Predicting mortality in critically ill patients with hypertension using machine learning and deep learning models.

Area of Science:

• Critical Care Medicine
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Accurate mortality prediction in Intensive Care Unit (ICU) patients with hypertension is crucial for clinical decision-making.
• Traditional prognostic tools often lack the sophistication to capture complex clinical variable interactions in this population.
• Machine learning (ML) and deep learning (DL) offer advanced capabilities for developing more accurate predictive models.

Purpose of the Study:

• To evaluate the performance of various ML and DL models in predicting mortality among critically ill hypertensive patients.
• To identify key clinical predictors of mortality in this patient group.
• To compare the effectiveness of different ML and DL models for mortality prediction.

Main Methods:

• Retrospective analysis of 30,096 critically ill hypertensive ICU patients.
• Comparison of traditional ML models (logistic regression, decision trees, SVM) with DL models (1D CNNs, LSTMs).
• Evaluation using Area Under the Receiver Operating Characteristic Curve (AUC) and SHapley Additive exPlanations (SHAP) for predictor identification.

Main Results:

• The 1D Convolutional Neural Network (CNN) model achieved the highest AUC (0.7744), outperforming other ML and DL models.
• Key predictors of mortality identified across models included the Acute Physiology and Chronic Health Evaluation III (APS-III) score, patient age, and length of ICU stay.
• SHAP analysis confirmed the significant impact of these predictors on mortality risk assessment.

Conclusions:

• Deep learning models, particularly 1D CNNs, show superior predictive accuracy for mortality in critically ill hypertensive patients compared to traditional ML models.
• Integrating DL models into clinical workflows can improve early identification of high-risk patients for targeted interventions.
• Further research is needed for prospective validation and ethical considerations of DL model implementation in clinical practice.