Peer counselling (PC) programmes have been shown to improve breast-feeding outcomes in populations at risk for early discontinuation. Our objective was to describe associations between programme components (individual and combinations) and breast-feeding outcomes (duration and exclusivity) in a PC programme for low-income women.

Secondary analysis of programme data. Multivariable-adjusted Cox proportional hazards models were used to examine associations between type and quantity of peer contacts with breast-feeding outcomes. Types of contacts included in-person (hospital or home), phone or other (e.g. mail, text). Quantities of contacts were considered ‘optimal’ if they adhered to standard programme guidelines.

Programme data collected from 2005 to 2011 in Michigan’s Breastfeeding Initiative Peer Counseling Program.

Low-income (n 5886) women enrolled prenatally.

For each additional home, phone and other PC contact there was a significant reduction in the hazard of discontinuing any breast-feeding by 6 months (hazard ratio (HR)=0·90 (95 % CI 0·88, 0·92); HR=0·89 (95 % CI 0·87, 0·90); and HR=0·93 (95 % CI 0·90, 0·96), respectively) and exclusive breast-feeding by 3 months (HR=0·92 (95 % CI 0·89, 0·95); HR=0·90 (95 % CI 0·88, 0·91); and HR=0·93 (95 % CI 0·89, 0·97), respectively). Participants receiving greater than optimal in-person and less than optimal phone contacts had a reduced hazard of any and exclusive breast-feeding discontinuation compared with those who were considered to have optimum quantities of contacts (HR=0·17 (95 % CI 0·14, 0·20) and HR=0·28 (95 % CI 0·23, 0·35), respectively).

Specific components of a large PC programme appeared to have an appreciable impact on breast-feeding outcomes. In-person contacts were essential to improving breast-feeding outcomes, but defining optimal programme components is complex.