Strategies for handling large-scale missing blood markers in longitudinal studies

Project Summary

Collecting blood markers in large, long-term studies like the Canadian Longitudinal Study on Aging (CLSA) is critical. However, the COVID-19 pandemic disrupted blood sample collection, resulting in missing data for about two-thirds of CLSA participants during the 2018-2020 follow-up. Missing data, if not managed correctly, can bias study results; commonly used techniques, such as excluding incomplete cases, often yield unreliable findings. While machine learning methods have shown promise in addressing missing data, they may not fully capture changes over time essential in longitudinal research. Our goal is to provide a comprehensive strategy of variable selection for imputation models and to develop effective methods for managing substantial amount of missing blood biomarker data in longitudinal studies.