Analyzing primary care data to characterize inappropriate emergency room use

Area of Science:

• Health Services Research
• Biomedical Informatics
• Public Health

Background:

• Primary care data in Ontario offers a comprehensive biopsychosocial patient profile.
• Secondary analysis of this data is a recent development.
• Understanding factors influencing emergency room (ER) use is crucial for healthcare system optimization.

Purpose of the Study:

• To investigate the relationship between biopsychosocial concepts from primary care data and inappropriate ER use.
• To demonstrate a novel method for analyzing de-identified primary care records.
• To identify specific patient factors associated with high ER utilization.

Main Methods:

• Extraction of de-identified primary care data from Ontario.
• Application of natural language processing (NLP) to extract Unified Medical Language System (UMLS) codes.
• Statistical analysis using logistic regression to correlate UMLS codes with ER use patterns.

Main Results:

• Pain and mental health concepts were statistically significant predictors of inappropriate ER use.
• The NLP and statistical approach successfully identified key biopsychosocial factors.
• The study highlights the potential of primary care data for health services research.

Conclusions:

• Primary care data, when analyzed with NLP, can reveal significant associations with healthcare utilization patterns.
• Pain and mental health are key areas to address to potentially reduce inappropriate ER visits.
• This methodology can be adapted for analyzing system use in diverse healthcare contexts.