Automated healthcare-associated infection surveillance using an artificial intelligence algorithm.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Infection Control

Background:

• Healthcare-associated infections (HAIs) are a significant cause of adverse events in hospital settings.
• Effective surveillance is crucial for preventing and controlling HAIs.

Purpose of the Study:

• To evaluate the efficacy of artificial intelligence (AI) algorithms for detecting healthcare-associated infections (HAIs).
• To assess the performance of a multilayer perceptron neural network for HAI surveillance.

Main Methods:

• A cohort study was conducted using AI algorithms for infection surveillance.
• A multilayer perceptron neural network model was developed and evaluated.
• Performance metrics included area under the receiver operating curve (AUROC), specificity, and sensitivity.

Main Results:

• The AI model correctly identified 67 out of 73 patients with HAIs.
• The final model achieved an AUROC of 90.27%, specificity of 78.86%, and sensitivity of 88.57%.
• Respiratory infections demonstrated superior detection rates with AUROC values of 93.47% or higher.

Conclusions:

• AI algorithms are a feasible and effective method for HAI surveillance.
• AI has the potential to enhance the accuracy and efficiency of hospital-wide infection surveillance.
• Implementing AI can improve overall infection prevention performance and reduce adverse events.