Abstract
This paper introduces a discrete conditional survival model (DC-S) with a classification component for predicting patient outcome and survival component for predicting length of stay in hospital. The DC-S model consists of two components; the conditional component which utilises a classification tree and the survival component which models the survival distribution. The survival component of the model is conditioned on the discrete conditional component, the classification tree. The DC-S model with classification tree is applied to a healthcare scenario where the length of stay of babies in neonatal wards in Northern Ireland (United Kingdom) is modelled using the baby characteristics known on the first day of admission. The resulting model can accurately predict length of stay of babies and thus has the potential to be used in bed planning. Hospitals could use such good estimates for the length of stay of patients (determined on the day of arrival) to plan ahead to make the correct provisions available during their stay. Not only does this have resource implications, it can also help patient families. The resulting model can also predict the occurrence (or otherwise) of late onset sepsis, which has implications on a patients stay.