Evaluating robustness of a generalized linear model when applied to electronic health record data accessed using an...

Area of Science:

• Environmental Health
• Biomedical Informatics
• Epidemiology

Background:

• The Integrated Clinical and Environmental Exposures Service (ICEES) offers regulatory-compliant access to integrated clinical and environmental data.
• Existing validations exist for ICEES, but regulatory constraints on its OpenAPI may affect data integrity for complex analyses.
• Multivariate analysis is crucial for understanding multifactorial health conditions like asthma exacerbations.

Purpose of the Study:

• To investigate the robustness of the ICEES OpenAPI for multivariate analysis.
• To compare the impact of data loss from the ICEES OpenAPI on predictive modeling of asthma exacerbations.

Main Methods:

• A comparative analysis was performed using a generalized linear model (GLM).
• GLM was applied to both ICEES OpenAPI data and constraint-free source data.
• Predictive factors for asthma exacerbations were examined.

Main Results:

• Key predictors of asthma exacerbations identified by both datasets included sex, prednisone, race, obesity, and airborne particulate exposure.
• Data loss through the ICEES OpenAPI impacted the quality of GLM model fit.
• The impact of data loss was specific to certain interaction terms within the model.

Conclusions:

• The ICEES OpenAPI is capable of supporting multivariate analysis.
• Users must be aware of potential data loss due to regulatory constraints, which can affect model performance.
• Further considerations are needed to mitigate data loss for optimal analytical outcomes.