Development and Validation of a Risk Score to Differentiate Viral and Autoimmune Encephalitis in Adults

Abstract

Abstract Background Encephalitis represents a challenging condition to diagnose and treat. To assist physicians in considering autoimmune encephalitis (AE) sooner, we developed and validated a risk score. Methods The study was conducted as a retrospective cohort of patients with a diagnosis of definite viral encephalitis (VE) and AE from​​ February 2005 to December 2019. Clinically relevant and statistically significant features between cases of AE and VE were explored in a bivariate logistic regression model and results were used to identify variables for inclusion in the risk score. A multivariable logistic model was used to generate risk score values and predict risk for AE. Results were externally validated. Results A total of 1310 patients were screened. Of the 279 enrolled, 36 patients met criteria for definite AE and 88 criteria for definite VE. Patients with AE compared with VE were more likely to have a subacute to chronic presentation (odds ratio [OR] = 22.36; 95% confidence interval [CI], 2.05–243.7), Charlson comorbidity index <2 (OR = 6.62; 95% CI, 1.05–41.4), psychiatric and/or memory complaints (OR = 203.0; 95% CI, 7.57–5445), and absence of robust inflammation in the cerebrospinal fluid defined as <50 white blood cells/µL and protein <50 mg/dL (OR = 0.06; 95% CI, .005–0.50). Using these 4 variables, patients were classified into 3 risk categories for AE: low (0–1), intermediate (2–3), and high (4). Results were externally validated and the performance of the score achieved an area under the curve of 0.918 (95% CI, .871–.966). Discussion This risk score allows clinicians to estimate the probability of AE in patients presenting with encephalitis and may assist with earlier diagnosis and treatment.