Stochastic models of geriatric patient bed occupancy behaviour

This paper considers a census approach to the modelling of the time that geriatric patients spend in hospital and subsequently in the community by using a stochastic compartmental Markov model. A selection process is developed using maximum likelihood estimation to fit the model to daily census data on the duration spent in the hospital or the community. Census data are used as they are easy to collect and therefore maximize the usability of the model. The model is fitted to an extensive 16-year data-set and shown to provide realistic estimates of movements of patients by using only a single day's census result.