Measuring Quality in the Paediatric Intensive Care Unit

Lahn Straney (2010). Measuring Quality in the Paediatric Intensive Care Unit PhD Thesis, School of Population Health, The University of Queensland.

       
Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
s4129568_PhD_abstract.pdf Abstract application/pdf 65.51KB 0
s4129568_PhD_totalthesis.pdf Final Thesis Lodgement application/pdf 1.56MB 17
Author Lahn Straney
Thesis Title Measuring Quality in the Paediatric Intensive Care Unit
School, Centre or Institute School of Population Health
Institution The University of Queensland
Publication date 2010-07
Thesis type PhD Thesis
Supervisor Archie Clements
Anthony Slater
Total pages 175
Total black and white pages 175
Subjects 11 Medical and Health Sciences
Abstract/Summary ABSTRACT INTRODUCTION Variation in patient outcomes among health care providers may be indicative of differences in the quality of care. In order to make objective comparisons of patient outcomes between providers and over time it is necessary to adjust for differences in patient case-mix. Statistical tools known as risk-adjustment models are employed for this purpose. Disparities in adjusted patient outcomes can help to drive and direct quality improvement. Paediatric intensive care units (PICUs) play a crucial role in caring for the sickest of children. It is of interest to ensure that these units can effectively identify potential areas of deficiency relative to peers to facilitate continuous quality improvement. However, limited tools are available for contrasting outcomes in the PICU. Further, those models that have been developed have become outdated as the relationships between patient factors and outcomes change over time. Mortality, length of stay (LoS) and duration of respiratory support are important indicators of clinical performance and unit efficiency. Mortality is a common, easily understood and useful indicator of clinical performance. LoS provides insight into the efficiency of resource use and may also act as a surrogate for quality of care. Likewise duration of respiratory support may provide insights into the efficiency of resource use and also clinical practice regarding ventilator use . The aim of this research was to examine the relationships between patient characteristics at the time of admission with clinical outcomes and develop statistical methods that could adjust for differences in case-mix that would permit objective comparison of outcomes over time and among units. METHODS Statistical models for objectively contrasting patient outcomes in the PICU were developed. The models used patient characteristics at the time of admission as covariates in the models. Two methods were developed for contrasting variation in LoS among PICUs. The first model was a mixed-effects gamma regression model built using 47,068 admissions between 1997 and 2006 in Australian and New Zealand (ANZ) ICUs that accept paediatric admissions. The second model was a two compartment mixed-effects gamma regression model that was developed using 12,763 ANZ PICU admissions to quantify variation in mean LoS among short and long stay patients separately. A log-normal regression model for contrasting patient duration of respiratory support was constructed using a subset of the LoS data. A revised Paediatric Index of Mortality (PIM3) was developed using a logistic regression model and 45,706 admissions to Australian, New Zealand and UK PICUs between 2007 and 2008. RESULTS AND INTERPRETATION Site-level variation in patient outcomes was found. Among the 20 units which accepted paediatric ICU admissions, the first LoS model revealed 6 units that had an adjusted mean LoS that was significantly longer than expected and 5 that were significantly shorter. Cessation of respiratory support is likely to represent ‘readiness to discharge’. LoS over and above this point may represent unit inefficiency; however it may be confounded by bed block or unavailability of appropriate step-down facilities. Concurrent analysis of respiratory support duration and LoS contextualised the variation in LoS and indicated that bed block is not a significant confounder of variation in LoS in this study population. The two-compartment model for estimating site-level variation in short and long stay patients revealed differences in the site effect among short and long stay patients. Of the 8 PICUs studied, 4 sites had a statistically significant effect on long stay patients and 5 sites had a statistically significant effect on short stay patients. PIM3, the updated model for assessing mortality risk, provides more accurate assessments of mortality risk and unit performance, particularly among low risk patients. CONCLUSIONS Good quality assessment must encompass a broad range of indicators to paint a holistic picture of unit performance. The methods described enable units to assess clinical performance within the context of resource use. Clinical outcomes including LoS and duration and respiratory support may be adjusted for differences in case-mix among units allowing objective comparisons of performance to be made. In addition the results show that variation in patient outcomes exist among PICUs in Australia and New Zealand. These finding demonstrates that units may not be performing equally and provides direction for reviews of the practice of care.
Keyword Pediatrics
Quality of Health Care
Risk Adjustment
Regression Analysis
Intensive care
Length of Stay
Mortality

 
Citation counts: Google Scholar Search Google Scholar
Created: Fri, 19 Nov 2010, 04:06:08 EST by Mr Lahn Straney on behalf of Library - Information Access Service