Risk factors and development of a clinical prediction model for carbapenem-resistant Enterobacterales bacteraemia: A...

To identify risk factors for carbapenem resistance in Enterobacterales bacteraemia and predictors of 14- and 28-d mortality.

Adult patients (≥18 y) with Enterobacterales bacteraemia admitted to a tertiary-care centre between January 2015 and January 2024 were retrospectively analysed. Isolates were classified as carbapenem-resistant Enterobacterales (CRE), extended-spectrum β-lactamase-producing (ESBL), or non-ESBL (nESBL). Independent predictors of CRE were identified using multivariable logistic regression. Model calibration, internal validation and a nomogram were used to assess predictive performance. Cox proportional hazards models identified predictors of 14- and 28-d mortality.

Of 895 patients, 14.9% had CRE bacteraemia. Independent predictors of CRE were intensive care unit (ICU) admission (OR 4.87, 95% CI 3.18-7.57), hospitalisation for ≥14 d before index blood culture (OR 2.67, 95% CI 1.74-4.09) and prior carbapenem exposure (OR 2.79, 95% CI 1.84-4.25). Twenty-eight-day mortality was higher in the CRE than ESBL and nESBL groups (63.9% vs. 32.3% vs. 28.3%; P < 0.001). ICU admission, sepsis and inappropriate empirical antimicrobial therapy independently predicted mortality. The prediction model showed good discrimination (Harrell's C-index, 0.79).

CRE bacteraemia was associated with high mortality and occurred predominantly in critically ill patients with prolonged hospitalisation and prior carbapenem exposure. Mortality was driven primarily by illness severity and inappropriate empirical antimicrobial therapy rather than carbapenem resistance itself. Early appropriate empirical therapy independently improved survival regardless of resistance profile.