Measuring efficiency in Australian and New Zealand paediatric intensive care units

Purpose: To develop a measure of paediatric intensive care unit (PICU) efficiency and compare the efficiency of PICUs in Australia and New Zealand. Methods: Separate outcome prediction models for estimating clinical performance and resource usage were constructed using patient data from 20,742 admissions between 2005 and 2007. A standardised mortality ratio was calculated using a recalibrated Paediatric Index of Mortality 2 model. A random effects length of stay (LoS) prediction model was used to provide an indicator of unit-level variation in resource use. A modified Rapoport-Teres plot of risk-adjusted mortality versus unit mean LoS provided a visual representation of efficiency. To account for potential differences in admission threshold, the calculation of performance measures was repeated on patients receiving mechanical respiratory support and compared to those estimated for all patients. Results: The modified plot provides a useful tool for visualising ICU efficiency. Two units were identified as potentially inefficient with higher SMR and risk-adjusted mean LoS at the 95% level. One unit had a significantly lower SMR and significantly higher risk-adjusted mean LoS. The measures for both SMR and risk-adjusted mean LoS showed good agreement between all patients and those who received mechanical respiratory support. Conclusion: There is significant variation in efficiency among PICUs in Australia and New Zealand. Two units were designated as inefficient and one unit was considered to be effective at the expense of high resource use. Application of these methods may help to identify ineffi-ciencies in units located in other countries or regions.