Artificial intelligence for healthcare-associated infection surveillance and control: a scoping review of evidence and implementation readiness

Background Healthcare-associated infections (HAIs) impose substantial clinical and resource burdens, while surveillance depends on fragmented clinical data and expert review. The value of automation depends on the surveillance step at which its output is used, what data it requires, and whether staff can review and act on it. We mapped how artificial intelligence (AI) and automation have been used for HAI surveillance and how ready these approaches are for implementation. Methods We conducted a scoping review following methodological guidance for scoping reviews and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR). Five databases were searched for English-language studies published from January 2019 to April 2026. Of 4,252 screened records, 102 full-text reports were assessed and 80 studies were included. Studies were classified using prespecified workflow and implementation-readiness domains. A second reviewer independently verified 16 key fields in all 80 studies; initial exact agreement was 85.2%. Results Surgical site infection (SSI) was the most frequent target (38/80, 47.5%). The main positions in the surveillance workflow were case-definition support (23/80), triage or prioritization (20/80), and candidate generation (13/80). Fifty-one studies were retrospective offline evaluations and 11 were prototypes or proofs of concept; 9 described a live pilot or routine use. Five studies reported external validation: three at a single external site and two across multiple external sites. Workload during operational use was quantified in 2 studies, governance was operationally specified in 2, and transferability was evaluated across sites or networks in 6. Conclusions Current evidence indicates that AI and automation can support several stages of HAI surveillance, particularly candidate generation, review prioritization, evidence extraction, and case-definition support within supervised workflows. Prospective evaluations should establish their value in routine workflows and across settings, while assessing workload and defining responsibility for decisions and system oversight.