Confronting models with data

Abstract

Until now, most of the mathematical modelling work on nosocomial infections has used simple models that have permitted qualitative, but not reliable quantitative predictions about the likely effect of different interventions. Increasingly, researchers would like to use models to provide reliable quantitative answers to both scientific and policy questions. This requires confronting models with data. Here, we discuss the importance of this confrontation with data with reference to previous modelling work, and outline the standard methods for doing this. We then describe a powerful new set of tools that promises to allow us to provide better answers to such questions, making far greater use than current methods of the information content of highly detailed hospital infection datasets. These tools should allow us to address questions that would have been impossible to answer using previous analytical techniques.