Transfusion · 2025
Machine learning-based prediction of bleeding risk in extracorporeal membrane oxygenation patients using transfusion as a surrogate marker.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Kamio Tadashi, Ikegami Masaru, Mizuno Megumi, Ishii Seiichiro, Tajima Hayato, Machida Yoshihito, Fukaguchi Kiyomitsu
Аннотация
BACKGROUND: The increasing use of extracorporeal membrane oxygenation (ECMO) has highlighted challenges in managing bleeding complications. Optimal transfusion strategies remain uncertain for this diverse patient group, necessitating accurate predictive tools. This study developed and validated a machine learning (ML) algorithm to predict bleeding complications in patients with ECMO, using red blood cell (RBC) transfusion as a surrogate marker.
METHODS: Data from the Tokushukai Medical Database (2018-2022), covering 71 hospitals, were used. An ML approach was employed to predict bleeding complications, using RBC transfusion events as a surrogate marker. Model performance was evaluated using precision, recall, F1 score, and accuracy. SHapley Additive exPlanations (SHAP) analysis was conducted to identify key factors influencing model predictions.
RESULTS: Out of 470 ECMO-treated intensive care unit patients, 357 were included for model development. Forty-seven variables were used, with the light gradient boosting machine (LightGBM) and random forest models performing better than the other models, with receiver operating characteristic (ROC) area under the curve (AUC) above 0.7 for both (accuracy: 70.5%, ROC AUC: 0.703, recall: 0.784, and ROC AUC: 0.705, respectively). Models such as extreme gradient boosting performed similarly, while support vector classification had the lowest performance. SHAP analysis identified circulating blood volume, hemoglobin, and weight as the most important predictive factors.
DISCUSSION: The LightGBM and Random Forest models effectively predict bleeding complications in patients with ECMO, using RBC transfusion as a surrogate marker. This tool can support early identification of high-risk patients and improve overall transfusion management.
METHODS: Data from the Tokushukai Medical Database (2018-2022), covering 71 hospitals, were used. An ML approach was employed to predict bleeding complications, using RBC transfusion events as a surrogate marker. Model performance was evaluated using precision, recall, F1 score, and accuracy. SHapley Additive exPlanations (SHAP) analysis was conducted to identify key factors influencing model predictions.
RESULTS: Out of 470 ECMO-treated intensive care unit patients, 357 were included for model development. Forty-seven variables were used, with the light gradient boosting machine (LightGBM) and random forest models performing better than the other models, with receiver operating characteristic (ROC) area under the curve (AUC) above 0.7 for both (accuracy: 70.5%, ROC AUC: 0.703, recall: 0.784, and ROC AUC: 0.705, respectively). Models such as extreme gradient boosting performed similarly, while support vector classification had the lowest performance. SHAP analysis identified circulating blood volume, hemoglobin, and weight as the most important predictive factors.
DISCUSSION: The LightGBM and Random Forest models effectively predict bleeding complications in patients with ECMO, using RBC transfusion as a surrogate marker. This tool can support early identification of high-risk patients and improve overall transfusion management.