Risk factors and predictive model for pressure injuries in ICU patients with a Braden score ≤9 based on initial serum biomarkers at admission.

Abstract

Objective: To explore the association between initial serum biomarkers at admission and the development of pressure injuries (PIs) in intensive care unit (ICU) patients with a Braden score ≤9, and to construct a predictive model for PI risk. The goal is to support early identification and personalized nursing interventions for high-risk individuals.

Methods: A retrospective study of 290 ICU patients with a Braden score ≤9 admitted to six tertiary hospitals in Shanghai between January 2020 and January 2024. Patients were divided into a PI group (n = 115) and a non-PI group (n = 175) based on whether they developed a pressure injury. Demographic characteristics, clinical interventions, and laboratory data on the day of admission were collected. Univariate and multivariate logistic regression analyses were used to identify independent risk factors and to build a predictive model, with model performance evaluated via receiver operating characteristic (ROC) analysis.We developed a bedside-applicable logistic prediction model using admission-day variables and, after model estimation, derived risk thresholds from the ROC coordinate table to enable three-tier risk stratification.

Results: Independent risk factors for pressure injury included older age, lower serum sodium and calcium levels, elevated white blood cell count, increased glutamate and alanine levels, positive intestinal pathogen colonization, use of mechanical ventilation, and application of physical restraints (all P < 0.05). The final logistic regression model showed good fit (Hosmer-Lemeshow test, P = 0.807) and strong discrimination (AUC = 0.910, 95 % CI: 0.877-0.943), with a sensitivity of 81.7 % and specificity of 86.9 %.The model showed excellent discrimination (AUC 0.910) and was operationalized into low (P < 0.3426), intermediate (0.3426≤P < 0.5720), and high risk (P ≥ 0.5720), each mapped to specific bedside actions.

Conclusion: Among ICU patients with a Braden score ≤9, initial serum biomarkers combined with key clinical features can effectively identify individuals at high risk of pressure injuries. The predictive model developed in this study demonstrates strong clinical applicability and discriminative power, providing scientific support for precision nursing interventions.This study provides a clear, clinically actionable risk-to-action protocol for ICU patients with Braden scores ≤9.