Caregivers Attitude Detection From Clinical Notes.

Area of Science:

• Natural Language Processing
• Healthcare Informatics
• Clinical Psychology

Background:

• Caregiver attitudes significantly influence healthcare quality and disparities.
• Clinical notes contain specialized, ambiguous language requiring domain expertise.
• Negative language in notes doesn't always indicate a negative attitude.

Purpose of the Study:

• To address the challenge of detecting caregiver attitudes from clinical notes.
• To leverage advanced language models for analyzing clinical text.
• To improve patient satisfaction and identify care provider well-being issues.

Main Methods:

• Annotation of MIMIC clinical notes.
• Training state-of-the-art language models from Hugging Face.
• Evaluation in zero-shot, few-shot, and fully-trained scenarios within the Neonatal Intensive Care Unit.

Main Results:

• The RoBERTa model achieved an F1-score of 0.75 in identifying caregiver attitudes.
• Demonstrated effectiveness of language models in understanding nuanced clinical language.
• Successfully applied to the Neonatal Intensive Care Unit setting.

Conclusions:

• Automated detection of caregiver attitudes from clinical notes is feasible.
• This approach can enhance patient satisfaction and support care provider well-being.
• Potential for early detection and prevention of caregiver syndromes like burnout.