al. A distributional approach to obtain adjusted comparisons of proportions of a population at risk. Emerg Themes Epidemiol. 2016;13:8. doi: 10.1186/s12982-016-0050-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Bin Y.S. Is sleep quality more important than sleep duration for public health? Sleep. 2016;39:1629–1630. doi: 10.5665/sleep.6078. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Spoormaker V.I., van den Bout J. Depression and anxiety complaints; relations with sleep disturbances. Eur Psychiatry. 2005;20:243–245. doi: 10.1016/j.eurpsy.2004.11.006. [DOI] [PubMed] [Google Scholar]
- 51.Mohammadian M., Khosravi A., Nohi S., Mousavi S.A. Factor associated with self-reported sleep quality in adults-a population based study. Knowledge And Health. 2018;12:1–6. [Google Scholar]
- 52.Asghari A., Farhadi M., Kamrava S.K., Ghalehbaghi B., Nojomi M. Subjective sleep quality in urban population. Arch Iran Med. 2012;15:95–98. [PubMed] [Google Scholar]
- 53.Lee S.Y., Ju Y.J., Lee J.E., Kim Y.T., Hong S.C., Choi Y.J., et al. Factors associated with poor sleep quality in the Korean general population: Providing information from the Korean version of the Pittsburgh Sleep Quality Index. J Affect Disord. 2020;271:49–58. doi: 10.1016/j.jad.2020.03.069. [DOI] [PubMed] [Google Scholar]
- 54.Du J., Liu Z., Zhang X., Shao P., Hua Y., Li Y., et al. Occupational Stress and Insomnia Symptoms Among Nurses During the Outbreak of COVID-19 in China: The Chain Mediating Effect of Perceived Organizational Support and Psychological Capital. Front Psychiatry. 2022;13 doi: 10.3389/fpsyt.2022.882385. [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]
- 55.Lo K., Woo B., Wong M., Tam W. Subjective sleep quality, blood pressure, and hypertension: a meta-analysis. J Clin Hypertens (Greenwich) 2018;20:592–605. doi: 10.1111/jch.13220. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Surani S., Brito V., Surani A., Ghamande S. Effect of diabetes mellitus on sleep quality. World J Diabetes. 2015;6:868–873. doi: 10.4239/wjd.v6.i6.868. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Siegel J.M. The neurotransmitters of sleep. J Clin Psychiatry. 2004;65(Suppl. 16):4–7. [PMC free article] [PubMed] [Google Scholar]
- 58.Hamidovic A., de Wit H. Sleep deprivation increases cigarette smoking. Pharmacol Biochem Behav. 2009;93:263–269. doi: 10.1016/j.pbb.2008.12.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Vaisi-Raygani A., Mohammadi M., Jalali R., Ghobadi A., Salari N. The prevalence of obesity in older adults in Iran: a systematic review and meta-analysis. BMC Geriatr. 2019;19:371. doi: 10.1186/s12877-019-1396-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated and/or analyzed during the current study are not publicly available due the commitment of the authors to the administrators of the ShECS in the use and publication of data but are available from the corresponding author on reasonable request.