Development and validation of a prediction model for in-hospital mortality in patients with sepsis.

Abstract

Background: Sepsis, a life-threatening condition marked by organ dysfunction due to a dysregulated host response to infection, involves complex physiological and biochemical abnormalities.

Aim: To develop a multivariate model to predict 4-, 6-, and 8-week mortality risks in intensive care units (ICUs).

Study design: A retrospective cohort of 2389 sepsis patients was analysed using data captured by a clinical decision support system. Patients were randomly allocated into training (n = 1673) and validation (n = 716) sets at a 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) regression identified variables incorporated into a multivariate Cox proportional hazards regression model to construct a prognostic nomogram. The area under the receiver operating characteristic curve (AUROC) assessed model accuracy, while performance was evaluated for discrimination, calibration and clinical utility.

Results: A risk score was developed based on 11 independent predictors from 35 initial factors. Key predictors included minimum Acute Physiology and Chronic Health Evaluation II (APACHE II) score as having the greatest impact on prognosis, followed by days of mechanical ventilation, number of vasopressors, maximum and minimum Sequential Organ Failure Assessment (SOFA) scores, infection sources, Gram-positive or Gram-negative bacteria and malignancy. The nomogram demonstrated superior discriminative ability, with AUROC values of 0.882 (95% confidence interval [CI], 0.855-0.909) and 0.851 (95% CI, 0.804-0.899) at 4 weeks; 0.836 (95% CI, 0.798-0.874) and 0.820 (95% CI, 0.761-0.878) at 6 weeks; and 0.843 (95% CI, 0.800-0.887) and 0.794 (95% CI, 0.720-0.867) at 8 weeks for training and validation sets, respectively.

Conclusion: A validated nomogram and web-based calculator were developed to predict in-hospital mortality in ICU sepsis patients. Targeting identified risk factors may improve outcomes for critically ill patients.

Relevance to clinical practice: The developed prediction model and nomogram offer a tool for assessing in-hospital mortality risk in ICU patients with sepsis, potentially aiding in nursing decisions and resource allocation.