AI-Driven Models Enhance Prediction of In-Hospital Mortality for ICU Cancer Patients
Machine learning models, including the CatBoost classifier, show high accuracy in predicting in-hospital mortality among ICU patients with lymphoma, offering valuable insights for personalized risk assessment.
Recent advances in machine learning have demonstrated promising potential in predicting the risk of in-hospital death among ICU patients with lymphoma. A comprehensive study published in PLOS ONE details the development and validation of multiple machine learning models, utilizing data from the Medical Information Mart for Intensive Care IV database. The research involved 1,591 adult ICU patients diagnosed with lymphoma, of whom 342 (21.5%) experienced in-hospital mortality.
The study compared fifteen different machine learning algorithms through receiver operating characteristic and area under the curve (AUC) analyses. Among these, the CatBoost classifier exhibited the highest predictive accuracy with an AUC of 0.7766. Significant predictors identified by the models included blood urea nitrogen (BUN), platelet count, prothrombin time, heart rate, systolic blood pressure, activated partial thromboplastin time, oxygen saturation, and bicarbonate levels.
Particularly noteworthy was the prominent role of BUN, emphasized through SHapley Additive exPlanation (SHAP) analysis, which ranked it as the most influential factor in predicting mortality. The models also provided personalized risk assessments using SHAP force plots, enabling identification of high-risk patient subgroups.
According to the study authors, these machine learning models, especially the CatBoost classifier, could serve as valuable tools for clinicians in estimating mortality risk, supplementing traditional assessment methods. Their interpretability via SHAP analysis offers insights into patient-specific risks, potentially informing more targeted interventions and improving ICU care for lymphoma patients.
This innovative approach highlights the significant role of AI in personalized medicine, ultimately aiming to enhance patient outcomes and clinical decision-making in critical care settings.
Source: https://medicalxpress.com/news/2025-08-ai-hospital-mortality-icu-patients.html
Stay Updated with Mia's Feed
Get the latest health & wellness insights delivered straight to your inbox.
Related Articles
Addressing the Gap: Accessibility Features for ALS Patients and Their Caregivers in Everyday Technology
A recent study reveals a significant gap in awareness and use of accessibility features in digital devices among ALS patients and caregivers, highlighting opportunities for enhanced support and education to improve quality of life.
New Insights into Cellular Processes Behind Drug-Resistant Cancers
Researchers have uncovered how cancer cells bypass DNA damage checkpoints to continue dividing, offering new avenues for resistant cancer therapies. Learn about groundbreaking findings on centrosomes and PLK1 enzyme involvement.
Urgent Call for Global Action to Prevent Future Pandemics Through One Health Approach
Urging global action, health experts emphasize the importance of the One Health approach to prevent future pandemics through coordinated efforts across human, animal, and environmental health sectors.
Aging Reduces Overactive Immune Genes, Leading to Improvement in Lupus Symptoms in Older Adults
Aging appears to diminish the activity of immune-related genes in lupus patients, potentially leading to symptom improvement in older adults. This discovery opens new possibilities for age-specific treatments and understanding disease progression.