Development and validation of nomograms for predicting the risk probability of carbapenem resistance and 28-day all-cause mortality in gram-negative bacteremia among patients with hematological diseases

Objectives Gram-negative bacteria (GNB) bloodstream infections (BSIs) are the most widespread and serious complications in hospitalized patients with hematological diseases. The emergence and prevalence of carbapenem-resistant (CR) pathogens has developed into a considerable challenge in clinical practice. Currently, nomograms have been extensively applied in the field of medicine to facilitate clinical diagnosis and treatment. The purpose of this study was to explore risk indicators predicting mortality and carbapenem resistance in hematological (HM) patients with GNB BSI and to construct two nomograms to achieve personalized prediction. Methods A single-center retrospective case-control study enrolled 244 hospitalized HM patients with GNB-BSI from January 2015 to December 2019. The least absolute shrinkage and selection operator (LASSO) regression analysis and multivariate logistic regression analysis were conducted to select potential characteristic predictors of plotting nomograms. Subsequently, to evaluate the prediction performance of the models, the prediction models were internally validated using the bootstrap approach (resampling = 1000) and 10-fold cross validation. Results Of all 244 eligible patients with BSI attributed to GNB in this study, 77 (31.6%) were resistant to carbapenems. The rate of carbapenem resistance exhibited a growing tendency year by year, from 20.4% in 2015 to 42.6% in 2019 (p = 0.004). The carbapenem resistance nomogram constructed with the parameters of hypoproteinemia, duration of neutropenia ≥ 6 days, previous exposure to carbapenems, and previous exposure to cephalosporin/β-lactamase inhibitors indicated a favorable discrimination ability with a modified concordance index (C-index) of 0.788 and 0.781 in both the bootstrapping and 10-fold cross validation procedures. The 28-day all-cause mortality was 28.3% (68/240). The prognosis nomogram plotted with the variables of hypoproteinemia, septic shock, isolation of CR-GNB, and the incomplete remission status of underlying diseases showed a superior discriminative ability of poorer clinical prognosis. The modified C-index of the prognosis nomogram was 0.873 with bootstrapping and 0.887 with 10-fold cross validation. The decision curve analysis (DCA) for two nomogram models both demonstrated better clinical practicality. Conclusions For clinicians, nomogram models were effective individualized risk prediction tools to facilitate the early identification of HM patients with GNB BSI at high risk of mortality and carbapenem resistance.