Prediction of infection caused by extended-spectrum beta-lactamase-producing Enterobacteriaceae: development of a clinical decision-making nomogram

Objective: This study aimed to assess the population at risk of infection by extended-spectrum beta-lactamase (ESBL)-producing organisms, using clinical criteria. Materials and methods: All urine cultures positive for Enterobacteriaceae in a Spanish hospital department from January 2010 to 2014 were reviewed. All isolates with ESBL-positive strains were collected, and isolates received during the first week of each month with ESBL-negative strains from symptomatic patients hospitalized or admitted to the emergency room. Multivariate analysis of the factors involved was undertaken and a nomogram developed to predict the probability of infection by ESBL-producing microorganisms. Results: The study included 1524 patients with urinary tract infection (UTI): 416 ESBL-positive and 1108 ESBL-negative. In univariate analysis, risk factors were: male gender (p = 0.036), age (p < 0.0001), nursing home (p < 0.0001), previous antimicrobial therapy (p < 0.0001) or hospitalization (p < 0.0001), diabetes (p < 0.0001), chronic renal insufficiency (p < 0.0001), severe underlying disease (p < 0.0001), neoplasia (p = 0.0005), urological (p < 0.0001) and non-urological invasive procedure (p = 0.0003), recurrent UTI (p < 0.0001), urological (p < 0.0001) or abdominal surgery (p < 0.0001) and permanent urethral catheter (p < 0.0001). In multivariate analysis, the data set was split into a development cohort of 1067 patients and a validation cohort of 457 cases. A nomogram was developed to predict the probability of infection by ESBL-producing bacteria, which included seven variables: age (p < 0.0001), gender (p = 0.004), nursing home (p < 0.0001), previous antimicrobial therapy (p = 0.04) or hospitalization (p < 0.0001), recurrent UTI (p < 0.0001) and non-urological invasive procedure (p = 0.005). The discriminative accuracy was 0.79 (95% confidence interval 0.77-0.83). Conclusions: A nomogram was developed that predicts the risk of infection by ESBL-producing Enterobacteriaceae with reasonable accuracy. It could improve clinical decision making and enable more efficient empirical treatment.