Retrospectively Collected EQ-5D-5L Data as Valid Proxies for Imputing Missing Information in Longitudinal Studies.

Area of Science:

• Health Outcomes Research
• Patient-Reported Outcomes

Background:

• Missing data in longitudinal studies using the EQ-5D-5L instrument presents significant challenges.
• Validated methods for handling missing EQ-5D-5L data are scarce in existing literature.

Purpose of the Study:

• To investigate the feasibility of using retrospective EQ-5D-5L data collection as a proxy for missing longitudinal data.
• To compare the performance of retrospective data collection against longitudinal single imputation.

Main Methods:

• Participants prospectively completed the EQ-5D-5L at 3 months postdischarge.
• Retrospective EQ-5D-5L data were collected at 6, 9, or 12 months postdischarge.
• Intraclass correlation coefficients and weighted kappa statistics assessed concordance between in-the-moment, retrospective, and imputed data.

Main Results:

• High agreement (weighted kappa: 0.72-0.95) was found for mobility, self-care, and usual activities dimensions between in-the-moment and retrospective EQ-5D-5L data.
• Concordance for utility indices (ICC: 0.79-0.81) and quality-adjusted life-years (ICC: 0.93-0.96) was good to excellent.
• Retrospective collection outperformed longitudinal single imputation.

Conclusions:

• Retrospective EQ-5D-5L data collection demonstrates high concordance with in-the-moment data.
• This method is a viable alternative for imputing missing EQ-5D-5L data in longitudinal studies.
• Further research is needed to generalize these findings across diverse populations and disease areas.