Development and validation of a nomogram for pressure injury risk prediction in stroke patients: a retrospective...

Retrospective cohort study.

This study aimed to identify independent risk factors for pressure injury (PI) during the post-stroke recovery phase, develop and validate a nomogram prediction model to facilitate the identification of high-risk individuals for PI, and establish a theoretical framework for optimizing clinical intervention strategies.

Retrospective clinical data were collected from 284 hospitalized stroke patients in the recovery phase (including 85 PI cases) at the Affiliated Hospital of North Sichuan Medical College between January 2018 and December 2022. Participants were randomly allocated into training (70%) and internal validation (30%) cohorts. An external validation cohort comprising 60 stroke patients (30 PI cases) from Zigong First People's Hospital (January 2023-January 2024) was additionally analyzed. Univariate analysis and LASSO regression were utilized to screen independent PU risk factors, followed by nomogram construction. Model performance was evaluated using the C-index, calibration curves, and Decision Curve Analysis (DCA). Comparative analyses were conducted against the Braden scale (Model 2) and a combined model incorporating the Braden scale (Model 3).

Independent risk factors for PI in post-stroke recovery patients included hemorrhagic stroke subtype, advanced age, hypoalbuminemia, elevated leukocyte counts, and low Activities of Daily Living (ADL) scores. The nomogram model incorporating these five predictors demonstrated AUC values of 0.902 (training cohort), 0.935 (internal validation), and 0.936 (external validation), exceeding the predictive capacity of individual variables: stroke type (AUC = 0.642), age (AUC = 0.756), albumin level (AUC = 0.754), leukocyte count (AUC = 0.712), and ADL score (AUC = 0.839). Calibration curves indicated strong concordance between predicted and observed outcomes, while DCA confirmed substantial clinical net benefit. The Braden scale (AUC = 0.817) exhibited inferior predictive performance compared to our model, and the combined model (AUC = 0.901) showed no significant improvement, underscoring the parsimony and clinical utility of the proposed nomogram.

The nomogram developed in this study for predicting PIs in stroke recovery patients demonstrates high accuracy and discrimination, facilitating the early identification of high-risk individuals and aiding in the formulation of personalized intervention strategies.