Understanding Primary and Secondary Concerns from Patient Portal Messages through Clinical Data Annotation, Analysis,...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Natural Language Processing

Background:

• Efficient patient portal message (PPM) management is crucial for patient-centered care.
• Understanding patient concerns within PPMs is vital for timely intervention.
• The increasing volume of PPMs poses a challenge to healthcare providers.

Purpose of the Study:

• To annotate and analyze patient concerns in 2,239 patient portal messages.
• To develop and evaluate an AI-driven approach for automated patient concern identification and classification.
• To enhance the accuracy of primary patient concern identification using integrated AI models.

Main Methods:

• Manual annotation of 2,239 patient portal messages to identify primary and secondary concerns.
• Leveraging pretrained language models for binary classification to detect all patient concerns.
• Implementing multi-class classification with integrated convolutional neural networks for primary concern identification.

Main Results:

• The AI approach successfully identified and classified patient concerns within PPMs.
• Binary classification effectively discerned the presence of patient concerns.
• Multi-class classification, enhanced by CNNs, showed significant potential in identifying primary patient concerns.

Conclusions:

• AI demonstrates significant potential in managing the growing volume of patient portal messages.
• Automated analysis facilitates prompt identification and addressing of patient healthcare needs.
• This approach can lead to more effective and timely medical interventions, improving patient care.