Table 3. Quality assessment of included studies, includes cohort studies and controlled before and after studies.

Cohort Studies
	Selection Bias	Allocation Bias	Confounding	Blinding	Data Collection Objective	Attrition Bias	Intervention Integrity	Statistics	Overall Strength
Fung et al, 2010 [46]	Unclear – could be a health worker effect present – the non copayment group derives coverage from employers whereas the copayment group must obtain coverage individually.	No	Propensity score matched. Ethnicity not included – sensitivity analysis ruled out importance	Not mentioned	Yes	Unclear	Yes	Yes	Weak
Polinski et al, 2011[51]	Unexposed group is a composite group made of three separate groups. Retirees appear to use more medicines than exposed groups. There were some differences in age, gender and ethnicity.	No	Propensity matched on a wide range of variables. Authors split study into an early and established group because there was no baseline data for the early group.	Not mentioned	Yes	No	Yes	Yes	Weak
Wang et al, 2011 [49]	No, although may be education differences between groups 1 and 8. See Doshi.	No	Propensity matched, but some differences in groups prevailed. Income/neighbourhood proxy for income not included.	Not Mentioned	Yes	Unclear	Yes	Yes	Weak
Maciejewski et al, 2010 [50]	No – see Doshi and Wang	No	Propensity matched, but some differences in groups prevailed Income/neighbourhood proxy for income not included.	Not mentioned	Yes	Unclear	Yes	Yes	Weak