A hot-deck multiple imputation procedure for gaps in longitudinal recurrent event histories

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Longitudinal studies often have missing data in event histories.
• Gaps complicate the analysis of recurrent and terminal events, such as asthma episodes and death, or menstrual cycles and menopause.
• Existing methods for handling missing data gaps have significant limitations.

Purpose of the Study:

• To propose a novel regression-based hot-deck multiple imputation method for imputing missing data in longitudinal studies.
• To accurately estimate the onset time of a marker event and the duration to a terminal event.
• To address limitations of simple gap-filling or case-dropping methods.

Main Methods:

• A regression-based hot-deck imputation technique is proposed.
• Predictive mean matching is employed to integrate longitudinal data and final event times.
• Multiple imputation is utilized to account for imputation uncertainty.

Main Results:

• The method effectively imputes information within data gaps.
• It improves the estimation of marker event onset and subsequent duration.
• The procedure was successfully applied to menopausal transition data.

Conclusions:

• The proposed hot-deck imputation method offers a robust solution for missing data in longitudinal studies with recurrent and terminal events.
• This approach enhances the accuracy of timing and duration analyses.
• The method shows promise for applications in women's health and other fields.