Evaluation of predicted and actual length of stay in 22 Scottish intensive care units using the APACHE III system

The most recent edition of the Acute Physiology and Chronic Health Evaluation provides a prediction of intensive care unit length of stay in addition to the probability of hospital mortality. Intensive care length of stay is an important determinant of intensive care costs and may be an important indicator of quality of care. Data were collected from 22 Scottish intensive care units over a 2-year period to allow comparison of actual intensive care unit length of stay with that predicted by the Acute Physiology and Chronic Health Evaluation III system. Correlation between actual and predicted stay for individual patients was poor. However, performance of the model for patients, grouped either by predicted length of stay or by intensive care unit, indicated that the model stratified patient groups appropriately while demonstrating a consistent bias. Length of stay in Scottish intensive care units was found to be consistently lower than that predicted by a model which is based on intensive care practice in the USA. Variations in severity of illness in intensive care unit populations cannot readily explain differences in intensive care unit length of stay. The availability of a model capable of predicting length of intensive care stay, based on data reflecting practice in the UK, would compliment current methods of assessing effectiveness of intensive care.