Intensive care unit (ICU)–acquired infections with antibiotic-resistant bacteria have been associated with substantial health and economic costs. Moreover, southern Europe has historically reported high levels of antimicrobial resistance.
We studied a cohort of adult patients admitted to the ICU who developed an ICU-acquired infection related to an invasive procedure in a sample of Spanish, Italian, and Portuguese hospitals between 2008 and 2016, using data from The European Surveillance System (TESSy) released by the European Centers for Disease Control (ECDC). We analyzed the association between infections with selected antibiotic-resistant bacteria of public health importance and excess LOS using regression, matching, and time-to-event methods. We controlled for several confounding factors as well as time-dependent biases. We also computed the associated economic burden of excess resource utilization for each selected country.
In total, 13,441 patients with at least 1 ICU-acquired infection were included in the analysis: 4,106 patients (30.5%) were infected with antimicrobial-resistant bacteria, whereas 9,335 patients (69.5%) were infected with susceptible bacteria. The unadjusted association between resistance status and excess LOS was 7 days (95% CI, 6.13–7.87; P < .001). Fully adjusted models yielded significantly lower estimates: 2.76 days (95% CI, 1.98–3.54; P < .001) in the regression model, 2.60 days (95% CI, 1.66–3.55; P < .001) in the genetic matching model, and a hazard ratio of 1.15 (95% CI, 1.11–1.19; P < .001) in the adjusted Cox regression model. These estimates, alongside the prevalence of resistance, translated into direct hospitalization attributable costs per ICU-acquired infection of 5,224€ (95% CI, 3,691–6,757) for Spain, 4,461€ (95% CI, 1,948–6,974) for Portugal, and 4,320€ (95% CI, 1,662–6,977) for Italy.
ICU-acquired infections associated with antibiotic-resistant bacteria are substantially associated with a 15% increase in excess LOS and resource utilization in 3 southern European countries. However, failure to appropriately control for significant confounders inflates estimates by ∼2.5-fold.