Development and Validation of a Nomogram for Mortality Prediction in Septic Patients with Prolonged or Chronic...

Background/Objectives: Patients with prolonged or chronic critical illness (PCI/CCI) represent a subgroup characterized by extended stays in an intensive care unit (ICU), persistent organ dysfunction, and increased susceptibility to recurrent sepsis episodes. Current sepsis prognostic tools have not been specifically tailored for this high-risk population. This study aimed to develop and validate a prognostic nomogram for predicting mortality in septic ICU patients with PCI/CCI. Methods: Data were obtained from the Russian Intensive Care Dataset (RICD). Eligible patients had confirmed sepsis episodes according to Sepsis-3 criteria. The cohort was randomly split into training and testing sets in a 7:3 ratio. Multivariable Cox regression identified predictors of mortality, which were incorporated into a prognostic nomogram. Predictive accuracy was assessed using Harrell's C-index, and horizon-specific area under the receiver operating characteristic curve (AUROC). Results: A total of 336 septic patients were analyzed, with an overall ICU mortality of 14.0%. Median ICU length of stay was 44 days, and median time to sepsis onset was 10 days. Recurrent sepsis episodes occurred in 28.6% of patients. In multivariable analysis, four predictors of mortality were identified: age, SOFA score at sepsis onset, type 2 diabetes mellitus, and time to sepsis onset. The nomogram demonstrated a C-index of 0.787 (95% confidence interval [CI] 0.669; 0.890) in the training set and one of 0.715 (95% CI 0.584; 0.836) in the testing set. In the testing set, the horizon-specific AUROCs were 0.898, 0.741, and 0.703 for 14-, 28-, and 42-day survival prediction, respectively. Conclusions: The prognostic nomogram, specifically tailored for PCI/CCI septic patients, demonstrated a testing-set C-index of 0.715, with higher 14-day predictive performance, whereas predictive accuracy decreased at later time points. Prospective multicenter validation is necessary before clinical implementation.