Irish WD, Sherrill EH. An econometric approach to generating population cost estimates for event-time data: an example using renal transplant failure data. Poster presented at the International Society for Pharmacoeconomics and Outcomes Research Annual Meeting; 2003. [abstract] Value Health. 2003 May; 6(3):178-9.

OBJECTIVES: Generation of population cost estimates for event-time data requires a sophisticated approach to account for the probability of incurring the event over time. This research investigated the relationship of donor and recipient factors with all-cause renal graft failure and estimates of the cost of failure from a Medicare perspective.


METHODS: A two-part econometric approach was used to determine Medicare claims attributable to all-cause graft failure (including deaths). First, type-specific hazard functions were estimated with Cox proportional hazards models. Using data from USRDS for primary renal transplants in adults between 1993-1998, we developed separate predictive models for transplants from living and cadaveric donors after identifying covariates associated with graft loss. Models were stratified by transplant year and included donor and recipient characteristics plus clinical variables including immunosuppression therapies. Next, the log-transformed costs for patients who experienced the event were modeled against the covariates to estimate costs specifically associated with failure at a given time point. For patients who did not experience the event, predicted costs were generated based on the model coefficients and individual covariates. Retransformation of the log costs included an adjustment using residual smearing. The expected Medicare claims associated with graft failure were calculated by combining the estimated cumulative hazards of graft failure with the smeared estimate of the claims associated with the event.


RESULTS: For living donor transplants (N = 5831), expected Medicare claims attributed to renal transplant graft failure were approximately $13,073 ± $14,318 (median = $8,933; range = $560–$191,169) at 3 years post-transplant. For cadaveric donor transplants (N = 22,941), the expected Medicare claims were approximately $15,075 ± $13,149 (median = $11,540; range = $699–$214,184).


CONCLUSIONS: These estimates provide groundwork for population-based studies to address the cost-effectiveness of various treatments to delay or prevent graft loss. The method allows policymakers to assess population costs after taking into account the probability of event occurrence.

Share on: