A machine learning-based prediction of hospital mortality in mechanically ventilated ICU patients
Hexin Li1, Negin Ashrafi1, Chris Kang1
1Department of Industrial and Systems Engineering, University of Southern California (USC), Los Angeles, CA, United States of America.
Background:
Mechanical ventilation (MV) is vital for critically ill ICU patients but carries significant mortality risks. This study aims to develop a predictive model to estimate hospital mortality among MV patients, utilizing comprehensive health data to assist ICU physicians with early-stage alerts.
Methods:
We developed a Machine Learning (ML) framework to predict hospital mortality in ICU patients receiving MV. Using the MIMIC-III database, we identified 25,202 eligible patients through ICD-9 codes. We employed backward elimination and the Lasso method, selecting 32 features based on clinical insights and literature. Data preprocessing included eliminating columns with over 90% missing data and using mean imputation for the remaining missing values. To address class imbalance, we used the Synthetic Minority Over-sampling Technique (SMOTE). We evaluated several ML models, including CatBoost, XGBoost, Decision Tree, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Logistic Regression, using a 70/30 train-test split. The CatBoost model was chosen for its superior performance in terms of accuracy, precision, recall, F1-score, AUROC metrics, and calibration plots.
Results:
The study involved a cohort of 25,202 patients on MV. The CatBoost model attained an AUROC of 0.862, an increase from an initial AUROC of 0.821, which was the best reported in the literature. It also demonstrated an accuracy of 0.789, an F1-score of 0.747, and better calibration, outperforming other models. These improvements are due to systematic feature selection and the robust gradient boosting architecture of CatBoost.
Conclusion:
The preprocessing methodology significantly reduced the number of relevant features, simplifying computational processes, and identified critical features previously overlooked. Integrating these features and tuning the parameters, our model demonstrated strong generalization to unseen data. This highlights the potential of ML as a crucial tool in ICUs, enhancing resource allocation and providing more personalized interventions for MV patients.
Insights
This study developed a machine learning model to predict hospital mortality in patients on mechanical ventilation (MV). The CatBoost model achieved high accuracy, offering early alerts for ICU physicians.
Area of Science:
- Critical Care Medicine
- Health Informatics
- Machine Learning in Healthcare
Background:
- Mechanical ventilation (MV) is essential for critically ill ICU patients but is associated with significant mortality risks.
- Developing predictive models for hospital mortality in MV patients is crucial for timely interventions.
- Comprehensive health data can be leveraged to create early warning systems for ICU physicians.
Purpose of the Study:
- To develop and validate a machine learning (ML) framework for predicting hospital mortality in intensive care unit (ICU) patients receiving mechanical ventilation (MV).
- To identify key clinical features predictive of mortality in MV patients.
- To enhance early-stage alerts for ICU physicians to improve patient outcomes.
Main Methods:
- Utilized the MIMIC-III database to identify 25,202 eligible patients on MV.
- Employed backward elimination and Lasso for feature selection, identifying 32 critical features.
- Applied data preprocessing techniques including handling missing data and Synthetic Minority Over-sampling Technique (SMOTE) for class imbalance.
- Evaluated multiple ML models, with CatBoost demonstrating superior performance.
Main Results:
- The CatBoost model achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.862, outperforming existing literature.
- Demonstrated high predictive accuracy with an accuracy of 0.789 and an F1-score of 0.747.
- Showcased superior calibration and performance compared to other evaluated ML models.
Conclusions:
- The developed ML model, utilizing systematic feature selection and the CatBoost algorithm, offers a robust tool for predicting hospital mortality in MV patients.
- The preprocessing methodology effectively identified critical features, simplifying analysis and improving model generalization.
- Highlights the potential of ML in ICUs for optimizing resource allocation and enabling personalized patient interventions.
More Related Videos
04:05Author Spotlight: Unveiling Prognostic Indicators in Heart Failure - The Role of Phase Angle and Bioelectrical Impedance Analysis
Published on: June 30, 2023
09:44Manual Muscle Testing: A Method of Measuring Extremity Muscle Strength Applied to Critically Ill Patients
Published on: April 12, 2011
Related Concept Videos
01:29Mechanical Ventilation I: Indication and Settings
01:23Mechanical Ventilation II: Invasive Ventilation
Negative-Pressure Ventilators
Negative-pressure ventilators create a vacuum around the chest or body to draw air into the lungs, simulating breathing. This method does not require an...
01:23Mechanical Ventilation III: Noninvasive Ventilation
Noninvasive Positive-Pressure Ventilation...
01:29Acute Respiratory Failure-V
Ensure that patients are monitored continuously for their response to therapy, including changes in...
01:20Assessment of Ventilation I: Respiratory Rate
A Ventilation assessment is critical for monitoring a patient's health status. Respiration, one of the most accessible vital signs, provides insights into the function of numerous body systems and can indicate serious health issues, such as brainstem injuries from head trauma.
Critical Guidelines for Assessing Ventilation: