Prediction of In-Hospital Mortality in Non-ST-Segment Elevation Myocardial Infarction, Based on Interpretable Machine Learning DOI Creative Commons

Mengru Cao,

Chunhui Li

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(8), P. 4226 - 4226

Published: April 11, 2025

This study sought to establish machine learning models for forecasting in-hospital mortality in non-ST-segment elevation myocardial infarction (NSTEMI) patients, and focused on model interpretability using Shapley additive explanations (SHAP). Data were gathered from the Medical Information Mart Intensive Care—IV database. The synthetic minority over-sampling technique Edited Nearest Neighbors used address class imbalance. Four algorithms employed, including Adaptive Boosting (AdaBoost), Random Forest (RF), Gradient Decision Trees (GBDT), eXtreme (XGBoost). SHAP was utilized improve transparency credibility. all-features RF demonstrated optimal performance, with an accuracy of 0.8513, precision 0.9016, AUC 0.8903. summary plot revealed that Acute Physiology Score III, lactate dehydrogenase, three most crucial characteristics, higher values indicating a greater risk. demonstrates applicability learning, particularly RF, predicting NSTEMI use enhancing providing clinicians clearer insights into feature contributions.

Language: Английский

Clinicopathological Prognostic Model for Survival in Adult Patients With Secondary Hemophagocytic Lymphohistiocytosis DOI Open Access
Pongthep Vittayawacharin, Benjamin J. Lee,

Ghayda' E'leimat

et al.

European Journal Of Haematology, Journal Year: 2025, Volume and Issue: unknown

Published: March 15, 2025

ABSTRACT Background Data on bone marrow (BM) findings in secondary hemophagocytic lymphohistiocytosis (sHLH) and their association with overall survival (OS) are limited. Objectives This study aimed to develop a prognostic model incorporating BM clinico‐laboratory factors affecting OS. Methods We retrospectively evaluated 50 adults sHLH developed clinicopathological based survival‐associated factors. Results Most patients demonstrated normocellular (46.3%) mild activity (44.2%). Factors associated multivariable analyses (MVA) were age above 70 years (hazard ratio [HR] 3.89, p = 0.016), infection‐related (HR 4.62, 0.006), hemoglobin < 7 g/dL 5.21, 0.001), hypocellular 3.07, 0.04). A HLH risk assigned 1 point each MVA‐identified factor, categorizing into low‐ (score 0–1), intermediate‐ 2–3), high‐risk 4) groups. The 6‐month OS from bootstrapping internal validation among the low‐, intermediate‐, groups 84.2%, 55.6% ( 0.001) 7.7% respectively. area under receiver operating characteristic curve (AuROC) was 0.87. Conclusions stratified three distinct outcomes, potentially guiding future therapy.

Language: Английский

Citations

0
Prediction of In-Hospital Mortality in Non-ST-Segment Elevation Myocardial Infarction, Based on Interpretable Machine Learning DOI Creative Commons

Mengru Cao,

Chunhui Li

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(8), P. 4226 - 4226

Published: April 11, 2025

This study sought to establish machine learning models for forecasting in-hospital mortality in non-ST-segment elevation myocardial infarction (NSTEMI) patients, and focused on model interpretability using Shapley additive explanations (SHAP). Data were gathered from the Medical Information Mart Intensive Care—IV database. The synthetic minority over-sampling technique Edited Nearest Neighbors used address class imbalance. Four algorithms employed, including Adaptive Boosting (AdaBoost), Random Forest (RF), Gradient Decision Trees (GBDT), eXtreme (XGBoost). SHAP was utilized improve transparency credibility. all-features RF demonstrated optimal performance, with an accuracy of 0.8513, precision 0.9016, AUC 0.8903. summary plot revealed that Acute Physiology Score III, lactate dehydrogenase, three most crucial characteristics, higher values indicating a greater risk. demonstrates applicability learning, particularly RF, predicting NSTEMI use enhancing providing clinicians clearer insights into feature contributions.

Language: Английский

Citations

0