The Economic Impact of the Utilization of Liver Allografts with High Donor Risk Index


*Corresponding author: David A. Axelrod,


The disparity between the organ supply and the demand for liver transplantation (LT) has resulted in the growing utilization of ‘marginal donor’ organs. While economic outcomes for subsets of ‘marginal’ organs have been described for renal transplantation, similar analyses have not been performed for LT. Using UNOS data for 17 710 LTs performed between 2002 and 2005, we assessed the relationship between recipient model for end-stage liver disease (MELD) score, organ quality as defined by donor risk index (DRI, Feng et al. 2005) and hospital length of stay (LOS). Single-center cost-accounting data for 338 liver transplants were then analyzed with a multivariate linear regression model to determine the estimated cost associated with a day of LOS. Overall, 8.4% of donor organs were classified as high risk (DRI > 2–2.5) and 1.9% as very high risk (DRI > 2.5). In the lowest MELD group (0–10), the LOS difference between ‘ideal’ donors (DRI < 1.0) and very high risk (DRI > 2.5) was 10.6 days which was associated with an estimated incremental cost of $47 986. For patients with MELD >35, the average LOS increased from 23.2 to 41.8 days when very high DRI donors were used, resulting in an estimated increase in cost of nearly $84 000. We conclude that the use of marginal liver grafts results in increased hospital costs independent of recipient risk factors.


The cost of liver transplantation (LT) continues to rise as a result of current allocation systems for deceased donor organs that require that the ‘sickest’ be transplanted first, while reimbursement is static or declining leading to significant financial risk for the nation's transplant centers (1–5). Furthermore, despite a reduction in wait-list mortality following the implementation of model for end-stage liver disease (MELD), patients continue to die awaiting an organ leading to national efforts to expand access to transplantation through the utilization of ‘marginal donors’ (6). Over the past decade, the number of LTs performed using allografts from donors older than 65 has doubled (5% to 9.8%) and the use of organs from donors after cardiac death (DCD) is 10-fold higher (0.3% to 4.0%) (7).

While no exact definition of expanded criteria donors (ECD) exists for liver allografts as has been defined for renal allografts, it is widely appreciated that a variety of donor factors have been associated with worse outcomes. The donor risk index (DRI) described by Feng and colleagues is one measure of organ quality (8). The DRI incorporates multiple aspects of donor quality including age, cause of death, race, height and DCD; as well as organ specific factors: partial or split allograft, location (local, regional or national sharing) and cold ischemic time. Increasing DRI has been strongly correlated with decreasing graft and patient survival. While these ‘marginal’ organs can be successfully used in the proper patient, they are at higher risk of graft dysfunction, graft failure and potentially decreased patient survival (9–12).

Analysis of the impact of donor factors on cost of LT has been largely limited to assessment of donor age. Allografts from donors >60 years old have been associated with a significant increase in resource utilization (2). However, little data are available on the financial impact or other significant donor risk factors on the cost of transplant which are included in the DRI. Furthermore, while it is intuitively clear that the use of marginal organs in profoundly ill recipients is likely to increase transplant costs, this relationship has not been closely examined.

The purpose of this study is to estimate the financial impact of increased donor risk factors on resource utilization following LT. Because no large, universal data source includes both cost and clinical data, this analysis combines cost-accounting data from a single academic medical center with national clinical and length of stay (LOS) data in order to estimate the impact of marginal donors on overall transplant cost.

Materials and Methods

Clinical outcome data

Clinical data for 17 710 liver transplants performed from 2002 to 2005 were analyzed using data from the United Network for Organ Sharing (UNOS). The following patients were excluded from this analysis: status 1 patients and live donor recipients. Donor data were abstracted and used to calculate the DRI as described by Feng et al. (8). Recipient data were used to determine the LOS posttransplant and to adjust for recipient characteristics in the multivariate model. The actual laboratory (calculated) MELD/PELD score was used in all analyses to assess the degree of physiological illness among transplant recipients.

Multivariate analyses were performed following the elimination of certain subgroups which might bias the analyses. Patients transplanted by MELD exemption were considered as one group at risk of higher than average costs given comorbid conditions (e.g. pulmonary hypertension). Because there is no specific variable for this group, patients were identified if their MELD score at transplant differed from their calculated MELD score by more than 1 standard deviation. This method identified 1016 potential recipients in this category, representing 5.7% of the total population. We also performed the multivariate analysis both with and without recipients of combined liver-kidney analyses and with and without adjustment for patient death within 30 days of transplant. Finally, we presented univariate data on DCD, split liver and donor following brain death (DBD) separately for both primary and retransplant recipients, respectively.

Statistical analyses, including one-way analysis of variance and chi-square analysis, were used to determine the impact of DRI and recipient characteristics on hospital LOS following transplant. LOS was defined as the duration of hospitalization from the day of transplant to the day of transplant discharge, excluding days prior to transplant. All MELD and DRI categories were determined a priori to reduce the risk of bias. The data were subsequently reanalyzed using a linear regression analysis to determine the independent impact of increasing DRI. To account for the natural right skew in LOS data, a logarithmic transformation was performed. All beta coefficients were then reconverted to allow easier interpretation. A variable was included for UNOS region to control for potential regional differences in the impact of organ quality. Unfortunately, center identified data were not available for use in this analysis.

Financial data

Hospital cost-accounting data were retrospectively reviewed for 338 consecutive adult liver transplant procedures performed at Northwestern Memorial Hospital from 2000 to 2005. Patients undergoing living donor transplants and those with fulminant hepatic failure were excluded from this analysis.

Patient and donor characteristics were determined from chart review and UNOS records for all patients. Cost-accounting data included all inpatient costs associated with the initial hospitalization. For the small number of patients retransplanted during their initial hospitalization, all costs were combined into a single analysis. Calculated MELD score was used in all analyses.

Multivariate linear regression was used to estimate the incremental cost of each additional day of LOS on overall hospital costs. Regressions were performed using both cost and logarithm of cost without significant differences in the resulting beta coefficient for LOS. Thus, the untransformed results are reported for ease of interpretation.

Combined analysis

The beta coefficient determined from the cost-accounting data was used as an estimate of the incremental resources associated with each additional day of hospitalization. This coefficient was then multiplied by the incremental LOS associated with very high DRI organs to estimate the incremental cost of utilizing these organs in recipients within given MELD strata.

IRB approval

This study was approved by the Institutional Review Boards of the Feinberg School of Medicine at Northwestern University and the Saint Louis University School of Medicine.


National MELD and DRI results

Recipient characteristics for patients undergoing deceased donor LT are summarized in Table 1. High-risk recipients with MELD scores greater than 30 constituted nearly 20% of transplants performed. There were statistically significant differences in demographic characteristics (age, gender and race) of liver transplant recipients across MELD categories, although it is unlikely that these small differences were clinically significant.

Table 1.  Characteristics of recipients by MELD category
 MELD category
(n = 836)(n = 5291)(n = 7528)(n = 1430)(n = 1609)
  1. MELD = model for end-stage liver disease.

Age (mean)41.2 (21.8)50.0 (13.7)51.3 (12.4)50.1 (12.9)49.8 (11.2)
Male (%)54.365.971.267.868.7
White (%)75.879.672.867.264.1
 Chronic liver disease78.285.774.781.183.9

Although the majority of donor organs fell into a low donor risk category, 8.4% of organs fell into a high-risk group with DRI between (2.0–2.5) and 1.9% into the very high-risk group with DRI greater than 2.5 (Table 2). These organs were more likely to come from non-white, DCD donors and split donors (p < 0.001). Other high-risk factors included a substantially greater number of regional or nationally shared organs (p < 0.001) with significantly longer cold ischemic time (p < 0.001). As shown in Figure 1, the high DRI organs were more likely to come from the extremes of age, either very young or very old. Over the period of this study, the percent of donor organs in the high-risk group (DRI > 2.0) was found to have increased over the period of analysis from 8.3% in 2002 to 12.8% in 2005 (p < 0.001), although LOS overall fell across MELD categories for patients receiving these DRI organs (Table 3).

Table 2.  Characteristics of donor organs by donor risk index category
 Donor risk index
(n = 1686, 11.5%)(n = 7337, 50.0%)(n = 4156, 28.3%)(n = 1233, 8.4%)(n = 274, 1.9%)
  1. DCD = donation after cardiac death; CIT = cold ischemia time.

DCD (%)
White (%)99.571.965.952.338.7
Split (%)
Regional/National (%)4.020.832.862.985.0
CIT (h)7.3 (3.1)7.6 (3.6)7.9 (3.8)8.5 (3.9)8.9 (3.4)
Height (cm)181.6 (6.1)172.1 (10.8)166.1 (16.2)155.6 (28.6)122.9 (46.8)
Figure 1.

Donor age characteristics of liver donors as categorized by donor risk index score.

Table 3.  Average length of stay (with DRI ≥ 2.0)
MELD categoryTransplant year
(n = 266, 8.3%)(n = 434, 10.2%)(n = 466, 10.2%)(n = 341, 12.8%)
  1. MELD = model for end-stage liver disease; DRI = donor risk index.

0–1026.3 (43.9)20.5 (33.8)14.5 (12.9)13.0 (6.5)
11–2021.2 (23.4)16.4 (17.7)18.6 (24.0)16.4 (13.0)
21–3023.3 (28.2)21.8 (29.0)20.6 (23.9)16.4 (16.1)
31–3534.4 (39.0)19.9 (15.5)28.2 (27.1)26.0 (17.6)
36+32.7 (27.3)39.3 (50.7)37.8 (47.0)25.1 (23.0)

As expected overall LOS increased as both recipient MELD score and donor DRI increased. This effect, however, was not confined to high MELD or high DRI organs. Nationally, increasing DRI organs were found to be associated with a significant increase in hospital LOS within each MELD group studied (p < 0.001) (Figure 2). The incremental LOS associated between the best organs (DRI < 1.0) and the worst organs (DRI > 2.5) ranged from 10.6 days for patients with MELD score <10 to 18.6 days for the patients with MELD scores greater than 35.

Figure 2.

Relationship between the recipient model for end-stage liver disease score, liver allograft donor risk index and hospital length of stay among liver recipients transplanted from 2002 to 2005.

The impact of DRI on hospital LOS remained largely consistent for various high-risk groups (Table 4–6). For DBD donors, in low MELD patients (<30), increasing DRI resulted in an extension of LOS from 12.7 days in low-risk donors to 28.1 days among the highest risk donors. The impact of DRI was even more profound in the high MELD patients (>30) in which very high DRI organs were associated with a doubling in the average LOS. For DCD donors, the DRI gradient was apparent only in the low MELD patients, perhaps reflecting the small number of DCD organs being used in high MELD patients. Finally, in patients undergoing retransplant, there was a dramatic increase in the LOS associated with very high DRI organs.

Table 4.  Impact of DRI on length of stay by transplant type (with and without previous transplant)
Transplant typeDonor risk index
  1. MELD = model for end-stage liver disease; DBD = donation after brain death; DCD = donation after cardiac death; LT = liver transplant; LKT = liver and kidney transplant.

MELD 0–30
 DBD whole LT12.714.415.916.928.1
 DCD whole LT13.717.419.0
 Split LT19.719.617.522.0
 LKT (except DCD LKT)18.316.916.738.6
MELD 31+
 DBD whole LT21.624.022.729.942.1
 DCD whole LT37.424.326.67.0
 Split LT22.
 LKT (except DCD LKT)20.422.033.526.0
Table 5.  Impact of DRI on length of stay by transplant type (primary transplants only)
Transplant typeDonor risk index
MELD 0–30
 DBD whole LT12.614.115.516.227.7
 DCD whole LT11.917.419.1
 Split LT19.718.617.022.0
 LKT (except DCD LKT)17.417.316.941.1
MELD 31+
 DBD whole LT20.623.821.930.942.1
 DCD whole LT41.125.420.07.0
 Split LT22.123.316.025.0
 LKT (except DCD LKT)20.822.131.127.1
Table 6.  Impact of DRI on length of stay by transplant type re-transplant receipients
Transplant typeDonor risk index
MELD 0–30
 DBD whole LT15.019.424.038.062.5
 DCD whole LT43.317.316.5
 Split LT20.054.522.7
 LKT (except DCD LKT)24.713.815.811.0
MELD 31+
 DBD whole LT31.625.330.216.5
 DCD whole LT11.019.535.3
 Split LT22.3
 LKT (except DCD LKT)19.121.547.916.0

Multivariate regression analysis was then performed to assess the independent impact of increasing DRI on the hospital LOS controlling for recipient characteristics and clustering by UNOS region (Table 7). As noted, patients transplanted by MELD exemption points other than HCC were excluded. In this analysis, when compared with donors with DRI 1.0–1.5, donors in the lowest risk group (DRI < 1.0) were associated with a 6.5% reduction in LOS (p < 0.001). In comparison, donors organs with a high DRI (2.0–2.5) were associated with a 9% increase in LOS and very high DRI donors (>2.5), which comprise the greatest risk, were associated with a 30% increase (p < 0.001 for both). The impact of this result was similar to that of female recipients and older recipients. These results did not differ when patients receiving a liver-kidney transplant were excluded nor when recipient death was included as an independent variable (data not shown).

Table 7.  Multivariate regression analysis
Variable% Increase in LOSp-Value
  1. DRI = donor risk index; MELD = model for end-stage liver disease; LOS = length of stay.

 0.0–1.0−6.5 (−10.3, −2.7)<0.001
 1.5–2.03.7 (0.9, 6.4)0.009
 2.0–2.59.0 (4.5, 13.4)<0.001
 2.5+29.7 (20.7, 38.7)<0.001
Recipient age (years)Reference 
 0–2419.3 (13.5, 25.1)<0.001
 25–34−2.9 (−10.0, 4.1)0.417
 35–443.2 (−0.8, 7.1)0.116
 55–647.7 (4.8, 10.5)<0.001
 65+7.0 (2.7, 11.3)0.002
Recipient gender
 Male−8.4 (−11.0, −5.9)<0.001
Recipient race
 Black−1.3 (−5.5, 2.8)0.532
 0–10−12.4 (−18.0, −6.7)<0.001
 11–20−13.4 (−16.1, −10.6)<0.001
 31–3541.1 (36.9, 45.4)<0.001
 36+77.6 (73.3, 81.8)<.001
Cause of liver disease
 Noncholestatic5.0 (1.6, 8.5)0.005
 Metabolic14.4 (7.9, 20.8)<0.001
 Malignancy−18.9 (−23.9, −13.9)<0.001
 Other20.6 (15.0, 26.2)<0.001
Previous transplant
 Yes19.8 (14.6, 24.9)<0.001

Institutional cost analysis

The demographic characteristics of the 338 patients transplanted at single academic medical center were similar to national data. The average age was 53 years and 66% were male. The mean calculated MELD score at transplant was 22, and 15% of the patients had MELD scores greater than 35. Forty-four percent were transplanted for HCC and 22% required combined liver/kidney transplants.

Donor characteristics also reflected national trends. The average age at donation was 37 years. A cerebral vascular accident was the cause of donor death in 26%, 9% died from anoxic injuries, while the remainder died as a result of trauma, CNS tumors or other causes. DCD donor livers represented 3% of the transplanted organs.

Average LOS in this population was 14 days. Multivariate analysis of perioperative hospital costs revealed three major cost drivers: hospital LOS, MELD score and a diagnosis of HCC. Overall, hospital costs were found to increase by $4527 per day of LOS. MELD score was associated with an increase in cost of $1138 per MELD point, while a diagnosis of HCC decreased hospital costs by $9674. The reduction in the cost of care for HCC patients reflects their relatively improved physiologic status made possible by MELD upgrades for patients with this malignancy. Reestimation of the cost per day of transplant after the exclusion of patients receiving a combined liver/kidney transplant was minimally changed to $4387 per day of LOS. In this data set MELD score was no longer predictive of overall cost once LOS was controlled for.

Combined analysis

To estimate the incremental cost of care associated with the use of high-risk organs, the incremental LOS associated with very high-risk organs (DRI > 2.5) and the increased cost of care associated with longer hospitalization were combined. As shown in Table 8, the estimated incremental costs associated with a longer LOS varied by MELD group studied. For low MELD patients, the organs with a DRI >2.5 compared to a DRI <1.0 can be expected to add nearly $50 000 to the cost of the transplant. For high-risk recipients (MELD > 35), this incremental cost may be as much as $84 000, which represents an increase of nearly 60% over the mean cost of transplant.

Table 8.  Estimated impact of highest DRI organs on overall hospital costs
MELD categoryLOS (mean days)Estimated increased cost
Low DRI (0.0–1.0)Highest DRI (2.5+)
  1. DRI = donor risk index; MELD = model for end-stage liver disease; LOS = length of stay.

0–1011.7 (7.2)22.3 (38.1)$47986
11–2012.2 (11.2)26.0 (28.5)$62473
21–3013.5 (14.0)29.0 (35.0)$70169
31–3519.5 (17.3)33.3 (22.2)$62473
36+23.2 (24.8)41.8 (53.4)$84202


LT remains the sole therapeutic option for patients with end-stage liver disease. Many patients continue to die from their illness while waiting for transplant, leading to important efforts to expand the number and use of available organs. Use of these organs can be expected to improve survival when used in appropriate patients (11,13,14). However, the cost of using marginal organs as defined by the DRI, appears likely to increase resource utilization, hospital LOS, and therefore, hospital costs. This trend is consistent across MELD categories and appears to increase with higher DRI.

Nationally, the severity of illness among patients reaching transplantation has been rising. Following the implementation of the MELD system of organ allocation, the number of recipients with a MELD score greater than 30 has increased from 10% to 14% (15). By transplanting patients most likely to die without a transplant, the MELD system has been very successful in achieving the goal of lowering wait-list mortality. Unfortunately, this rising severity of illness is also likely to increase the overall cost of LT, particularly in regions in which increased competition and demand for organs results in a higher MELD score at transplantation (1,5,16).

Although no exact definition exists for ‘marginal liver grafts’, clinical results utilizing donors with less than optimal donors have been gratifying. While older donors were initially rejected, recent large series have reported excellent outcomes in the nonhepatitis C population (12,14,17). In the non-HCV group, there was no demonstrable difference in survival between older donor livers and standard livers. Likewise, in carefully selected patients, livers from DCD donors can also be used successfully (11). Mateo and colleagues recently reported that in ‘low-risk patients’, use of ‘low-risk DCD livers’ in which the cold ischemic time was less than 10 h and warm ischemic time is less than 30 min, resulted in equivalent graft and patient survival rates to standard livers (10). The authors argue for targeted use of this new source of donor livers. However, transplant centers may need to adjust their clinic practice to permit safe use of these organs (e.g. decreasing cold ischemic time). Other markers of high donor risk including positive viral serologies, a history of high-risk behavior or the presence of neurologic malignancy may also need to be considered as they are not captured using the DRI to define a ‘extended’ or ‘marginal’ liver.

Our analysis suggests that although excellent outcomes can be achieved with marginal liver allografts, the overall cost of this care is likely to be significantly higher. As has been shown in the kidney literature with ECD allografts, the use of marginal donors may increase the upfront costs, as well as potentially increase the long-term cost of transplantation (18,19). In renal transplantation, increased costs reflect the prevalence of delayed graft function in the ECD organs (20). The cause of the longer LOS following LT demonstrated in this analysis is likely multifactorial. Marginal donors may have a higher incidence of primary nonfunction. This is particularly true of organs with longer ischemic times and those from anatomic variants. These grafts may also have primary dysfunction resulting in longer ICU stays, a greater requirement for blood products and resuscitation and increased risk of infection.

It is likely that this analytical approach underestimates the true incremental cost of the use of these high DRI organs. In their analysis, Feng and colleagues demonstrated that the use of organs with a DRI greater than 2.0, resulted in a reduction in 3-year graft survival of 20% (80% for DRI < 1.0 to 60% for DRI > 2.0) (8). This early graft failure will result in increased need for retransplantation which has dramatically increased cost compared to primary transplant procedures (21,22). Other high cost complications which occur with greater frequency with high-risk donors include biliary leaks in both DCD (23) and anatomic variant grafts (24), early aggressive recurrence of hepatitis C in older liver grafts (12) and higher incidence of kidney allograft failure among recipients of liver/kidney transplants from marginal donors (25).

Our analysis has several potential limitations. The use of hospital LOS, although highly correlated, is not a direct measure of true hospital costs. However, LOS has been found in previous reports to be an accurate reflection of resource utilization in LT (2). The use of institutional cost data provides an estimate of the order of magnitude of the cost implications of the use of marginal donors. Furthermore, complicated cases that require longer operating times, more blood transfusions and a higher utilization of resources in general are likely to result in a longer LOS. Thus, we believe that LOS constitutes an excellent marker of clinical acuity and a reliable proxy for resource utilization. To more accurately estimate costs, a larger study including cost data from multiple institutions in a variety of regions including posthospitalization care would be needed to perform a more complete analysis. However, such studies are complex, expensive and rarely per-formed.

The second limitation is that the national data include results from a variety of centers, undoubtedly at varying positions along the learning curve in the use of marginal, and in particular DCD, organs. Clearly, there appears to be a national learning curve reflected in reductions in LOS observed across MELD and DRI categories over the course of the 5 years of this analysis. Recent data suggest that more experience and proper selection of DCD organs in particular may help to avoid many of the complications leading to high upfront cost (10,13,14). Whether or not the cost differential inherent in the use of these organs will diminish over time is an empirical question which can only be answered by further analysis as the use of these organs continues to grow.

Given that transplant center profitability is determined by the difference between reimbursement and cost, it is imperative that the economic impact of high-risk donors be considered in any financial evaluation of LT (26). Since the case rates which dictate transplant reimbursement often include organ acquisition costs, we propose that a consideration should be given to a discounted price from OPOs for marginal organs in general, and liver allografts with very high DRIs in particular, to avoid a serious financial disincentive for their use. Alternatively, reimbursement policies will need to be better correlated with donor and recipient risk so that they can be more aligned with cost.

In conclusion, expansion of the donor pool remains a vital activity for the entire transplant community. However, the success of the organ cooperative and other efforts to expand donation is likely to be limited if transplant centers are economically disadvantaged by the aggressive use of marginal organs. Currently, reimbursement is not indexed by the quality of the donor or, in general, the severity of illness of the recipient. We have demonstrated that both DRI and recipient MELD score are closely correlated with hospital costs. Therefore, public policy reform may be needed to ensure that transplant centers can continue to accept and utilize this new organ supply in increasingly sick recipients without incurring an undue economic burden.