Construction and Validation of a Machine Learning-Based Prediction Model for Short-Term Mortality in Critically Ill Patients with Liver Cirrhosis - 02/12/24
Highlights |
• | We employed feature importance scoring techniques from ten different machine learning models to rank the variables based on their contribution to the predictive performance of the model. This process enabled us to identify and select key variables for modeling. |
• | Subsequently, we constructed a Stacking ensemble model utilizing multiple machine learning algorithms, which demonstrated superior performance compared to individual models. |
• | To validate our findings, we utilized several external validation datasets, and the model exhibited commendable performance and generalizability across these datasets. |
• | Furthermore, we visualized the model to facilitate clinical decision-making and employed the SHAP (SHapley Additive exPlanations) method to assess the importance of the variables. This approach not only enhances the interpretability of the model but also provides valuable insights into the key factors influencing predictions, thereby supporting clinicians in making informed decisions at the bedside. |
Abstract |
Objective |
Critically ill patients with liver cirrhosis generally have a poor prognosis due to complications such as multiple organ failure. This study aims to develop a machine learning-based prediction model to forecast short-term mortality in critically ill cirrhotic patients in the intensive care unit (ICU), thereby assisting clinical decision-making for intervention and treatment.
Methods |
Machine learning models were developed using clinical data from critically ill cirrhotic patients in the MIMIC database, with multicenter validation performed using data from the eICU database and Qinghai University Affiliated Hospital(QUAH). Various machine learning models, including a Stacking ensemble model, were employed, with the SHAP method used to enhance model interpretability.
Results |
The Stacking ensemble model demonstrated superior predictive performance through internal and external validation, with AUC and AP values surpassing those of individual algorithms. The AUC values were 0.845 in the internal validation set, 0.819 in the eICU external validation, and 0.761 in the QUAH validation set. Additionally, the SHAP method highlighted key prognostic variables such as INR, bilirubin, and urine output. The model was ultimately deployed as a web-based calculator for bedside decision-making.
Conclusion |
The machine learning model effectively predicts short-term mortality risk in critically ill cirrhotic patients in the ICU, showing strong predictive performance and generalizability. The model's robust interpretability and its deployment as a web-based calculator suggest its potential as a valuable tool for assessing the prognosis of cirrhotic patients.
El texto completo de este artículo está disponible en PDF.Keywords : Critically ill patients with liver cirrhosis, machine learning, prediction model, short-term mortality, intensive care unit (ICU)
Esquema
Bienvenido a EM-consulte, la referencia de los profesionales de la salud.
El acceso al texto completo de este artículo requiere una suscripción.
¿Ya suscrito a @@106933@@ revista ?