Objectives To identify the socioeconomic and geographical groups in which the recent under-5 mortality increase observed in several African countries was most pronounced, and to explore the contribution of a number of proximate determinants of under-5 mortality.

Methods Time trends in under-5 mortality were assessed with Cox Proportional Hazards regression analysis, using Demographic and Health Surveys data for Burkina Faso, Cameroon, Côte d'lvoire, Kenya and Zimbabwe for the late 1980s – 1990s. We tested for differences in time trends between socioeconomic and rural/urban subgroups, and described the inequalities in time trends in living conditions, malnutrition and health care use.

Results Under-5 mortality increased substantially (ranging from 25% to 71% in 10 years) within the five countries. In Kenya, the increase was the largest among children born to less educated mothers (test for difference between educational groups: P = 0.074) and in rural areas (P = 0.090). In Cameroon, the increase was the largest among the higher educated (P = 0.013), and in Zimbabwe among the higher educated (P = 0.098) and in urban areas (P = 0.093). For Burkina Faso and Côte d'Ivoire, we did not observe statistically significant differences between educational and rural/urban subgroups. The decline in skilled delivery attendance in Zimbabwe and Kenya was similar among the less and higher educated. The decline in immunization coverage during the mid-1990s in Zimbabwe was the largest in the group with the highest mortality increase, but in Kenya it was as large among the less and higher educated. Whereas in Kenya the increase in malnutrition was the largest in the group with the highest mortality increase, this was not the case in Zimbabwe.

Conclusions The recent increase in under-5 mortality in some African countries was highly concentrated in specific population subgroups. Exactly which groups were most affected was highly variable. It cannot be assumed that lower socioeconomic groups are always most vulnerable. Strategies to halt the under-5 mortality increase should be based on disaggregate information for individual countries.