Predicting Patient Resource Utilization and Identifying Drivers of Healthcare Costs among Brain Tumor Patients. (16th November 2020)

• Record Type: Journal Article
• Authors: Jimenez, Adrian; Khalafallah, Adham M; Huq, Sakibul; Patel, Palak; Mukherjee, Debraj
• Is Part Of: Neurosurgery. Volume 67(2010)Supplement 1
• Publication Date: 2020-11-16
• Subjects: Nervous system -- Surgery -- Periodicals; 617.48005
• DOI: 10.1093/neuros/nyaa447_204
• Languages: English
• ISSNs: 0148-396X

Abstract

Abstract: INTRODUCTION: Identifying specific drivers of healthcare cost is necessary for efficiently allocating resources, reducing unnecessary spending, and providing high-value surgical care. METHODS: Patients who underwent brain tumor surgery between 2017–2019 at a single academic institution were included in our study. Patients at risk for high healthcare resource utilization were defined as those who had total hospital charges above the 75th percentile (upper quartile) of all patients in our dataset (i.e. > $46, 939.70). Five machine learning algorithms were trained on 70% of available data and evaluated on the remaining 30% to predict both patients at risk for high healthcare resource utilization and to estimate total hospital charges. RESULTS: A total of 720 brain tumor patients were included in our final analysis. Our patient cohort was majority male (52.8%) and Caucasian (75.4%). The neural network algorithm was chosen as the optimal classification algorithm (c-statistic = 0.86, Brier score = 0.12) and the elastic net linear regression algorithm was chosen as the optimal regression algorithm (R 2 = 0.44). The three most important variables for the classification model were ICU length of stay, surgery duration, and Karnofsky Performance score (KPS), while the three most important variables for the regression model were surgery duration, KPS, and surgical wound infection. Both models were incorporated into an online calculator (https://ajimene8.shinyapps.io/cost_calculator/ ). CONCLUSION: Our online calculator may serve as a practical tool for providing high-value healthcare to brain tumor patients.