Outcome prediction in critical care: the ICNARC model

To describe the background to, rationale for, and structure and performance of the Intensive Care National Audit & Research Centre risk prediction model.

The Intensive Care National Audit & Research Centre model was published in 2007 as a new risk prediction model, predicting risk of death before ultimate discharge from acute hospital for all admissions to adult, general critical care units in the UK. It was developed using a high-quality clinical database of over 200,000 admissions to 163 critical care units and prospectively validated in over 30,000 admissions to 20 different units. The Intensive Care National Audit & Research Centre model was designed to address limitations of preexisting models, particularly exclusion of certain patient groups, and varying effects of physiological derangement in different underlying conditions.

The Intensive Care National Audit & Research Centre model performs well in comparison with preexisting models when evaluated in independent validation data from UK critical care units. The use of interactions between the physiology score and diagnostic category produces better fit within individual diagnostic groups. The elimination of model exclusion criteria, for example age less than 16 years, means that the model compares the observed and expected outcomes for all patients admitted to a critical care unit providing a fairer method for comparative audit.