European journal of medical research · 2025
Online machine learning model for predicting delirium risk in elderly patients with chronic kidney disease: development and preliminary validation.
СтатьяВалидационное 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.
Gao Lang, Wang Guangdong, Yang Xingyi, Ge Yuanshuo, Tong Shijun, Huang Weijie
Аннотация
BACKGROUND: Delirium frequently complicates elderly chronic kidney disease (CKD) patients due to multifactorial vulnerability. Early detection in geriatric intensive care unit (ICU) settings is challenged by traditional assessments' communication deficits. Machine learning refines predictions through multidimensional pattern recognition. We develop an interpretable online tool for timely high-risk delirium identification, guiding interventions to enhance outcomes.
METHODS: Elderly chronic kidney disease patients were selected using International Classification of Diseases (ICD) codes, with delirium defined per Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Baseline traits, vitals, lab parameters, comorbidities, and clinical scores were collected. Dual feature selection employed Least Absolute Shrinkage and Selection Operator (Lasso) regression and Boruta. Eight machine learning models underwent assessment via Decision Curve Analysis (DCA) and calibrated curves. The superior model was interpreted using SHapley Additive exPlanations (SHAP) values and deployed as a web-based risk calculator.
RESULTS: Ten key predictors were identified through dual feature selection. The Gradient Boosting Machine (GBM) model demonstrated good calibration and sustained high predictive accuracy in both internal and external validation. SHAP analysis revealed Glasgow Coma Scale (GCS) score, Sequential Organ Failure Assessment (SOFA) score, and sedative usage as clinically significant predictors. The finalized model was deployed as a clinically applicable web-based risk calculator.
CONCLUSION: This study's prediction model accurately assesses delirium risk in elderly CKD patients, showing robust performance and clinical utility. Its web-based dynamic calculator supplies personalized risk evaluation in ICU settings, supporting early identification and proactive interventions to enhance clinical management.
METHODS: Elderly chronic kidney disease patients were selected using International Classification of Diseases (ICD) codes, with delirium defined per Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Baseline traits, vitals, lab parameters, comorbidities, and clinical scores were collected. Dual feature selection employed Least Absolute Shrinkage and Selection Operator (Lasso) regression and Boruta. Eight machine learning models underwent assessment via Decision Curve Analysis (DCA) and calibrated curves. The superior model was interpreted using SHapley Additive exPlanations (SHAP) values and deployed as a web-based risk calculator.
RESULTS: Ten key predictors were identified through dual feature selection. The Gradient Boosting Machine (GBM) model demonstrated good calibration and sustained high predictive accuracy in both internal and external validation. SHAP analysis revealed Glasgow Coma Scale (GCS) score, Sequential Organ Failure Assessment (SOFA) score, and sedative usage as clinically significant predictors. The finalized model was deployed as a clinically applicable web-based risk calculator.
CONCLUSION: This study's prediction model accurately assesses delirium risk in elderly CKD patients, showing robust performance and clinical utility. Its web-based dynamic calculator supplies personalized risk evaluation in ICU settings, supporting early identification and proactive interventions to enhance clinical management.