SURG-41. MACHINE LEARNING PREDICTS THE INFLUENCE OF FRAILTY ON THE CLINICAL MANAGEMENT OF NEUROFIBROMATOSIS TYPE 1: A MIXED-EFFECTS MODELING STUDY USING THE NATIONWIDE READMISSIONS DATABASE

Abstract

Abstract Recent literature has highlighted the association of frailty with significantly higher rates of morbidity and mortality in patients with CNS neoplasms. However, there is a paucity of research regarding the effects of frailty as it relates to neurocutaneous disorders, namely NF1. In this study we evaluated the role of frailty in NF1 patients and compared its predictive utility to that of the Elixhauser Comorbidity Index (ECI). Namely, we used the Nationwide Readmissions Database from 2016 to 2017 to select for those patients with a diagnosis of NF1 who underwent neurosurgical resection of an intracranial tumor. Patient frailty was queried through the use of the Johns Hopkins Adjusted Clinical Groups (JHACG) frailty-defining indicator. ECI scores were collected for these patients for measurement of comorbidities. Propensity score matching was performed for age, sex, ECI, insurance type, median income by ZIP code, and yielded 60 frail and 60 non-frail patients. Receiver operating characteristic (ROC) curves were created for relevant complications, including mortality, non-routine discharge, cost in top quartile, length of stay [LOS] in top quartile, and readmission, using comorbidity indices as predictor values. The area under the curve (AUC) of each ROC served as a proxy for model performance. Following matching of the groups, frail patients still had an increased average hospital cost ($85,441.67±$59,201.09) compared to non-frail patients ($49,321.77±$50,705.80)(p=0.010). Similar trends were also found for LOS between frail (23.1±14.2 days) and non-frail patients (10.7±10.5 days)(p=0.0020). For each complication of interest, ROC curves plotted showed frailty performing as well as ECI and the combination of Frailty+ECI (p >0.05). Frailty+ECI (AUC: 0.924)outperformed the model using only ECI (AUC: 0.842, p=0.027). These findings suggest that frailty is as efficacious as ECI in predicting complications including mortality, non-routine discharge, readmission, and high cost in patients who underwent resection of intracranial lesions associated with NF1.