For decades, women's employment has been seen as crucial for achieving greater autonomy and empowerment for women, and for promoting better health and nutrition outcomes for children, particularly in low- and middle-income countries (LMIC). However, numerous empirical studies of the relationship between women's work and child outcomes have shown mixed results. Our study tests the assumptions of a model that suggests loss of maternal care during working hours may produce negative health outcomes for children. We use longitudinal data collected from traditionally semi-nomadic, boat-dwelling Shodagor families in Matlab, Bangladesh to determine the importance of maternal care as a mechanism influencing the relationship between women's work and child illness. We use Bayesian linear mixed models to assess the influence of occupation and amount of care on average days of child illness per month, and also to examine the role that allomothers play in buffering against potential negative impacts of lost maternal care on child illness. Results show that children who receive more care from mothers experience fewer days of illness, and that availability of high-quality alloparents mediates the relationship between maternal work and child health. These results indicate that both the care and resources provided by mothers influence children's biological outcomes. This has important implications for policy and aid interventions in LMIC, which have been developed to capitalize on an assumed positive relationship between maternal work and child health and nutrition.