Inverse intensity weighting in generalized linear models as an option for analyzing longitudinal data with triggered...

Area of Science:

• Epidemiology
• Biostatistics
• Health Services Research

Background:

• Longitudinal studies with irregularly observed categorical outcomes pose analytical challenges.
• Generalized linear models (GLMs) have limitations regarding missing data and observation contribution.
• Participant loss to follow-up can introduce bias in longitudinal data analysis.

Purpose of the Study:

• To present a triggered sampling design and inverse intensity weighting analysis for longitudinal studies.
• To address biases from missing data and unequal observation contributions in epidemiological studies.
• To illustrate the utility of these methods using data from the Longitudinal Examination of Attitudes and Preferences (LEAP) Study.

Main Methods:

• Utilized a triggered sampling design, collecting data after specific health status declines.
• Employed inverse intensity weights calculated from an Anderson-Gill recurrent-event regression model.
• Applied these methods within a generalized linear model (GLM) framework.

Main Results:

• The combination of triggered sampling and inverse intensity weighting effectively mitigates bias in longitudinal data.
• This analytical approach helps equalize the contributions of irregularly spaced observations.
• The study demonstrates the practical application and benefits of these methods in real-world health research.

Conclusions:

• Triggered sampling and inverse intensity weighting offer a robust solution for analyzing longitudinal data with irregular observations.
• These methods enhance the accuracy and reliability of findings in epidemiological and health services research.
• The approach facilitates a more comprehensive assessment of longitudinal study designs and their impact on data analysis.