Atsushi Kawaguchi1, Dominic Cave1, Qi Liu2, Yutaka Yasui2. 1. Department of Pediatrics, Pediatric Critical Care Medicine, University of Alberta, Alberta, Canada. 2. School of Public Health, University of Alberta, Edmonton, Alberta, Canada.
Abstract
OBJECTIVE: To identify patients at high risk for failed early extubation in the operating room (OR) following the Fontan procedure and generate a predictive model to allow improved clinical decision making. DESIGN: The success of an early-extubation strategy (extubation in the OR) was reviewed in patients aged 0 to 17 years old, undergoing the Fontan procedure between 2008 and 2011. Patients who required reintubation following primary extubation in the OR were compared with those who did not. Logistic regression with a backward variable selection was used to develop a predictive model in two stages: first, using pre-/perioperative predictors and then using postoperative predictors among the first-stage positive. SETTING: Canadian quaternary-care university children's hospital PICU. The treatment policy was changed from the routine extubation in PICU to extubation in the OR in January 2008. RESULTS: A total of 75 patients met our inclusion criteria: 8 patients required reintubation. Patients' average body weight was 14.5 kg (standard deviation [SD] 3.7), average age was 3.5 (SD 1.9) years, and average preoperative transcutaneous arterial saturation was 80.9% (SD 6.8). The first-stage predictive model contained three predictors: concomitant procedure (odds ratio [OR] >999, 95% confidence interval [CI] 15.7-infinity, p < 0.001), total bypassing time (cutoff; ≥99 minutes) (OR >999, 95% CI 6.5-infinity, p < 0.001), and absence of fenestration for pre/operative variables (OR >999, 95% CI 9.5-infinity, p < 0.001). The second-stage model included chest-tube fluid loss (CTFL ≥9.9 mL/kg/first 6 h). Our sequential prediction model had net sensitivity of 87.5% and specificity of 77.6%. CONCLUSION: We produced a predictive model for failed early extubation in Fontan patients. The sensitivity and specificity values are in the range of clinical utility. The model should be validated with an independent sample with a larger sample size. Georg Thieme Verlag KG Stuttgart · New York.
OBJECTIVE: To identify patients at high risk for failed early extubation in the operating room (OR) following the Fontan procedure and generate a predictive model to allow improved clinical decision making. DESIGN: The success of an early-extubation strategy (extubation in the OR) was reviewed in patients aged 0 to 17 years old, undergoing the Fontan procedure between 2008 and 2011. Patients who required reintubation following primary extubation in the OR were compared with those who did not. Logistic regression with a backward variable selection was used to develop a predictive model in two stages: first, using pre-/perioperative predictors and then using postoperative predictors among the first-stage positive. SETTING: Canadian quaternary-care university children's hospital PICU. The treatment policy was changed from the routine extubation in PICU to extubation in the OR in January 2008. RESULTS: A total of 75 patients met our inclusion criteria: 8 patients required reintubation. Patients' average body weight was 14.5 kg (standard deviation [SD] 3.7), average age was 3.5 (SD 1.9) years, and average preoperative transcutaneous arterial saturation was 80.9% (SD 6.8). The first-stage predictive model contained three predictors: concomitant procedure (odds ratio [OR] >999, 95% confidence interval [CI] 15.7-infinity, p < 0.001), total bypassing time (cutoff; ≥99 minutes) (OR >999, 95% CI 6.5-infinity, p < 0.001), and absence of fenestration for pre/operative variables (OR >999, 95% CI 9.5-infinity, p < 0.001). The second-stage model included chest-tube fluid loss (CTFL ≥9.9 mL/kg/first 6 h). Our sequential prediction model had net sensitivity of 87.5% and specificity of 77.6%. CONCLUSION: We produced a predictive model for failed early extubation in Fontan patients. The sensitivity and specificity values are in the range of clinical utility. The model should be validated with an independent sample with a larger sample size. Georg Thieme Verlag KG Stuttgart · New York.