This is a cross sectional study using data from two independent cohorts of University students: i-Share in France and the Kyoto University students’ cohort in Japan.
The i-Share cohort study
The Internet-based Students Health Research Enterprise (i-Share) project is a prospective population-based cohort study of students of French speaking Universities. It was initiated by the Universities of Bordeaux and Versailles Saint-Quentin, where active recruitment started in February 2013. The project has been further extended, on a voluntary basis, to all Universities and higher education institutes in France and is still ongoing. To be eligible to participate, a student has to be at least 18 years of age and be able to read and understand French. Students are informed about the purpose of the project by flyers, information stands at University campuses, during lectures, and via social media and newsletters. The i-Share project consists of a web-based baseline questionnaire asking voluntary participants information on their physical and mental health status, socio-demographic characteristics and lifestyle habits. Within three months after completion of the baseline questionnaire, a supplementary web-based questionnaire concerning mental health is addressed to students to explore more precise information on psychological factors including depressive symptoms. Like the baseline questionnaire, the mental health questionnaire is completed on a voluntary basis. For this comparative study, we used data available as of September 3rd, 2018. The i-Share project is carried out in accordance with the Declaration of Helsinki and has been approved by the National Commission of Informatics and Liberty (Commission Nationale de l’Informatique et des Libertés - CNIL), approval number DR-2013-019. Every student is requested to sign an online informed consent form before completion of the baseline questionnaire.
The Kyoto University students’ cohort
Kyoto University Health Service carries out an annual health checkup addressed to all students in April. This checkup consists of anthropometric measures including height, weight, blood pressure, urinalysis, and a self-administrated web-based questionnaire including items about lifestyle, physical and mental health status, past and present medical history, sleep satisfaction and self-rated health. Concerned students come from 10 undergraduate school departments and 20 graduate school departments covering all fields of study such as humanities, sciences, health studies, law and economy. Data for this comparative study refer to the year 2017. We excluded all students who answered the web-based questionnaire, but did not participate in the annual health checkup. All analyses and procedures were anonymous and in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the Kyoto University Ethical Committee (Ethics Code No: R1732). Individual informed consent was waived for the use of de-identified clinical data according to the national ethical guidelines of Japan.
Outcome, Exposure and Confounding variables
Depressive symptoms
Depressive symptoms were assessed using the first two items of the 9-item Patient Health Questionnaire (PHQ-9) [25], also known as the PHQ-2 [26]. These two items enquiry about the frequency of the symptoms of depressed mood (feeling down, depressed or hopeless) and anhedonia (little interest or pleasure in doing things), scoring each as 0 (not at all), 1 (several days), 2 (more than half the days) and 3 (nearly every day). Thus, the total score of the PHQ-2 can range from 0 to 6. As previously recommended [27], we used the cutoff score of 3, by dichotomizing this variable into two categories: “low depressive symptoms” (0-2) and “high depressive symptoms” (3-6). PHQ-2 has demonstrated good reliability and validity to assess major depression among primary care patients [26].
Self-rated health
To assess health status, we used a five-point Likert scale of self-rated health corresponding to the question “How would you rate your health in general?”. Answers were “very good,” “good,” “fair”, “bad” or “very bad” [16]. This variable was further dichotomized into two categories: “poor” (including “bad” or “very bad”) and “good” (including “fair”, “good” or “very good”) self-rated health.
Socio-demographic characteristics
We collected the following variables: sex (male, female), age (quantitative variable), and year of study (the first year of undergraduate versus other years).
Sleep quality
In the i-Share cohort, sleep quality was assessed through a single item from the Pittsburgh Sleep Quality Index [28]. Students were asked to categorize their sleep quality during the three previous months preceding the survey as “good”, “somewhat good”, “neither good nor bad”, “somewhat bad” or “bad”. In the Kyoto cohort, sleep quality was assessed by categorizing sleep satisfaction as “very satisfied”, “satisfied”, “dissatisfied”, or “very dissatisfied”. In order to compare the results from these two different questions, we further dichotomized the answer options from the two questionnaires into two categories: “bad” (including “somewhat bad” or “bad” for i-Share, and “dissatisfied” or “very dissatisfied” for Kyoto) and “good” (including “good”, “somewhat good” or “neither good nor bad” for i-Share, and “very satisfied” or “satisfied” for Kyoto) sleep.
Body mass index (BMI)
In the i-Share cohort, height and weight were self-reported in the web-based baseline questionnaire. In the Kyoto cohort, when having health checkup, students’ height and weight were measured. In both cohorts, we calculated body mass index (BMI) as weight in kg divided by height in m squared (kg/m²) and categorized it in four groups: underweight (<18.5), normal weight (≥18.5, <25), overweight (≥25.0, <30), and obesity (≥30).
Health habits
In the i-Share cohort, students were asked whether they were regularly practicing one or more sport activities (“yes” or “no”). In the Kyoto cohort, students were asked how often they usually exercised in the last month (“everyday”, “sometimes, or “rarely”). The physical activity categories were dichotomized into “regular” (corresponding to “yes” in i-Share, “everyday” or “sometimes” in Kyoto), and “occasional or less” (corresponding to “no” in i-Share and “rarely” in Kyoto). For smoking status, in both studies the categories were dichotomized into “yes” (corresponding to “daily or occasionally smoking”) and “no” (corresponding to “never smoking”). Alcohol consumption was assessed as current frequency of consumption of alcoholic beverages (i.e. beer, wine, whisky, vodka, tequila, cocktails). More precisely, in the i-Share questionnaire the categories were divided into “never”, “from once a year to once per month”, “from several times per month to once per week”, and “several times per week”. In the Kyoto questionnaire, the categories were divided into “never”, “sometimes”, “20 g/day” and “with trouble”. Finally, to compare data from the two studies, the categories for alcohol consumption were dichotomized into “never” and “at least sometimes” drinking.
Statistical analysis
We performed general descriptive statistics for each cohort and calculated comparative p values between the two cohorts. Chi-square test for categorical variables and t test for continuous variables were used respectively. Then, we performed the same description by the two categories of depressive symptoms (0-2 and 3-6 scores) in order to obtain the univariate associations with all covariates, including self-rated health. Multivariate logistic regression models were performed for each cohort to further explore the association between depressive symptoms and all covariates, including self-rated health. For each multivariate logistic regression model, we tested potential cofounding factors (interactions) and/or effect modifiers (relative variation of the ORs).
Main analyses were performed in male and female students separately. Missing values were included in the model as a separate category. All tests were 2-tailed, and p values of <0.05 were considered as statistically significant. SAS version 9.4 (SAS Institute Inc, Cary, NC USA) and R version 3.5.1 for Windows were used for all analyses.