*3.3. Predictors of Depression in the Nine Countries*

Multivariate logistic regression was performed to explore significant predictors of depression among a set of variables that were previously included in the relationship analysis using Pearson's χ<sup>2</sup> test. An estimation was assessed separately for each country: Colombia (Table S3), Czechia (Table S4), Germany (Table S5), Israel (Table S6), Poland (Table S7), Russia (Table S8), Slovenia (Table S9), Turkey (Table S10), and Ukraine (Table S11). All estimates of the multivariate logistic regressions are shown in Figure 3.

The regression model performed for depression among Colombian university students showed a significant effect only for gender; estimate = 0.76, 95% *CI* = 0.004, 1.516, *SE* = 0.39, OR = 2.14, *Z* = 1.97, Wald's χ2(1) = 3.88, *p* < 0.05 (Table S3). However, the bias corrected accelerated bootstrapping (BCa) method did not confirm gender to be a significant predictor of depression in Colombian students; bootstrap BCa 95% *CI*<sup>B</sup> = −0.092, 1.539. The regression model was found to be significant, χ2(143) = 21.591, *p* < 0.01, *R*<sup>2</sup> CS = 0.13, *R*2 McF = 0.10, *R*<sup>2</sup> <sup>N</sup> = 0.18.

For Czech students (Table S4), logistic regression showed two predictors of depression, namely exposure to COVID-19 (estimate = 0.66, 95% *CI* = 0.052, 1.267, *SE* = 0.31, OR = 1.93, *Z* = 2.13, Wald's χ2(1) = 4.53, *p* < 0.05) and comparative health (estimate = 1.301, 95% *CI* = 0.396, 2.207, *SE* = 0.46, OR = 3.67, *Z* = 2.81, Wald's χ2(1) = 7.93, *p* < 0.01). On the other hand, bootstrap confirmed only comparative self-rated health as a significant predictor of depression among Czech students (BCa 95% *CI*<sup>B</sup> = 0.119, 2.223). This model of regression was significant, χ2(296) = 38.639, *p* < 0.001, *R*<sup>2</sup> CS = 0.12, *R*<sup>2</sup> McF = 0.12, *R*<sup>2</sup> <sup>N</sup> = 0.18.


**Figure 3.** Logistic regression estimates heatmap for depression symptoms among university students from Colombia, Czechia, Germany, Israel, Poland, Russia, Slovenia, Turkey, and Ukraine. Positive estimates are marked in red, negative estimates are marked in blue. PIC = Perceived Impact of COVID-19 on Students' Well-being. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

Among German students (Table S5), depression can be predicted by the total perceived impact of the coronavirus on daily life (estimate = 1.33, 95% *CI* = 0.483, 2.170, *SE* = 0.43, OR = 3.77, *Z* = 3.08, Wald's χ2(1) = 9.51, *p* < 0.01) and physical health (estimate = 1.38, 95% *CI* = 0.419, 0.2.345, *SE* = 49, OR = 3.98, *Z* = 2.81, Wald's χ2(1) = 7.91, *p* < 0.01). These findings were confirmed by the bootstrapping method for both variables, total PIC (BCa 95% *CI*<sup>B</sup> = 0.351, 2.101) and PH (BCa 95% *CI*<sup>B</sup> = 0.232, 2.395). The regression model showed adequate significance, χ2(253) = 60.82, *p* < 0.001, *R*<sup>2</sup> CS = 21, *R*<sup>2</sup> McF = 0.17, *R*<sup>2</sup> <sup>N</sup> = 0.28.

For Israeli participants (Table S6), excluding study level and physical health as predictors of depression, χ2(189) = 49.79, *p* < 0.001, *R*<sup>2</sup> CS = 0.22, *R*<sup>2</sup> McF = 0.18, *R*<sup>2</sup> <sup>N</sup> = 0.30, the regression model was found to be significant. Among two significant predictors of depression, namely, exposure to the coronavirus (estimate = 0.86, 95% *CI* = 0.092, 1.633, *SE* = 0.39, OR = 2.37, *Z* = 2.19, Wald's χ2(1) = 4.81, *p* < 0.05) and total PIC (estimate = 1.35, 95% *CI* = 0.302, 2.394, *SE* = 0.53, OR = 3.85, *Z* = 2.53, Wald's χ2(1) = 6.39, *p* < 0.05), only total PIC was confirmed by the bootstrapping procedure (BCa 95% *CI*<sup>B</sup> = 0.147, 2.524).

When the regression was performed for Polish students (Table S7), three variables revealed sufficient significance level using both classical and bootstrapping methods, namely gender (estimate = 0.78, 95% *CI* = 0.190, 1.377, *SE* = 0.30, OR = 2.19, *Z* = 2.59, Wald's χ2(1) = 6.70, *p* < 0.01; BCa 95% *CI*<sup>B</sup> = 0.115, 1.393), total PIC (estimate = 1.00, 95% *CI* = 0.299, 1.700, *SE* = 0.36, OR = 2.72, *Z* = 2.80, Wald's χ2(1) = 7.82, *p* < 0.01; BCa 95% *CI*<sup>B</sup> = 0.197, 1.653), and comparative PH (estimate = 1.27, 95% *CI* = 0.286, 2.250, *SE* = 0.50, OR = 3.55,

*Z* = 2.53, Wald's χ2(1) = 6.41, *p* < 0.05; BCa 95% *CI*<sup>B</sup> = 0.126, 2.323). The regression model was found to be significant, χ2(288) = 62.46, *p* < 0.001, *R*<sup>2</sup> CS = 0.19, *R*<sup>2</sup> McF = 0.15, *R*<sup>2</sup> <sup>N</sup> = 0.25.

In the Russian sample of university students (Table S8), the following predictors of depression were found: gender (estimate = 0.93, 95% *CI* = 0.300, 1.560, *SE* = 0.32, OR = 2.53, *Z* = 2.89, Wald's χ2(1) = 8.37, *p* < 0.01), exposure to the coronavirus (estimate = 0.76, 95% *CI* = 0.080, 1.443, *SE* = 0.35, OR = 2.14, *Z* = 2.19, Wald's χ2(1) = 4.80, *p* < 0.05), PICqualification (estimate = 0.90, 95% *CI* = 0.175, 1.624, *SE* = 0.37, OR = 2.46, *Z* = 2.43, Wald's χ2(1) = 5.92, *p* < 0.05), PIC-social relationships (estimate = 1.39, 95% *CI* = 0.681, 2.100, *SE* = 0.36, OR = 4.02, *Z* = 3.84, Wald's χ2(1) = 14.76, *p* < 0.001), physical activity (estimate = −0.78, 95% *CI* = −1.375, −0.176, *SE* = 0.31, OR = 0.46, *Z* = −2.54, Wald's χ2(1) = 6.43, *p* < 0.05), and physical health (estimate = 1.39, 95% *CI* = 0.472, 2.304, *SE* = 0.48, OR = 4.01, *Z* = 2.97, Wald's χ2(1) = 8.83, *p* < 0.01). In addition, bootstrap showed a significant effect for gender (BCa 95% *CI*<sup>B</sup> = 0.209, 1.590), PIC-qualification (BCa 95% *CI*<sup>B</sup> = 0.102, 1.636), PIC-social relationships (BCa 95% *CI*<sup>B</sup> = 0.195, 2.157), and PH (BCa 95% <sup>B</sup> = 0.125, 2.305). The regression model was significant, χ2(271) = 69.38, *p* < 0.001, *R*<sup>2</sup> CS = 0.22, *R*2 McF = 0.18, *R*<sup>2</sup> <sup>N</sup> = 0.29.

The regression model conducted in the Slovenian sample of students (Table S9) showed a good fit, χ2(197) = 77.13, *p* < 0.001, *R*<sup>2</sup> CS = 0.31, *R*<sup>2</sup> McF = 0.29, *R*<sup>2</sup> <sup>N</sup> = 0.43. Among variables, exposure to COVID-19 (estimate = 1.14, 95% *CI* = 0.258, 2.021, *SE* = 0.45, OR = 3.125, *Z* = 2.534, Wald's χ2(1) = 6.423, *p* < 0.05), PIC-qualifications (estimate = 1.26, 95% *CI* = 0.337, 2.183, *SE* = 0.47, OR = 3.52, *Z* = 2.68, Wald's χ2(1) = 7.154, *p* < 0.01), and comparative PH (estimate = 1.61, 95% *CI* = 0.402, 2.825, *SE* = 0.62, OR = 5.02, *Z* = 2.61, Wald's χ2(1) = 6.81, *p* < 0.01) were found to be significant predictors of depression. The bootstrapping method confirmed the significance of all three variables, namely exposure to the coronavirus (BCa 95% *CI*<sup>B</sup> = 0.058, 2.114), PIC-qualifications (BCa 95% *CI*<sup>B</sup> = 0.155, 2.076), and comparative PH (BCa 95% *CI*<sup>B</sup> = 0.165, 2.916).

Although three variables were shown as significant predictors of depression among Turkish students (Table S10), namely gender (estimate = 0.91, 95% *CI* = 0.398, 1.413, *SE* = 0.26, OR = 2.48, *Z* = 3.50, Wald's χ2(1) = 12.24, *p* < 0.001), PIC-social relationships (estimate = 0.61, 95% *CI* <sup>=</sup> −0.001, 1.213, *SE* = 0.31, OR = 1.83, *<sup>Z</sup>* = 1.96, Wald's <sup>χ</sup>2(1) = 3.84, *p* < 0.05), and comparative PH (estimate = 1.37, 95% *CI* = 0.090, 2.583, *SE* = 0.64, OR = 3.81, *Z* = 2.10, Wald's χ2(1) 4.42, *p* < 0.05), only gender was confirmed using the bootstrapping method (BCa 95% *CI*<sup>B</sup> = 0.350, 1.429). The model's fit for the Turkish sample was good, χ2(294) = 42.17, *p* < 0.001, *R*<sup>2</sup> CS = 0.13, *R*<sup>2</sup> McF = 0.10, *R*<sup>2</sup> <sup>N</sup> = 0.18. Among Ukrainian participants (Table S11), depression can be predicted by gender (estimate = 0.80, 95% *CI* = 0.150, 1,456, SE = 0.33, OR = 2.23, *Z* = 2.41, Wald's χ2(1) = 5.81, *p* < 0.05), exposure to COVID-19 (estimate = 0.87, 95% *CI* = 0.060, 1.686, SE = 0.42, OR = 2.39, *Z* = 2.11, Wald's χ2(1) = 4.43, *p* < 0.05), and PA (estimate = −0.85, 95% *CI* = −1.413, −0.288, SE = 0.29, OR = 0.427, *<sup>Z</sup>* <sup>=</sup> −2.97, Wald's <sup>χ</sup>2(1) = 8.79, *<sup>p</sup>* < 0.01). Gender (BCa 95% *CI*<sup>B</sup> = 0.070, 1.512) and PA (BCa 95% *CI*<sup>B</sup> = −1.396, −0.218) were also found to be significant predictors when the bootstrapping method was used. The regression model presented a good fit, χ2(298) = 45.65, *p* < 0.001, *R*<sup>2</sup> CS = 0.14, *R*<sup>2</sup> McF = 0.12, *R*<sup>2</sup> <sup>N</sup> = 0.20.
