Farm children face unique health risks due to sharing their residential environment with hazardous machinery and materials. Causes of mortality among farm children have not been comprehensively described. OBJECTIVE: In the Agricultural Health Study (AHS) cohort, we examined causes of mortality among 21,360 children in Iowa and North Carolina between 1975 and 1998. METHODS: We matched identifying information for children provided by mothers on self-administered questionnaires to state death registries (1975-1998). Data on farm and family characteristics were provided by parents via enrollment questionnaires (1993-1997). Standardized mortality ratios (SMRs) were calculated, using state mortality data to generate expected deaths. We used logistic regression to examine parent, child and farm characteristics associated with injury mortality. RESULTS: There were 162 deaths in Iowa (SMR=0.69; 95% confidence interval (CI)=0.60, 0.81) and 26 deaths in North Carolina (SMR=0.42; 95%CI=0.28, 0.61) in children aged 0-19 years. This deficit was largely due to deaths in the first year of life. Although deaths from overall unintentional injury were not increased, excess agricultural machinery mortality was observed in Iowa (SMR=9.25; 95% CI=5.12, 16.70). In case-control comparisons, maternal age less than 25 years at child's birth (OR=2.17; 95%CI=1.05, 4.49) and having more than 2 children in the family (OR=2.79; 95%CI=1.47, 5.30) were associated with increased child injury mortality. For children under 14 years, participation in farm work was associated with increased risk of agricultural machine-related mortality (OR=3.92; 95% CI=1.04, 14.78). CONCLUSIONS: Parent and child characteristics associated with child injury mortality could be used to target farm safety interventions.