Sleep quality and its psychological correlates among university students in Ethiopia: a cross-sectional study

A cross sectional study was conducted in two universities (Haramaya University, in Eastern Ethiopia and University of Gondar in North West Ethiopia) in Ethiopia. Study participants were undergraduate regular students from second year to final year. First year students were excluded from the study because they were not yet admitted or fully registered at the start of the present study.

A multistage sampling design by means of probability proportional to size (PPS) was used to select departments for participation in the research. This approach was performed for both universities, and all students in selected departments were invited to participate. Students who expressed an interest in participating in the study were invited to meet in their classroom where they were informed about the purpose of the study and asked to participate in the survey. A random sample of these students was then selected and asked to complete a self-administered individual survey after providing consent. There was no set time limit for completing the survey. Students who could not read the survey (i.e., were blind) were excluded, as were those enrolled in correspondence, extension, or night school programs. A total of 2,817 undergraduate students participated in the study, the participation rate was 94%. After excluding students with incomplete questionnaires and missing sleep quality scores, the final analyzed sample consisted of 2,551 students. Based on the information provided, students excluded from analysis had similar characteristics to those considered. All completed questionnaires were anonymous, and no personal identifiers were used.

A bi-lingual self-administered questionnaire was used to collect information for this study. The questionnaire was initially developed in English and was then translated by Ethiopian linguist experts into Amharic, the lingua franca for the country. The final bi-lingual questionnaire was then evaluated by experts fluent in both Amharic and English. Finally, the questionnaire was pilot tested in students at other universities in Ethiopia before being deployed at the participating universities. The questionnaire ascertained demographic information including age, sex, and education level. Behavioral factors such as alcohol consumption were assessed using the WHO STEPS instrument [27] while khat use (a stimulant plant widely used in East Africa and Arabian Peninsula) [28] was measured adopting a set of questions.

Sleep quality was assessed using the previously validated Pittsburgh Sleep Quality Index (PSQI) [2]. The PSQI consists of 24 questions which generate seven component scores each ranging from 0 to 3 (0 score equals better and 3 is worst) and one global sleep quality score. The global score > 5 designates poor sleep quality while score ≤ 5 is considered good quality sleep [2]. The instrument has been validated in different population including university students [2, 7, 11].

Depression, anxiety, and stress symptoms were measured using the Depression Anxiety and Stress Scale (DASS 21) which consists 21 questions; seven under each of the three negative affective states. The questions inquire the experience of the items in the past week and each item is scored from 0 (did not apply to me at all) to 3 (applied to me very much). Finally the values obtained were multiplied by 2. The reliability of these instruments was checked using the Cronbach α and it was 0.82 for depression, 0.78 for anxiety and 0.79 for stress components. Table 1 depicts previously suggested cut of values for each component of DASS-21 score [29]. Categorization of study participants according to symptom levels of stress, depression and anxiety were made using cut off points suggested in the literature [29]. See Table 1.

The Perceived Stress Scale (PSS) was also used to measure the perception of stress. The PSS is a 10 item instrument that include questions about ones feelings and thoughts in the last one month (how often one felt in a certain way) with a 5 point Likert scale (0 = never, and 4 = very often). The score ranges from 0 to 40 and higher score represented higher level of perceived stress [30]. The psychometric property of this instrument was also studied in other studies and favorable results were documented [31, 32]. In this study, reliability of the instrument was also measured using Cronbach α and it was 0.76.

Frequency, median, mean with standard deviation and 95% confidence intervals were used to characterize sleep quality and duration characteristics and other demographic and behavioral characteristics. Analysis of variance (ANOVA) procedures were used to assess mean differences in sleep quality, stress, and depression and anxiety scores according to number of years at university. A post-hoc analysis using Bonferroni correction methods was made for those with a statistically significant difference.

Multi-level analysis techniques were used to measure associations of demographic and psycho-social factors with the measures of sleep quality [33]. The group level variable considered was year of stay in the university. Fitting the selected variable on the multilevel mixed model, the intra cluster correlation coefficient was calculated and only 4.5% of the variance in the PSQI score was attributed to the group level variation. Thus, binary logistic regression was used to evaluate the association between the set of demographic and psychological factors and poor quality sleep. The adjusted odds ratio with the 95% CI and the P-values are reported.

All completed questionnaires were anonymous, and no personal identifiers were used. Approval to conduct this study was obtained from the deans of all participating universities. The procedures used in this study were approved by the Addis Continental Institute of Public Health (ACIPH) and University of Gondar Institutional Review Board (IRBs), also from University of Washington Human Subject Committee. Permission was also obtained from Gondar and Haramaya Universities. Informed consent was obtained from each participant.

The response rate in this study was 94%. After excluding participants with missing information for sleep quality, a total of 2,551 students remained for analysis. More than fifty percent (1,298) of students were from Haramaya University and more than seventy five percent (1,940) were males. The median age was 21 years [range 17 to 35 years]; and 88.6% (2,465) of students were between 20 and 24 years of age. Most students were sophomores and juniors 49.4% (1,258) or 42.2% (1,074) respectively. See Table 2.

Approximately 50.8% (1,294) of the students reported mild to extremely severe depressive symptoms, 58% (1,369) of students reported mild to extremely severe levels of anxiety symptoms, and 34.1% (864) of students reported mild to extremely severe levels of stress seven days prior to the study. Using the perceived stress scale score, 48% (1,217) were in the third and fourth quartile of the scale four weeks prior to the survey (Table 2). Mean scores of perceived stress, anxiety and depression were statistically significantly reduced with increasing numbers of years in university (P-values <0.005 for both stress and anxiety; P-value = 0.005 for depression) (Figure 1). According to the post hoc analysis, significant mean differences in depression scores were accounted for by differences between second and third year students with the P-value of 0.006. For perceived stress, the differences were between second and third year, second year and fourth year and third year and fourth year (p<0.001). For anxiety, the significant difference between second year and third year; second year and fourth year were important with the following p-values 0.008 and 0.004 respectively.

The mean global PSQI score (computed using the component scores) was 6.23, 95% CI (6.12, 6.35). Overall, 55.8% (1,424) of students were classified as having poor sleep quality. The mean duration of sleep per night was reported to be 6.79 ± 1.9 hours with 95% CI (6.7, 6.8). Of note, and only 37.6% (960) reported sleeping > 7 hours. Mean PSQI scores were statistically significantly associated with years of study at university (P <0.005) (Figure 1); the differences were between second and third year, second year and fourth year and third year and fourth year (p<0.001). The median amount of sleep latency was 30 minutes with inter-quartile range of 20 minutes. Of note, when assessment was limited to the component item that measured self-report of poor sleep quality, only 12.4% (314) of the students specifically stated that their sleep quality was fairly bad or very bad. Self-reports of day time dysfunction was less frequent, with 6.1% (155) of students in the highest difficulty category while the habitual sleep efficiency of less than or equal to 74% was reported by 19.4% (418) of the participants. The second highest sleep disturbance score was also reported by 26.9% (686) of the participants and use of medication in the last one month was reported by 8.7% (245) of the participants. Table 3.

Binary logistic regression was used to assess factors associated with sleep quality. Three multivariable models were used to explore these factors. The first model considered the demographic variables such as sex, age and other miscellaneous variables as years of study and university. The second model considered psychological factors such as perceived stress level, as well as anxiety and depression. The last model included both demographic and psychological factors as well as behavioral factors such as coffee consumption, cigarette, alcohol and khat use.

Female sex was not significantly associated with poor sleep quality in the first model but in the last model with AOR = 1.23 95% CI: (1.00, 1.57). Compared to students in 4^th year or above, second year [AOR = 2.91 95% CI: (2.10, 4.02)] and third year students [AOR = 2.25, 95% CI: (1.62, 3.12)] had higher odds of poor sleep quality. The odd of poor sleep quality has shown reduction as the students year of study increase; besides, the magnitude of effect reduced slightly for second year students and increased for third years in the final model. Students from Gondar University had 36% higher odds of poor sleep quality than Haremaya University students [AOR = 1.36, 95% CI: (1.13, 1.63)]. The magnitude of effect also showed increment in the last model.

The level of perceived stress scale has shown significantly increasing odds of poor sleep quality across the quartiles with the following odds ratio (95% CI) [1.35 (1.10, 1.70)], [1.55 (1.23, 1.96)] and 1.90 (1.46, 2.46)] respectively. Similarly, for depression, the odds ratio increased as the level of depression increased from mild to extremely severe with the following odds ratio (95% CI) [1.36 (1.06, 1.75)], [1.64 (1.27, 2.11)], [1.64 (1.11, 2.42)], [2.65 (1.56, 4.49)]respectively; the magnitude of effect has shown slight increment in the final model. Lastly, for anxiety, moderate to extremely severe levels of anxiety were significantly associated with poor sleep quality compared to the normal level with [AOR = 1.27 95% CI: (1.002, 1.61)] for moderate, [AOR 1.62 95% CI: (1.17, 2.24)] for severe and [AOR = 1.76 95% CI: (1.27, 2.45)] for extremely severe levels of anxiety. The magnitude of effect was attenuated in the final model for all categories. See Table 4.