Multiple imputation methods for longitudinal blood pressure measurements from the Framingham Heart Study

Area of Science:

• Genetics
• Biostatistics
• Longitudinal Studies

Background:

• Missing data is a significant challenge in longitudinal studies, potentially leading to biased or inefficient analyses.
• Missingness can be an indicator of adverse outcomes, particularly concerning in genetic research.
• Excluding incomplete observations can compromise the integrity of genetic analyses.

Purpose of the Study:

• To evaluate the impact of missing data on genetic analyses.
• To compare case-wise deletion with propensity score and regression imputation methods in SAS.
• To assess the efficiency and potential bias introduced by different missing data handling techniques.

Main Methods:

• Comparison of case-wise deletion with two multiple imputation methods (propensity score and regression) in SAS.
• Utilized both real and simulated datasets for analysis.
• Focused on genetic analyses within longitudinal study contexts.

Main Results:

• Propensity score and regression imputation methods yielded results similar to case-wise deletion for both real and simulated data.
• Estimates of heritability were substantially lower with case-wise deletion and the two imputation methods compared to complete data in simulated datasets.
• A potential bias was observed when missingness patterns were correlated within families.

Conclusions:

• Standard imputation methods may produce biased heritability estimates if missingness is correlated within families.
• The choice of method for handling missing data is critical in genetic analyses.
• Further research into imputation methods that account for within-family correlations is warranted.