Drivers of the cost of liver transplantation: Is it the donor, the recipient, or the center?1

Liver transplantation is now well accepted as the most effective modality for the management of end-stage liver disease. As debates over its effectiveness have diminished, most of the current controversy has focused on issues of equity of organ distribution, appropriate prioritization, and cost effectiveness. With the implementation of the Model for End-Stage Liver Disease (MELD)–based allocation scheme, the sickest-first prioritization schema has ameliorated equity concerns at the local organ distribution level. However, there is still substantial regional variation in MELD scores at transplant across the United Network for Organ Sharing (UNOS) regions, so much of the allocation debate has focused on whether broader regional distribution should be implemented to equalize access across the nation. Similarly, increasing MELD scores at the time of transplant have raised questions about the increasing costs of transplantation and particularly about the impact of renal impairment, the need for renal replacement therapy, and the increased rates of liver-kidney transplantation. With many transplants being paid on a capitated single-case basis, the need to understand the costs of transplantation and the drivers of these costs is critically important. DRI, donor risk index; ICU, intensive care unit; LOS, length of stay; MELD, Model for End-Stage Liver Disease; UNOS, United Network for Organ Sharing. In this issue of Liver Transplantation, Washburn et al.1 provide important data to help us with this debate. By comparing the length of stay (LOS) and the predictors of LOS at 2 large transplant programs in different UNOS organ distribution areas, they show us not only which drivers of cost are similar across the UNOS regions but also the differences that exist between areas with different degrees of donor scarcity. Therefore, the first interesting observation from their article concerns the differences at the time of transplant between the 2 centers. The recipients at center A had higher MELD scores at the time of transplant, which were driven mostly by higher creatinine levels and rates of pretransplant dialysis, and tended to have a higher rate of liver-kidney transplantation. There was also significantly higher use of liver donor liver transplantation, there were more organ imports at regional and national levels, and the donor age and donor risk index (DRI) were higher at center A despite the higher MELD scores at transplant. Even though patients were sicker at center A versus center B, LOS post-transplant, which the authors chose as their measure of cost, was similar. This supports previous data showing that center differences are quite important in determining LOS.2 At both centers, the MELD score was perhaps the largest determinant of LOS, with a significant increase in LOS per additional MELD point at the time of transplantation. The impact was a 30% to 40% increase in LOS per 10-point increase in the MELD score. Further evidence of the importance of Center effects is seen in their multivariate model where the Hazard Ratio was 0.88 or 12% difference in LOS after controlling for MELD and other factors. Outside of MELD, all other predictors of cost were different at the 2 centers, including both donor factors (age, weight, cold ischemia time, and nonlocal donors) and recipient variables (age, gender, and retransplantation). In addition to the strong center effects on overall resource utilization, it is interesting that over the time period of the study, center A had similar LOS, DRIs, and MELD scores at transplant, whereas at center B, LOS declined despite increases in the MELD score and DRI over time. This argues for the important impact of a learning curve for managing the sickest patients in order to minimize LOS. Whether this also reflects improved donor-recipient matching at center B is unclear, as there was no correlation between MELD and DRI in their analyses. One would expect a negative correlation between DRI and MELD (ie, high-DRI organs are going into lower MELD score patients), although this practice may have evolved over time, masking an overall correlation. Despite the lack of correlation between the MELD score and DRI, the lower impact of DRI on LOS in comparison with recipient factors leads me to believe that the highest risk organs were not being put into critically ill patients. Additionally, it is likely that only the rare marginal organ that contributes to early graft dysfunction or nonfunction would increase cost. One interesting finding was the marked impact of the transplant number on LOS only at center B, despite similar numbers of retransplants at each center. Retransplantation has been shown to be a strong driver of cost in most previous analyses2, 3 and is seen in the combined model; thus, there may have been a lack of power to see this difference at center A, where the recipients were sicker and the overall LOS was higher for primary transplants. A hazard ratio of 1.45 indicates that a retransplant has an impact on LOS equivalent to a greater than 10-point increase in the MELD score at the primary transplant. What conclusions can we draw from this article? Although the MELD score does not appear to have a marked impact on posttransplant survival, accumulating data suggest that high MELD scores at transplant are associated with higher costs at a given center.4 This has implications for managed care contracting and points to the need to explore tiered contract pricing. Second, center-specific differences in resource utilization after transplantation are marked. These differences can overcome the impact of the MELD score on LOS. The declining LOS at center B despite increases in the MELD score and DRI suggests that experience and best practices can lead to reduced costs, particularly over time. Finally, the marked disparities that currently exist in terms of the MELD score at transplant across the country have broad implications. In addition to the impact of geographic differences on patients with respect to the risk of death before transplantation, the differences in MELD scores at transplant have an economic impact on transplant centers. As a transplant community, we should strive to ameliorate these differences over time. The major weakness of this article is the use of LOS as a surrogate for cost. Although LOS is accepted as a surrogate of cost, it has been shown in previous articles that resource utilization per day does vary between centers.2 It is also likely that the intensity and thus cost of care per day vary between the sickest patients and those that are less sick. This may lead to even larger differences in cost for the higher MELD patients and make the similarities in LOS between the 2 centers less meaningful if resource utilization per day was higher at center A. At the very least, I would have liked to see LOS broken down into intensive care unit (ICU) LOS and non-ICU LOS, as the cost of ICU care and renal replacement therapy can have a marked impact on the total cost. This additional data would likely have increased the impact of pretransplant MELD on cost. None of their models include the cost of readmission or pretransplant hospitalization, both of which are likely higher for the sickest patients. Given that discharge to a skilled nursing facility or rehabilitation unit was higher for older patients and those with higher MELD scores, the bias would be toward a greater impact of the MELD score and age on cost. Pretransplant costs are also higher in this group. Thus, their analyses greatly underestimate the societal cost of the higher MELD score at transplant seen in some regions. Finally, it would also be interesting to see speculation on what are thought to be the major determinants of the center effects on cost, particularly as both centers have mature, large-volume transplant programs. This would allow the adoption of cost-saving approaches to transplanting the sickest patients.