Machine learning reveals neutrophil-to-lymphocyte ratio as a crucial prognostic indicator in severe Japanese encephalitis patients

Abstract

Japanese encephalitis (JE) is a severe infectious disease affecting the central nervous system (CNS). However, limited risk factors have been identified for predicting poor prognosis (PP) in adults with severe JE. In this study, we analyzed clinical data from thirty-eight severe adult JE patients and compared them to thirty-three patients without organic CNS disease. Machine learning techniques employing branch-and-bound algorithms were used to identify clinical risk factors. Based on clinical outcomes, patients were categorized into two groups: the PP group (mRs ≥ 3) and the good prognosis (GP) group (mRs ≤ 2) at three months post-discharge. We found that the neutrophil-to-lymphocyte ratio (NLR) and the percentage of neutrophilic count (N%) were significantly higher in the PP group compared to the GP group. Conversely, the percentage of lymphocyte count (L%) was significantly lower in the PP group. Additionally, elevated levels of aspartate aminotransferase (AST) and blood glucose were observed in the PP group compared to the GP group. The clinical parameters most strongly correlated with prognosis, as indicated by Pearson correlation coefficient (PCC), were NLR (PCC 0.45) and blood glucose (PCC 0.45). In summary, our findings indicate that increased serum NLR, N%, decreased L%, abnormal glucose metabolism, and liver function impairment are risk factors associated with poor prognosis in severe adult JE patients.

Keywords: Japanese encephalitis; cognitive impairment; machine learning; neutrophil-to-lymphocyte ratio; percentage of neutrophilic.