A Longitudinal Investigation of Stage 2 Pressure Injury Outcomes With Machine Learning Technique to Identify Relevant...

Area of Science:

• Biomedical Engineering
• Clinical Medicine
• Data Science

Background:

• Pressure injuries (PIs) represent a significant global health concern with substantial socioeconomic impact.
• While numerous factors influence PIs, those specifically driving disease aggravation remain incompletely understood.
• Predictive modeling is crucial for identifying at-risk individuals and implementing targeted interventions.

Purpose of the Study:

• To employ machine learning techniques to identify variables strongly correlated with the aggravation of pressure injuries.
• To analyze patient data to uncover specific risk factors contributing to PI worsening.
• To develop a predictive model for PI aggravation.

Main Methods:

• An observational study involving 71 patients with Stage 2 PIs from May 2018 to June 2021.
• Patients were categorized into an aggravated group (Group A) and a healed group (Group B).
• A Random Forest with hyperensemble approach, analyzing 24 factors, was utilized to determine variable importance via mean decrease accuracy.

Main Results:

• Machine learning analysis identified serum albumin, Braden Scale score, hemoglobin, wound size, serum blood urea nitrogen, body mass index, serum protein, and serum creatinine as highly associated with PI aggravation.
• Conversely, end-stage renal disease, sex, and myocardial infarction showed less association with PI worsening.
• Group A comprised 14 patients, while Group B included 57 patients.

Conclusions:

• The developed PI prediction model shows potential as a valuable tool for PI prevention.
• These identified factors can inform the development of targeted strategies to mitigate the risk of pressure injury aggravation.
• Further research can refine predictive accuracy and clinical applicability.