A multilevel model for infant health outcomes: maternal risk factors and geographic variation

Abstract

This paper considers the interaction between place of residence and maternal characteristics for infant health outcomes. It illustrates how indices of performance specified in terms of the varying incidence of adverse events or the effect of maternal risk on such outcomes may be sensitive to sampling unreliability, and how this may be clarified in a Bayesian analysis. An adjustment for sampling errors, together with a correction for the interplay of various risk factors and the effect of area deprivation, may considerably modify impressions of relative performance gained from comparing simple perinatal mortality rates.