Risk Prediction Models for Invasive Mechanical Ventilation in Patients with Autoimmune Encephalitis: A Retrospective Cohort Study

Abstract

Background and Objectives. Timely identification of developing severe respiratory failure in patients with autoimmune encephalitis (AE) is crucial to ensure prompt treatment with invasive mechanical ventilation (IMV), which can potentially improve the outcome. We aimed to develop a nomogram for requiring IMV based on easily available clinical characteristics. Methods. A multivariate predictive nomogram model was developed using the risk factors identified by LASSO regression and assessed by receiver operator characteristics (ROC) curve, calibration curve, and decision curve analysis. Results. The risk factors predictive of severe respiratory failure were male gender, impaired hepatic function, elevated intracranial pressure, and higher neuron-specific enolase. The final nomogram achieved an AUC of 0.770. After validation by bootstrapping, a concordance index of 0.748 was achieved. Conclusions. Our nomogram accurately predicted the risk of developing respiratory failure needing IMV in AE patients and provide clinicians with a simple and effective tool to guide treatment interventions in the AE patients.