A clinical use case to evaluate the i2b2 Hive: predicting asthma exacerbations

Area of Science:

• Clinical informatics
• Biomedical data science
• Respiratory medicine research

Background:

• The i2b2 Hive platform facilitates clinical data analysis.
• Asthma exacerbation prediction requires comprehensive clinical data.
• Extracting specific data like smoking status can be challenging.

Purpose of the Study:

• To assess the i2b2 Hive's utility for querying, visualizing, and extracting clinical data.
• To evaluate the performance of the i2b2 NLP cell for extracting smoking status.
• To analyze the feasibility of using i2b2 for asthma exacerbation prediction studies.

Main Methods:

• A use case from the i2b2 airways diseases driving biology project (asthma exacerbations prediction) was selected.
• Structured clinical data (diagnosis, demographics, vitals) and unstructured clinical notes were analyzed.
• The i2b2 NLP cell was employed to extract smoking status from history and physical exam reports.

Main Results:

• The i2b2 workbench could query approximately half of the required clinical data elements.
• A commercial database management system client was necessary for querying the remaining structured data.
• Automated extraction of smoking status achieved a mean precision of 0.79 and a mean specificity of 0.90.

Conclusions:

• The i2b2 Hive demonstrates potential for clinical data querying and extraction in respiratory research.
• The integrated NLP capability shows promise for extracting key information like smoking status from clinical notes.
• Further optimization may be needed for comprehensive data querying within the i2b2 environment.