Identifying a Bayesian Network for the problem Hospital and Families : The analysis of patient satisfaction with their stay in hospital.

The Bayesian network originally developed as a knowledge representation formalism with a human expert. Construction of a Bayesian network with a domain expert is difficult and a time consuming task. As a result experts have problems making the relations between variables explicit, which hinders the construction of network structures and even if the human experts are good enough at judging qualitative relations, they are not usually so good at quantizing these relations. During the late 1980s people realised that the statistical foundation of Bayesian networks make it possible to learn from data rather than from an expert. The present work is part of a wider project of work and study intended to research and describe variables, indicators, and their relationships associated with the experience of families which, together or through one of their members, live through exceptional experiences, such as for example hospitalisation, or much more commonly and on a more daily level, access to the most elementary social and health services. The database used in the work project is the one obtained in the sample surveys which ISTAT (Central Institute of Statistics) organised in Italy on health conditions and the use made of social and health services by Italian families. The results obtained with this initial application find substantial and interesting confirmation with what was obtained in other ways and with other statistical methods of analysis.