<![CDATA[Background: Stroke is the third leading cause of death globally, and early prediction of in-hospital mortality risk is essential for improving clinical outcomes and optimizing resource allocation. However, no in-hospital mortality system applicable to both ischemic and hemorrhagic stroke currently exists in Indonesia. This study aims to identify significant predictors of in-hospital mortality among stroke patients, to develop an in-hospital mortality prediction score system, and to assess the performance of the score. Methods: This retrospective cohort study included stroke patients hospitalized at Dr. Saiful Anwar Hospital, Malang, Indonesia in January-December 2024, with data obtained from electronic medical records. Multivariate logistic regression with backward stepwise elimination was performed to identify predictors significantly associated with in-hospital mortality. Scoring weights were derived from regression coefficients. Model performance was assessed using accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the area under the curve (AUC). Results: Eight significant predictors were identified: age ≥60 years, male sex, National Institute of Health Stroke Scale (NIHSS) ≥16, leukocytosis, hyperglycemia, hemorrhagic stroke type, pneumonia, and sepsis. These predictors were incorporated into an in-hospital mortality scoring system. The scoring system demonstrated an accuracy of 72.92%, sensitivity 80%, specificity 71.77%, PPV 31.37%, NPV 95.7%, and an AUC of 0.784. Conclusions: Eight out of thirteen proposed predictors were significantly associated with in-hospital mortality among stroke patients. Based on these predictors, a prognostic scoring system for in-hospital mortality was developed, demonstrating fair discriminative performance.]]>
