Empirical therapies among adults hospitalized for community-acquired upper urinary tract infections: A decision-tree analysis of mortality, costs, and resistance

Background Poor outcomes occur when patients with serious infections receive antibiotics to which the organisms are resistant. Methods Decision trees simulated in-hospital mortality, costs, incremental cost-effectiveness ratio per life year saved, and carbapenem resistance according to 3 empirical antibiotic strategies among adults hospitalized for community-acquired (CA) upper urinary tract infections (UTIs): ceftriaxone (CRO) plus gentamicin (GM) in the intensive care unit (ICU), imipenem (IMP), and individualized choice (IMP or CRO) based on clinical risk factors for CA- extended-spectrum β -lactamase (ESBL). Results The estimated prevalence of CA-ESBL on admission was 5% (range, 1.3%-17.6%); 3% and 97% were admitted to the ICU and medical ward (MW), respectively. In the ICU, CRO plus GM was dominated; IMP was cost-effective (incremental cost-effectiveness ratio: €4,400 per life year saved compared with individualized choice). In the MW, IMP had no impact on mortality and was less costly (−€142 per patient vs CRO, −€38 vs individualized choice). The dominance of IMP was consistent in sensitivity analyses. Compared with CRO, colonization by carbapenem-resistant pathogens increased by an odds ratio of 4.5 in the IMP strategy. Conclusion Among the ICU patients, empirical IMP therapy reduces mortality at an acceptable cost. Among MW patients, individualized choice or CRO is preferred to limit carbapenem resistance at a reasonable cost.