Generalizability of population-based studies on AIDS: a comparison of newly and continuously surveyed villages in rural southwest Uganda.

BACKGROUND: Population-based studies are thought to provide generalizable epidemiological data on the human immunodeficiency virus type 1 (HIV-1) epidemic. However, longitudinal studies are susceptible to bias from added attention caused by study activities. We compare HIV-1 prevalence in previously and newly surveyed villages in rural southwest Uganda. METHODS: The study population resided in 25 neighbouring villages, of which 15 have been surveyed for 10 years. Respondents (>/=13 years) provided socio-demographic and sexual behaviour data and a blood sample for HIV-1 serology in private after informed consent. We tested the independent effect of residency: (1) original versus new villages; (2) proximity to main road; and (3) proximity to trading centre on HIV-1 serostatus of respondents using multivariate logistic regression. RESULTS: There were 8,990 adults censused, 68.3% were from the original villages, 48.2% were males and 6111 (68.0%) were interviewed and had definite HIV-1 serostatus. The HIV-1 prevalence was 6.1% overall, 5.7% in the new, and 6.4% in the original villages (P = 0.25). Residency in the new or original villages did not independently predict HIV-1 serostatus of respondents (P = 0.46). Independent predictors of HIV-1 serostatus were education (primary or higher, odds ratio [OR] = 1.7 and 1.4, respectively), being separated or widowed OR = 4.2, reported previous use of a condom OR = 1.8, or reported genital ulceration OR = 3.3, and age group 25-34 and 35-44 years OR = 5.8 and OR = 4.8 (all P