A Validated Nomogram That Predicts Prognosis of Autoimmune Encephalitis: A Multicenter Study in China

Abstract

The aim of this retrospective study was to derive and validate a reliable nomogram for predicting prognosis of autoimmune encephalitis (AE). A multi-center retrospective study was conducted in four hospitals in China, using a random split-sample method to allocate 173 patients into either a training (n = 126) or validation (n = 47) dataset. Demographic, radiographic and therapeutic presentation, combined with clinical features were collected. A modified Rankin Scale (mRS) at discharge was the principal outcome variable. A backward-stepwise approach based on the Akaike information criterion was used to test predictors and construct the final, parsimonious model. Multivariable analysis was conducted using logistic regression to develop a prognosis model and validate a nomogram using an independent dataset. The performance of the model was assessed using receiver operating characteristic curves and a Hosmer-Lemeshow test. The final nomogram model considered age, viral prodrome, consciousness impairment, memory dysfunction and autonomic dysfunction as predictors. Model validations displayed a good level of discrimination in the validation set: area under the Receiver operator characteristic curve = 0.72 (95% Confidence Interval: 0.56–0.88), Hosmer–Lemeshow analysis suggesting good calibration (chi-square: 10.33; p = 0.41). The proposed nomogram demonstrated considerable potential for clinical utility in prediction of prognosis in autoimmune encephalitis.