Economic causes and effects of AIDS in South African households

To investigate the magnitude and temporal directionality of associations between illness and death, and income and expenditure, in households affected by HIV/AIDS.

A cohort study with repeated measures carried out in 405 households (1913 occupants), known to have HIV-infected occupants and their neighbours, in one rural and one urban area of South Africa.

Monthly adult equivalent income and expenditure. Illness episodes and deaths attributed to HIV/AIDS, tuberculosis and pneumonia.

Interview surveys of household heads were conducted at baseline and five more times, biannually, providing information on household economics, illnesses and deaths. Regression analyses used marginal structural models and 'before-after' models to analyse changes.

In marginal structural models, current or previous AIDS illness was independently associated with 34% [95% confidence intervals (CI) 23-43%] lower monthly expenditure, and current or recent poverty was associated with 1.74 (95% CI 0.94-3.2) times higher odds of an AIDS death. In before-after models, each AIDS death was independently associated with a 23% (95% CI 11-34%) greater expenditure decline over 3 years, and a 100 US dollars higher monthly expenditure at baseline was associated with 0.31 (95% CI 0.13-0.74) times as many AIDS deaths and with 0.41 (95% CI 0.27-0.64) times as many AIDS illness episodes over 3 years.

AIDS deaths and illnesses predicted declining expenditure and poverty predicted AIDS, suggesting that both welfare and effective treatment are needed.