Including deaths when measuring health status over time

Area of Science:

• Health Services Research
• Biostatistics
• Epidemiology

Background:

• Measuring longitudinal health status is complicated by mortality.
• Deaths represent known health states, unlike typical missing data.
• Existing health measures lack defined values for deceased individuals.

Purpose of the Study:

• To evaluate eight distinct strategies for incorporating mortality data into health status analyses.
• To determine how different methods of handling deaths affect the interpretation of health changes over time.
• To provide recommendations for selecting appropriate methods based on study objectives.

Main Methods:

• Analysis of three health status measures from two distinct datasets.
• Examination of eight strategies, ranging from discarding deaths to using mortality as the sole measure.
• Computer simulations to thoroughly investigate the impact of mortality on health trend analyses.

Main Results:

• Strategies assigning less weight to deaths showed more positive health trends, potentially favoring groups with higher mortality.
• Strategies assigning greater weight to deaths demonstrated more negative health trends, favoring groups with lower mortality.
• The chosen method significantly influences the perceived change in health status over time.

Conclusions:

• The optimal strategy for incorporating deaths depends on the specific goals of the health intervention or study.
• For health promotion initiatives, recoding health variables to predict future health probability (e.g., 2-year survival) is recommended.
• Careful consideration of mortality handling is essential for accurate longitudinal health status assessment.