Objective: To model the effect of socio-economic deprivation and a few transmission-related indicators of the tuberculosis (TB) incidence at small area level, to discuss the potential of each indicator in targeting places for developing preventive action.
Methods: Ecological spatial study of TB incidence in Olinda, a city in the north-east of Brazil, during the period 1996-2000. Three socio-economic indicators (mean number of inhabitants per household; percentage of heads of household with <1 year's formal education; percentage of heads of households with monthly income lower than the minimum wage) and two transmission-related indicators (number of cases of retreatment; number of households with more than one case during the period under study), all calculated per census tract, were used. We adopted four different full hierarchical Bayesian models to estimate the relative risk of the occurrence of TB via Markov chain Monte Carlo.
Results: The best specified model includes all the selected covariates and the spatially structured random effect. The gain in goodness-of-fit statistic when the spatial structure was included confirms the clustered spatial pattern of disease and poverty. In this model, the covariates within the non-zero credibility interval were the number of persons per house, the number of cases of retreatment and the number of households with more than one case (all with relative risk > or = 1.8) in each census tract.
Conclusions: The possibility to estimate in the same framework both the contribution of covariates at ecological level and the spatial pattern should be encouraged in epidemiology, and may help with establishing Epidemiological Surveillance Systems on a territorial basis, that allows rational planning of interventions and improvement of the Control Programme effectiveness.