S. Faisal1, A. G. Cuenca1, S. D. Larson1, D. W. Kays1, S. Islam1 1University Of Florida,Gainesville, FL, USA
Introduction: The development of prognostic metrics are especially important in the identification of disease states that may rapidly worsen, such as necrotizing enterocolitis (NEC). While many such predictors have been reported and are thought to be associated with NEC, none have been validated. The purpose of this study was to attempt to create a model that could help predict NEC based on clinical, physiologic, and lab parameters.
Methods: We retrospectively collected clinical data on 108 patients with NEC as well as 38 age-matched controls from 2000 to 2009. We performed multiple logistic regression and developed receiver operator curves based on the clinical data collected to determine if any metrics that have been reported as well additional parameters including the presence of cardiac or hepatic dysfunction could be important for the development of NEC and further generalized to the age-matched controls.
Results: Using Univariate analysis, we found significant differences (p < 0.05) in the birth weight, bandemia, sodium concentration, percent lymphocytes, hemoglobin, method of delivery, and mean arterial pressure (MAP), however we did not note differences in pH, absolute neutrophil count, platelet count, presence of cardiac dysfunction (See Table). Logistic regression was then performed on the significant variables. Surprisingly, only bandemia, MAP, and hemoglobin concentration at the time of clinical suspicion of NEC was found to be significant in our population, with OR of 1.29, 1.13 and 0.7, respectively (see Table). It is possible that missing data and selection bias may confound our model. Therefore, ROC were performed on the variables collected. Bandemia, MAP, and hemoglobin concentration were again found to have the greatest areas under the curve, respectively 0.86, 0.79, and 0.72.
Conclusion: While bandemia is already considered an important clinical variable, these data suggest that we may be able to improve on already recognized clinical parameters by including decreased hemoglobin concentration and elevated MAP in the clinical algorithm currently used for the identification of NEC in at risk patients. Surprisingly, pH, cardiac dysfunction, or ANC were not found to be predictive of NEC in our patient population. This model is to be tested and validated prospectively in the future.
