Modeling infant mortality in a hierarchical Bayesian framework: spatio-temporal convergence in Italy from 1990 to 2010

Fedele Greco, Università di Bologna
Francesco Scalone, Università di Bologna

The main purpose of this paper is to define a statistical method to model infant mortality rates for provincial areas in Italy by using a Bayesian approach. Results of this analysis are used to detect the convergence process of infant mortality at territorial level. As a matter of fact, when working on single small areas, direct estimates of infant mortality rate (henceforth IMR) can be affected by large variances. Since provincial sub-populations in Italy can widely vary, direct estimates of mortality rates can show evident uncertainty when considering smaller provinces. In a preliminary descriptive analysis, we clearly demonstrated the existence of specific spatiotemporal patterns which can be easily incorporated in a comprehensive Bayesian hierarchical model. Therefore, we propose a statistical model that allows area-level estimates of infant mortality to borrow strength from each other by exploiting spatial association of provincial IMRs and taking into account temporal correlation. This approach has become very popular in the disease mapping literature but, to our knowledge, it has not been employed for modeling IMRs in a demographic framework. Adopting a Bayesian approach, Markov Chain Monte Carlo methods are used to fit the model and to sample from the posterior predictive distribution. As a result, it appears that model based estimates are less variable than direct estimates. Indicators for assessing convergence and inequalities in infant mortality across provinces and time are also calculated. So measures of variability on direct and model-based estimates are taken into account. These preliminary results show the persistence of infant mortality inequalities, since an increasing trend of the coefficients of variation is observed.