Transplant center outcome is being increasingly scrutinized, so it is critical to have a consistent approach to data analysis. Standard practice has been to include death with function as a graft loss. But doing so may obscure other important risk factors and make it difficult to compare centers. To document this data analysis problem, we studied half-life and risk factors for long-term graft survival in 2230 kidney transplant recipients who had ≥1 year of function. Four separate Cox regression analyses were done, differing in how death with function is considered: death with function considered a graft loss (analysis 1); all deaths censored (analysis 2); definitively non-transplant-related deaths censored, i.e., deaths from infection, malignancy, or cardiac problems analyzed as a graft loss (analysis 3); and definitively non-transplant-related as well as cardiac deaths censored (analysis 4). For each analysis, variables included immunosuppressive protocol, age at transplant, donor source, diabetes, gender, transplant number (primary vs. retransplant), and HLA ABDR mismatches (0 vs. ≥1 mismatch). There were important differences in risk factors, depending on how death with function is considered. For example, when all deaths are considered a graft loss, age >50, cadaver donor source, diabetes, retransplantation, and >0 antigen mismatch were found to be risk factors for long-term graft survival. However, when all deaths are censored, age >50, cadaver donor source, retransplantation, and diabetes were no longer risk factors. In fact, age >50 was associated with significantly better graft survival when all deaths are censored (analysis 2), suggesting that the increased graft loss seen in these patients is nonimmunologic. Similarly, t 1/2 is markedly different for different patient subgroups depending on how death with function is considered. For example, nondiabetic living donor (non-HLA- identical) recipients >50 have a t 1/2 of 9±1 years when death with function is considered a graft loss; for the same group, t 1/2 is 62±28 years when death with function is considered. For diabetic patients ≤50, when death with function is considered a graft loss, t 1/2 is 9±0.9 years for living donor recipients and 7±0.7 years for cadaver donor recipients. For the same patients, when death with function is censored, t 1/2 is 27±4 years for living donor recipients and 24±4 years for cadaver donor recipients. Our analysis suggests that death with function needs to be considered in analyzing kidney transplant outcomes. It is important for transplant centers to develop a consistent approach to this data analysis problem, so that results can be meaningfully compared. We propose that all transplant graft survival data be presented in two ways, with and without death censored.