Factors Affecting University Students’ Sleep Quality during the Normalisation of COVID-19 Epidemic Prevention and Control in China: A Cross-Sectional Study

: Background: Insufﬁcient and poor-quality sleep have signiﬁcant negative health consequences for university students in China. In this study, we aimed to assess the subjective sleep quality of university students during the normalisation of COVID-19 epidemic prevention and control in China and to identify key factors affecting their sleep quality. Materials and Methods: A cross-sectional survey was conducted among 1326 university students from Hubei Province, China. Latent proﬁle analysis was conducted on the results of class-difference tests of sleep patterns. Multinomial logistic regression was used to explore the relationship between the inﬂuencing factors and three classes of sleep quality. Results: The overall score of sleep quality (9.18 ± 3.22) among university students was assessed by using the PSQI scale, and 427 (32.20%) students reported poor sleep quality. Three distinct classes of sleep patterns were identiﬁed, namely, good sleepers (Class 1, 70.44%), poor sleep quality with less medication use (Class 2, 26.55%) and poor sleepers (Class 3, 3.01%). Conclusions: Compared with ‘good sleepers’, students having ‘poor sleep quality with less hypnotic drug use’ were inﬂuenced by their education stage, smoking habits, physical activity, depression and anxiety. Meanwhile, ‘poor sleepers’ may be affected by their age, origins, smoking habits, mental stress, depression and anxiety. Signiﬁcant heterogeneity was conﬁrmed in the sleep patterns of university students. Their behavioural lifestyles and mental health-related factors demonstrated different relationship patterns with sleep quality. Multiple sleep promotion interventions, including moderate aerobic exercises, psychological counselling and mindfulness training, should be regularly performed in groups to improve their sleep quality.


Introduction
A lack of sufficient and high-quality sleep has become common for the public.According to the China Sleep Research Report (2022), 57.41% of respondents reported having 1-7 days of insomnia in the past month.Over the past decade, the Chinese slept for an average of 8.5 h in 2012, which was reduced to 7.06 h in 2021.In addition, nearly 64.75% of respondents reported sleeping for less than 8 h a day [1].Sleep is fundamental to mental and physical health.However, among university students, the most common sleep disorders include inadequate sleep hygiene (ISH), staying up late, daytime dysfunction and insomnia [2].Empirical studies suggest that young adults are vulnerable to negative distress and other psychological problems.Their stress mainly comes from heavy academic pressure, severe employment difficulties [3], emotional regulation [4] and COVID-19 lockdown measures [5].As many students fail to properly cope with their stress, they start to feel helpless and develop negative expectations for their future, leading to psychological disorders or other diseases.Specifically, depression and mental stress share many cognitive components that can have a strong negative connection to sleep disturbance [6][7][8].
Apart from negative emotions, other factors that affect sleep duration include problematic mobile phone or internet use and long working hours, both of which crowd out sleep time [9,10].Moreover, in addition to throwing their psychology and lifestyle off balance, the COVID-19 epidemic has spurred fears of catching the virus or dramatically shifted the learning patterns of university students.Normalisation of COVID-19 prevention and control meant that China enacted persistent, strict, high-standard epidemic prevention and control measures to fight against the virus in daily life for a long time.Thus, most university students were restricted inside their dormitories or houses, which exacerbated their overuse of mobile devices and resulted in sleep disorders, such as insomnia and disturbed sleep rhythm [11,12].Notably, irregular and unhealthy living habits tend to have adverse effects on the endocrine system, immune system and metabolism and in turn disturb the underlying mechanisms of normal sleep [13].Research shows that high sleep quality and quantity are associated with an appropriate amount of physical activity (PA) [14].Other researchers contend that inadequate PA and sedentary behaviour can lead to mental health issues and poor sleeping quality among students [15].Jaehne observed a higher frequency of insomnia symptoms and less restorative sleep among smokers than non-smokers [16].To further understand the relationships between these variables in a population of university students, health-related lifestyles and habits should be considered when analysing the factors that influence their sleep quality.
Previous studies that evaluated the sleep quality of university students mainly categorised their participants into 'good' and 'poor' sleepers based on substantive domains of assessment [17].However, whether the sleep patterns and sleep-related potential risk factors of university students have various manifestations remains unknown.In addition to their complexities, the factors that influence sleep quality show obvious group heterogeneity.Following a 'person-centred' approach, in this study, we employed latent profile analysis (LPA) to examine the distinctiveness of the sleep quality sub-types of university students and to identify their potential sleep categories based on their symptoms.We aimed to assess the sleep status of university students during the normalisation of COVID-19 epidemic prevention and control in China and to identify the factors affecting their sleep quality.We hypothesised that students' poor sleep was associated with their unhealthy behavioural lifestyle, depression and mental strain caused by lockdown measures on campus.

Study Design and Participants
A cross-sectional study was designed to assess the sleep quality of university students in Hubei Province, China.Wuhan City was taken as the survey anchor, and university students from the cities of Xiaogan and Huanggang (about half an hour away from Wuhan) were recruited via convenience sampling.A 60-item self-designed questionnaire was constructed under the guidance of previous studies and related theories.This questionnaire comprised the following parts: (1) basic information, which asked for the gender, age, origins, education stage, smoking status and drinking status of the participants; (2) the PSQI; (3) the 10-item version of the Centre for Epidemiological Studies Depression Scale (CESD-10); and (4) a PA measurement.All participants provided their informed consent before answering the questionnaire.The following inclusion criteria were adopted in the sampling: (1) ≥16 years of age; (2) no severe mental disorders (clinically diagnosed mental illness); and (3) pursuing an undergraduate degree.

Pittsburgh Sleep Quality Index (PSQI)
The sleep quality of the participants was assessed using the Chinese version of the PSQI (revised by Liu Xianchen) [18].This scale contains 18 items and 7 components, namely, subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbance, hypnotic drug use and daytime dysfunction.Each component was weighted equally on a scale of 0 to 3, with a total score ranging from 0 to 21.A higher score corresponded to poorer sleep quality.The PSQI scale had a Cronbach's alpha of 0.728.

Depression
CES-D10 was used to measure the depressive symptoms of the participants.This 10-item scale is the shortened version of the original 20-item CES-D scale [19], developed by Andresen, Malmgren, Carter and Patrick [20].The participants were required to report how many days they experienced depressive symptoms during the previous week on a 3-point scale, ranging from 0 ('less than 1 day') to 3 ('5 to 7 days').The total score ranged from 0 to 30, with a higher score indicating more depressive symptoms.CES-D10 has reported good internal reliability and validity in previous studies [21].The Cronbach's alpha of the scale in this study was 0.78, indicating that it has good internal reliability.

Physical Activity
Self-report items were used to measure the PA of the participants.The items 'Over the last week, how often have you exercised for at least 30 min per day?', 'How long do you usually exercise per day last week?' and 'How intense are your physical activities?' were rated on a 5-point Likert scale.This PA measurement had a Cronbach's alpha of 0.790.An exploratory factor analysis was performed to test the validity of this scale, and a KMO value of 0.753 and significant Bartlett's sphericity (p < 0.001) were obtained.The validity and reliability of this measure were also supported in previous research [22].

Data Collection
This survey was carried out online from December 2021 to January 2022.Participants were recruited from 9 universities (including science and technology universities, regular universities, medical universities and comprehensive universities) in the cities of Wuhan, Huanggang and Xiaogan of Hubei Province, China.Each school was first stratified by grade level and then grouped by major.At least 100 students were selected from each sample site.They completed the questionnaires voluntarily through a WeChat QR code or Wenjuanxing web-link for cluster sampling [23].Dedicated quality controllers were trained to eliminate invalid questionnaires by inspecting their time taken, content quality and data format.A total of 1361 questionnaires were received.After quality control procession, 1326 were deemed valid.The response rate was 97.43%.

Sample Calculation
The sample size calculation was performed using the following formula: p is the estimated value of the expected probability.According to previous research, the prevalence of sleep disorder was p = 56.87%[24], and DEEF = 1, δ = 0.03 [25].We calculated the required sample size to be 1047.Indeed, the total valid sample size obtained in this study was 1326, which met the required sample size.

Statistical Analyses
Data were entered into Microsoft Excel 2016, and the descriptive statistics were analysed using SPSS 24.00.The LPA on the results of class difference tests of sleep quality was conducted using Mplus 7.0.To determine the optimal number of profiles in the data, the LPA model evaluation indicators were used, including the Akaike information criterion (AIC), Bayesian information criterion (BIC) and adjusted BIC (aBIC).A smaller value of these indices showed a better model fit [26].The entropy index is commonly used in research to evaluate the classification quality of a model.The LPA model obtained an entropy index of over 0.80, indicating a classification accuracy of over 90%.Meanwhile, the results of the Lo-Mendell-Rubin (LMR) likelihood ratio test and bootstrap-based likelihood ratio test (BLRT) were p < 0.05, indicating that the model with K categories was superior to that with K-1 categories [27].Multinomial logistics regression was performed on potential types of sleep quality using the R3.6.0 software, and the difference was statistically significant at p < 0.05.
Figure 2 presents the results of the multinomial logistic regression for predicting the three latent profiles.To prevent and reduce the detrimental effects of PSQI, Class 1 was designated as the reference cluster and compared with the other two groups.Compared with Class 1, Class 2 was closely associated with students' education stage, smoking habits, physical activity, mental stress, depression and school lockdown anxiety.The university students with a below-undergraduate education level (OR = 0.65, p = 0.012), smoking habits (OR = 1.91, p < 0.001), regular physical activity (OR = 1.61, p = 0.001), depression (OR = 4.06, p < 0.001) and anxiety (OR = 1.91, p < 0.001) were significantly more likely to be in Class 2.
The sleep quality of the participants in Class 3 was affected by age, place of origin, smoking habits, mental stress, depression and school lockdown anxiety (Figure 3).University students aged ≥18 years were more likely to be in Class 3 compared with students aged below 18 years (OR = 3.08, p = 0.011).University students coming from urban areas (OR = 0.39, p = 0.013) and with unhealthy behavioural and psychological lifestyles, such as smoking (OR = 9.94, p < 0.001), mental stress (OR = 3.78, p = 0.001), depression (OR = 3.15, p = 0.012) and school lockdown anxiety (OR = 3.54, p < 0.001), also showed a high likelihood of belonging to this class.versity students aged ≥18 years were more likely to be in Class 3 compared with students aged below 18 years (OR = 3.08, p = 0.011).University students coming from urban areas (OR = 0.39, p = 0.013) and with unhealthy behavioural and psychological lifestyles, such as smoking (OR = 9.94, p < 0.001), mental stress (OR = 3.78, p = 0.001), depression (OR = 3.15, p = 0.012) and school lockdown anxiety (OR = 3.54, p < 0.001), also showed a high likelihood of belonging to this class.

Discussion
A sleep disorder detection rate of 32.20% was recorded among the participants, and this value is higher than the sleep disorder rate (13.22%) of university students previously surveyed in Tianjin, China [28].Meanwhile, the overall score of sleep quality (9.18 ± 3.22) was above the norm (5.26 ± 2.38) [29].These results underscore the prominence of sleep disorders among the surveyed university students.By further analysing the homogeneous groups of individuals, the LPA method identified three latent profiles, namely, good sleepers (70.44%), poor sleep with less hypnotic drug use (26.55%) and poor sleepers (3.01%).Of the five components in the PSQI, actual sleep time (1.95 ± 1.43) and sleep efficiency (2.43 ± 1.13) obtained the highest scores, which may be ascribed to the multiple internal dimensions of sleep problems [30].Hypnotic drug use (0.17 ± 0.53) was also uncommon among the university students; that is, the surveyed students showed low dependence on hypnotic drugs for improving their sleep quality.This

Discussion
A sleep disorder detection rate of 32.20% was recorded among the participants, and this value is higher than the sleep disorder rate (13.22%) of university students previously surveyed in Tianjin, China [28].Meanwhile, the overall score of sleep quality (9.18 ± 3.22) was above the norm (5.26 ± 2.38) [29].These results underscore the prominence of sleep disorders among the surveyed university students.By further analysing the homogeneous groups of individuals, the LPA method identified three latent profiles, namely, good sleepers (70.44%), poor sleep with less hypnotic drug use (26.55%) and poor sleepers (3.01%).Of the five components in the PSQI, actual sleep time (1.95 ± 1.43) and sleep efficiency (2.43 ± 1.13) obtained the highest scores, which may be ascribed to the multiple internal dimensions of sleep problems [30].Hypnotic drug use (0.17 ± 0.53) was also uncommon among the university students; that is, the surveyed students showed low dependence on hypnotic drugs for improving their sleep quality.This finding may be ascribed to the strict regulation policy of psychotropic drugs [31] for preventing sleeping drug abuse and addiction.
Senior students demonstrated a higher likelihood of having poor sleep quality compared with the lower-level students, possibly due to the fact that these students are facing direct pressure to look for jobs or prepare for postgraduate or doctoral exams.For instance, the number of regular Chinese university graduates in 2021 was reported to be 9.09 million, which was 350,000 greater than the number reported in the previous year.Faced with new challenges and uncertainties in a new period, university graduates face fierce competition in the labour market [32].Moreover, the COVID-19 epidemic has reduced the number of employment opportunities [33], hence pushing more university students to pursue higher education instead, hence intensifying the competition for graduate studies.These pressures increased the anxiety of senior students, thereby leading to their poor sleep quality.
Students from rural areas (OR = 0.39; 95% CI −0.36, 0.72) are not prone to having sleep problems, which contradicted the results of a previous study [9].This finding may be ascribed to their affective coping styles.Specifically, urban students prefer to suppress their modes of expression or regulate their emotions; in other words, instead of focusing on their emotional states, they avoid interpersonal communication [34].Accordingly, urban university students excessively use their mobile phones or play computer games at night to eliminate negative emotions, and the excessive amount of time they spend on these activities leads to inefficient and inadequate sleep [35].
A common factor shared by the Class 2 and Class 3 students is school lockdown anxiety during the normalisation of COVID-19 epidemic prevention and control.The university students with anxiety are more likely to be in either Class 2 (OR = 1.91; 95% CI 1.60, 2.22) or Class 3 (OR = 3.54; 95% CI 2.70, 4.50).Poor sleep quality is not only a subjective wakefulness disorder but is also caused by psychological disorders [36,37].Sleep quality is uniquely associated with perceived stress and psychological factors [38,39].At the end of 2021, colleges in Wuhan city conducted strict lockdown management on campus in accordance with the requirements of epidemic prevention and control.As a result, college students were unable to enter and leave campus freely and had to stay in their dormitories for a long time.Students' worries about the quality of life and worse service support system may have become the main source of tension, and the risk perception in special situations then becomes the medium of strengthening pressure and conducting mental tension, which negatively impact the mental health of college students [40][41][42].These students face challenges in adapting to the psychologically stressful scenario of COVID-19 insecurity and negative emotions [43,44].This might explain their poor sleep quality.
Students with depression have a high likelihood of being in Class 2 (OR = 4.06; 95% CI 3.76, 4.36).Students with depressive symptoms or mental stress (OR = 3.78; 95% CI 2.99, 4.86) may suffer from poorer sleep quality than those without mental health problems.This view was supported by previous studies and may be ascribed to the fact that depression weakens the regulation function of the human serotonin system, which reduces sleep quality by suppressing slow-wave sleep, increasing wakefulness and introducing difficulties in falling asleep or sleep maintenance disorders [45].Stephen highlighted the importance of evaluating multiple mental health symptoms simultaneously, given that the associated mental health symptom dimensions strongly interact with one another [46].A higher level of mental stress corresponds to a worse quality of sleep [47].Mental stress can activate the brain in its high excitatory state, and, as a result of long-term tension, may consume many body functions continuously, hence affecting the quality of sleep [48].
Sleep quality is also associated with unhealthy behaviours.Students who smoke (OR = 9.94; 95% CI 8.92-10.26)tend to be in Class 3.An experiment among young smokers in Germany revealed that some changes in sleep patterns depend on the amount of daily tobacco consumption [16], proving that the nicotine from tobacco can influence the sleepregulating mechanisms of the body.These effects stimulate the release of dopamine, norepinephrine, serotonin and acetylcholine, all of which define the arousal levels of wakefulness [49].
Physical activity plays a proactive role in sleep quality.The reduction in physical activity time of college students due to COVID-19 is an obstacle to sleep quality.Those who seldom take part in PA (OR = 1.61; 95% CI 1.32, 1.90) tend to demonstrate poor sleep quality.Oliveira proposed that regular PA will increase the body temperature and in turn stimulate the hypothalamus to control the heat dissipation mechanism of the body, hence triggering and increasing slow-wave sleep, which is conducive to entering the deep sleep stage quickly and increasing the need for sleep to achieve anabolism [50,51].These results altogether show that psychological well-being and behaviours are vital in sleep regularity and satisfaction.

Limitation
This study has several limitations.Firstly, using a cross-sectional design did not produce very precise or convincing results, hence preventing this research from establishing a causal conclusion.Therefore, in-depth studies should be carried out in the future.Secondly, all information presented in this study was assessed using self-reported questionnaires, and survey respondents may overreport their normative activity to bring into congruence the ideal identity, which is prone to subjective bias.Future research may combine self-reported questionnaires with objective measurements or diagnosis results to improve the scientificity of the findings.

3. 3 .
Distinct Subgroups of the Sleep Quality of University Students Three profiles of PSQI (Class 1, Class 2 and Class 3) are shown in Figure 1.Profile 1, labelled as 'Good sleepers' (N = 934, 70.44%), is characterised by a low probability of responding 'high' to any sleep component except for 'sleep efficiency'.Profile 2, labelled as 'Poor sleep quality with less hypnotic drug use' (N = 352, 26.55%), is characterised by a relatively low probability of responding 'high' to all sleep components except for 'daytime dysfunction'.Profile 3, labelled as 'Poor sleepers' (N = 40, 3.01%), is characterised by relatively higher scores for each sleep component compared with those in the two other profiles, especially for 'hypnotic drug use'.3.3.Distinct Subgroups of the Sleep Quality of University Students Three profiles of PSQI (Class 1, Class 2 and Class 3) are shown in Figure 1.Profile 1, labelled as 'Good sleepers' (N = 934, 70.44%), is characterised by a low probability of responding 'high' to any sleep component except for 'sleep efficiency'.Profile 2, labelled as 'Poor sleep quality with less hypnotic drug use' (N = 352, 26.55%), is characterised by a relatively low probability of responding 'high' to all sleep components except for 'daytime dysfunction'.Profile 3, labelled as 'Poor sleepers' (N = 40, 3.01%), is characterised by relatively higher scores for each sleep component compared with those in the two other profiles, especially for 'hypnotic drug use'.

Figure 1 .
Figure 1.Sleep quality patterns of the three profiles.

Figure 1 .
Figure 1.Sleep quality patterns of the three profiles.

Figure 2 .
Figure 2. Forest plot of logistic regression analysis results of latent profiles (Class 2 vs. Class 1).Figure 2. Forest plot of logistic regression analysis results of latent profiles (Class 2 vs. Class 1).

Figure 3 .
Figure 3. Forest plot of logistic regression analysis results of latent profiles (Class 3 vs.Class 1).

Figure 3 .
Figure 3. Forest plot of logistic regression analysis results of latent profiles (Class 3 vs.Class 1).

Table 2 .
Latent class model fit indicators of the Pittsburgh sleep quality index (N = 1326).

Table 3 .
Univariate analysis of different sleep types of university students.