Matrix of intervention activities.

Breastfeeding is a cost-effective intervention vital to child health, yet rates remain suboptimal, especially in fragile settings. Primary care providers are critical to improving breastfeeding outcomes but face multilevel barriers. While hospital-based interventions have shown success, adaptations to primary care are limited, and understanding of the mechanisms driving provider behavior change remains insufficient. In Kosovo, where exclusive breastfeeding is declining alongside increasing infant mortality, we piloted a behavioral science intervention to improve primary care provider practices. A before-and-after design (baseline June–July 2019; endline June–July 2021) using mixed-methods assessed changes in provider behavior, maternal experience, and the institutional environment across five municipalities. Validation compared maternal reports with clinical observations. The intervention targeted provider behavior with emotionally engaging activities, job aids and institutional strategies, and was co-designed with system actors. Data sources included direct observation of consultations, exit interviews with mothers, provider interviews and facility assessments. Across 609 consultations with matched exit interviews, we recorded a 14.9 percentage point increase in the mean frequency of breastfeeding-friendly clinical behaviors by providers (95% CI: 6.4-23.4) and a 13.3 percentage point increase in interpersonal behaviors (95% CI: 5.2-21.3) after versus before intervention. Gains were strongest for providers explaining ability to breastfeed (+39.0 points), inviting mother to ask questions (+21.8 points) and explaining follow-up required (+18.7 points). Gaps between observer and maternal reports narrowed across most indicators, indicating stronger alignment between care delivered and perceived. Mothers reported greater confidence and better care experiences. Providers’ knowledge and attitudes improved and shifts in leadership engagement and facility norms were seen, though broader system integration was limited. A behaviorally informed, co-designed intervention can measurably improve breastfeeding counseling behaviors and maternal experiences in Kosovo’s primary care settings. Findings underscore the value of targeting deeper behavioral drivers, not just knowledge gaps, within fragile health systems.