Alzheimer's & dementia : the journal of the Alzheimer's Association · 2025
Detection of emergency department patients at risk of dementia through artificial intelligence.
СтатьяМногоцентровое исследованиеНаблюдательное исследованиеНазвание и аннотация: Перевод в очереди
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.
Cohen Inessa, Taylor Richard Andrew, Xue Haipeng, Faustino Isaac V, Festa Natalia, Brandt Cynthia, Gao Emily, Han Ling, Khasnavis Siddarth, Lai James M, Mecca Adam P, Sapre Atharva Vinay, Young Juan, Zanchelli Michael, Hwang Ula
Аннотация
INTRODUCTION: The study aimed to develop and validate the Emergency Department Dementia Algorithm (EDDA) to detect dementia among older adults (65+) and support clinical decision-making in the emergency department (ED).
METHODS: In a multisite retrospective study of 759,665 ED visits, electronic health record data from Yale New Haven Health (2014-2022) were used to train three supervised and semi-unsupervised positive-unlabeled machine learning models (XGBoost, Random Forest, LASSO). A separate test set of 400 ED encounters underwent adjudicated chart review for validation.
RESULTS: EDDA achieved an area under the receiver-operating characteristic curve (AUROC) of 0.85 in the test set and 0.93 in the validation set. Positive-unlabeled learning improved performance. Agreement between EDDA and clinician-adjudicated dementia diagnoses was moderate (kappa = 0.50), with 17% of EDDA-positive patients having undiagnosed probable dementia.
DISCUSSION: EDDA enhances dementia detection in the ED, with potential for real-time implementation to improve patient outcomes and care transitions.
HIGHLIGHTS: Developed a machine learning algorithm using electronic health record data to detect dementia in the emergency department (ED). Algorithm designed to balance detection accuracy with ease of ED implementation. Parsimonious model with limited but predictive variables selected for rapid ED use. Focused on real-time application, optimizing ED workflows, and clinician support. Aims to enhance ED dementia detection, patient safety, and care coordination.
METHODS: In a multisite retrospective study of 759,665 ED visits, electronic health record data from Yale New Haven Health (2014-2022) were used to train three supervised and semi-unsupervised positive-unlabeled machine learning models (XGBoost, Random Forest, LASSO). A separate test set of 400 ED encounters underwent adjudicated chart review for validation.
RESULTS: EDDA achieved an area under the receiver-operating characteristic curve (AUROC) of 0.85 in the test set and 0.93 in the validation set. Positive-unlabeled learning improved performance. Agreement between EDDA and clinician-adjudicated dementia diagnoses was moderate (kappa = 0.50), with 17% of EDDA-positive patients having undiagnosed probable dementia.
DISCUSSION: EDDA enhances dementia detection in the ED, with potential for real-time implementation to improve patient outcomes and care transitions.
HIGHLIGHTS: Developed a machine learning algorithm using electronic health record data to detect dementia in the emergency department (ED). Algorithm designed to balance detection accuracy with ease of ED implementation. Parsimonious model with limited but predictive variables selected for rapid ED use. Focused on real-time application, optimizing ED workflows, and clinician support. Aims to enhance ED dementia detection, patient safety, and care coordination.