Physician's information customizer (PIC): using a shareable user model to filter the medical literature

Area of Science:

• Medical Informatics
• Information Science

Background:

• Physicians require efficient information management for medical practice.
• Current medical information tools lack personalization, hindering user acceptance and effectiveness.
• User models representing physician information needs are rarely integrated into these tools.

Purpose of the Study:

• To develop the Physician's Information Customizer (PIC), a system for generating shareable user models.
• To demonstrate the utility of user models in customizing medical informatics applications.
• To enhance the filtering and ranking of medical literature based on individual physician preferences.

Main Methods:

• Developed PIC to elicit stable physician attributes (specialty, research focus, interests, patient characteristics, practice locale).
• Integrated PIC-generated user models into a medical informatics application.
• Custom-filtered and ranked articles from Medline based on user model data.

Main Results:

• PIC successfully generates shareable user models from easily acquired physician attributes.
• Preliminary evaluation showed PIC accurately ranks 66% of articles according to user preferences.
• Demonstrated the feasibility of using user models to filter relevant articles from large collections.

Conclusions:

• User models can significantly improve the customization and effectiveness of medical information management tools.
• PIC provides a feasible approach to personalizing medical literature retrieval.
• Further research is needed to optimize the user model and filtering algorithms.