Evans 2011.

Methods	Interrupted time series study; triple differences with state fixed‐effect method; 1993‐2001 and 1999‐2004
Participants	127,209 participants of 9 surveys of the Behavioral Risk Factor Surveillance System (1993‐2001); working‐age adults (21 to 40 years); women; 1‐parent and 2‐parent families (127,209 observations) and 1‐parent families (64,033 observations); US 2683 to 3090 participants of 4 waves of the National Health and Nutrition Examination Survey III (1999‐2004); working‐age adults (21 to 40 years); women; US
Interventions	Increase in value of the Earned Income Tax Credit in 1996
Outcomes	Primary outcomes: self rated general health; mental health or psychological distress Secondary outcomes: physical health (number of bad physical health days over the last 30 days; number of risky biomarkers for inflammation, cardiovascular disease and metabolic conditions); change in employment
Notes	Intervention context: permanent in‐work tax credit intervention for families introduced in 1975; means‐tested to low‐income groups (phase out starts at 17% to 42% of average income from wages (depending on family type); Immervoll 2009); amongst the most generous in‐work tax credit interventions internationally (up to 7% of an average income from wages for families with 1 dependent child; up to 11% of an average income from wages for families with 2 dependent children; Immervoll 2009), distributed nearly USD 62 billion to over 27 million individuals in 2011; had a high (approximately 80%) up‐take and low (less than 1%) administration costs; follow‐up for intervention to outcome measure was 2 years (1996‐18998); follow‐up for intervention to outcome measure was 5 years for analyses of the Behavioral Risk Factor Surveillance System (1996‐2001) and 8 years for analyses of the National Health and Nutrition Examination Survey III (1996‐2004) Equity impact estimated for the following PROGRESS variables: level of education
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Unclear risk	Unclear whether random sampling strategy employed; nationally representative sample achieved
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Survey participants not allocated to the intervention by the researchers; secondary analysis of survey data collected for a different purpose than estimating the impact of in‐work tax credit for families on health in adults (that is, blinding of participants and personnel neither feasible nor necessary)
Blinding of outcome assessment (detection bias) All outcomes	High risk	Survey participants not allocated to the intervention by the researchers; secondary analysis of survey data collected for a different purpose than estimating the impact of in‐work tax credit for families on health in adults (that is, blinding of outcome assessors neither feasible nor necessary); high risk from misclassification bias of the exposure due to crude exposure assessment
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Survey response rates not reported; unclear how missing outcome data (if any) were addressed
Selective reporting (reporting bias)	Low risk	Significant and non‐significant effects reported
Other bias	High risk	The intervention was unlikely to alter data collection. The point of analysis was the point of intervention. The baseline data from the National Health and Nutrition Examination Survey III was for 1999 (3 years after the intervention had been implemented in 1996). The study adjusted for trends over time in the health outcome between women eligible for a large increase in EITC in 1996 (women with 2 or more dependent children) and women not eligible for the increase in EITC (women with 1 dependent child) within the group of potentially EITC‐eligible women (low‐education mothers), and controlled for trends in never EITC‐eligible women (high‐education mothers); however, it assumed that trends in the health outcome in the potentially and the never EITC‐eligible women were comparable, meaning that the study had a high risk of bias from unmeasured or unadjusted confounding. Reverse causation not controlled for