Key points are not available for this paper at this time.
Immunosuppressed patients carry a disproportionate infectious risk driven by underlying disease, biologic or immunomodulatory therapies, glucocorticoid exposure, comorbidities, latent infections, antimicrobial resistance, and fragmented follow-up. Artificial intelligence may support infection risk management by integrating clinical, microbiological, pharmacological, and operational data; however, it should not be implemented as an autonomous solution or a substitute for clinical judgment. This technical report proposes a clinical-operational framework for the responsible use of artificial intelligence in infection risk management among immunosuppressed patients. The framework includes seven domains: baseline infection risk stratification, screening and follow-up of latent infections, alerts for infectious deterioration, support for rational antimicrobial use, antimicrobial resistance surveillance, traceable clinical documentation, and continuous performance auditing. To reduce risks related to bias, false reassurance, alert fatigue, limited transportability, performance degradation, and excessive reliance on model outputs, the framework emphasizes local validation, human oversight, proportional explainability, institutional governance, longitudinal monitoring, adverse event review, and formal updating mechanisms. The objective is not to claim that artificial intelligence independently reduces infections or improves outcomes, but to define minimum conditions for its use as a supervised support tool within integrated programs for prevention, antimicrobial stewardship, patient safety, and risk management. This approach may be useful when prevention of latent tuberculosis, bacteremia, sepsis, opportunistic infections, and antimicrobial resistance requires coordinated clinical, microbiological, pharmacological, technological, quality, and audit processes.
Bobadilla et al. (Wed,) studied this question.