Statistical Analysis of Highly Granular Longitudinal Data on Hemodialysis Patients

• Activity Number: 49
• Type: Invited
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: ENAR
• Author(s): Maurice A. Brookhart*+
• Companies: University of North Carolina at Chapel Hill

Abstract

Patients receiving chronic hemodialysis are at very high mortality risk and suffer from frequent infections and vascular events. These patients typically receive three sessions of hemodialysis per week at which times labs are drawn, clinical measurements are recorded, and treatments are administered. Many existing analyses of treatment effects in hemodialysis patients use aggregate data in which cumulative doses of medications over prolonged periods of time are associated with risk of outcomes of interest. Given the dynamic health status of these patients, these approaches are subject to strong selection bias and confounding. Increasingly, rich clinical data are available on hemodialysis patients that provide detailed treatment and clinical measurements recorded at each dialysis session. We discuss approaches and challenges to the analysis of these data at the dialysis session level. Using treatment for anemia as an example, we compare results of conventional study of aggregated treatment data to an analysis of granular, dialysis-session level data.