This ecological study examines the variations in diarrhoea-specific infant mortality rates among municipalities in the State of Ceará, north-east Brazil, using data from a community health workers’ programme. Diarrhoea is the main cause of postneonatal deaths in Ceará, and diarrhoea mortality rates vary substantially among municipalities, from 7 to 50 per thousand live births. To determine the inter-relationships between potential predictors of diarrhoea-specific infant mortality, eleven variables were classified into proximate determinants (i.e. adequate weight gain and exclusive breast-feeding in first 4 months) and underlying determinants (i.e. health services and socioeconomic variables). The health services variables included percentage with prenatal care up-to-date, participation in growth monitoring and immunization up-to-date, while the socioeconomic factors included female illiteracy rate, per capita gross municipality product and percentage of households with low income, percentage of households with inadequate water supply and inadequate sanitation, and urbanization. Using linear regression analysis variables were included from each group to build regression models. The significant determinants of variability in diarrhoea-specific infant mortality between municipalities were prevalence of infants exclusively breast-feeding, percentage of infants with adequate weight gain, percentage of pregnant women with prenatal care up-to-date, female illiteracy rate and inadequate water supply. These findings suggest that community-based promotion of exclusive breast-feeding in the first 4 months and care-giving behaviours that prevent weight faltering, including weaning practices and feeding during and following diarrhoea episodes, may further reduce municipality-level diarrhoea-specific mortality. Primary heath care strategies addressing these two proximate determinants provide only a partial solution to reducing diarrhoeal disease mortality. Improvements in municipal health services (prenatal care) and socioeconomic status variables, including water supply and maternal education, can also contribute to reduction of infant mortality due to diarrhoea. These results may be used by government health officials to set priorities by considering not only the strength of the association between selected risk factors and diarrhoea mortality rates, but also the prevalence of the risk factors being considered at the municipality level. Finally, the methods used are applicable to other settings with community-based primary health care decentralized to the state or municipal level.