Background Populations in subSaharan Africa are ageing, despite the devastating effect of the Human immunodeficiency virus (HIV) epidemic in many countries. In South Africa the ageing population and the consequent increase in the prevalence of the chronic diseases of ageing is forming part of a complex and rapid health and social transition. To understand better the health and mortality experience of older people, we have carried out a 3 year follow-up of a cohort of people aged 50 years and older in rural South Africa.

Design The study took place in the Agincourt Health and Socio-demographic Surveillance Site (DSS) in the rural North East of South Africa, which has had an annual census almost every year since 1992. Basic demographic data (pregnancy outcome, deaths, migration) are collected every year to update the census information, and complementing socio-economic data is collected at regular intervals. In 2006 individuals aged 50 plus living in the Agincourt DSS, were invited to answer an adapted version of the World Health Organisation (WHO) - Study on Global Ageing and Adult Health (SAGE) questionnaire. The 4,047 respondents from 2006 study were then followed until 31st July 2009 or until outmigration or death. Cox regression analysis was used to determine factors related to survival.

Results By the end of July 2009 there were 374 deaths. When we calculated death rates by age group we found that death rates appeared to fall at later ages and then rise again (a twin peak effect), so that similar death rates were seen in men aged 55–59 and those aged 75–79 years old. There was a similar, although less striking, pattern for women. We explored this further using the full demographic surveillance dataset, and found a steep rise in death rates in people aged over 50 in more recent years. Further analysis suggested that the steep rise may be driven by increasing rates of death from HIV and tuberculosis, indicating that the HIV epidemic is also affecting older people who are often disregarded in commentaries on the subject. In addition we found that women had lower mortality than men (HR for death 0.35, 95% CI 0.28 to 0.44) and that that higher mortality was associated with individuals living without a partner, those from households in lower socio-economic status quintiles, those with higher levels of disability and those reporting poorer quality of life.