ObjectiveTo establish a Nomogram prediction model for inpatient mortality in intensive care units(ICU), and to provide guidance for reducing inpatient mortality in ICU.

MethodsClinical data of 1 133 inpatients in ICU were analyzed retrospectively.Death risk factors of inpatients in ICU were screened by logistic regression analysis.Nomogram prediction model was constructed.Decision curve analysis(DCA) was used to compare simple and complex evaluation model.

ResultsMultivariate logistic regression analysis showed that hospital infection(OR=1.876, 95% CI: 1.037-3.043), hypertension(OR=1.133, 95%CI: 1.090-1.177), diabetes meillitus (OR=1.141, 95% CI: 1.064-1.209), blood transfusion(OR=1.357, 95%CI: 1.102-3.421), low GCS score(OR=0.953, 95%CI: 0.917-0.991), APACHE Ⅱ score(OR=2.638, 95%CI: 0.794-8.692), tracheotomy(OR=3.973, 95%CI: 2.386-6.615), endotracheal intubation(OR=1.562, 95%CI: 1.163-2.266), arteriovenous cannulation(OR=1.365, 95%CI: 1.067-3.172), days of arteriovenous cannulation(OR=1.825, 95%CI: 1.224-2.979), urinary catheter intubation(OR=2.016, 95%CI: 1.050-3.870), and days of urinary catheter intubation(OR=2.689, 95%CI: 1.724-4.195) were independent risk factors for death of hospitalized ICU patients(P < 0.05).Nomogram model was established according to multivariate logistic regression results, and it was verified that the prediction model had good consistency(C-index=0.748, P < 0.05).DCA showed that the threshold probability was within the range of(0.09-0.49), and the net benefit of the complex model was higher than that of the simple model.The threshold probability was within the range of(0.49-0.87), the net benefit of the simple model was higher than that of the complex model.

ConclusionsThe Nomogram model is successfully established to predict the death of ICU inpatients.