Nomogram for predicting mortality in hospitalized patients with infective endocarditis

Abstract

This study aimed to develop a nomogram for accurately predicting in-hospital mortality in patients with infective endocarditis (IE). We conducted a retrospective analysis of clinical, echocardiographic, and laboratory data from IE patients admitted between January 2010 and September 2024. 252 IE patients from the Second Hospital of Lanzhou University were included in the training cohort, while 65 IE patients from the First Hospital of Lanzhou University were enrolled for external validation. The least absolute shrinkage and selection operator (LASSO) regression method was used to identify factors associated with in-hospital mortality. A nomogram was constructed using multivariate logistic regression. Model performance was assessed using receiver operating characteristic (ROC) curve and calibration curve. Clinical utility was evaluated through decision curve analysis (DCA) and clinical impact curve (CIC). The nomogram included five independent risk factors: embolic events, vegetation size ≥ 10 mm, moderate or higher pulmonary hypertension, hydropericardium, and surgery. The area under the curve (AUC) of the nomogram in the training cohort was 0.850 (95% CI: 0.794-0.906), and external validation cohort was 0.819 (95% CI: 0.693-0.946). The calibration plot demonstrated excellent prediction consistency. Both DCA and CIC confirmed the clinical utility of the nomogram. We developed and validated a nomogram for predicting in-hospital mortality in patients with IE. The model demonstrated excellent performance and provided a useful tool to assist clinicians in decision-making and patient management.

Keywords: Infective endocarditis; Mortality; Nomogram; Prediction model.

Fig. 1

Nomogram for estimating the risk of in-hospital mortality. To use the nomogram, following these three steps: (i) Identifying the patient’s value for each predictor on its corresponding axis and draw a vertical line upward to the “Points” axis to determine the score assigned to that variable. (ii) Summing the individual scores to obtain the “Total Points”. (iii) Locating the total score on the “Total Points” axis and draw a vertical line downward to estimate the corresponding probability of in-hospital mortality. PH, pulmonary hypertension.

Fig. 2

Dynamic nomogram interface for predicting in-hospital mortality. The Shiny-based web application enables clinicians to input patient-specific variables and receive an immediate estimate of in-hospital mortality risk. Shown here is an example of a 47-year-old male with embolism, vegetation size < 10 mm, moderate-to-high pulmonary hypertension, hydropericardium, and no surgical intervention. The resulting predicted risk is 73.47%, displayed numerically and graphically. The dynamic nomogram is available at https://ayq-2025.shinyapps.io/nomogram_app/. PH, pulmonary hypertension.

Fig. 3

Best match factor selection by LASSO regression. (a) is the LASSO regression path diagram. (b) Plot of the best matching factors selected by the ten-fold cross validation method.

Fig. 4

Forest plots of independent risk factors for in-hospital mortality by multivariate analysis.

Fig. 5

(A) Receiver operating characteristic (ROC) curve, (B) Calibration curve, (C) Clinical decision curve analysis (DCA), and (D) Clinical impact curve (CIC) in the training cohort.

Fig. 6

(A) Receiver operating characteristic (ROC) curve, (B) Calibration curve, (C) Clinical decision curve analysis (DCA), and (D) Clinical impact curve (CIC) in the external validation cohort.