Development of a Novel Tool

A Nomogram for Predicting In-hospital Mortality of Patients in Intensive Care Unit After Percutaneous Coronary Intervention

Miao Yuan; Bin Cheng Ren; Yu Wang; Fuxian Ren; Dengfeng Gao


BMC Anesthesiol. 2023;23(5) 

In This Article

Abstract and Introduction


Backgrounds: Increased risk of in-hospital mortality is critical to guide medical decisions and it played a central role in intensive care unit (ICU) with high risk of in-hospital mortality after primary percutaneous coronary intervention (PCI). At present, most predicting tools for in-hospital mortality after PCI were based on the results of coronary angiography, echocardiography, and laboratory results which are difficult to obtain at admission. The difficulty of using these tools limit their clinical application. This study aimed to develop a clinical prognostic nomogram to predict the in-hospital mortality of patients in ICU after PCI.

Methods: We extracted data from a public database named the Medical Information Mart for Intensive Care (MIMIC III). Adult patients with coronary artery stent insertion were included. They were divided into two groups according to the primary outcome (death in hospital or survive). All patients were randomly divided into training set and validation set randomly at a ratio of 6:4. Least absolute shrinkage and selection operator (LASSO) regression was performed in the training set to select optimal variables to predict the in-hospital mortality of patients in ICU after PCI. The multivariate logistical analysis was performed to develop a nomogram. Finally, the predictive efficiency of the nomogram was assessed by area under the receiver operating characteristic curve (AUROC), integrated discrimination improvement (IDI), and net reclassification improvement (NRI), and clinical net benefit was assessed by Decision curve analysis (DCA).

Results: A total of 2160 patients were recruited in this study. By using LASSO, 17 variables were finally included. We used multivariate logistic regression to construct a prediction model which was presented in the form of a nomogram. The calibration plot of the nomogram revealed good fit in the training set and validation set. Compared with the sequential organ failure assessment (SOFA) and scale for the assessment of positive symptoms II (SAPS II) scores, the nomogram exhibited better AUROC of 0.907 (95% confidence interval [CI] was 0.880–0.933, p < 0.001) and 0.901 (95% CI was 0.865–0.936, P < 0.001) in the training set and validation set, respectively. In addition, DCA of the nomogram showed that it could achieve good net benefit in the clinic.

Conclusions: A new nomogram was constructed, and it presented excellent performance in predicting in-hospital mortality of patients in ICU after PCI.