ESBL-PRODUCING ENTEROBACTERALES BLOODSTREAM INFECTIONS: ANTIMICROBIAL THERAPY, EARLY DETECTION AND OUTCOMES

Background. Extended-spectrum β-lactamase–producing Enterobacterales (ESBL-E) bloodstream infections (BSI) are increasingly common both in healthcare settings and in the community and pose a major public-health challenge due to limited therapeutic options and poor outcomes. ESBL-E are associated with a higher rate of inappropriate empirical therapy, which in turn leads to increased clinical failure, higher mortality, and prolonged hospitalization. At the same time, their rising prevalence in the community has also driven overuse of second-/third-line broad-spectrum agents, with suboptimal de-escalation once culture results are available. Early detection of patients at risk of ESBL-E infection could improve prescribing but remains difficult. Furthermore, uncertainty persists about whether ESBL-E BSI can be treated with piperacillin/tazobactam as a carbapenem-sparing option. Methods. This doctoral research comprised three phases. First, a retrospective observational study of a five-year cohort (2019–2023) of Enterobacterales BSI admitted to the Emergency Department of Niguarda Hospital evaluated empirical/definitive prescribing and in-hospital mortality. The same cohort was used to externally validate existing risk scores to early identify patients with ESBL-E BSIs and to derive a local prediction tool. Finally, a nine-year retrospective cohort (2015–2023) of ESBL-E BSI compared empirical piperacillin/tazobactam with carbapenems. Results. ESBL-E BSIs were prevalent (26.1%; 95% CI, 23.0–29.3%); 33% of them received inappropriate empirical therapy. Neither ESBL-E aetiology nor inappropriate empirical therapy was associated with higher in-hospital mortality, unlike older age, higher Charlson Comorbidity Index, immunosuppression, sepsis/septic shock, and pulmonary or soft-tissue infection. Piperacillin/tazobactam was the most common empirical agent in both ESBL-E and non-ESBL-E patients (53% overall), yet de-escalation occurred in only ~25% of non-ESBL-E BSI. Four external risk scores showed reduced discrimination versus original reports: Tumbarello AUROC 0.64 (95% CI: 0.60–0.68), Duke 0.68 (95% CI: 0.64–0.72), Kim 0.67 (95% CI: 0.63–0.71), and Jones 0.55 4 (95% CI: 0.51–0.59). The ESBL-E score was locally derived and internally validated achieving a moderate discrimination (AUROC 0.70; 95% CI, 0.65–0.74). Empirical piperacillin/tazobactam was associated with higher in-hospital mortality than empirical carbapenems (14.3% vs 3.5%; adjusted risk difference, 10%; 95% CI, 2–19%), whereas the two definitive therapies showed no difference (6% vs 9%; risk difference, −2.5%; 95% CI, −9.2 to 4.2). Conclusions. Better early risk stratification for ESBL-E—potentially combining clinical factors with rapid microbiological testing—is needed to reduce inappropriate empiric therapy and unnecessary broad-spectrum agents use. Antibiotic-stewardship interventions should strengthen de-escalation. Further, larger studies are required to define the patient subgroups and infections in which piperacillin/tazobactam is non-inferior to carbapenems.