OBJECTIVES: This study aimed to better understand reasons why children in South Africa die at home, including caregivers' care-seeking experiences, decision-making, choice of treatment provider and barriers to accessing care during a child's final illness. DESIGN: This qualitative study included semi-structured in-depth interviews and focus group discussions with caregivers of children who died below the age of 5 years. Data were thematically analysed, and key findings compared with the Pathways to Survival Framework-a model frequently used in the study of child mortality. An adapted model was developed. SETTING: Two rural health and demographic surveillance system (HDSS) sites in South Africa-the Agincourt HDSS and the Africa Health Research Institute. PARTICIPANTS: Thirty-eight caregivers of deceased children (29 participated in in-depth interviews and 9 were participants in two focus group discussions). Caregivers were purposively sampled to ensure maximum variation across place of death, child age at death, household socioeconomic status, maternal migration status and maternal HIV status. FINDINGS: Although caregivers faced barriers in providing care to children (including insufficient knowledge and poor transport), almost all did seek care from the formal health system. Negative experiences in health facilities did not deter care-seeking, but most respondents still received poor quality care and were not given adequate safety-netting advice. Traditional healers were only consulted as a last resort when other approaches had failed. CONCLUSION: Barriers to accessing healthcare disrupt the workings of previously accepted care-seeking models. The adapted model presented in this paper more realistically reflects care-seeking experiences and decision-making during severe childhood illness in rural South Africa and helps explain both the persistence of home deaths despite seeking healthcare, and the impact of a child's death on care-seeking in future childhood illness. This model can be used as the basis for developing interventions to reduce under-5 mortality.