Table 3. Multivariate logistic regression models to predict clinically relevant psychological distress (GHQ-12 ≥ 3) at T1, T2, T3 and all timepoints.

Models were selected through stepwise forwards and backwards regression to obtain the lowest Akaike Information Criterion (AIC). In the final mixed effects model, participant ID was included as a random effect. OR: odds ratio. 95% CI: 95% confidence interval. AIC: Akaike Information Criterion. ROC: receiver operating curve. ICC: intraclass correlation coefficient. For weekly hospital shifts, the OR is the increase in risk per additional shift worked.

	T1		T2		T3		All timepoints
Variables	OR (95% CI)	p-value	OR (95% CI)	p-value	OR (95% CI)	p-value	OR (95% CI)	p-value
Sex: Male	0.66 (0.36, 1.20)	0.2	0.59 (0.24, 1.43)	0.3	0.36 (0.14, 0.89)	0.033	0.43 (0.22, 0.85)	0.016
Profession: Physician	—	—	—	—	—	—	—	—
Profession: RN	2.54 (1.34, 4.84)	0.005	1.21 (0.45, 3.30)	0.7	1.26 (0.54, 2.98)	0.6	1.69 (1.11, 2.58)	0.014
Profession: RT	1.45 (0.68, 3.14)	0.3	1.19 (0.35, 4.04)	0.8	0.82 (0.25, 2.48)	0.7	1.56 (0.65, 3.73)	0.32
Profession: Allied health	1.11 (0.48, 2.60)	0.8	1.92 (0.59, 6.36)	0.3	0.91 (0.29, 2.76)	0.9	1.54 (0.60, 3.92)	0.37
Median number of weekly hospital shifts			0.76 (0.59, 0.98)	0.037
Median number of weekly hospital shifts with COVID-19 exposure					1.33 (1.09, 1.64)	0.006
Median number of COVID-19 patients in the ICU			1.06 (1.00, 1.13)	0.070
Timepoint: T2							0.24 (0.15, 0.40)	<0.001
Timepoint: T3							0.16 (0.09, 0.27)	<0.001
No. of observations	309		166		190		695
AIC	392.5		230.0		239.7		862.7
Area under the ROC	0.64		0.64		0.68		0.69
ICC							0.40