Journal of critical care · 2025
Machine learning survival analysis for predicting kidney disease progression in patients with acute kidney injury undergoing continuous kidney replacement therapy: An analysis of the LINKA database.
СтатьяПоддержка исследования: вне правительства СШАНазвание 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.
Yun Donghwan, Hong Ari, Kim Kwangsoo, Lee Jeonghwan, Kim Yaerim, Jin Kyubok, Kim Ji Eun, Ahn Shin Young, Ko Gang-Jee, Park Seokwoo, Kim Sejoong, Jung Hee-Yeon, Cho Jang-Hee, Park Sun-Hee, Koh Eun Sil, Chung Sungjin, Lee Jung Pyo, An Jung Nam, Kim Sung Gyun, Kim Dong Ki, Han Seung Seok
Аннотация
PURPOSE: The progression of acute kidney injury (AKI) to end-stage kidney disease (ESKD) poses challenges due to high risks of comorbidities and poor outcomes. This study aimed to develop and validate machine learning survival models for predicting ESKD in patients receiving continuous kidney replacement therapy (CKRT) for AKI.
METHODS: A total of 1444 AKI patients who survived beyond one week after CKRT were included. Data from six hospitals were used to develop the model, and data from two hospitals formed the validation cohort. A comprehensive set of 122 clinical and laboratory variables was utilized to construct prediction models, including CoxBoost, Elastic-Net Cox, random survival forest, and Cox proportional hazards models. Model performance was assessed using the concordance index (C-index).
RESULTS: The CoxBoost model demonstrated superior performance, with a C-index of 0.811 (95 % confidence interval, 0.756-0.865) in the internal validation cohort and 0.742 (0.700-0.788) in the external validation cohort. This model reduced the variable set to 23 key parameters, with 24-h urine output on day 7 of CKRT, preexisting chronic kidney disease, and day 7 kidney function and systemic laboratory measures identified as the most critical predictors. A simplified scoring system derived from six binarized variables effectively stratified patients into low-, intermediate-, and high-risk groups for ESKD progression.
CONCLUSION: This machine learning survival approach highlights a set of critical, readily measurable predictors of ESKD risk and may support targeted risk stratification, bedside decision-making, and more efficient allocation of post-CKRT surveillance and kidney-care resources.
METHODS: A total of 1444 AKI patients who survived beyond one week after CKRT were included. Data from six hospitals were used to develop the model, and data from two hospitals formed the validation cohort. A comprehensive set of 122 clinical and laboratory variables was utilized to construct prediction models, including CoxBoost, Elastic-Net Cox, random survival forest, and Cox proportional hazards models. Model performance was assessed using the concordance index (C-index).
RESULTS: The CoxBoost model demonstrated superior performance, with a C-index of 0.811 (95 % confidence interval, 0.756-0.865) in the internal validation cohort and 0.742 (0.700-0.788) in the external validation cohort. This model reduced the variable set to 23 key parameters, with 24-h urine output on day 7 of CKRT, preexisting chronic kidney disease, and day 7 kidney function and systemic laboratory measures identified as the most critical predictors. A simplified scoring system derived from six binarized variables effectively stratified patients into low-, intermediate-, and high-risk groups for ESKD progression.
CONCLUSION: This machine learning survival approach highlights a set of critical, readily measurable predictors of ESKD risk and may support targeted risk stratification, bedside decision-making, and more efficient allocation of post-CKRT surveillance and kidney-care resources.