Derivation and validation of a predictive mortality model of in-hospital patients with Acinetobacter baumannii nosocomial infection or colonization

Abstract

Purpose: Acinetobacter baumannii (Ab) is a Gram-negative opportunistic bacterium responsible for nosocomial infections or colonizations. It is considered one of the most alarming pathogens due to its multi-drug resistance and due to its mortality rate, ranging from 34 to 44,5% of hospitalized patients. The aim of the work is to create a predictive mortality model for hospitalized patient with Ab infection or colonization.

Methods: A cohort of 140 sequentially hospitalized patients were randomized into a training cohort (TC) (100 patients) and a validation cohort (VC) (40 patients). Statistical bivariate analysis was performed to identify variables discriminating surviving patients from deceased ones in the TC, considering both admission time (T0) and infection detection time (T1) parameters. A custom logistic regression model was created and compared with models obtained from the "status" variable alone (Ab colonization/infection), SAPS II, and APACHE II scores. ROC curves were built to identify the best cut-off for each model.

Results: Ab infection status, use of penicillin within 90 days prior to ward admission, acidosis, Glasgow Coma Scale, blood pressure, hemoglobin and use of NIV entered the logistic regression model. Our model was confirmed to have a better sensitivity (63%), specificity (85%) and accuracy (80%) than the other models.

Conclusion: Our predictive mortality model demonstrated to be a reliable and feasible model to predict mortality in Ab infected/colonized hospitalized patients.

Keywords: Acinetobacter baumannii; Acinetobacter baumannii mortality; Colonization; Infection; Multidrug- resistance; Predictive mortality model.

Fig. 1

ROC-curves for each predictive mortality model, tested on the “Training cohort”

Fig. 2

Percentage of correctly classified patients among the models in the TC and VC