Back
For best output, select "Paper Size" as "A4" and "Margin" as "0" or "None".
To save or print to PDF, please select Print Destination > Save as PDF, enable Background Graphics under "More Settings", then click "Save".
During the congress, E-Posters will be accessible to all participants on the congress website 24/7, as well as in the E-poster stations in the congress center.
Preparing your E-Poster
Please review the E-Poster format requirements carefully when preparing your E-Poster. Should your E-Poster not meet the mentioned requirements, it may not be displayed as described above.
E-Poster Submission Deadline
Please prepare and upload your E-Poster no later than March 14, 2026 11.59PM CET. After this date, you will no longer be able to prepare and upload your E-poster and it will not be displayed and accessible on the congress website.
Please follow the instructions below to input your abstract title.
Abstract titles should be brief and reflect the content of the abstract.
Hyponatremia is the most common electrolyte disorder in hospitalized patients and is an established independent predictor of in-hospital mortality. Although this high mortality rate is particularly associated with profound hyponatremia, a long-standing clinical question has been whether this poor prognosis is caused by hyponatremia itself or if it merely reflects severe underlying comorbidities. To address this question, our study aimed to develop an accurate machine learning model for predicting in-hospital mortality in patients with profound hyponatremia and to identify key predictors through model interpretation.
This retrospective cohort study included adult patients with profound hyponatremia (serum sodium ≤120 mEq/L) admitted to a tertiary medical center in Japan between 2014 and 2024. From 49 clinical features extracted from medical records, we selected 20 features using three methods: LASSO, recursive feature elimination (RFE), and SelectKBest. These features were used with five classifiers—logistic regression, support vector machine, random forest, XGBoost, and LightGBM— to build and evaluate a total of 15 machine learning models. The best-performing model, identified by the highest area under the receiver operating characteristic curve (AUC-ROC), was interpreted using SHAP (Shapley Additive Explanations) to quantify the contribution of each feature.
Five hundred ninety-seven eligible patients were included in this study, with a median age of 76 years and a median baseline serum sodium of 118 mEq/L, and 143 (24.0%) died during their hospital stay. The model combining RFE and a random forest classifier demonstrated the highest predictive performance on the test data (ROC-AUC = 0.841; 95% CI, 0.776–0.899). SHAP analysis revealed that the most influential predictors of in-hospital mortality were, in order: the Charlson Comorbidity Index (CCI), serum albumin, performance status, and the presence of metastatic malignancy. Specifically, a high CCI score (≥3), a low serum albumin level (<3.0 g/dL), and a poor performance status (≥3) were each associated with an increased risk of in-hospital mortality. Conversely, the serum sodium level at diagnosis ranked eighth in importance, and an extremely low level (<115 mEq/L) paradoxically has an association with survival.
Figure1. Predictive performance of machine learning models.
Figure 2. SHAP dependence plots for key predictive features.
We developed a machine learning model that accurately predicts in-hospital mortality in patients with profound hyponatremia. The key predictors reflect the patient's comorbidities, nutritional status, and general condition, suggesting that the patient's fundamental health state strongly determines the prognosis. The observation that extremely low serum sodium levels were linked to a better prognosis strongly supports the hypothesis that profound hyponatremia is associated with high mortality, not because it is inherently lethal, but because it serves as a marker reflecting the severity of the underlying disease.
