The COVID-19 International Student Well-being Study (C19 ISWS) is used, which collected information on student well-being through an online survey during the first wave of the COVID-19 pandemic in 133 HEIs in 26 countries. Except South Africa, Israël and Russia, all included countries are located in Europe. Details about the study procedures can be found in the study protocol [17]. For this study, a subsample of the data was used to cover each participating country during a period with relatively stable policy measures (see Tables A1 and Table A2 in the Appendix). A random selection of 1,000 cases was drawn in countries with a larger size sample in order to correct for an overrepresentation of these countries in the total sample. This resulted in an analytical sample consisting of 20,103 respondents (73,9% is female, 78,3% is below the age of 26). Descriptive statistics are presented in Table A3.

An eight-item version of the Center for Epidemiologic Studies–Depression Scale (CES-D-8) scale was used to measure the frequency and severity of depressive feelings [18, 19]. Respondents were asked to indicate how often in the week previous to the survey they felt or behaved in a certain way (felt depressed, felt that everything was an effort, slept poorly, felt lonely, felt sad, could not get going, enjoyed life, or felt happy – last two items are reverse-coded). Response categories forming a 4-point Likert scale ranged from none or almost none of the time (0) to all or almost all of the time (3). Scale scores were assessed using a non-weighted summed rating and ranged from 0 to 24, with higher scores indicating a higher frequency and severity of depressive feelings. In the C19 ISWS sample, the country-specific Cronbach’s alphas ranged between 0.85 and 0.90 [17].

A measure of heavy episodic drinking (HED), defined in line with the definition of the WHO (2022) was used. HED was defined as drinking at least 60 grams of pure alcohol on one single occasion in the last seven days. Students reporting that they ‘never’ or ‘less than once a week’ drank six glasses of alcohol (as an approximation of 60 grams of pure alcohol) on a single occasion scored ‘0’ on the dichotomous variable HED and those answering ‘once a week’, ‘more than once a week’ or ‘(almost) daily’ scored ‘1’.

In addition to gender, age, migrant background, parental level of education and study program were taken into account. Age was included as a dichotomous variable to avoid a strong overlap with study program, which distinguished first-year students from other students. The highest level of education attained by either parent was used as a proxy of their socioeconomic position [20].

In order to assess the level of academic stress during the COVID-19 pandemic, respondents were asked to what degree they agreed with the statements: (1) My university/college workload has significantly increased since the COVID-19 outbreak; (2) I know less about what is expected of me in the different course modules/units since the COVID19 outbreak; (3) I am concerned that I will not be able to successfully complete the academic year due to the COVID-19 outbreak; and (4) The change in teaching methods resulting from the COVID − 19 outbreak has caused me significant stress. Each item was rated on a 5-point Likert scale ranging from 0 (strongly disagree) to 4 (strongly agree). The scale had a Cronbach’s alpha 0,7 and country-specific reliability indices are reported elsewhere (17). Worries to get (re)infected with COVID-19 was measured by a scale ranging from 0 (no worries) to 10 (very worried about getting (re)infected by COVID-19). Change in financial situation was based on the difference in answers to the statement ‘I had sufficient financial resources to cover my monthly costs’, thereby comparing their situation before the COVID-19 outbreak with that of their situation during the week prior to filling out this survey. The variable consists of four categories: (1) no change: not struggling, (2) no change: struggling with financial resources, (3) worse than before the COVID-19 outbreak, and (4) better than before the COVID-19 outbreak. Finally, participation in social activities was included as a categorical variable differentiating between students who had (1) online and face to face (FTF) social activities (= ref), (2) no social activities, (3) only FTF, and (4) only online activities.

To assess the stringency of COVID-19 protective measures implemented in each country, we used national measures of the University of Oxford coronavirus government response tracker (OxCGRT) stringency index, and regional measure for New Jersey and Quebec [20]. To take the strength and timing (in relation to the survey period) of the pandemic into account, we calculated the country’s level of excess mortality (p-score) during the period of the data collection, using data from Eurostat (2020) or national or regional statistics bureaus, and additionally used this variable to calculate the timing of the survey in relation to the peak of the first wave of the COVID-19 pandemic with three categories (0) before the peak, (1) during and (2) after the peak of the first wave of the COVID-outbreak. In addition, we included the control variable real GDP growth rate 2020 as an indicator of the socioeconomic condition of the country. An overview of these macrolevel indicators is provided in Appendix Table A4.

A hierarchical three-level model was constructed which clustered respondents within HEIs (N = 125), which were again clustered within countries (N = 26). Multilevel linear and logistic regression analyses were performed with respectively depressive feelings and HED as dependent variables, and with a random slope for gender included. In the first model, gender differences were estimated, controlled for the relevant individual and country level control variables. In the second model, a cross-level interaction between gender and the stringency index was included to estimate whether the gender differences in depression and HED varied by the stringency of the COVID-19 measures. In a third model, the individual stressors were added to examine whether they mediated the relation between gender, the stringency index, and depression or HED.

Sensitivity analyses were performed with the number of glasses of alcohol (e.g., a glass of wine, a shot, or a glass of beer between 25 and 33 cl) on average per week during the COVID-19 outbreak as dependent variable. As this is a count variable, negative binomial models were estimated (Table A5 in the appendix).

Data preparation and descriptive statistics were done in SPSS® version 26, and the multilevel analyses were performed in MLwiN Version 3.05.