OBJECTIVES: To determine the prevalence and incidence of anaemia in HIV-positive and negative individuals; to identify risk factors for anaemia, prior to the introduction of HAART; and to determine the validity of the clinical diagnosis of anaemia. METHODS: Between 1990 and 2003, we followed a rural population based cohort of HIV-infected and uninfected participants. Prevalence and incidence of anaemia were determined clinically and by laboratory measurements. The sensitivity, specificity and predictive values of clinical diagnosis were calculated. RESULTS: The prevalence of anaemia at enrolment was 18.9% among HIV-positive and 12.9% among HIV-negative participants (P = 0.065). Incidence of anaemia increased with HIV disease progression, from 103 per 1000 person-years of observation among those with CD4 counts >500 to 289 per 1000 person-years of observation among those with CD4 counts <200. Compared to laboratory diagnosis, the clinical diagnosis of anaemia had a sensitivity of 17.8%, specificity of 96.8%, a positive predictive value of 50.6% and a negative predictive value of 86.4%. Being female, low CD4 cell counts, HIV-positive, wasting syndrome, WHO stage 3 or 4, malaria, fever, pneumonia and oral candidiasis were associated with prevalent anaemia. CONCLUSIONS: Anaemia prevalence and incidence were higher among HIV-positive than negative participants. Compared to laboratory diagnosis, clinical detection of anaemia had a low sensitivity. Clinicians working in settings with limited laboratory support must be conscious of the risk of anaemia when managing HIV/AIDS patients, particularly when using antiretroviral drugs which by themselves may cause anaemia as a side effect. We recommend that haemoglobin should be measured before starting ART and monthly for the first three months.