Impact of Recipient MELD Score on Resource Utilization


*Corresponding author: Kenneth Washburn,


The model for end stage liver disease (MELD) system prioritizes deceased donor organs to the sickest patients who historically require higher healthcare expenditures. Limited information exists regarding the association of recipient MELD score with resource use. Adult recipients of a primary liver allograft (n = 222) performed at a single center in the first 27 months of the MELD system were analyzed. Costs were obtained for each recipient for the 12 defined categories of resource utilization from the time of transplant until discharge. True (calculated) MELD scores were used. Inpatient transplant costs were significantly associated with recipient MELD score (r= 0.20; p = 0.002). Overall 1-year patient survival was 85.0% and was not associated with MELD score (p = 0.57, log rank test). Recipient MELD score was significantly associated with costs for pharmacy, laboratories, radiology, dialysis and physical therapy. Multivariate linear regression revealed that MELD score was most strongly associated with cost compared to other demographic and clinical factors. Recipient MELD score is correlated with transplant costs without significantly impacting survival.


Liver transplantation (OLT) has become a viable therapeutic option for patients with advanced liver disease. Yet, this is one of the most complex, resource intensive and costly procedures currently available (1–4). Previous reports have suggested that patient outcomes and resource utilization are related to severity of recipient disease (2,3,5). Measurement of recipient disease severity has undergone an evolution in the recent past. The current, and arguably most accurate, system used is that of the Model for End-Stage Liver Disease, or MELD (6–8), which prioritizes deceased donor hepatic allografts to candidates with the highest risk of mortality without a liver transplant. The early results of this change in allocation have shown a decrease in additions to the liver transplant waiting list and a decrease in waiting list mortality (9).

High acuity patients with increased pre-transplant mortality risks would intuitively incur increased rates of resource utilization and potentially decreased outcomes with OLT. The ability of MELD, a pretransplant mortality risk score, to predict posttransplant mortality has been less accurate (10). Axelrod et al. (11) demonstrated that costs under the MELD system are higher than prior to the use of MELD and Medicare reimbursements are inadequate resulting in a net loss for the transplant center. Little information exists regarding resource utilization in relation to recipient MELD score since the inception of this new system. This report evaluates the influence of recipient disease severity, reflected by the recipient MELD score, on pre- and posttransplant use of hospital resources.

Materials and Methods

Patient population and data acquisition

Institutional Review Board approval was obtained for this study. Hospital charge data were obtained for adult patients undergoing OLT at the University Hospital-University of Texas Health Science Center at San Antonio for the first 27 months after implementation of the MELD system (February 28, 2002, to May 21, 2004). The analysis included all adult recipients of a deceased donor whole organ allograft, a right trisegmental split allograft or a right lobe adult living donor allograft. Adult recipients of a liver transplant were included, including those requiring retransplantation (n = 4) during the index admission. Recipients who were status 1 at the time of the primary transplant were excluded. Recipient information was collected including age, disease etiology, gender, presence of hepatocellular carcinoma (HCC) and pre-transplant MELD score.

Financial data

Financial data were obtained from a single source and included all inpatient charges for the index admission for the transplant procedure. Length of stay (LOS) was recorded for hospitalization 90 days prior to transplant and for 90 days after discharge from the transplant admission as well as LOS from the time of surgery to discharge for the transplant admission. To our knowledge, most if not all hospital admissions occurring within 90 days after transplant were at University Hospital, San Antonio, TX. Local charge data for 12 different categories of utilization were collected. Cost to charge ratios were obtained for each of the 12 different utilization categories (anesthesia, 0.994; operating room, 0.542; blood products, 0.483; dialysis, 0.340; intensive care unit, 0.843; radiology, 0.232; inpatient ward, 0.51; laboratory, 0.319; medical/surgical supplies, 0.516; organ acquisition, 1.0; pharmacy, 0.50 and respiratory services, 0.466). These ratios were calculated from the financial data on the specific patients involved in this cohort. Cost data were categorized from the day of transplant until the time of discharge. Cost data were collected on all adult patients who underwent OLT. Four patients required retransplantation during the index admission and these charges were assigned to the first transplant with the associated MELD score.

Total cost data used represent both direct and indirect costs. Overall, indirect costs represented 19.9% of the total cost. The allocated indirect costs for each of the resource categories ranged from 11.6% to 37.3%.

Outpatient costs for the 90 days prior to transplant were captured in the same fashion as the inpatient costs. All outpatient care is delivered through the transplant clinic located in University Hospital. The same financial accounting system is used for the inpatient and outpatient environment. Captured costs included those for clinic use, laboratory, pharmacy, radiology, cardiology, gastroenterology and endoscopy services and emergency room visits.

Costs for rehabilitation stays posttransplant at University Hospital were captured in the same fashion as described above. The majority of patients (66%) receiving rehab care was sent to other healthcare systems and the costs or charges were not available for analysis. We used the average costs per day of rehab at University Hospital to estimate the costs for rehab care in other healthcare systems. These costs were added to the post-transplant cost prior to analysis.


MELD exception points were noted but were not used in evaluation. The MELD score used was the calculated MELD (12) to avoid artificial inflation of this score by exception points granted by Regional Review Boards for HCC and other disease entities. Donor information recorded included age and cold ischemic time.

Statistical analysis

Survival was analyzed using Kaplan-Meier product-limit estimates. Groups were compared using the log rank test. A general linear modeling approach was used to assess the relationship between MELD score and costs. Cost data were log transformed to normalize their distribution. The association between covariates and cost was represented as the regression slope and a test of significance for the slope. The magnitude of association is reported as the correlation coefficient. We constructed multivariate models to determine the impact of demographic and clinical covariates on the relationship between MELD and total inpatient costs incurred from the day of transplant until discharge. Contingency tables were constructed to examine the relationship between categorical variables. Exact tests of association were performed using the StatXact 7 software package (Cytel Corp., Cambridge, MA), all other statistical analyses were performed using the SAS (version 9.1.3, SAS Institute, Cary, NC).


Recipient demographics

Between February 28, 2002, and May 21, 2004, a total of 224 adult liver transplants were performed at our institution. Two patients were transplanted as a status 1 as the primary listing and were excluded from analysis. Of the 222 patients used for the cost analysis, there were 199 whole organ deceased donor grafts, 18 right trisegmental deceased donor grafts and 5 right lobe living donor grafts. A total of 15 combined liver/kidney grafts were performed. For descriptive purposes only, recipients were equally divided into MELD score quartiles (6–14, 15–20, 21–27 and 28–40). Review Board exceptions were granted to 68 (30.6%) patients in the study group. Of the patients transplanted with exceptions, 59/68 had calculated MELD scores at transplant of 20 or less. HCC was the indication for exception requests for the overwhelming majority of cases. Donor and recipient ages were comparable across MELD scores. Hepatitis C was more prevalent in patients with MELD scores 20 or less, likely a consequence of the high percentage of HCC in this group. Cold ischemic time was similar across MELD scores (Table 1).

Table 1.  Donor and recipient characteristics by MELD score quartile
 Recipient MELD
6–1415–2021–2728–40p Value
  1. RRB = regional review board; CIT = cold ischemic time; HCV = hepatitis C.

LRD3 (5%)2 (4%)0 (0%)0 (0%)0.035
Split graft4 (7%)7 (12%)3 (5%)5 (9%)0.40
Liver/Kidney0 (0%)2 (4%)5 (9%)9 (17%)0.005
Whole organ50 (88%)47 (84%)54 (95%)47 (92%)0.020
RRB exceptions39 (70%)20 (40%)9 (12%)0 (0%)0.001
Recipient age53.453.652.854.10.70
Donor age37.946.239.436.60.010
CIT (min)3353543713560.50
HCV33 (57%)21 (38%)25 (44%)18 (35%)0.079

Recipient length of hospitalization and costs

Average hospitalization within 90 days prior to OLT (LOS-pre OLT) was significantly higher (5.2 days) in the MELD 28–40 group than in any other group (Table 2). The LOS after OLT (LOS-post OLT) was significantly longer (15.7 days) for the MELD 28–40 than any other group. Significantly fewer patients in the MELD 28–40 group were discharged to home (65%) than any other group. Once discharged from the primary admission for OLT, the LOS for readmissions to the hospital in the first 90 days after OLT (post-OLT LOS) was similar across all groups. Mean total hospital costs for the transplant (day of surgery to day of hospital discharge) were significantly higher for the MELD 28–40 compared to all other groups which had similar costs.

Table 2.  Hospitalization and disposition by MELD score quartile
 Recipient MELD
6–1415–2021–2728–40p Value
  1. LOS-pre OLT = hospitalized days within 90 days prior to transplant; LOS-post OLT = hospitalized days from day of transplant until release from the hospital; Post-OLT LOS = readmission hospitalized days within 90 days after discharge from the original transplant.

Mean LOS-pre OLT00.10.55.2<0.001
Mean LOS-post OLT10.69.110.815.70.006
Disposition (% home)848081650.001
Mean post-OLT LOS5.
Mean total cost ($)69 56070 23168 34494 7630.002

Hospital costs related to resource use and MELD

Hospital charges were obtained for 12 different resource utilization categories and converted to cost with the appropriate cost-charge ratio for that category. The overall hospital cost-charge ratio for this cohort of patients was 61.8%. The payor mix for the 222 patients in this study included Medicare (32.9%), Medicaid (16.6%) and commercial carriers (50.5%). Log transformed inpatient total costs for the transplant procedure until discharge are plotted against the corresponding MELD score in Figure 1. The scatter plot shows a statistically significant association between recipient MELD score and transplant costs (r= 0.20; p = 0.002). It is also apparent from Figure 1 that there is a wide variation of costs associated with recipient MELD score. MELD score is associated with total costs; however, the association of MELD with various individual resource area costs is highly variable (Table 3). Costs related to pharmacy (r= 0.229; p < 0.0001), laboratory (r= 0.137; p = 0.042), radiology (r= 0.171; p = 0.011), dialysis (r= 0.267; p < 0.0001) and physical therapy (r= 0.171; p = 0.011) were significantly correlated with recipient MELD score. When costs from 90 days prior to transplant, inpatient cost and costs for 90 days after discharge were evaluated in relation to MELD at transplant there was an incremental increase in correlation (r= 0.22; p = 0.0008) from that shown in Table 3.

Figure 1.

Scatter plot showing the relationship between recipient MELD score at time of transplant and log of total costs incurred from surgery until discharge (r = 0.20; p = 0.002).

Table 3.  Relationship between recipient MELD score and transplant costs incurred from surgery until discharge
VariableR95% CIp Value
  1. CI = confidence interval; ICU = intensive care unit.

  2. The correlation coefficient and its 95% confidence interval and p-value are shown.

Total costs0.2000.080.330.002
Medical/Surgical supplies−0.077−0.2070.0560.252
Operating room−0.055−0.1850.0780.417
Blood bank0.076−0.0560.2060.259
Respiratory services0.042−0.0910.1720.537
Organ acquisition−0.091−0.2200.0410.176
Vascular laboratory0.017−0.1150.1480.803
Physical therapy0.1710.0400.2950.011

Costs prior to transplant and after discharge

Costs were collected for the 90 days prior to transplant and 90 days after discharge. There is no significant correlation between pretransplant (r= 0.09; p = 0.27) costs and MELD. When pretransplant costs were divided between inpatient and outpatient, there was no significant correlation with MELD. There was no significant correlation between posttransplant costs and MELD score at the time of transplant (r= 0.04; p = 0.52).

The impact of MELD and other variables on inpatient transplant costs

To assess the possible confounding effect of other variables described in Table 1 and MELD on total costs, multivariate linear regression models were constructed to statistically adjust for other clinical and demographic variables which might influence costs incurred during the inpatient hospitalization after the transplant procedure (Table 4). To determine the effect of other potentially explanatory factors, alternate models were constructed including one or more covariates in regression models after forcing in the MELD score. For these multivariate analyses, the only covariate significantly associated with Log (total costs), in addition to MELD score, was liver + kidney transplant (p = 0.0018), although this had a minimal impact on the significance of the MELD covariate (now p = 0.0197). The regression models included donor age, recipient age, donor sex, recipient sex, split (p = 0.93) and alternate models that included these covariates plus interaction terms. The inclusion of other covariates did not explain the significant association of MELD with Log (total costs).

Table 4.  Multivariate general linear model regression summary
Regression model variablesType III SS1Fp-Value
  1. 1SS = sum of squares.

  2. The type III sum of squares along with the test of association for each covariate in the model is shown for different regression models. In no instance does the inclusion of other variables in the model alter the conclusions regarding the relationship between MELD score and the logarithm of total costs incurred from surgery until discharge (adjusted p-value for MELD score in each model is shown in bold).

Liver & kidney1.52810.020.002
Split liver0.0010.010.930
Liver and kidney0.4142.740.099
Split liver0.0020.010.919
MELD × Liver & kidney0.7965.280.023
MELD × Split liver0.0030.020.897
Recipient sex0.0330.210.890
Donor sex0.0230.150.865
Recipient sex0.0370.230.877
Donor sex0.0280.170.841
Recipient age0.0170.110.741
Donor age0.3222.040.155
Recipient age0.0010.000.951
Donor age0.3051.920.167

Overall survival

Overall patient survival for the entire study group without censoring early deaths was 85.0% at 1 year. Survival for each MELD group is shown in Figure 2. One-year overall survival was 87.7%, 85.7%, 86.0% and 82.7% for the MELD 6–14, MELD 15–20, MELD 21–27 and MELD 28–40 groups, respectively. There were no significant differences in survival between these groups (p = 0.57, log rank test).

Figure 2.

Patient survival in relation to MELD score at the time of transplant (p = 0.57, log rank test).


The MELD scoring system has shown itself to be an accurate predictor of pre-transplant mortality. This has allowed it to be used in stratifying patients for OLT based on disease severity with a resultant shift of deceased donor organs to the sickest recipients. MELD is a predictor of pre-transplant mortality and the ability of MELD to predict posttransplant outcomes has been less accurate. Earlier work has shown that the predictive value of MELD for post-transplant mortality is not nearly as accurate as it is for predicting pretransplant mortality (10,13). Previous studies have shown that sicker patients require more healthcare resources (2,3,5). Yet, little information exists to look at resource use in relation to MELD score in pre- and post-transplant periods.

Limited economic data exist in the literature in relation to the use of the MELD system for allocation in OLT. The utility of the MELD system for distribution of deceased donor livers has generally been accepted as a strong move forward in allocation by decreasing wait-list mortality. Extensive analysis of single center and national data has shown the utility of the system in this regard. Incumbent upon the transplant community is the responsibility to continue to systemically assess the impact of the MELD system across all aspects of OLT. Economic comparisons to the previous allocation scheme utilizing the Child-Pugh score might be useful to demonstrate the financial impact on centers as a consequence of the MELD system (11). Our analysis assesses the impact of varying MELD scores on resource use within the MELD system. The economic implications of increased resource use with higher MELD recipients are potentially significant for transplant centers and third-party payers.

Our analysis uses direct costs for the 222 adult recipients. Furthermore, costs for 12 different areas of resource use were determined using cost-charge ratios specific for these areas and these recipients. Complete data were available on all recipients for the inpatient transplant admission providing us with a highly specific data set. Similarly, the distribution of MELD score for our recipients is close to a normal Gaussian distribution (data not shown).

Intuitively, one would surmise that sicker patients require more resources before and immediately after OLT. Previous work from the University Health System Consortium showed that indeed the use of hospital-based resources is strongly related to the recipient disease severity (14). Based on our analysis, there does appear to be a correlation of recipient MELD at transplant with inpatient transplant cost. Patients with increasing MELD scores do appear to utilize greater resources. However, the 12 individual studied resource categories were not consistently associated with MELD score. Costs for pharmacy, laboratories, radiology, dialysis and physical therapy were significantly correlated with MELD scores. The relationship of MELD and inpatient cost is statistically significant, yet with r= 20, this relationship in this data set may not be as strong as one might have anticipated. These results could imply that other clinical factors not necessarily associated with the recipient MELD score may influence and impact cost. In our data set, once patients are out of the hospital utilization of resource as measured by hospital readmissions, cost does not appear to be strongly associated with MELD score at the time of transplant. The addition of estimates of rehab care did not significantly impact this finding. Despite this observation, higher MELD score recipients were much more likely to need ongoing rehab care once discharged from the initial inpatient hospitalization. Some of these patients also received home rehab care which could not be captured. The cost of rehab care is substantially less than that of inpatient hospitalization care. Therefore, these patients may receive institutionalized long-term rehab care but the costs are not as substantial as that of the inpatient hospitalization. The higher MELD score patients (28–40) had higher inpatient costs, a longer LOS, and more often required transition to rehab care following transplantation.

Hospitalized days prior to transplant for patients with MELD scores over 28 where significantly higher when compared to other groups. A similar finding has been observed and reported previously with work from the Scientific Registry of Transplant Recipients (SRTR). In that report the authors found that hospitalization rates per patient year were significantly decreased after OLT for all MELD groups except for those in the 6–9 group (15). In that study, patients in the MELD group 30–40 had the largest decrease in hospitalization rates compared to all other groups. The findings in our report and those from the SRTR may indicate that though MELD is an accurate predictor of pre-transplant mortality, it may not correlate well with posttransplant hospitalization. Our data did not show a significant correlation between pretransplant costs and MELD score at the time of transplant even though pretransplant hospitalization rates were higher for the sicker patients. Costs were collected for 90 days prior to transplant and the MELD score used is that at the time of transplant, not necessarily the same as 90 days prior to transplant. Costs incurred prior to this time frame were not collected and may not truly reflect the total pretransplant resource use. Likewise, some costs may be attributed to transplant evaluation for some patients and not others and we do not have an accurate way to segregate these costs. The recorded LOS prior to transplant may be a better surrogate marker for costs in this cohort.

Total inpatient transplant costs are significantly affected by recipient MELD score at the time of transplant. A combined liver and kidney transplant does appear to influence costs as one might expect; however, the overall impact on the significance of the MELD score in the regression models was relatively small. Other demographic and clinical variables such as the use of split liver grafts, donor or recipient age or sex or cold ischemic time had minimal effect on costs. There may be other variables not collected that could influence costs; however, in this study, recipient MELD score influences total costs significantly more than any of the other variables evaluated.

The association of recipient MELD score at the time of transplant and the cost associated with the care rendered may indicate the need for risk adjustment in regards to third-party reimbursements. Clinical outcomes are incrementally risk-adjusted in order to compare programs with differing recipient and donor demographics. However, the associated financial outcomes are not usually risk-adjusted resulting in increased financial exposure for both the payor and provider. Outlier protection and contracts that spread the financial risk are important, and may soften the financial impact of high-risk cases. However, charges in the outlier phase of many contracts are reimbursed at only a fraction and often without profit (16). These results, if validated, could support discussions emphasizing the need to financially risk-adjust reimbursements for liver transplant centers in a way comparable to that used for clinical outcomes.

Survival in the sickest group (MELD 28–40), though lower, was not significantly different than the remaining patient groups. Our survival results are very similar to the national results for these groups of patients (17). Our results might indicate that higher MELD score patients (>27) require more resources prior and shortly after transplant, but once recovered from the procedure, they achieve acceptable outcomes without a continuous need for higher resource utilization as measured by hospitalization rates. The alternative for this group of patients is most certainly death as the 3-month mortality for a candidate with a MELD score over 27 ranges from 50% to 90%.

Caution is needed when interpreting single-center data, this study is no exception. Actual resource use following discharge from the hospital may vary considerably given the disposition of the recipient. Significantly, more patients with higher MELD scores transitioned to some form of inpatient rehabilitation facility prior to going home. As there were a number of different facilities, complete financial data are not practically obtainable. Likewise, some hospitalizations may have occurred at other hospitals prior to transplant that we did not know about which would not be captured in the current analysis of pretransplant resource use. Cost accounting and charge masters vary across transplant centers thus potentially limiting the generalization of our findings. Likewise, individual institutions may utilize different methods for allocating direct and indirect costs which should be taken into account if one tries to generalize these findings to other institutions with similar or different economic environments.

Practice patterns do vary across centers based on experience and patient disease acuity. MELD score variations between centers at time of transplant (18,19) could produce results different from what we report. However, with a number of pre-MELD studies reporting a relationship between disease severity and cost of OLT, our findings could be generalized to most adult liver transplant centers.

In conclusion, as reported prior to the implementation of the MELD system, recipient disease severity, as measured by mortality risk, does appear to significantly impact costs for transplant. These higher acuity patients have higher costs and more hospitalized days prior to and after OLT which may indicate the need for risk-adjusted reimbursements for this patient group. Other factors can influence total costs, but not to the same degree as recipient MELD. There may be other variables not studied here that also impact costs. These results should be taken in the context of a single-center study and validated in other studies before using the results to make recommendation in regards to the practice of OLT.