Detecting hospital-acquired infections: A document classification approach using support vector machines and gradient...

Area of Science:

• Medical Informatics
• Computational Linguistics
• Machine Learning

Background:

• Hospital-acquired infections (HAIs) present a substantial risk to patient well-being.
• Current surveillance methods for HAIs are labor-intensive for hospital staff.

Purpose of the Study:

• To develop an automated surveillance system for detecting potential HAIs in patient records.
• To reduce the manual workload associated with HAI surveillance.
• To apply and evaluate machine learning techniques for HAI detection in Swedish patient records.

Main Methods:

• Text classification algorithms, specifically Support Vector Machines (SVM) and Gradient Tree Boosting (GTB), were employed.
• The study focused on applying these methods to Swedish patient records, an area where they had not been previously utilized.
• Preprocessing techniques and parameter tuning were utilized to optimize model performance.

Main Results:

• Both SVM and GTB demonstrated encouraging results in detecting HAIs.
• Gradient Tree Boosting achieved the highest performance metrics: 93.7% recall, 79.7% precision, and 85.7% F1 score, particularly when using stemming.
• The findings indicate that effective preprocessing and parameter tuning can yield high recall with suitable precision for this application.

Conclusions:

• Machine learning, specifically Gradient Tree Boosting, shows significant promise for automated HAI surveillance.
• The developed system can effectively screen patient records, aiming for high recall to minimize missed cases.
• This approach offers a viable solution to reduce the burden on hospital staff while enhancing patient safety through improved infection detection.