Deep Learning-Based Risk Prediction of 30-Day Readmission in Patients with Malignant Peripheral Nerve Sheath Tumors (MPNSTs): A Nationwide Analysis

Introduction: Malignant Peripheral Nerve Sheath Tumors (MPNST) are aggressive soft tissue sarcomas with poor prognosis due to their invasive nature, high recurrence, and metastasis rates. Surgical resection, often combined with radiation and chemotherapy, is the primary treatment. Hospital 30-day readmission is a key clinical parameter reflecting post-operative conditions and potentially predicting long-term outcomes. This study aims to analyze characteristics of patients with and without readmission to identify risk factors and validate their predictive power for long-term outcomes.

Methods: Patients diagnosed with MPNST between 2004 and 2017 were queried from the National Cancer Database (NCDB) using ICD-O codes 9540-3, and were categorized into two groups based on 30-day readmission status. Univariate analysis was used to compare demographic, treatment, and clinical outcome data. We evaluated the predictive performance of multiple machine learning survival models through Receiver operating characteristic (ROC) with the Area Under the Curve (AUC) calculated.

Results: The mean age of the 30-day readmission group was significantly higher (39.04 ± 20.77 years) than the no-readmission group (34.12 ± 23.02 years, p=0.031), while no significant difference in racial or residential distributions were found. The readmission group had a higher proportion of patients with CDCC score ≥2 (8.9% vs. 2.8%, p=0.003). More patients in the readmission group received radiation (17.8% vs. 10.2%, p=0.051) and had earlier radiation initiation (60.4 ± 27.0 days vs. 93.4 ± 133.5 days, p=0.009). Mortality at 30 and 90 days was also significantly higher in the readmission group (p < 0.001). The Random Forest Classifier model was found to be the best performing model (AUC=0.980). Multivariate logistic regression identified a CDCC score of 2 or greater as a significant predictor of 30-day readmission, as well as other factors such as age, and the total dose and length of radiation.

Conclusion : Older age and radiation treatment are associated with higher readmission rates following MPNST surgery, and 30-day readmission correlates with increased mortality risk. A CDCC of 2 or greater is a significant predictor of 30-day readmission, highlighting the need for targeted post-operative care for patients with higher comorbidity scores to reduce readmission and improve outcomes.