Mixed effects logistic regression models for multiple longitudinal binary functional limitation responses with...

Area of Science:

• Gerontology
• Biostatistics
• Longitudinal Data Analysis

Background:

• Analyzing longitudinal functional limitations in older adults presents challenges due to informative drop-out and confounding baseline outcomes.
• Understanding heterogeneity in risk factor associations is key for effective interventions.

Purpose of the Study:

• To investigate heterogeneity in functional limitation outcomes using data from the Longitudinal Study of Aging (LSOA).
• To develop a statistical model that accounts for informative drop-out and baseline confounding in longitudinal functional limitation data.

Main Methods:

• Utilized an extended nested random effects logistic model with autoregressive structure to analyze longitudinal functional limitation data.
• Incorporated shared random effects to model drop-out and baseline outcomes, estimating parameters via maximum likelihood with numerical integration.
• Assessed model robustness by varying assumptions regarding random effects, drop-out, and baseline outcome inclusion.

Main Results:

• Observed significant heterogeneity in associations between functional limitation outcomes and risk factors like prior limitations and physical activity.
• Found less heterogeneity in time-level random effects variance components across outcomes and time.
• Demonstrated that omitting baseline outcomes leads to bias under an autoregressive structure.

Conclusions:

• The proposed shared parameter selection model effectively addresses informative drop-out and baseline confounding in longitudinal functional limitation analysis.
• Accounting for baseline functional status is essential to prevent bias in longitudinal aging studies.
• The findings underscore the complexity of functional limitation trajectories and the need for sophisticated statistical approaches.