Table 2. Risk of bias.

Potential cause of bias	Optimal characteristics of an unbiased study	Risk of bias
		Bara 2009 (18)	Berthelsen 2015 (25) Flo 2014 (24) Thun 2014 (21)	Bildt 2002 (17)	Bohle 1989 (16)	d‘Errico 2010 (19)	Driesen 2001 (20)	Lin 2012 (22)	Nabe-Nielsen 2011 (26)	Norder 2015 (23)
1. Study participants	Study participants are representative of the study population on which the conclusion is to be drawn.	Low	Medium	Low	Medium	Medium	Low	Medium	Low	Low
2. Loss of study participants	The available study data, i.e. data on the study participants not lost to follow-up, is representative of all the original study participants.	Medium	Medium	Medium	Not stated	High	Medium	Medium	Medium	Low
3. Measurement of cause variable	Shift work was recorded in a valid way, and equally for all participants.	Low	Low	Low	Low	Medium	Low	Low	Low	Low
4. Measurement of target parameter	Depression was recorded in a valid way, and equally for all participants.	Low	Low	Low	Low	Low	Low	Low	Low	Low
5. Adjustment for confounding factors	Major confounding variables were appropriately adjusted for.	Low	Low	Low	Low	Low	Low	Low	Low	Medium
6. Statistical analysis and reporting of findings	Statistical analysis is appropriate, and all primary target parameters are reported.	Low	Low	Low	Low	Low	Low	Low	Medium	Low

Cause 1: Risk was classified as low if the participation rate was at least 80% or nonparticipation was not selective.

Cause 2: Risk was classified as low if the number of participants at short-term (<1 year) follow-up was at least 80% of the number at baseline, or if the number at long-term (= 1 year) follow-up was 70% of the number at baseline, or if there was information to the effect that there were no differences between participants and those lost to follow-up that might lead to selection.