OBJECTIVE: We aimed to assess Child Death Overview Panel (CDOP) data validity, and cause of death classification, by comparison with information from a local birth cohort study (Born in Bradford, BiB), and another cause of death coding system (causes of death and associated conditions-CODAC). We then aimed to use CDOP data to calculate ethnic-specific infant mortality rates (IMRs), and compare characteristics of infants who died of congenital anomalies (CA) with those who died from other causes (non-CA). DESIGN: Retrospective cohort study. SETTING: Bradford Metropolitan District. PATIENTS: All infant deaths, 2008 to 2013. MAIN OUTCOME MEASURES: Infant mortality rates from CA and non-CA causes. RESULTS: 315 infant deaths were included, 56 of whom were BiB recruits. Agreement between CDOP and BiB was moderate to perfect for all characteristics except ethnicity, which showed weak agreement (kappa=0.58). The same deaths (27/56) were classified as CA by CDOP and CODAC. IMRs (per 1000 live births, 2009-2013) were highest in Pakistani infants (all causes 9.8, CA cause 5.5) compared with white British (all causes 4.3, CA cause 1.3) and other infants (all causes 5.1, CA cause 1.4). In multivariate analysis, infants who died of CA cause were more likely to have been born at term (OR 3.18) and to consanguineous parents (OR 3.28) than infants who died of non-CA cause. CONCLUSIONS: Excess Pakistani mortality appears to be partly explained by an excess of deaths from CA, which in this population appears associated with a greater prevalence of consanguinity.