Table 5.

Summary of risk of bias of included studies for non randomised trials, controlled before after studies, interrupted time series, and historically controlled trials (1=strong, 2=moderate, 3=weak, -=Not applicable)

Study author, Year	Selection bias	Design	Confounders	Blinding	Data collection methods	Withdrawals/ dropouts	Global rating
Brenner et al, (2011) [92]	1	1	3	2	1	-	2
Chandir et al, (2010) [83]	2	2	1	2	-	3	2
Fatugase et al, (2013) [58]	2	2	1	3	2	1	2
Finocchario-Kessler et al, (2014) [37]	1	3	1	2	-	1	2
Galasso et al, [35]	2	3	3	2	1	3	3
Kundu et al, (2012) [85]	1	3	3	2	1	1	3
McCollum et al, (2012) [34]	2	2	3	2	-	1	2
Oche et al, (2011) [67]	1	2	3	2	-	1	2
Ridde et al, (2013) [93]	2	2	3	2	-	3	2
Robinson et al, (2001) [46]	3	2	3	2	-	-	3
Ryman et al, (2011) [47]	2	2	3	2	-	-	2
Schwartz et al, (2004) [52]	2	2	3	2	-	-	3
Simonyanm et al, (2013) [42]	1	2	1	2	-	1	1

Note: Data collection methods “not applicable” when validity and reliability were not of tools were not a concern (e.g. self-report or medical records). Withdrawals/dropouts “not applicable” when surveys were conducted on a different group of people at baseline and follow-up