Assessing sustainment of health worker outcomes beyond program end: Evaluation results from an infant and young child feeding intervention in Bangladesh

**Background:** Optimal infant and young child feeding (IYCF) practices strongly influence child nutrition, growth, and survival, yet interventions often struggle with sustainment after external funding ends. The Alive & Thrive (A&T) initiative, supported by the Bill and Melinda Gates Foundation, was implemented in Bangladesh from 2009-2014 as a demonstration project to improve IYCF behaviors through interpersonal counseling, community mobilization, mass media, and policy advocacy. This study examines whether improvements in health worker outcomes persisted three years after program conclusion, contributing to the limited empirical literature on program sustainability.

**Methods:** Data come from a cluster-randomized controlled trial with repeated cross-sectional surveys of frontline health workers in 20 randomly selected rural upazillas (10 intervention, 10 comparison) across 5 divisions of Bangladesh. Surveys were conducted at baseline (2010, n=290), endline (2014, n=511), and post-endline (2017, n=600). The primary outcome was quality of IYCF service delivery (count of topics discussed during counseling, range 0-14, measured at endline and post-endline). Intermediate outcomes included IYCF knowledge score (0-14), job satisfaction (binary: very satisfied vs. not), and job readiness (binary: adequate training for job). Program activities assessed as effect modifiers were comprehensiveness of refresher training (proportion of topics recalled, 0-1) and receipt/amount of financial incentives. Multivariable difference-in-difference linear regression models with robust standard errors clustered at the upazilla level were used, adjusting for health worker cadre, age, years in job, and education level.

**Key Results:** At endline, health workers in intervention areas discussed significantly more IYCF topics than those in comparison areas (4.93 vs. 3.97 topics, p<0.001), but by post-endline this gap was no longer significant (3.97 vs. 3.34 topics, p=0.067), indicating a "voltage drop" in service delivery quality. IYCF knowledge showed sustained improvement: at baseline there was no difference between groups (9.51 vs. 9.59, p=0.91), but by endline intervention-area workers had significantly higher scores (12.80 vs. 11.09, p<0.001), and this gap persisted at post-endline (12.65 vs. 11.13, p<0.001). Job satisfaction increased in intervention areas during the program (from 69% to 79% satisfied) but declined to 48% by post-endline, not significantly different from comparison areas (38%, p=0.10). Job readiness improved in intervention areas by endline (96% vs. 86% in comparison, p=0.0021) but regressed to baseline levels by post-endline (83% vs. 86%, p=0.28). Program activities also declined: refresher training comprehensiveness remained significantly higher in intervention areas at post-endline (0.15 vs. 0.13, p=0.018) but incentive receipt dropped dramatically from 90% to 5% in intervention areas. Comprehensive refresher trainings appeared protective against service delivery deterioration—workers with more comprehensive trainings delivered more IYCF messages at post-endline.

**Clinical Implications:** This study demonstrates that while health worker knowledge gains from IYCF programs can be sustained, improvements in service delivery quality may not persist without continued program support. The findings highlight a "know-do gap" where knowledge alone is insufficient to maintain practice changes. Comprehensive refresher trainings may protect against deterioration in service quality, suggesting that even modest continued investment in training could help sustain program effects. The substantial decline in financial incentives (from 90% to 5% receipt in intervention areas) likely contributed to declines in job satisfaction and readiness. Programs of limited duration should prioritize identifying and investing in protective factors like refresher training to encourage sustained impact, and future research should collect post-endline data to empirically test sustainability concepts.