Background: Bloodstream infection (BSI) due to carbapenem-resistant organisms (CROs) has emerged as a worldwide problem associated with high mortality. This study aimed to evaluate the risk factors associated with mortality in HM patients with CROs BSI and to establish a scoring model for early mortality prediction.
Methods: We conducted a retrospective cohort study at our hematological department from January 2018 to December 2021, including all HM patients with CROs BSI. The outcome measured was death within 30-day of BSI onset. Survivor and non-survivor subgroups were compared to identify predictors of mortality. Univariate and multivariate Cox regression analyses were used to identify prognostic risk factors and develop a nomogram.
Results: In total, 150 HM patients were included in the study showing an overall 30-day mortality rate of 56%. Klebsiella pneumonia was the dominant episode. Cox regression analysis showed that pre-infection length of stay was > 14 days (score 41), Pitt score > 4 (score 100), mucositis (score 41), CAR (The ratio of C-reactive protein to albumin) > 8.8 (score 57), early definitive therapy (score 44), and long-duration (score 78) were positive independent risk predictors associated with 30-day mortality, all of which were selected into the nomogram. Furthermore, all patients were divided into the high-risk group (≥ 160 points) or the low-risk group based on the prediction score model. The mortality of the high-risk group was 8 times more than the low-risk group. Kaplan-Meier analysis showed that empirical polymyxin B therapy was associated with a lower 30-day mortality rate, which was identified as a good prognostic factor in the high-risk group. In comparison, empirical carbapenems and tigecycline were poor prognostic factors in a low-risk group.
Conclusion: Our score model can accurately predict 30-day mortality in HM patients with CROs BSI. Early administration of CROs-targeted therapy in the high-risk group is strongly recommended to decrease mortality.
Keywords: carbapenem-resistant organisms, predictive model, nomogram, hematological malignancies, bloodstream infection