Timing is everything: Using time-varying binary indicators for evaluating post-transplant risk factors.

Risk factors often emerge after transplantation, yet many analyses treat them as fixed at the time of transplant, producing inaccurate or counterintuitive results due to immortal time bias. Using infection‑related hospitalizations after heart transplantation as an example, we show how incorporating a time‑varying binary indicator (TVBI) in a Cox model for such hospitalizations properly aligns the timing of the exposure with survival follow‑up and yields more credible effect estimates. We summarize key assumptions, diagnostic checks, and limitations of the TVBI approach, and highlight complementary visualization tools. Together, these methods offer a clear framework for estimating the impact of post‑transplant exposures on survival after heart and lung transplantation.

Duke Scholars

Author Megan Lee Neely Biostatistics & Bioinformatics, Division of Biostatistics