Artificial intelligence enhances genomic surveillance in healthcare outbreak investigations

Abstract

BACKGROUND: Outbreak investigation and control are critical for preventing the spread of infectious diseases in healthcare settings. Traditional methods rely on manual processes, which are time-consuming and limited in scope. Whole genome sequencing (WGS) surveillance improves outbreak detection but still requires extensive manual chart reviews to identify transmission routes. Integrating artificial intelligence (AI) may enhance the efficiency and accuracy of these investigations. METHODS: We evaluated an AI tool developed to streamline healthcare outbreak investigations detected by the Enhanced Detection System for Healthcare-associated Transmission (EDS-HAT). For outbreaks detected between November 2021 and November 2023, multiple data elements were extracted from electronic health records (EHR) for all patients. The AI algorithm was applied to identify transmission routes, and its performance was assessed against expert manual reviews. Key measures included additional transmission routes identified and sensitivity. RESULTS: Data from 172 outbreaks involving 476 case patients were analyzed. The AI tool identified 37 transmission routes that were missed by manual review, including procedures and provider routes. The algorithm achieved a sensitivity of 76.0% (95% confidence interval [CI] 71.1%-81.1%) compared to manual review, increasing to 91.7% (95% CI 87.7%-94.7%) after accounting for transmission at other facilities and downstream exposures. CONCLUSION: The EDS-HAT AI tool significantly improved outbreak investigations by automating the identification of transmission routes, both with concordant findings of manual review as well as finding additional routes of transmission missed by traditional chart review. AI with genomic surveillance has the potential to optimize outbreak detection and investigation to streamline interventions in healthcare settings.

Ähnliche Arbeiten