Table 3.

Methodological Quality of Included Studies

Study	Allocation methods	Attrition	Other potential limitations
Coronado et al., 2016	Random allocation met	Low attrition (<1%)	Mammography van locations differed by clinic site and affected mammography completion rates. Study was underpowered to examine clinic-level effects or moderators.
Elder et al., 2017	Random allocation met	Not reported	Self-report for cancer screening outcomes. Cluster effects of churches not explained thoroughly. Small sample size of comparison group.
Fernandez et al., 2009	Random allocation unmet	High attrition (37% in intervention arm; 31% in control arm)	Screening records only located for 58.3% of women in mammography cohort. Potential for nonresponse bias by different response rates across sites. Loss to follow-up was 33.1%. There was differential attrition between study arms. Limited generalizability of results since study was conducted in the rural U.S.-Mexico border region.
Jandorf et al., 2014	Random allocation unclear	High attrition in 8-month follow-up (41%) compared to 2-month follow-up (24%)	Bias from self-reported screening exams. Potential bias in use of medical record review, particularly for medically underserved individuals. Different study sites (New York compared to Arkansas) makes comparisons between study sites difficult because of different compositions of Hispanic subgroups (e.g., Mexican compared to Puerto Rican women).
Nuño et al., 2011	Random allocation met	Low attrition (4% in intervention arm; 3% in control arm)	One-third of mammograms reported were self-reported only. Possible contamination between study groups. Limited generalizability of results since study was conducted in the rural U.S.-Mexico border region.