2021 Trauma/Critical Care Presentations

Rathnayaka M K D Gunasingha, MD1; Hannah Hensman, BS2; Ravi Sanka, MS3; Seth Schobel, PhD3; Patrick Benoit, MD1; Rondi B Gelbard, MD4; Eric Elster, MD1; Matthew J Bradley, MD1; 1Walter Reed National Military Medical Center; 2DecisionQ Corporation; 3Uniformed Services University of the Health Sciences; 4The University of Alabama at Birmingham

Background: An open abdomen (OA) is a complex wound that requires a multidisciplinary team and a large amount of resources. Indications for an open abdomen range from a damage control laparotomy for a trauma patient to a second look after laparotomy for bowel ischemia. Not only does it put the patient at risk for serious complications directly related to OA, but the physiological stress of critical illness increases the risk of other complications, such as ventilator-associated pneumonia (VAP) and venous thromboembolism (VTE). There have been studies that have shown a relationship between pneumonia and thrombosis. Our aims to characterize respiratory and thrombotic complications in an OA population and to describe any relationship between the two using clinical and biomarker data in predictive modeling of VTE.

Methods: Adults patients who underwent an exploratory laparotomy after abdominal trauma at a Level 1 trauma center were identified from a prospectively collected database with clinical and biomarker data points from 2014 to 2019. Clinical and biomarker data were collected and analyzed at initial assessment. A Random Forest (RF) model was developed and variables were selected with backwards elimination. The model was validated with leave-one-out. Continuous variables were evaluated with Kruskal-Wallis.

Results: Three hundred and ninety-five patients with an open abdomen were identified, of which 43 (10.9%) patients developed at least one infectious respiratory complication (VAP or hospital-acquired pneumonia) and 29 (7.3%) patients developed VTE during their hospital course. Of those who had pneumonia, 8 (18.6%) also had VTE. Patients with higher injury severity score (ISS) and more transfusions were more likely to develop VAP (p<0.001, p<0.01) and VTE (p<0.001, p<0.05). While multiple cytokines were elevated, VAP, and VTE both had significant elevations in IL-1RA and IL-6. Using complete clinical data and 11 available biomarkers, a RF model predicting VTE was developed with 118 patients. The model selected four features—total blood products received in first 24 hours, presence of leg fracture, IL-6, VEGF—with an AUC 0.86, sensitivity 0.84, and specificity 0.83.

Conclusion:  Both infectious respiratory complications and VTE have elevations in IL-1RA and IL-6, which suggest a similar inflammatory process. Clinical and biomarker data in a RF model can significantly aid in the diagnosis of VTE in this high risk trauma population. This model can be generalized and externally validated in a larger trauma population.