Mark S Wainwright1, Remigiusz Lewandowski. 1. Department of Pediatrics, Northwestern University Feinberg School of Medicine, Chicago, IL, USA. m-wainwright@northwestern.edu
Abstract
INTRODUCTION: Multivariate data analyses have the potential to enrich the use of the complex plethora of data gathered in the care of critically ill patients. We sought to apply hierarchical cluster analysis to investigate factors affecting outcome in children with acute brain injury requiring ICP monitoring. METHODS: We performed a retrospective analysis on patients admitted to the pediatric intensive care unit with ICP monitoring for management of acute brain injury between 2003 and 2008. We performed cluster analysis comparing survivors and non-survivors. We compared the area the under the curve (time and duration above a determined value) between groups. RESULTS: We identified 32 patients among whom 8 died during hospitalization. Cluster analysis identified an association among CPP, PaCO(2), and end-tidal CO(2), and among ICP, central venous pressure, and lactate, in addition to other unexpected associations. Values (mean ± SD) for ICP in survivors (13.7 ± 7.9 mmHg) were lower than (42.7 ± 36.2) in non-survivors. Mean (59.4 vs 41.3), and minimum (47.3 vs 28.1) CPP were higher in survivors. The AUC for ICP was greater in the subjects who died and cluster analyses supported the importance of ICP and CPP related to outcome. CONCLUSIONS: The dose of abnormal ICP may contribute to outcome in insults that involve increased ICP. These results are proof of principle of the potential application of hierarchical clustering to the clinical practice of pediatric neurocritical care.
INTRODUCTION: Multivariate data analyses have the potential to enrich the use of the complex plethora of data gathered in the care of critically illpatients. We sought to apply hierarchical cluster analysis to investigate factors affecting outcome in children with acute brain injury requiring ICP monitoring. METHODS: We performed a retrospective analysis on patients admitted to the pediatric intensive care unit with ICP monitoring for management of acute brain injury between 2003 and 2008. We performed cluster analysis comparing survivors and non-survivors. We compared the area the under the curve (time and duration above a determined value) between groups. RESULTS: We identified 32 patients among whom 8 died during hospitalization. Cluster analysis identified an association among CPP, PaCO(2), and end-tidal CO(2), and among ICP, central venous pressure, and lactate, in addition to other unexpected associations. Values (mean ± SD) for ICP in survivors (13.7 ± 7.9 mmHg) were lower than (42.7 ± 36.2) in non-survivors. Mean (59.4 vs 41.3), and minimum (47.3 vs 28.1) CPP were higher in survivors. The AUC for ICP was greater in the subjects who died and cluster analyses supported the importance of ICP and CPP related to outcome. CONCLUSIONS: The dose of abnormal ICP may contribute to outcome in insults that involve increased ICP. These results are proof of principle of the potential application of hierarchical clustering to the clinical practice of pediatric neurocritical care.
Authors: Kendra S Woods; Christopher M Horvat; Sajel Kantawala; Dennis W Simon; Jaskaran Rakkar; Patrick M Kochanek; Robert S B Clark; Alicia K Au Journal: Pediatr Crit Care Med Date: 2021-02-01 Impact factor: 3.971