Related Experiment Video
Updated: Jan 14, 2026
Objective Nociceptive Assessment in Ventilated ICU Patients: A Feasibility Study Using Pupillometry and the Nociceptive Flexion Reflex
Published on: July 4, 2018
Comparative Evaluation of Pressure Injury Risk Assessment Tools and Machine Learning Models in Postoperative Surgical
Touran Bahrami Babaheidari1, Mansooreh Tajvidi2, Mahmood Bakhtiyari1,3
1Department of Community Medicine, School of Medicine, Alborz University of Medical Sciences, Karaj, Iran.
Background:
Pressure injuries are a significant and preventable complication in critically ill patients, especially in those admitted to intensive care units (ICUs) following orthopaedic, gastrointestinal and neurosurgical procedures.
Aim:
This study aimed to evaluate the predictive accuracy of commonly used pressure injury risk assessment tools (Braden, modified Braden, Waterlow and Cubbin-Jackson scales) and to compare their performance with machine learning based predictive models among patients admitted to the surgical ICU.
Study Design:
This prospective, single-centre cohort was conducted between February 2020 and May 2023 at a tertiary care surgical hospital in Karaj, Iran to investigate the potential association between the questionnaire scores and clinical and laboratory measurements with pressure injury occurrence within the ICU. We also explored the utility of machine learning models to improve prediction accuracy.
Results:
The results showed that out of 256 included patients, 33 patients (12.9%) developed pressure injuries, with the modified Braden demonstrating the highest predictive accuracy (Area under ROC curve, AUC: 0.90), followed by the Braden (AUC: 0.79), Waterlow (AUC: 0.72) and Cubbin-Jackson (AUC: 0.71) scales. The key predictors of pressure injury development were a history of pressure injuries (OR: 14.74, 95% CI: [4.04-53.76]), low haemoglobin (OR: 0.82, 95% CI: [0.71-0.94]), and diabetes mellitus (OR: 4.44, 95% CI: [1.71-11.51]). Machine learning models using clinical predictors showed a marginal improvement in accuracy but did not outperform the modified Braden scale alone.
Conclusions:
This study suggests that the modified Braden scale is the most reliable tool for predicting pressure injuries in surgical ICU patients, with limited added benefit from machine learning models or other study predictors. Future research should focus on incorporating broader clinical measures to explore further strategies for prevention.
Relevance To Clinical Practice:
The results of the current study suggest that the modified Braden questionnaire is the single most accurate predictor of pressure injury development in the surgical ICU setting.
