D. E. Wang1, Y. Fang2, S. E. Sherman2, E. Newman1, G. Ballantyne1, H. Pachter1, M. Melis1 1New York University School Of Medicine,Department Of Surgery,New York, NY, USA 2New York University School Of Medicine,Department Of Population Health,New York, NY, USA
Introduction: Currently, there are no standard criteria for admission to an intensive care unit (ICU) following colorectal surgery and rates of ICU admission fluctuate widely across various healthcare settings. Undertriage can lead to preventable complications. Over-triage can cause iatrogenic harm to the patient alongside unnecessary costs and misutilization of resources. We sought to identify predictors for appropriate ICU triage following colorectal surgery.
Methods: We performed a retrospective analysis of patients undergoing colorectal surgery (2011-2015). Objective criteria (e.g. prolonged hypotension, reintubation, new onset of cardiac arrhythmia) were used to identify appropriate ICU admissions. Patient demographics, underlying conditions, and common surgical risk scores such as Charlson Comorbidity Index (CCI), Surgical Apgar Score (SAS), and Revised Cardiac Risk Index (RCRI) were utilized in a decision tree model to identify factors correlating with objective criteria for ICU admission. The accuracy and overall quality of the model was assessed by examining the test misclassification rate achieved after separating the dataset into training and validation sets. The model was then compared to other approaches including bivariate analysis and multiple logistic regression analysis.
Results: Events requiring ICU care were observed in 41 of 104 patients who underwent colorectal surgery. Decision tree analysis identified the lowest intraoperative heart rate, CCI, and eGFR as the 3 most important predictors for ICU admission with an overall misclassification rate of 21.2% (Figure 1a). SAS, CCI, age, history of solid tumor, and preoperative hematocrit were also predictors of need for ICU. Using 65 patients who were randomly selected as training set to build a model and the remaining 39 patients to test the model, a simpler decision tree was achieved utilizing CCI and lowest intraoperative heart rate with an overall misclassification rate of 20.5% (Figure 1b). Bivariate analysis of the top 9 predictors identified through the decision tree model revealed significant differences (p<0.05) in lowest heart rate, CCI, SAS, RCRI, and history of a solid tumor but insignificant differences in eGFR, age, and preoperative hematocrit. Multiple logistic regression analysis of these 9 variables resulted in an overall misclassification rate of 23.1%.
Conclusion: Many patients are unnecessarily admitted to the ICU following colorectal surgery. Using decision tree modeling, we found that lowest intraoperative heart rate and CCI represent the two most important predictors for ICU admission. Further analysis with larger datasets will be necessary to develop clinical tools to enhance postoperative triage after colorectal surgery.
