Objective: To analyse the contribution of maternal survival and HIV status to child (under-5 years) mortality in a rural population cohort in South-west Uganda.

Methods: Approximately 10 000 people residing in 15 neighbouring villages were followed between 1989 and 2000 using annual censuses and serological surveys to collect data on births, deaths, and adult HIV serostatus. Mother-child records were linked, child mortality risks (per 1000 births) and hazard ratios (HRs) for child mortality according to maternal HIV serostatus were computed, allowing for time-varying covariates.

Results: A total of 3727 children were born, of whom 415 died during 14 110 child years of follow-up. Mother's HIV status at birth was ascertained unambiguously for 3004 children, of whom 218 were born to HIV-positive mothers. Infant mortality risk was higher for HIV seropositive than seronegative mothers (225 versus 53) as was child mortality risk (313 versus 114). Child mortality risk was also higher for mothers who died (571) than for surviving mothers (128). After controlling for child's age and sex, independent predictors of mortality in children were: mother's terminal illness or death (HR = 3.8); mother being HIV positive (HR = 3.2); child being a twin (HR = 2.0); teenage motherhood (HR = 1.7) and maternal absence (HR = 1.7).

Conclusion: Maternal survival and HIV status are strong predictors of child survival. The higher mortality in HIV-infected women compounds mortality risks for their children, regardless of children's HIV status. Programmes aimed at the welfare of children should take into account the independent effect of mothers' HIV and vital status.