Automatic infection detection based on electronic medical records.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support

Background:

• Accurate patient care decisions are challenging in emergency departments.
• Electronic Medical Records (EMRs) offer potential for automated diagnosis.
• Infection detection is a critical aspect of emergency medicine.

Purpose of the Study:

• To develop and evaluate machine learning models for automatic infection detection using EMRs.
• To identify key data features within EMRs that contribute to accurate infection prediction.

Main Methods:

• Utilized a dataset of EMRs from an emergency department.
• Applied five categories of patient information: personal details, admission notes, vital signs, diagnostic test results, and medical image diagnoses.
• Trained and evaluated machine learning models for infection detection.

Main Results:

• Machine learning models achieved an Area Under the Receiver Operator Characteristic Curve (AUC) of 0.88 for infection detection.
• Admission notes, in text format, provided the highest predictive value with an AUC of 0.87.
• The system demonstrated decent performance in automatic infection detection.

Conclusions:

• The study presents a novel EMR processing system for automated medical decision-making.
• Machine learning models effectively identify infection indicators within EMRs.
• This approach enhances the potential for early and accurate infection diagnosis in clinical settings.