Seixas 2011.

Methods	Both cluster and individually randomised RCT, first 4 work sites got baseline training, then these were cluster‐randomised to tool‐box or no tool‐box training and then individuals were randomised to noise level indicators or no indicators
Participants	Construction workers; various trades n = 176 USA
Interventions	Many comparisons possible, we choose to compare two interventions considered to be most relevant for practice Intervention 1: baseline training plus noise 'tool box' on‐site training (n = 44) Intervention 2: baseline training plus noise 'tool box' on‐site training plus personal noise level indicator (n = 41) Control: baseline training (n = 46)
Outcomes	Noise level measured as Leq at 2 and 4 months' follow‐up
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
14. Blinding (subjects)	High risk	different interventions at the same site visible
15. Blinding (outcome assessors)	Unclear risk	not reported
16. Retrospective unplanned subgroup analyses	Low risk	no subgroup analyses
17. Follow‐up	Low risk	similar follow‐up
18. Statistical tests	Unclear risk	differences in outcome not tested
19. Compliance	Low risk	NLI checked every week, participating in training course or excluded
20. Outcome measures	Low risk	Leq according to NIOSH criteria
21. Selection bias (population)	Low risk	different intervention groups
22. Selection bias (time)	Low risk	same time
23. Randomization	Unclear risk	methods not reported
24. Allocation concealment	Unclear risk	not reported
25. Adjustment for confounding	Low risk	intention to treat
26. Incomplete outcome data	High risk	reported, no differences between groups, but loss to follow‐up ranged from 20%‐33%