Discovering associations between problem list and practice setting.

Area of Science:

• Health Informatics
• Clinical Decision Support
• Natural Language Processing

Background:

• The Health Information Technology for Economic and Clinical Health Act (HITECH) promotes electronic health records (EHRs) for improved patient outcomes.
• Meaningful Use criteria for EHRs emphasize problem lists for patient health management.
• Leveraging problem lists for tailored care across diverse practice settings remains a challenge due to unestablished associations.

Purpose of the Study:

• To investigate the associations between patient problem lists and clinical practice settings.
• To develop a method for generating prioritized problem lists relevant to specific practice environments.
• To explore the potential for NLP and topic modeling in uncovering these associations.

Main Methods:

• Utilized clinical documents from Mayo Clinic patients.
• Applied natural language processing (NLP) and topic modeling (Latent Dirichlet Allocations - LDA).
• Normalized problem lists and practice settings, employing statistical tests (chi-squared), TF-IDF, and enrichment analysis for concept selection. Evaluated using 5-fold cross-validation.

Main Results:

• Developed a method to generate prioritized problem lists associated with specific practice settings.
• Achieved high performance in practice setting prediction: recall up to 0.931, precision up to 0.931, and F1 score up to 0.931 at k=10.
• Demonstrated the effectiveness of LDA in modeling relationships between problem concepts and practice settings.

Conclusions:

• This study is the first to explore the association between problem lists and hospital practice settings.
• The findings provide a foundation for utilizing these associations to deliver more tailored patient care.
• Future research will focus on practical applications of these discovered associations for enhanced clinical decision-making.