Generalized functional linear model with a point process predictor.

Area of Science:

• Statistics
• Biostatistics
• Health Informatics

Background:

• Point process data are increasingly prevalent, particularly in electronic health records (EHR).
• Analyzing the association between point process predictors and scalar responses is crucial but challenging with existing methods.
• Current generalized functional linear regression models do not accommodate point process predictors.

Purpose of the Study:

• To propose a novel generalized functional linear regression model specifically designed for point process predictors.
• To address the limitations of existing methods that cannot handle point process data.
• To enable robust analysis of associations involving complex temporal event data.

Main Methods:

• Development of a joint modeling framework combining a log-Gaussian Cox process for the point process predictor and a generalized linear regression for the outcome.
• Implementation of a new algorithm for efficient model estimation using Gaussian variational approximation.
• Extensive simulation studies to validate the proposed method's performance.

Main Results:

• The proposed model demonstrates superior performance in simulation studies compared to existing methods.
• The method effectively handles the complexities of point process predictors in regression analysis.
• Successful application to a real-world electronic health records dataset from intensive care unit patients.

Conclusions:

• The novel generalized functional linear regression model provides an effective approach for analyzing point process data.
• This method advances the statistical toolkit for researchers working with electronic health records and similar data types.
• The proposed approach offers a valuable tool for understanding complex associations in health data.