Japanese encephalitis (JE) is a severe form of viral encephalitis characterized by acute inflammation of the central nervous system, primarily caused by the Japanese Encephalitis Virus (JEV) (1). Although the development of JE occurs in only a small percentage of JEV-infected individuals (0.1–1%), it still poses a significant public health burden, with approximately 68,000 to 96,000 reported cases and 15,000 deaths worldwide each year (2, 3). Among severe JE cases with neurological dysfunction, the fatality rate can reach 20–52%, and neurological sequelae have been reported in 44–60% of patients (2). Furthermore, nearly 50% of JE patients continue to experience neurological sequelae even 1 year after hospital discharge (4). Conducting clinical studies with large sample sizes is challenging due to the regional and seasonal incidence patterns of JE.

To date, limited information is available regarding the risk factors associated with poor prognosis in severe adult JE cases. The Japanese Encephalitis Virus can directly induce neuronal damage, and the resulting inflammation can exacerbate this effect (5). Previous studies have indicated a correlation between the innate immune response and fatal outcomes in flavivirus infections (6). However, the major risk factors contributing to poor prognosis in JE infections remain unclear. Our previous research focused on analyzing proteome profiling changes in the cerebrospinal fluid (CSF) of severe adult JE patients, identifying a subgroup with lower survival rates and a higher risk of cognitive impairment (7). Identifying key factors associated with poor prognosis through clinical indicators is an ongoing research topic in JE.

In the past, the lack of extensive samples and high-quality clinical studies has hindered a clear understanding of the specific clinical indicators associated with adult JE. The onset and progression of adult JE are characterized by rapidity, while the prognosis remains unfavorable. How to utilize commonly employed clinical indicators for early prediction of unfavorable prognosis, thereby enabling prompt implementation of targeted therapeutic interventions such as early tracheotomy and mechanical ventilation, holds the potential to significantly mitigate both mortality and disability rates.

Machine learning has emerged as a valuable tool in investigating clinical risk factors. Branch and bound algorithms, specifically designed for logistic regression analysis using R script packages, are commonly employed for analyzing large-scale biomedical data (8). The present study aims to employ machine learning techniques and statistical analysis tools to identify risk factors for poor prognosis in severe adult JE patients using extensive clinical data. By leveraging these advanced analytical approaches, we seek to gain valuable insights into the prognosis of severe JE and contribute to the understanding and management of this debilitating disease. To the best of our knowledge, this study represents the pioneering application of machine learning in analyzing risk factors associated with unfavorable outcomes in adults affected by JE.

JE is a significant infectious disease that poses a serious threat to human health. Studies conducted in South Korea have reported a notable increase in the proportion of adults affected by JE in recent years (13). Additionally, a retrospective study spanning over 15 years in Gansu, China, found distinct clinical manifestations between children and adults (14). Children with JE commonly present with symptoms such as vomiting, irritability, hypersomnia, convulsions, and spasms. On the other hand, adults experience symptoms such as changes in blood pressure, pupil size, positive meningeal stimulation signs, and positive pathological reflexes (hypertonia and Babinski sign). These findings suggest that adults tend to exhibit more severe clinical manifestations and higher mortality rates. The differences in manifestations may be attributed to the absence of JE vaccination during childhood or the higher average age of onset (13). Therefore, it is crucial to identify risk factors for poor prognosis in adult patients with JE and develop personalized treatment plans that can provide more effective supportive care based on different subpopulations.

In our study, all patients presented with fever, and some also experienced symptoms such as headache, nausea, vomiting, and other manifestations of high cranial pressure. Consciousness and mental symptoms were prevalent among the patients. The machine learning analysis identified WBC in CSF, total protein in CSF, N%, L%, RBC in blood, and Cl⁻ as the key factors associated with JEV infection. Furthermore, the correlation coefficient analysis, using the clinical prognosis (mRS) as a grouping variable, revealed that increased NLR, N%, and decreased L% were risk factors associated with the clinical outcome of JE. Abnormal glucose metabolism and liver function were also closely associated with poor prognosis in JE. Additionally, the proportion of lung infections was higher in the poor prognosis group.

Our findings align with previous research, which demonstrated an increase in WBC, N%, and CSF WBC in severe JEV infection (15). JEV induces an inflammatory response characterized by the accumulation of various immune cells around the spleen, lymph nodes, and in the blood. This leads to increased WBC and neutrophil counts in the peripheral blood (16). Viral infections often result in increased lymphocyte counts, with or without elevated WBC levels. The potential pathogenesis underlying these observations involves nervous stress response, tissue damage response, immune response disorder, and secondary or mixed bacterial infections (2, 15, 16). In our study, 65.8% of the patients had lung infections, which may explain the increased percentage of WBC and neutrophils. Specifically, the poor prognosis group exhibited lower GCS scores, more severe disturbance of consciousness, and a higher likelihood of ventilator-associated pneumonia following tracheal intubation. Although WBC levels were outside the normal range, no statistically significant difference was observed between the poor prognosis and good prognosis groups. There were no significant differences in electrolyte, metabolite, and blood gas analysis between the two groups. Apart from bacterial infection, the increase in N% may be attributed to excessive cellular immune defense.

The correlation coefficient analysis demonstrated opposite trends in N% and L% with regard to prognosis. Increased NLR, N%, or decreased L% indicated a potential poor prognosis for patients with JE. These immunological findings highlight the important role of the innate immune system in the recovery from JEV infection. Previous research has reported elevated levels of certain components that are associated with the prognosis of JEV infection (7). Neutrophils express IgG Fc receptors, complement C3b, and C5a receptors on their surface. The upregulation of complement levels in the CSF of patients with JE may impact chemotaxis, thereby promoting and enhancing neutrophil phagocytosis. However, chemotaxis and excessive activation of neutrophils can be detrimental to the recovery from JEV infection. The overall condition of patients also plays a significant role in the prognosis of the disease, with liver function and glucose metabolism contributing to the impairment (17).

Once the virus enters the central nervous system, uncontrolled viral proliferation may occur (2). Proinflammatory cytokines can trigger immune cell infiltration and clearance of infectious viral factors. However, an excess of proinflammatory cytokines can lead to tissue damage and systemic inflammation (18). Previous studies have demonstrated that JEV infection, especially after the virus enters the CNS through the blood–brain barrier, triggers unrestricted viral proliferation, leading to a series of inflammatory reactions and activation of the systemic immune response (19). Overactivation of inflammatory cells can result in severe cytokine storms and tissue damage (20). This may explain why increased N% is a risk factor for poor prognosis in JE. Therefore, future treatments for severe JE may focus on regulating the innate immune response induced by the JE virus within a specific range, maintaining antiviral function while avoiding excessive inflammatory reactions.

Currently, two articles on the subject of JE and machine learning were reported, including one of our previous research. In a study performed by Tehmina et al., deep proteomic networks and machine learning techniques were employed to investigate the hypothesis regarding the presence of JE diagnostic protein signatures. However, it is worth noting that this study compared the CSF proteomics result from JE samples with other CNS infection patients, but without incorporating an appropriate control group. It is important to consider that patients with various infections, including those caused by other flaviviruses, may exhibit similar CSF protein expression patterns.

The study identified significant risk factors associated with poor prognosis in adult patients with JE. These factors include WBC in CSF, total protein in CSF, N%, L%, RBC in blood, and Cl⁻. Correlation analysis further confirmed the association of increased NLR, N%, and decreased L% with adverse clinical outcomes. Additionally, abnormal glucose metabolism, liver function impairment, and a higher proportion of lung infections were observed in the poor prognosis group. The innate immune response appears to play a crucial role in the recovery from JEV infection, and maintaining a balanced immune response is crucial. Further research, including larger and prospective studies, is warranted to validate these findings and explore potential targeted therapies for severe JE.

In summary, our findings suggest that elevated serum NLR, N%, decreased L%, abnormal glucose metabolism, and liver function impairment may serve as potential prognostic indicators for severe adult patients with JE. However, further confirmation is required through prospective, multicenter randomized controlled trials with larger sample sizes. Additionally, conducting basic mechanisms research would provide deeper insights into these associations.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

The studies involving humans were approved by the Ethics Committee of Lanzhou General Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

PF, YH, YW, YH, and YL: conceptualization. YW, YL, MD, and HQ: data curation. YH, MD, and RL: formal analysis. YW, YH, YL, ZY, HQ, and RL: methodology. RY, PF, and YW: supervision. PF, YH, and MD: visualization. YW and MD: writing – original draft. RY and PF: writing – review and editing. All authors contributed to the article and approved the submitted version.