Related Experiment Video
Updated: Sep 12, 2026
Competing-Risk Nomogram for Predicting Cancer-Specific Survival in Multiple Primary Colorectal Cancer Patients after Surgery
Published on: September 27, 2024
A multidimensional nomogram for clinical risk stratification of bloodstream infection and mortality in hematologic
Lingqiong Lan1,2, Shaozhen Chen1,3,4, Haojie Zhu1,5
1The Second Department of Hematology, National Regional Medical Center, Binhai Campus of the First Affiliated Hospital, Fujian Medical University, Fuzhou, China.
Background:
Bloodstream infections (BSIs) are serious complications in patients with hematologic malignancies (HMs) and are associated with substantial morbidity and mortality. Effective risk stratification is crucial for optimizing empirical antimicrobial decisions. This study aimed to develop and internally validate nomograms for predicting BSI occurrence and post-BSI mortality among HM patients.
Methods:
This retrospective cohort study enrolled 3,014 HM patients, including 725 with BSI. The cohort was randomly divided into derivation and internal validation cohorts at a 6:4 ratio. The BSI occurrence model was developed using multivariable logistic regression, while the post-BSI mortality prediction model among patients with BSI was developed using Cox regression. Model performance was evaluated based on discrimination, calibration, and decision curve analysis.
Results:
The BSI occurrence model demonstrated acceptable discrimination, with area under the receiver operating characteristic curve (AUC) values of 0.775 in the derivation cohort and 0.761 in the internal validation cohort. The final nomogram incorporated 12 variables, among which neutrophil count ≤0.5 × 109/L received a relatively high point allocation. The post-BSI mortality prediction nomogram showed moderate discrimination, with C-indices of 0.721 and 0.713 in the derivation and internal validation cohorts, respectively. Nine variables were included in the final post-BSI mortality model, and polymicrobial BSI received a relatively high point allocation within the nomogram. Both models demonstrated acceptable calibration and potential clinical utility based on decision curve analysis. The BSI occurrence model was further assessed in a small independent cohort, whereas the mortality model requires additional external validation.
Conclusions:
The BSI occurrence model showed acceptable discrimination in the derivation and internal validation cohorts and was preliminarily assessed in an independent cohort. The post-BSI mortality model showed moderate discrimination in the internal validation cohort and requires further evaluation in larger and more representative populations.