JACC. Advances · 2025
Cardiac Output Estimation in the Intensive Care Unit.
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.
Palanques-Tost Eric, Pallarès-López Roger, Padrós-Valls Raimon, Song Steven, Reinertsen Erik, Churchill Timothy W, Stockwell Paige, Pomerantsev Eugene, Garasic Joseph, Sundt Thoralf M, Shah Pinak, Houstis Nicholas E, Aguirre Aaron D
Аннотация
BACKGROUND: Cardiac output (CO) is a quintessential property of the cardiovascular system, one whose estimation is vital to patient care in critical illness. The most common techniques for assessing CO, thermodilution (TD) and the estimated Fick (eFick) approximation, force tradeoffs that motivate a need for new methods.
OBJECTIVES: The purpose of this study was to novel CO estimators to fill key gaps in critical care medicine.
METHODS: Machine learning was used to estimate CO from physiology measurements made during routine clinical care in the intensive care unit (ICU) or cardiac catheterization lab. Models were trained and validated using a curated set of 13,172 ground-truth measurements of TD-CO from 4,825 patients. Model performance was evaluated using regression metrics, trajectory analysis, classification accuracy, and ΔCO tracking.
RESULTS: Three established eFick models all performed poorly in the ICU because their static estimates of oxygen consumption could not track the dynamics of critical illness. In the postcardiac surgery intensive care unit, the best eFick model erred in its CO predictions by 30% (mean absolute percentage error [MAPE]) with a coefficient of determination (R2) of -1.5. The best model derived here, labeled CORE (Catheter Optimized caRdiac output Estimation), predicted CO with an MAPE of 14% (P < 0.001 vs eFick) and an R2 of 0.58. These estimates could be calculated from measurements obtained with either a pulmonary artery catheter or a central venous catheter. The CORE model was also robust to the presence of moderate or severe tricuspid regurgitation, achieving an MAPE of 16% and R2 of 0.65 relative to a ground-truth determined by the direct Fick technique with measured oxygen consumption.
CONCLUSIONS: CO models that account for dynamic physiology in ICU patients were more accurate than widely used eFick models and more versatile than TD. The performance of these models combined with their adaptation to vascular access, broad applicability, ease of use, and ease of deployment should enable them to benefit patients across diverse ICU settings.
OBJECTIVES: The purpose of this study was to novel CO estimators to fill key gaps in critical care medicine.
METHODS: Machine learning was used to estimate CO from physiology measurements made during routine clinical care in the intensive care unit (ICU) or cardiac catheterization lab. Models were trained and validated using a curated set of 13,172 ground-truth measurements of TD-CO from 4,825 patients. Model performance was evaluated using regression metrics, trajectory analysis, classification accuracy, and ΔCO tracking.
RESULTS: Three established eFick models all performed poorly in the ICU because their static estimates of oxygen consumption could not track the dynamics of critical illness. In the postcardiac surgery intensive care unit, the best eFick model erred in its CO predictions by 30% (mean absolute percentage error [MAPE]) with a coefficient of determination (R2) of -1.5. The best model derived here, labeled CORE (Catheter Optimized caRdiac output Estimation), predicted CO with an MAPE of 14% (P < 0.001 vs eFick) and an R2 of 0.58. These estimates could be calculated from measurements obtained with either a pulmonary artery catheter or a central venous catheter. The CORE model was also robust to the presence of moderate or severe tricuspid regurgitation, achieving an MAPE of 16% and R2 of 0.65 relative to a ground-truth determined by the direct Fick technique with measured oxygen consumption.
CONCLUSIONS: CO models that account for dynamic physiology in ICU patients were more accurate than widely used eFick models and more versatile than TD. The performance of these models combined with their adaptation to vascular access, broad applicability, ease of use, and ease of deployment should enable them to benefit patients across diverse ICU settings.