The survival advantage of kidney transplantation (KT) compared with hemodialysis (HD) for patients with end-stage renal disease (ESRD) is well established.1-3 Similarly, from a public health policy standpoint, KT has been shown to be more economically sensible than HD, even for patients who receive a marginally acceptable graft.4,5 The economic benefit from KT has been shown to be highly correlated to characteristics inherent to the transplanted allograft (living vs deceased donor [DD], age, terminal creatinine).6 Poor allograft quality begets delayed graft function (DGF), slow graft function (SGF), a persistent lower estimated glomerular filtration rate (eGFR), and ultimately, an increased risk of graft failure (GF). The long-term economic consequences of poor graft quality has been estimated to be in excess of US $10 000 in the first 2 years post-KT.7 With extended-criteria donor (ECD) kidneys, the economic impact is more pronounced, with many complications, such as DGF and SGF, occurring during the initial hospitalization, potentially requiring HD immediately post-KT, that can substantially reduce the financial margin and lead to a net financial loss for the institution.8,9
In December 2014, a new kidney allocation system (KAS) was implemented that relies on the Kidney Donor Profile Index (KDPI), prioritizing historically disadvantaged populations.10,11 Kidney allocation system aims to distribute the highest-quality kidneys, measured by KDPI, to those candidates projected to live the longest.10 Additionally, there is increased regional and national organ sharing for highly sensitized patients and increased regional sharing of high KDPI (less ideal) organs.10 However, this redistribution of organs can have unintended negative consequences. Although KAS aims to maximize the equitable sharing of kidney allografts, concerns remain that its implementation through wider geographical sharing may lead to increased cold ischemia time (CIT) and CIT-associated adverse effects, namely DGF, SGF, and ultimately, transplant-related costs (TRC).
In this study, we sought to determine the effect, if any, of CIT on TRC using retrospective data in a large-volume KT center. In parallel fashion, we used registry data from the Scientific Registry of Transplant Recipients (SRTR) to extend our investigation to a larger multicenter cohort. Controlling for donor- and recipient-demographic characteristics, we modeled the effect that CIT would have on DGF, SGF, length of stay (LOS), and TRC. Our hypothesis was that longer CIT would increase DGF rates, which would in turn, increase TRC.
MATERIALS AND METHODS
This study used data from the SRTR. The SRTR data system includes data on all donors, waitlisted candidates, and transplant recipients in the United States, submitted by the members of the Organ Procurement and Transplantation Network, and has been described elsewhere. The Health Resources and Services Administration, US Department of Health and Human Services provides oversight to the activities of the Organ Procurement and Transplantation Network and SRTR contractors. Mortality and graft loss are established through linkage with the Social Security Death Master File, data from Center for Medicare Services, and waitlist data. Graft loss is defined as irreversible GF signified by return to renal replacement therapy, listing for KT, or retransplantation. For this study, all adult DD kidney-only transplant recipients in SRTR (SRTR cohort) between January 1, 2006, and December 31, 2014, were included. Patients with missing CIT and information on DGF were excluded. En-bloc pediatric and dual-kidney recipients were excluded.
As an institutional comparator, all DD kidney-only transplant recipients at the University of Minnesota (UMN cohort) during the same period with the same exclusion criteria were analyzed. All donor and recipient information is prospectively maintained in an Institutional Review Board–approved database. Multidisciplinary patient information is directly extracted from the electronic medical record. Pediatric transplants were excluded because of intrinsic differences in outcomes and cost structures. The data collected included donor age, gender, body mass index (BMI), race, ethnicity, date of donation, death status, cause of death, kidney function at time of death, CIT; recipient age, sex, allograft failure (if applicable), DGF (defined as the need for HD within 7 days post-KT), SGF (defined as a creatinine >3.0 on post-KT day 5), KT admission LOS, and KT admission cost.
Using a proprietary hospital accounting and analytics system (StrataJazz®; Strata Decision Technology, Chicago, IL), we collected financial information on all costs associated with the original KT admission (diagnosis-related group 652) for the UMN cohort. Costs associated with that admission accounting for direct (physician- and procedure-specific staff services, supplies used to deliver services, service-related equipment costs, and medications), indirect (utilization of existing infrastructure, administrative costs, revenue-sharing, amortization of existing debt), fixed (kidney acquisition cost), and variable (acuity of care) costs. Whereas direct costs are a function of an individual patient's condition and procedure, indirect costs are more consistently set from year to year with minimal variation.
We limited our analysis to inpatient costs accumulated at the initial KT admission, adjusted to 2006 dollars. Unlike Medicare-reimbursed rates, assigned costs more accurately capture the true cost of various procedures to the individual hospital system (vs charges made to the payer) and associated outcomes and the effect exerted on them by external factors (ie, wider geographical organ sharing, longer CIT, etc.).
Comparison of Cohorts
We summarized continuous covariates as the median value (25th, 75th percentile); categorical covariates, as the frequency (percentage) by cohort (UMN or SRTR). To test univariable differences between the cohorts, we used Wilcoxon rank-sum tests (continuous covariates) and Pearson χ2 tests (categorical covariates).
DGF, LOS, and Financial Cost Models
We assessed the effect of CIT on DGF/SGF using logistic regression, LOS using Poisson regression (allowing for overdispersion), and TRC using gamma regression. Across each of the models, we adjusted for recipient age, sex, race, preemptive transplant, time on HD, retransplantation, diabetes, peak PRA, and BMI; donor age, hypertension, creatinine, cause of death (stroke, anoxia, all other); donor and recipient number of HLA mismatched antigens; and use of machine perfusion. For the models of LOS and cost during the transplant admission, we fit models adjusting for only the recipient and donor characteristics above, as well as models that adjusted for DGF/SGF and LOS. Fitting these separate models allows us to estimate the effect of prolonged CIT on cost and LOS mediated through increased risk of DGF/SGF.
Regression models were selected that best addressed the parameters being studied. Briefly, Poisson regression is an appropriate regression model for count data (ie, number of days in the hospital). Exponentiating the regression coefficients gives the ratio of the estimated mean LOS for a 1-unit increase in the covariate of interest. Gamma regression assumes that the distribution of costs given covariates follows a gamma distribution (a distribution for positive right skewed variables) which is more appropriate than assuming a normal distribution (a distribution for symmetric variables with positive and negative random variables). In our analysis, we used the identity link so the coefficient estimates give the estimate increase in average cost for a 1-unit increase in the covariate.
In the models constructed with the SRTR cohort, we included a random effect for transplant center. For the UMN cohort, we considered a composite endpoint of DGF (HD in the first 7 days) or SGF (creatinine >3 mg/dL on postoperative day 5) to improve power. Because LOS has such a small defined window, the models for LOS are on the multiplicative scale to accentuate any meaningful difference.
All analyses were performed using SAS Version 9.4 (SAS System, Cary, NC) or R Version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria). All statistical tests were 2-sided tests with P < 0.05 indicating statistical significance. Code to reproduce the analyses is available as a Github repository at http://github.com/docvock.
Between 2006 and 2014, a total of 81 945 adult solitary DD KT were performed in the U.S. that met the inclusion criteria; 477 (0.6%) were performed at the UMN. The demographic characteristics of the recipients are listed in Table 1. There were 39.4% female recipients in the SRTR cohort and 40.0% in the UMN cohort with a median age of 54 and 55 years, respectively. The BMI of the SRTR cohort was slightly higher (27.8 kg/m2 vs 27.4 kg/m2). The demographic composition of the UMN cohort was primarily non-Hispanic white (71.3%), whereas only 43.6% non-Hispanic white recipients comprised the SRTR cohort; additionally, African-Americans and Hispanics were more prevalent in the SRTR cohort, whereas Asians were more prevalent in the UMN cohort (P < 0.0001). In the SRTR cohort, 10.0% recipients received a preemptive KT; 17.0% in the UMN cohort (P < 0.0001). The median time on HD for the SRTR cohort was 1241 and 871 days for the UMN cohort (P < 0.0001). The composition of ESRD causes among the 2 cohorts varied significantly with the most common cause of ESRD in the UMN cohort being diabetes type 2 (20.7%) and hypertension in the SRTR cohort (26.0%). Most recipients received their first KT (87.5% SRTR cohort; 75.3% UMN cohort) but there was a higher rate of retransplants in the UMN cohort (P < 0.0001). Peak (PRA) was comparatively higher in the UMN cohort (median of 26% vs 4%; P < 0.0001) as were HLA mismatches (P < 0.0001). Finally, DGF and CIT were higher in the SRTR cohort than in the UMN cohort (P < 0.0001).
The donor demographics are listed in Table 2. There were 39.8% female donors in the SRTR cohort and the UMN cohort with a median age of 41 and 40 years, respectively. The median BMI was 26.5 kg/m2 for both cohorts and the median terminal creatinine was 1.0 mg/dL. The majority of donors in the SRTR cohort (68.8%) and the UMN cohort (88.1%) were white but there were significant variations in the represented ethnicities (P < 0.0001). There was also a higher proportion of donors with HTN in the SRTR cohort (28.7% vs 21.4%; P = 0.0005). The majority of the donors in this cohort were brain-dead donors (85.1% SRTR cohort, 84.5% UMN cohort), and the cause of death in the overwhelming majority of donors was anoxia, cerebrovascular accidents, or head trauma. Machine perfusion was used more frequently in the UMN cohort (33.8% vs 38.8%; P = 0.023).
Regression Model for DGF/SGF
Overall, DGF was observed in 25.7% of the SRTR cohort and 17.8% of the UMN cohort, with an additional 20.3% experiencing SGF in the UMN cohort. The results of the multivariable logistic regression model of DGF are presented in Table 3. In the SRTR cohort, and consistent with prior literature, an increasing CIT was found to be associated with the development of DGF (odds ratio [OR], 1.41, 95% confidence interval [CI], 1.38-1.44; P < 0.0001) with a 10-hour increase in CIT associated with a 41% increase in the adjusted odds of DGF (95% CI, 38%-44%). Additional recipient characteristics associated with the development of DGF included male sex, nonwhite race, prolonged time on HD, retransplants, DM, increasing BMI, higher PRA, and more HLA mismatches. Additionally, donor characteristics associated with the development of DGF included increasing age, terminal creatinine, a history of hypertension, an anoxic or stroke-related death, donation after circulatory death. Preemptive transplants and the use of machine perfusion were associated with reduced rates of DGF. Similar results were obtained in the UMN cohort, albeit with less precision given the smaller cohort (OR, 1.33; 95% CI, 0.93-1.92; P = 0.112), where a 10-hour increase in CIT was associated with a 33% increase in the adjusted odds of DGF or SGF (95% CI, −7% to 92%).
Regression Model for Transplant LOS
Median LOS was 5 days in the SRTR cohort and 6 days in the UMN cohort. The results from the multivariable Poisson regression model of transplant admission LOS are given on Table 4. Without adjusting for DGF or other postoperative outcomes in the SRTR cohort, an increasing CIT was associated with a small, but significant, increased averaged LOS for the transplant admission (P < 0.0001). Specifically, a 10-hour increase in CIT was associated with a 4% increase in the average LOS adjusted for other donor and pretransplant recipient characteristics (95% CI, 2%-5%). Note that if everyone in the SRTR cohort increased the CIT by 10 hours, such a change would be associated with an increase in the mean LOS from 7.50 to 7.75 days. Other recipient characteristics associated with longer LOS included in the SRTR cohort: increasing age, time on HD, retransplants, DM, and increasing PRA and HLA mismatches. Additionally, donor characteristics associated with longer LOS included increasing age and terminal creatinine, a history of hypertension and donation after circulatory death. Preemptive transplants, more recent transplants, and use of machine perfusion were associated with a reduced LOS.
In the SRTR cohort, DGF was associated with a 60% increase the average transplant LOS (95% CI: 56% - 64%). When DGF is included in the regression model (Table 5), the effect of CIT was attenuated and marginally significant, suggesting that the effect of CIT on LOS is partially mediated by DGF. After additionally adjusting for DGF onset, a 10-hour increase in CIT was associated with a 1.2% increase in the average LOS (95% CI, −0.1% to 2.6%).
Effect estimates in the UMN cohort were generally of the same magnitude but often did not result in significant differences. Specifically, adjusting for donor and pretransplant recipient characteristics (but not DGF), a 10-hour increase in CIT was associated with a 4% increase in average LOS (95% CI −5% to 14.0%). In the UMN cohort, DGF was associated with a 71% increase the average transplant LOS (95% CI 50%-95%). After additionally adjusting for DGF or SGF onset, the effect of CIT on LOS in the UMN cohort was small and insignificant.
Model for TRCs
The results from the multivariable gamma regression model of TRC in the UMN cohort, adjusting for LOS, DGF/SFG, and recipient and donor characteristics are listed in Table 6. After adjusting for LOS, neither CIT nor DGF/SGF were significantly independently associated with increased TRC. Increased LOS resulted in an increase in TRC by US $3422 (95% CI US $3179 to US $3664) per additional day, indicating that most of the effect of CIT and DGF on TRC is mediated through increased LOS. An increase of US $3422 represents 3.9% of mean total cost in the UMN cohort. In particular, the average effect a 10-hour increase in CIT on TRC mediated through increased LOS is US $892 (95% CI US $533 to US $1237) or 1.0% of the average TRC in this cohort.
Analysis of Recipients With Extended LOS and Increased TRC
Because LOS or TRC did not increase with a commensurate increase in CIT in the UMN cohort, we questioned whether there were other factors affecting the overall cost model that could have overshadowed the impact of CIT. Recipients who had a significantly extended LOS and increased TRC from the mean (in the top 95th percentile) in our patient cohort were analyzed. The LOS, cost of KT admission, encountered complications during their post-KT hospital course and Clavien classification for these transplants are listed in Table 7.
The rationale for the implementation of a new KAS was to make KT more accessible to historically disadvantaged populations under the old allocation schema.10,11 Early reports of how KAS has affected recipient characteristics are illustrative of the nature of these changes.12-14 Transplants among donors and recipients with an age difference greater than 30 years declined by 23%; transplants for highly sensitized individuals with a PRA of 99% to 100% (2.4% pre-KAS vs 17.7% post-KAS) and those with 10 years or greater on HD (4.3% pre-KAS vs 18.6% post-KAS) increased sharply; transplants for black and Hispanic candidates and candidates aged 18 to 40 years increased, but transplants for candidates older than 50 years decreased.12-14 This analysis noted that kidneys are shipped over longer distances, leading to increased CIT (10.6% increase in kidneys transplanted with 24 to 36 hours CIT and a corresponding 18.7% decrease in those transplanted with <12 hours of CIT) and higher DGF rates (24.8% pre-KAS vs 29.9% post-KAS; P < 0.001).14,15 Although these early reports demonstrate a beneficial effect on those subgroups of patients,12-14 concerns remain on potential adverse effects to the individual center. We speculated that by increasing CIT, the cost of KT would increase by yielding CIT-related consequences, namely, DGF/SGF. We sought to explore this hypothesis by demonstrating that one of the unintended consequences of an increase in CIT is higher rates of DGF/SGF, prolonged LOS, and thereby, an increase in TRC. We used the statistical power of the large SRTR data set combined with the more granular and detailed analysis of our institutional database to test this association. The results presented here indicate that with both UMN and SRTR data, we can establish that an increase in CIT was associated with an increase in DGF/SGF, which was associated with an increase in LOS, and ultimately results in an increase in TRC (Figure 1).
In this study, we demonstrate a relationship between CIT, DGF/SGF, LOS, and ultimately TRC. The findings from our own center directly parallel the SRTR associations in direction and magnitude, but occasionally failed to demonstrate statistical significance, presumably due to large variation in the outcome variables (LOS and TRC) and the limited size of our institutional data set. Nonetheless, by using methods developed for mediation analysis, we can clearly establish that CIT has an effect on TRC by the influence it exerts on DGF/SGF and LOS (known as the “indirect” effect). Furthermore, the controlled direct effects (in mediation language, ie, the effect of CIT on LOS and TRC after adjusting for any mediators such as DGF16) are generally small in magnitude and nonsignificant. Similarly, CIT has little effect on TRC after adjusting for LOS and DGF/SGF. Taken together, this supports the conclusion that most (or all) of the increase in TRC as a result of an increased CIT is due to the increase in DGF/SGF and subsequent LOS.
Based on the high degree of variance in previous transplant cost analyses,17,18 one might predict a priori that we would not be able to demonstrate statistical association of CIT with the final outcome of TRC directly. That is, even if there is a direct impact of CIT on TRC, the large variabilities in individual costs would likely prevent achieving a statistical confirmation of this association. Instead, we would argue, based on clinical knowledge, previous reports, and the results of this study, that the controlled direct effect is likely small and positive. Therefore, measuring the indirect effect, that is, the effect of CIT through DGF/SGF and LOS, gives a lower bound on the total effect of CIT on LOS and TRC. Thus, we can extrapolate that an increase in CIT leads to increased LOS, which has the potential to increase TRC.
In this model, on a contemporary cohort of KT recipients in the SRTR database and an individual large volume transplant center, we demonstrate that an increased CIT per se does not lead to increased LOS and TRC. However, the financial repercussions of this increment are most readily apparent if recipients develop postoperative complications—either KT-related complications, such as DGF/SGF, or technical complications inherent to operating on an elderly, frail population with cardiovascular, diabetic, and renal comorbidities. Interestingly, we show that CIT in the SRTR cohort is predictive of DGF and this leads to increasing LOS. However, in our model, we are unable to show that increasing CIT leads to increased TRC directly. What is interesting is that in our model for TRC, retransplants and increasing recipient BMI seem to increase TRC—characteristics that primarily inherent to the recipient and not intrinsic to the allograft per se. These findings are corroborated by analyzing recipients who had an abnormally prolonged LOS with commensurate high TRC, in which postoperative care, LOS, and TRC were affected by either DGF/SGF or a medical/surgical complication after surgery. In many of these cases, the postoperative course was contingent on the recipient's baseline frailty and medical comorbidities, validating previous findings that operating on an elderly frail population leads to longer LOS and higher TRC.19-21
The effect of DGF and SGF on TRC has been previously studied.5,7,22-26 In an earlier series from the United States, Saidi et al25 examined the transplant admission cost difference between standard criteria (SCD) allografts versus ECD or donation after circulatory death (DCD) allografts. The ECD or DCD grafts were found to have a higher incidence of post-KT HD, longer LOS, more hospital readmissions due to poor or late onset graft function, which ultimately resulted in a US $20 000 to US $25 000 higher cost for their initial transplant admission—roughly a 30% increase in TRC compared to SCD grafts. By comparison, our results demonstrate that although DGF/SGF mediate a rise in TRC, these increases add less than 10% increase in TRC. It is important to stress a couple of points about this comparison. First, Saidi et al conducted their study over 10 years ago when DGF management was substantially more expensive, primarily because of the cost of inpatient HD post-KT. Today, the overwhelming majority of recipients who develop DGF receive HD in an outpatient basis, effectively swapping the cost from inpatient to outpatient costs. In our study, we only measured in-hospital charges, thereby excluding the charges of any HD recipients who may have received HD after discharge, as DGF patients at our institution are often treated. Clearly, for centers that may not have the infrastructure to take care of DGF as an outpatient, their inpatient HD costs will be higher and the effect of CIT on DGF rates will be much more pronounced. Second, hospital LOS and readmissions for a sizeable portion of surgical procedures in the same era have decreased substantially,27-32 a phenomenon that is driven by economic and administrative pressures.
In a more contemporary study, Schnitzler et al22 analyzed data for Medicare-insured kidney-only transplant recipients from 1995 to 2003 using the United States Renal Data System. They demonstrate that first anniversary eGFR is predictive of subsequent death-censored GF and patient mortality after KT. Furthermore, they show that an eGFR less than 45 mL/min per 1.73 m2 at 1 year post-KT anniversary predicted significantly increased subsequent costs even after controlling for the costs of GF and death.7 In a follow-up study, the same group calculated this added cost of a lower 1-year post-KT anniversary GFR to be between US $17 500 and US $18 200 in higher adjusted payments in the second and third post-KT years, respectively. What this series of studies suggests is that there is an inherent cost to transplanting inferior kidney allografts, into which CIT does not factor into appreciably—observations that are consistent with the findings from this study.
Finally, in a recent review of SRTR and the University Health System Consortium Database, Stahl et al26 compared greater than 19 500 adult DDKT recipients from 2009 to 2012, dividing recipients into cohorts by KDPI score (<85 vs >85) and analyzing LOS, 30-day readmission, discharge disposition, and DGF as indicators of resource use. Surprisingly, although LOS for the greater than 85 cohort was longer, 30-day readmission higher, DGF more prevalent, and discharge disposition more likely to not be home (ie, nursing home, assisted facility), index costs did not appear to be greater for >85 KDPI allografts. Pertinent to our study, the greater than 85 cohort had a significantly longer CIT (median 18.3 hours vs 16.1 hours), which parallels our findings that CIT does not in itself seem to play a role hastening immediate graft dysfunction, as measured by prolongation of short-term LOS or increased TRC.
Thus, how does our data fit into a more overarching context? According to the 2014 Milliman Research Report on US Organ and Tissue Transplant Cost Estimates, KT has the highest estimated costs per member per month for solid organ transplantation at US $1.29 for patients younger than 65 years and US $1.93 for patients older than 65 years, a respective 6% and 10% increase from 2011 estimates.33 Furthermore, of the 2014 billed charges per transplant, the hospital KT admission accounted for US $119 600 of the total US $334 300 average annual charge (35.8%). If according to the data presented herein, increased CIT and CIT-associated adverse effects from wider geographical sharing seem to not affect TRC directly, our model can serve to reassure KT programs that may be apprehensive or skeptical about the new KAS, which has the inadvertent effect of increasing CIT.
There are a number of limitations with our study. Most importantly, the cost data we present are from a single center with center-specific populations, treatment practices, and financial estimates. In particular, because the United States has a very unique healthcare payment structure compared to many other countries around the world and considerable practice variation exists among transplant centers even within the United States, the reproducibility and generalizability of this key outcome variable may be limited. Also, we collected costs within the original KT admission, allowing us to capture all costs related to the medical care conducted exclusively at our center and those reported to our hospital accounting system. Moreover, though the costs used for our model may have been charged by the hospital accounts payable, we do not have data on what was actually collected by accounts receivable. As previously mentioned, outpatient costs (for outpatient HD, medications, or other KT-associated care) are not accounted for in our model, thereby not capturing these KT-associated costs. Furthermore, even though our database is prospectively maintained, the retrospective nature of our analysis may limit our ability to directly control for confounding parameters that may affect our primary outcomes. Patient attrition and era effect in transplant also pose unique limitations to this study. Finally, the models presented herein have not been validated to have predictive qualities as they only use restrospective data. Prospective studies after the implementation of KAS are needed to establish the predictive value of these models.
Notwithstanding the aforementioned limitations, our study presents a statistical analysis on the effects of increasing CIT on a national KT cohort and at a large-volume KT center. We show that LOS and cost of the transplant admission is increased with longer CIT, but only when patients develop post-KT DGF/SGF. Methods to accurately predict which transplant recipients may develop DGF/SGF might identify cases in which reduced CIT would be economically advantageous. In this setting, strategies, such as prerecovery virtual or formal crossmatching and improved or more efficient organ transport, could reduce CIT. The additional costs incurred by an expedited transplant could be offset by savings in the early postoperative period.
1. Wolfe RA, Ashby VB, Milford EL, et al. Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med
2. Rabbat CG, Thorpe KE, Russell JD, et al. Comparison of mortality risk for dialysis patients and cadaveric first renal transplant recipients in Ontario Canada. J Am Soc Nephrol
3. Port FK, Wolfe RA, Mauger EA, et al. Comparison of survival probabilities for dialysis patients vs cadaveric renal transplant recipients. JAMA
4. Held PJ, McCormick F, Ojo A, et al. A cost-benefit analysis of government compensation of kidney donors. Am J Transplant
5. Snyder RA, Moore DR, Moore DE. More donors or more delayed graft function? A cost-effectiveness analysis of DCD kidney transplantation. Clin Transplant
6. Matas AJ, Schnitzler M. Payment for living donor (vendor) kidneys: a cost effectiveness analysis. Am J Transplant
7. Schnitzler MA, Johnston K, Axelrod D, et al. Associations of renal function at 1-year after kidney transplantation with subsequent return to dialysis, mortality and healthcare costs. Transplantation
8. Englesbe MJ, Dimick JB, Fan Z, et al. Case mix, quality and high-cost kidney transplant patients. Am J Transplant
9. Englesbe MJ, Ads Y, Cohn JA, et al. The effects of donor and recipient practices on transplant center finances. Am J Transplant
10. Israni AK, Salkowski N, Gustafson S, et al. New national allocation policy for deceased donor kidneys in the United States and possible effect on patient outcomes. J Am Soc Nephrol
11. Chopra B, Sureshkumar KK. Changing organ allocation policy for kidney transplantation in the United States. World J Transplant
12. Stewart DE, Kucheryavaya AY, Klassen DK, et al. Changes in deceased donor kidney transplantation one year after KAS implementation. Am J Transplant
13. Gebel HM, Kasiske BL, Gustafson SK, et al. Allocating deceased donor kidneys to candidates with high panel-reactive antibodies. Clin J Am Soc Nephrol
14. Massie AB, Luo X, Lonze BE, et al. Early changes in kidney distribution under the new allocation system. J Am Soc Nephrol
15. Hart A, Gustafson SK, Skeans MA, et al. OPTN/SRTR 2015 Annual Data Report: early effects of the new kidney allocation system. Am J Transplant
. 2017;17(Suppl 1):543–564.
16. Valeri L, Vanderweele TJ. Mediation analysis allowing for exposure-mediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros. Psychol Methods
17. Machnicki G, Lentine KL, Salvalaggio PR, et al. Kidney transplant Medicare payments and length of stay: associations with comorbidities and organ quality. Arch Med Sci
18. Machnicki G, Lentine KL, Salvalaggio PR, et al. Three-year post transplant Medicare payments in kidney transplant recipients: associations with pre-transplant comorbidities. Saudi J Kidney Dis Transpl
19. Moghani Lankarani M, Noorbala MH, Assari S. Causes of re-hospitalization in different post kidney transplantation periods. Ann Transplant
20. Naderi M, Aslani J, Hashemi M, et al. Prolonged rehospitalizations following renal transplantation: causes, risk factors and outcomes. Transplant Proc
21. Nemati E, Saadat AR, Hashemi M, et al. Causes of rehospitalization after renal transplantation; does age of recipient matter? Transplant Proc
22. Schnitzler MA, Gheorghian A, Axelrod D, et al. The cost implications of first anniversary renal function after living, standard criteria deceased and expanded criteria deceased donor kidney transplantation. J Med Econ
23. Almond PS, Troppmann C, Escobar F, et al. Economic impact of delayed graft function. Transplant Proc
24. Rosenthal JT, Danovitch GM, Wilkinson A, et al. The high cost of delayed graft function in cadaveric renal transplantation. Transplantation
25. Saidi RF, Elias N, Kawai T, et al. Outcome of kidney transplantation using expanded criteria donors and donation after cardiac death kidneys: realities and costs. Am J Transplant
26. Stahl CC, Wima K, Hanseman DJ, et al. Organ quality metrics are a poor predictor of costs and resource utilization in deceased donor kidney transplantation. Surgery
27. Han S, Smith TS, Gunnar W. Descriptive analysis of 30-day readmission after inpatient surgery discharge in the Veterans Health Administration. JAMA Surg
28. Jafari MD, Jafari F, Halabi WJ, et al. Colorectal cancer resections in the aging US population: a trend toward decreasing rates and improved outcomes. JAMA Surg
29. Kruszyna T, Niekowal B, Kraśnicka M, et al. Enhanced recovery after kidney transplantation surgery. Transplant Proc
30. Waits SA, Hilliard P, Sheetz KH, et al. Building the case for enhanced recovery protocols in living kidney donors. Transplantation
31. Song W, Wang K, Zhang RJ, et al. The enhanced recovery after surgery (ERAS) program in liver surgery: a meta-analysis of randomized controlled trials. Springerplus
32. Geubbels N, Bruin SC, Acherman YI, et al. Fast track care for gastric bypass patients decreases length of stay without increasing complications in an unselected patient cohort. Obes Surg
33. Hanson SG, Bentley TS. U.S. organ and tissue transplant cost estimates and discussion. Milliman Research Report Milliman. http://www.milliman.com/insight/research/health/2014-U_S_-organ-and-tissue-transplant-cost-estimates-and-discussion/