Xue-Lin Li,1,2,* Te Mi,1,2,* Cancan Liu,1,2 Mingchen Feng1,2 

1Department of Intensive Care Unit, Jining No. 1 People’s Hospital, Shandong First Medical University, Jining, People’s Republic of China; 2Jining Critical Care Diagnosis and Treatment Center, Jining, People’s Republic of China

*These authors contributed equally to this work

Correspondence: Mingchen Feng, Department of Intensive Care Unit, Jining No. 1 People’s Hospital, Shandong First Medical University, Jining, 272000, People’s Republic of China, Email sdjnfmch@163.com

Background: Septic shock is a high-mortality syndrome, particularly in patients aged 50 and older. Predicting mortality in this population is challenging due to clinical heterogeneity and limitations of traditional scoring systems like SOFA and APACHE II. This study aimed to develop and validate a predictive model using norepinephrine index (NEI)—a novel biomarker defined as the norepinephrine dose administered within the first 24 hours of ICU admission divided by BMI and 24 hours—and procalcitonin (PCT) to improve risk stratification and clinical decision-making.
Methods: A retrospective cohort of 94 patients aged ≥ 50 years with septic shock was analyzed. Key clinical variables within the first 24 hours were collected, and univariate and stepwise logistic regression identified predictors of 28-day mortality. The model’s performance was evaluated with ROC curves, AUC, and confusion matrices, alongside internal validation through stratified analysis, bootstrap resampling, and training-test splits. External validation was conducted in an independent cohort of 57 patients.
Results: The final model incorporating NEI and PCT achieved an AUC of 0.91, outperforming individual biomarkers (NEI: AUC = 0.86; PCT: AUC = 0.69). Nonlinear analysis identified NEI > 4mg· m² / (kg· 24h) and PCT < 50 ng/mL as critical thresholds for high mortality risk.
Conclusion: The NEI and PCT-based prognostic model provides a reliable tool for predicting 28-day mortality in septic shock patients aged 50 and above. However, as a single-center study with a relatively small sample size, the generalizability of these findings may be limited. Future multicenter studies with larger sample sizes are necessary to validate this model’s applicability across populations. This model holds potential to optimize clinical management, enabling timely interventions such as more intensive hemodynamic support and infection control.

Keywords: septic shock, mortality prediction, norepinephrine index, procalcitonin, machine learning, risk stratification