Pressure Ulcer Injury in Unstructured Clinical Notes: Detection and Interpretation

Area of Science:

• Nursing Quality Metrics
• Clinical Informatics
• Natural Language Processing

Background:

• Hospital-acquired pressure ulcer injury (PUI) is a key nursing quality indicator.
• Existing PUI detection methods rely on structured data and scales like Braden.
• Unstructured clinical notes, rich in information, have been underutilized for PUI detection.

Purpose of the Study:

• To develop an automated system for detecting pressure ulcer injury (PUI).
• To leverage unstructured clinical notes using a novel negation-detection algorithm.
• To improve the accuracy and efficiency of PUI detection in healthcare settings.

Main Methods:

• A novel negation-detection algorithm was applied to unstructured clinical notes.
• Text features generated by the algorithm were used for PUI detection.
• Performance was evaluated using logistic regression, random forests, and neural networks on the MIMIC-III dataset.

Main Results:

• The proposed algorithm significantly improved PUI detection accuracy compared to methods without negation detection.
• Key classifier features showed substantial overlap with established clinical attributes of PUI, enhancing interpretability.
• The framework demonstrated potential for on-demand, low-cost PUI detection.

Conclusions:

• Automated PUI detection using unstructured clinical notes with negation detection is feasible and effective.
• This approach can enhance the accuracy of PUI quality metrics.
• The method can reduce nursing assessment burden by automating the identification of most PUI cases.