Major Variation across Local Transplant Centers in Probability of Kidney Transplant for Wait-Listed Patients

Significance Statement

Geographic disparities in rates of kidney transplantation have been observed, but the role played by variations in practices at transplant centers versus differences in local organ supply and demand remains unclear. This retrospective national registry study compared the probability of receiving a deceased donor kidney transplant within 3 years of waiting list placement across centers. For the average patient, probability of transplant varied 16-fold between different centers across the United States; up to ten-fold variation persisted between centers working with the same local organ supply. Probability of transplant significantly associated with centers’ willingness to accept offers of organs for wait-listed patients. Large disparities between centers for likelihood of receiving a timely transplant may be related to center-level practice variations rather than geographic differences in underlying organ supply or patient case mix.

Abstract

Background

Geographic disparities in access to deceased donor kidney transplantation persist in the United States under the Kidney Allocation System (KAS) introduced in 2014, and the effect of transplant center practices on the probability of transplantation for wait-listed patients remains unclear.

Methods

To compare probability of transplantation across centers nationally and within donation service areas (DSAs), we conducted a registry study that included all United States incident adult kidney transplant candidates wait listed in 2011 and 2015 (pre-KAS and post-KAS cohorts comprising 32,745 and 34,728 individuals, respectively). For each center, we calculated the probability of deceased donor kidney transplantation within 3 years of wait listing using competing risk regression, with living donor transplantation, death, and waiting list removal as competing events. We examined associations between center-level and DSA-level characteristics and the adjusted probability of transplant.

Results

Candidates received deceased donor kidney transplants within 3 years of wait listing more frequently post-KAS (22%) than pre-KAS (19%). Nationally, the probability of transplant varied 16-fold between centers, ranging from 4.0% to 64.2% in the post-KAS era. Within DSAs, we observed a median 2.3-fold variation between centers, with up to ten-fold and 57.4 percentage point differences. Probability of transplantation was correlated in the post-KAS cohort with center willingness to accept hard-to-place kidneys (r=0.55, P<0.001) and local organ supply (r=0.44, P<0.001).

Conclusions

Large differences in the adjusted probability of deceased donor kidney transplantation persist under KAS, even between centers working with the same local organ supply. Probability of transplantation is significantly associated with organ offer acceptance patterns at transplant centers, underscoring the need for greater understanding of how centers make decisions about organs offered to wait-listed patients and how they relate to disparities in access to transplantation.

Kidney transplantation is the optimal treatment for patients with ESKD, providing superior survival, quality of life, and long-term cost compared with dialysis.^1–3 Despite these advantages, only a minority of patients with ESKD (13%) are wait listed, and even fewer receive a transplant in the United States, whereas the scarcity of donor organs results in the death or removal from the waiting list of nearly 9000 wait-listed candidates annually.^4,5

Transplant candidates without a living donor wait to receive a deceased donor kidney offer through an allocation system designed to objectively prioritize patients using factors including waiting time, age, immunologic matching, degree of sensitization, and geography.⁶ There is wide variation across centers in which patients are accepted for wait listing and which donor organs are accepted when offered.^7–9 These practice patterns and center criteria are not available for patients to review or compare when selecting a transplant center.¹⁰ Further, these practice variations potentially exacerbate existing differences in access to transplantation resulting from geographic heterogeneity in organ supply. The recent Executive Order on Advancing American Kidney Health highlighted the urgent need to improve kidney transplant rates through increased organ recovery and utilization.¹¹

The effect of center-level clinical practice variations on the probability of transplantation for wait-listed patients within the smallest geographic unit used for allocation—the donation service area (DSA)—is unknown, particularly under the new Kidney Allocation System (KAS) implemented in December 2014. There is a need to understand and separate the effect of transplant center practice from that of geographic variation in organ supply and incidence of ESKD, especially given recent proposals to revisit geographic boundaries to reduce disparities in access to transplant. We assess variation in the probability of transplantation across centers at the national, regional, and DSA levels and attempt to identify center characteristics contributing to the observed variation.

Methods

Study Population

This study used data from the Scientific Registry of Transplant Recipients (SRTR). The SRTR data system includes data on all donors, wait-listed candidates, and transplant recipients in the United States, submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The Health Resources and Services Administration, US Department of Health and Human Services provides oversight to the activities of OPTN and SRTR contractors.

All adults (age ≥18 years) added to the deceased donor kidney transplant waiting list in 2011 (pre-KAS cohort) or 2015 (post-KAS cohort) were included. In order to assess whether observed variations in probability of transplant have been present over time or were associated with the implementation of a new KAS in December 2014, analyses were performed on both a pre-KAS cohort and a post-KAS cohort. Candidates listed in 2011 were followed through December 3, 2014 and candidates listed in 2015 were followed through December 3, 2018 for the first of the following outcomes: deceased donor transplantation, living donor transplantation, death, or waiting list removal for reasons other than death or transplant. We excluded candidates at centers performing fewer than ten deceased donor transplants annually, on average, during the follow-up period and at centers listing fewer than ten candidates in the cohort waiting list year. We considered each addition to the waiting list a distinct opportunity for transplant; candidates who were delisted and subsequently relisted or who were newly listed at multiple centers during the cohort year could be included multiple times and would be counted at each center.

Cohort Characteristics

We examined candidate characteristics previously identified to influence likelihood of transplant, including listing age, sex, race, history of diabetes, obesity (body mass index [BMI] >30 kg/m²), blood type, maximum panel reactive antibodies (PRA) sensitization ≥98%, and dialysis vintage at listing. Data were complete for all candidate characteristics except history of diabetes, BMI, and PRA; <0.5% of the cohort was missing these data. Mode imputation was used to handle these missing categorical characteristics. We compared candidate characteristics between the pre-KAS and post-KAS cohorts and between OPTN regions for each cohort using chi-squared, t, Wilcoxon rank-sum, and Kruskal–Wallis tests.

Variation in Transplantation at the National, Regional, and Local Levels

To assess the likelihood of deceased donor kidney transplantation after wait listing in the presence of competing risks, we used a subdistribution modeling approach. Stratifying by OPTN region and KAS cohort, we calculated the cumulative incidence function for the probability of receiving a deceased donor kidney transplant for each center using Fine–Gray competing risk regression, treating living donor transplant, death, and waiting list removal as competing events. To account for differing candidate populations, we adjusted for factors taken into account during allocation (blood group, high PRA, dialysis vintage) and clinical/sociodemographic characteristics potentially affecting access to transplantation after listing (age at listing, race, sex, diabetes, obesity). For our primary analyses, we compared the adjusted probability of transplant within 3 years between centers for the “average patient” defined using the most frequently occurring category or mean continuous value within each OPTN region on the basis of the descriptive candidate analyses. The most frequently occurring category was the same across all OPTN regions (Supplemental Table 1); therefore, the average patient was defined as man, non-Black, nonobese, nondiabetic, blood group O, and PRA<98%, with the mean regional dialysis vintage and listing age within the OPTN region. For secondary analyses, we estimated probabilities for additional groups of patients varying one characteristic away from reference at a time: woman, Black, obese (BMI>30), diabetic, PRA≥98%, age 65 at listing, and time on dialysis at listing (national 25th, 50th, or 75th percentile of dialysis vintage).

Associations with Probability of Transplantation

We next explored associations between center-level practices, DSA-level organ supply, and the adjusted 3-year probability of transplantation in the post-KAS era. Center-level variables of interest included SRTR offer acceptance ratios, prevalent waiting list size, total transplant volume, number of deceased donor transplants, and proportion of living donor transplants from January 1, 2015 to December 3, 2018. Transplant center offer acceptance data were compiled from SRTR program-specific reports released July 2017 to July 2019, the only years that include offer metrics reporting how many organ offers a center accepted for transplant out of the total number of offers they received. Offer acceptance ratios are calculated by SRTR to compare offer acceptance between centers, with an offer acceptance ratio of one representing the national average practice. An offer acceptance ratio less than one indicates that a center is less likely to accept an offer than the national mean, and an offer acceptance ratio more than one indicates that a center is more likely to accept an offer. For each center, we calculated mean offer acceptance ratios from all available reports during our study period for each of five categories reported by SRTR: overall offer acceptance ratio for all offers, for high-risk kidneys (donors with a Kidney Donor Risk Index [KDRI] >1.75), for medium-risk kidneys (1.05<KDRI<1.75), for low-risk kidneys (KDRI<1.05), and for hard-to-place kidneys (kidneys offered to >100 candidates before being accepted for transplantation). It should be noted that the SRTR offer acceptance ratio as currently calculated does not include the offers that are bypassed in response to the center’s individual UNet preferences.¹² This likely results in an overestimation of the true organ offer acceptance rates at centers that utilize this option and may underestimate an association between true offer acceptance practice and the probability of transplantation.

Center waiting list characteristics examined included waiting list size, calculated as the number of candidates appearing on the waiting list at some point between January 1, 2015 and December 3, 2018, and the proportion of candidates who were initially listed with inactive status. To explore potential differences in selected patient case mix between centers, we also looked at the proportion of a center’s waiting list during this time period with the following characteristics: high maximum PRA (≥98%), long dialysis vintage (>3 years) at listing or at study follow-up entry (January 1, 2015 or listing date past January 1, 2015), Estimated Post-Transplant Survival score ≤20%, Black race, obese, history of diabetes, and age >65 at listing.

To represent varied local organ supply at the DSA level, we calculated the ratio of kidneys recovered for transplantation at each organ procurement organization to the prevalent number of candidates on the waiting lists of transplant centers falling within that DSA during the follow-up period. We calculated Spearman correlation coefficients between the center- or DSA-level variables and the center-level estimates for adjusted probability of transplantation within 3 years for the 2015 cohort.

Sensitivity Analyses—Inactive Status

Patients who are added to the waiting list but placed on inactive status are not eligible to get offers for deceased donor kidneys. Patients are inactivated on the waiting list for a variety of reasons, including a temporary illness that would preclude transplantation. Patients can sometimes change between inactive and active status frequently and for very short intervals, whereas some patients are listed with inactive status and never convert to active. These always-inactive individuals never receive organ offers while on the waiting list and thus, would decrease the overall probability of transplantation at a center when considering all wait-listed patients. In order to assess the potential effect of waiting list status on our estimates of probability of transplantation, we conducted an additional analysis excluding all individuals from our cohort who were inactive at the time of listing and remained inactive for the duration of follow-up in our study.

Analyses were conducted in Stata MP 15.1 (StataCorp, College Station, TX) and SAS 9.4 (SAS Institute Inc., Cary, NC). Statistical significance was determined at α=0.05. This study was approved by the Institutional Review Board at Columbia University Medical Center. All research activities were consistent with the principles of the Declaration of Istanbul.

Results

Cohort Characteristics

A total of 67,473 new wait-listing events were included in the analysis, representing 64,056 unique candidates. The 2011 (pre-KAS) cohort included 32,745 candidate listings at 190 centers, and the 2015 (post-KAS) cohort included 34,728 candidate listings at 193 centers. Nationwide candidate demographics were similar in the pre-KAS and post-KAS cohorts, although these characteristics varied by OPTN region, particularly for race and preemptive listing (Table 1). The proportion of candidates receiving a deceased donor transplant within 3 years of wait listing was slightly higher in the post-KAS group (22%) than the pre-KAS group (19%), although the proportion receiving living donor transplants was the same in both cohorts (14%) (Table 2).

Table 1.

Candidate characteristics of the pre-KAS cohort (wait listed in 2011) and the post-KAS cohort (wait listed in 2015)

Characteristics	Pre-KAS 2011 Cohort		Post-KAS 2015 Cohort
	Entire Cohort, n=32,745	Range among Regions	Entire Cohort, n=34,728	Range among Regions
Centers performing 10+ deceased donor kidney transplants per year, no.	190	8–26	193	8–25
Candidate characteristics
Age at listing, yra	52 (±13)	50–53	52 (±13)	50–54
Women,a %	39	36–41	37	35–39
Black race, %	30	8–49	30	8–48
History of diabetes, %	42	37–47	42	38–46
BMI>30 kg/m2,a %	36	29–42	37	30–42
Blood type O, %	48	43–55	48	46–53
PRA≥98%, %	8	3–9	7	4–10
Listed preemptively,a %	28	19–37	33	25–46
Median (IQR) dialysis vintage at listing if not preemptive, yra	1.21 (0.58–2.60)	0.86–1.39	1.31 (0.64–2.87)	1.02–1.48

^aIndicates that P value for chi-squared, t, or Wilcoxon rank-sum test comparing pre-KAS and post-KAS is <0.05.

Table 2.

Distribution of outcomes at 3 years since wait listing for a deceased donor kidney transplant before and after implementation of the new KAS

Outcomesa	Pre-KAS 2011 Cohort		Post-KAS 2015 Cohort
	Entire Cohort, n=32,745, %	Range among Regions, %	Entire Cohort, n=34,728, %	Range among Regions, %
Deceased donor transplant	19	12–29	22	17–32
Living donor transplant	14	10–20	14	10–22
Died	8	5–10	7	3–8
Removed from waiting list	17	12–22	20	13–27
Remaining on waiting list	42	35–48	37	29–46

^aThe 2011 (pre-KAS) cohort included 32,745 candidate listings at 190 centers, and the 2015 (post-KAS) cohort included 34,728 candidate listings at 193 centers.

National Variation

When comparing all centers nationally, there was a wide range in center-level adjusted probability of deceased donor kidney transplantation within 3 years of wait listing for the average patient, ranging from 0.8% (SEM: 0.007) to 74.5% (SEM: 0.050) in the pre-KAS era (Supplemental Figure 1) and from 4.0% (SEM: 0.008) to 64.2% (SEM: 0.077) in the post-KAS era (Figure 1)—>16-fold variation in the probability of transplantation after being wait listed. The median adjusted probability of transplant across all centers nationwide was 15% (interquartile range [IQR], 9%–25%) pre-KAS and 17% (IQR, 11%–26%) post-KAS.

Figure 1.

The distribution of adjusted center-level probability of deceased donor kidney transplant within 3 years of wait listing in 2015 varied considerably nationwide, ranging from 4.0% to 64.2%. Probabilities are for the average patient (man, non-Black, PRA<98%, no diabetes, nonobese, mean age and dialysis vintage at listing in the region). Each dot represents one center, and each color outline represents one OPTN region. Gray bars represent the 95% confidence intervals around each center’s probability. Nationwide PrTx is the probability of deceased donor transplant within 3 years for all kidney transplant candidates in the United States added to the waiting list in 2015 in a national model.

Highly sensitized patients (PRA≥98%) had lower probability of transplant at all centers nationwide compared with nonsensitized patients (PRA<98%) in the pre-KAS era, with a mean 2.19-fold lower probability of transplant within 3 years (Supplemental Figure 2). In the post-KAS era, the overall distribution of probability of transplant for highly sensitized patients started approaching that of the nonsensitized average patient, with a mean 1.05-fold lower probability of transplant for highly sensitized patients compared with nonsensitized patients, but center-level variations persisted (Supplemental Table 2).

In the pre-KAS era, the distributions of probabilities across centers for patients with different dialysis vintage at listing were similar: preemptively listed candidates had a mean 1.10-fold lower probability of transplant and candidates in the 75th percentile of dialysis vintage had a 1.08-fold higher probability of transplant within 3 years compared with the average patient with mean dialysis vintage at listing. In the post-KAS era, with increased importance of dialysis vintage in allocation, the distributions started to diverge: patients in the 75th percentile of dialysis vintage had a mean 1.25-fold higher probability of transplant and preemptive patients had 1.25-fold lower probability of transplant compared with the average patient (Supplemental Figure 2).

Regional Variation

We observed wide center-level variability in the adjusted probability of deceased donor kidney transplantation before and after KAS within OPTN regions (Figure 2). Each region had one or more centers with very low probability of transplant within 3 years of wait listing as well as center(s) with a higher probability than the national median of all centers. We found similar variability between centers within regions in the unadjusted model (Supplemental Figure 3). Region 4 (Texas and Oklahoma) had the largest absolute difference between centers in the adjusted probability of transplantation for the 2015 cohort (59.5%; 4.4%–63.9%), whereas the smallest difference (14.3%; 6.1%–20.4%) was observed in Region 1 (Maine, New Hampshire, Vermont, Connecticut, Rhode Island, and Massachusetts).

Figure 2.

The distribution of adjusted center-level probability of deceased donor kidney transplant within 3 years of wait listing in 2011 or 2015 varied within each OPTN region. Each box represents the 25th percentile to 75th percentile of adjusted probability of transplant for transplant centers within a region, with a line at the median. All regions had center(s) both above and below the national median probability of transplant both before (darker boxes) and after (lighter boxes) the 2014 change in the Kidney Allocation System.

When modeling different patient characteristics (highly sensitized, obese, higher dialysis vintage, etc.), the differences in the adjusted probability of transplant for the various types of patients were frequently smaller than the overall variation observed between the centers within each region with the highest and lowest probability of transplant for the average patient (Supplemental Figure 4).

Local (DSA-Level) Variation

Large differences in adjusted center-level probability of deceased donor transplantation were found even within DSAs, where all centers utilize the same organ procurement organization and local organ supply (Figure 3). For the 2015 cohort, there was a median 2.3-fold difference (IQR, 1.4–3.5) between the highest and lowest centers in each DSA when comparing the 43 of 58 DSAs with more than one included center. The largest absolute difference in adjusted probability of transplant (57.4%; 6.5%–63.9%) occurred in a DSA with seven transplant centers in OPTN Region 4, representing a 9.8-fold variation. The smallest absolute difference in probability of transplant (0.7%; 25.4%–26.1%) occurred in a DSA in Region 11 (Kentucky, Tennessee, Virginia, North Carolina, and South Carolina) with two centers. The 15 DSAs with a single center meeting inclusion criteria also had highly variable probability of transplantation ranging from 9.4% to 62.5% (median: 25.9%; IQR, 22.7%–37.3%), and these centers were not uniformly higher or lower than all other centers in their region (Figure 3).

Figure 3.

Range in adjusted center-level probability of deceased donor kidney transplant within 3 years of wait listing in 2015 showing significant variation within and across DSAs. Each point represents one transplant center, and each level along the y axis represents one of the 58 DSAs, grouped by OPTN region. Each horizontal line illustrates the range in probability of transplant between centers in a given DSA.

Associations with Probability of Transplant

We assessed potential factors associated with differences across centers nationwide and found that center offer acceptance ratio for hard-to-place kidneys displayed the highest correlation with adjusted probability of deceased donor transplantation for the 2015 cohort (r=0.55, P<0.001). Local organ supply also correlated with the adjusted probability of transplantation (r=0.44, P<0.001). There was no meaningful correlation between probability of deceased donor transplantation and center waiting list size (r=−0.17, P=0.02), total transplant volume (including both living and deceased donor transplants; r=0.13, P=0.08), or the proportion of wait-listed candidates starting with inactive status (r=−0.17, P=0.02) (Figure 4, Table 3). Stratifying by region, adjusted probability of transplant was significantly positively correlated with overall offer acceptance ratio in eight of 11 regions and with organ supply in three regions (Supplemental Table 3).

Figure 4.

Correlation scatterplots between center-level or DSA-level characteristics and the adjusted probability of deceased donor kidney transplant within 3 years of wait listing in 2015 demonstrate the strongest relationships with offer acceptance ratios and organ availability. The values presented in the lower right corners are the Spearman correlation coefficients for the relationship between the adjusted probability of transplantation within 3 years of wait listing and each center-level or DSA-level variable.

Table 3.

Correlations between center-level characteristics and the adjusted probability of deceased donor kidney transplant within 3 years of 2015 wait listing

Center Characteristics	Range between Centers, n=193	Spearman Correlation Coefficient	R2	P Value
SRTR offer acceptance ratios
Overall	0.26–8.48	0.54	0.29	<0.001
Low risk (KDRI<1.05)	0.40–3.19	0.52	0.27	<0.001
Medium risk (1.05≤KDRI≤1.75)	0.17–8.44	0.49	0.24	<0.001
High risk (KDRI>1.75)	0.09–8.91	0.48	0.23	<0.001
Hard to place (100+ offers)	0.06–21.41	0.55	0.30	<0.001
DSA-level recovered kidneys-wait list ratio	0.11–1.46	0.44	0.19	<0.001
Transplant center volumea
Total count of transplants	44–1385	0.13	0.02	0.08
Deceased donor transplant count	41–1043	0.24	0.06	<0.001
Living donor transplants, %	0.00–74.03	−0.50	0.25	<0.001
Center waiting list characteristicsa
Center waiting list size	140–8888	−0.17	0.03	0.02
Initially inactive, %	0.00–99.07	−0.17	0.03	0.02
With 3+-yr dialysis at listing, %	4.90–37.35	0.17	0.03	0.02
With 3+-yr dialysis at study follow-up entry (1/1/15 or listing), %	11.27–61.09	−0.21	0.04	0.003
With listing EPTS ≤20, %	11.78–52.36	−0.13	0.02	0.07
With maximum PRA ≥98, %	2.14–21.11	0.13	0.02	0.08
Black, %	0.26–84.53	−0.02	0.00	0.80
Obese, %	19.76–53.90	0.04	0.00	0.54
Diabetic, %	30.67–65.09	−0.14	0.02	0.05
Age >65 yr at listing, %	4.48–31.97	0.21	0.04	0.004

Each row represents a separate correlation analysis between adjusted probability of transplant and the row variable. EPTS, Estimated Post-Transplant Survival score.

^aTransplant center volume variables count the number of transplants performed at each center between January 1, 2015 and December 3, 2018, and center waiting list variables count prevalent candidates on the deceased donor kidney waiting list at any point during the same follow-up period.

Patients who receive living donor transplants are also added to the deceased donor waiting list in the United States even if they already identified a potential living donor; therefore, living donor transplant was considered a competing risk event in our model. As expected, there was an overall negative correlation between the probability of deceased donor transplantation and the number of living donor transplants performed at that center. The proportion of living donor transplants was significantly negatively correlated with probability of deceased donor transplant in five regions (Supplemental Table 3).

Finally, given heterogeneity in center waiting list candidate selection practices, we also assessed the effect of center-level waiting list composition on probability of transplant and found no meaningful correlation between the proportion of wait-listed patients with long (≥3-years) dialysis exposure at listing, PRA≥98%, obesity, diabetes, Black race, age >65 years, or Estimated Post-Transplant Survival score ≤20% and the adjusted probability of deceased donor kidney transplant at that center (Table 3, Supplemental Figure 5).

Sensitivity Analyses

Excluding candidates who never had active listing status during follow-up changed each center’s adjusted probability of transplantation by a mean increase of 1.7 percentage points; however, the probabilities from the full cohort and the cohort excluding always-inactive candidates were highly correlated (ρ=0.98, P<0.001), suggesting almost no change in the wide distribution of the adjusted probability of transplantation across transplant centers (Figure 5).

Figure 5.

Sensitivity analysis comparing center-level adjusted probability of transplant within 3 years of wait listing (2015) in the full cohort (n=34,728) and a cohort excluding candidates who never had active status during the follow-up period (January 1, 2015 to December 3, 2018; n=31,392) demonstrating limited impact of inactive status on a center's probability of transplant. The line of best fit (dotted line) comparing the adjusted probability of transplant for the full cohort and the cohort excluding always-inactive candidates follows the equation y=1.003x+0.0162, with an R² of 0.96.

Discussion

Geographic variations in access to transplantation have been previously described and have been historically attributed almost exclusively to differences in relative organ supply, allocation boundaries, and ESKD incidence across the country.^13–15 Our analysis demonstrates a 16-fold difference in the probability of transplantation for wait-listed candidates across centers nationwide even after controlling for patient characteristics, with large center differences persisting at the regional level and perhaps most importantly, even within individual DSAs. The persistence of 9.8-fold variation between centers within these small geographic areas with a shared organ supply underscores that these differences seemingly result from variations in center-level practice patterns, particularly the willingness to accept organ offers received for their patients, rather than underlying local organ supply. In contrast, the characteristics of wait-listed patients at a center do not seem to be associated with the adjusted probability of transplant, suggesting that unlike center selectivity in organ offer acceptance, selectivity in which patients are wait listed does not seem to significantly influence probability of deceased donor transplant after on the waiting list on the basis of the measured patient characteristics available within the registry data. However, because there are no national data on referral or evaluation for transplantation, it is unclear how the underlying population characteristics of referred or evaluated patients may differ from wait-listed patients.

There are substantial differences across centers in the criteria used to determine offer acceptance.^7,9 In addition to being able to decline individual organ offers for a given patient on the basis of center policy or clinical judgment, centers can also prespecify donor/organ characteristics for which offers will be automatically declined (i.e., bypassed) for all patients on their waiting list during regional and national allocation from the United Network for Organ Sharing Organ Center.^16–18 These bypass criteria are not publicly accessible, and patients are frequently unaware of instances when a center declines an organ that was offered specifically to them.¹⁹ The centers’ ability to decline donor organs, without patient participation or awareness, contributes to considerable variation in the proportion and characteristics of organ offers accepted between centers and adversely affects implementation of an otherwise objective allocation system.¹⁶

Most patients live within driving distance of at least two transplant centers but identify a paucity of information about differences between centers as a barrier to making an informed decision about which transplant center to choose.^20,21 Given that patients prioritize time to transplantation when choosing a center, information such as center-specific bypass criteria, detailed offer acceptance practices, and estimates of transplant probability, represents important patient-centered measures that could facilitate informed decision making if made available and accessible to potential candidates and their providers.²² However, to enable patients to make informed choices, either independently or as part of a shared decision-making effort with referring physicians, this information about centers needs to be presented in a way that is straightforward and responsive to the health literacy/numeracy of the intended audience. Additionally, such patient-centered measures support priorities of the recent Executive Order on Advancing American Kidney Health that includes expanding patient choice and improving access to transplantation and efficient, high-value care.¹¹ Understanding how center behavior affects probability of transplantation can help achieve these objectives as using even less than ideal–quality kidneys to hasten transplantation improves survival and quality of life for recipients while decreasing costs compared with dialysis, whereas organ offer refusals introduce inefficiency and potential bias into the allocation system.^{16,17,23–26} Greater transparency around selective use of organs by transplant centers and emphasizing the association between higher offer acceptance and higher probability of transplant for patients are likely to incentivize centers to reconsider their current clinical preferences in a manner that will increase organ offer acceptance.¹⁶ In addition, system-wide factors that disincentivize organ offer acceptance must be addressed. The increased cost of transplanting marginal kidneys and the concern for regulatory scrutiny increase reluctance toward accepting these offers, and centers that are undergoing performance evaluations have been shown to decrease utilization and transplant volume.^27,28 Restructuring payment models and regulatory metrics on the basis of post-transplant outcomes to remove these disincentives may improve offer acceptance and access to transplantation. An increased willingness on the part of transplant centers to use organs from less than ideal donors will decrease the current unacceptably high discard of deceased donor kidneys and potentially provide organ procurement organizations with the impetus to become more aggressive in their pursuit of organs for transplantation, thereby increasing overall kidney transplantation rates.^19,29,30

We also note that variation in the probability of transplantation across centers seems essentially unchanged following the introduction of the new KAS, except for subgroups of patients. Consistent with stated goals of the KAS, we observed an increase in probability of transplantation for patients with high PRA and the longest ESKD times. Although KAS has achieved its intended consequences, large overall differences between centers persist from the pre-KAS era, and geographic variation in the probability of transplantation remains largely unchanged despite the final rule mandate that geography should not affect access to transplantation.³¹

With continued geographic variation in transplantation after KAS implementation, there is renewed focus on the effect of both organ procurement organization practices and geographic allocation boundaries on the resulting organ supply.³² Our results demonstrate that the probability of transplant has a stronger association with transplant center practices, such as organ offer acceptance, than with local organ supply, suggesting that there are multiple targets for improving equitable access to transplant. Changes to organ distribution must be accompanied by efforts to increase organ utilization and organ offer acceptance broadly while recognizing potential differences in center-level expertise or logistic ability to accept and successfully transplant certain categories of less than ideal organs. A deeper exploration of the effect of patient preferences, specific center practices, and their interaction with wider system-level factors and proposed policy changes is warranted to better understand observed differences in access to transplantation.

Limitations of our study include short available follow-up time post-KAS, which restricted our analysis to patients listed in 2015 shortly after the introduction of KAS, as well as the potential for residual confounding from patient-level variables not captured in the registry or included in analysis. National data are not currently available for patients who were referred and subsequently declined for wait listing by a transplant center. Center-level practices, such as criteria for accepting referred patients onto their waiting list, may lead to variation in case mix between centers and may also influence the centers’ denominator for calculating probability of transplant. Our adjusted analyses account for allocation-level factors that affect access to transplantation after being listed and could help control for differences in selected case mix of wait-listed patients; however, there may be additional potential confounders at the patient level, such as changes in PRA over time or other comorbidities that are not reported in sufficient detail in the registry data. As the models were constructed regionally rather than within individual DSAs, there could be additional unmeasured differences between DSAs that contribute to some of the observed variation regionally and nationally. Additionally, given wide differences in how centers handle active versus inactive listing status (0%–100% of centers’ waiting lists started out inactive), differentiating between inactive status due to administrative practices versus actual patient health status is not feasible and precluded us from incorporating individual candidates’ listing status in our primary analysis. Our sensitivity analysis excluding always-inactive candidates found little change in the distribution of adjusted probabilities across centers, suggesting that center-level variations in practice in whether patients are listed as inactive or active at the outset are not a major contributor to the observed differences between centers.

In conclusion, we report large variation in the probability of deceased donor kidney transplantation even between transplant centers with shared local organ supplies, indicating that much of this observed variation is associated with center-level organ offer acceptance practices rather than geographic variations in organ supply or candidate case mix. A greater understanding of the effect of these center-level practices coupled with increased transparency on differences between centers would potentially facilitate patient selection of transplant centers that align with their own preferences. Such an approach would be more patient centered and might even influence transplant center practice to accept organs that they might otherwise decline for their patients.

Disclosures

D. Cohen reports personal fees from ITB Pharmaceuticals and Natera, outside the submitted work. S. Mohan reports grants from the National Institutes of Health and personal fees from Angion Biomedica and Kidney International Reports, outside the submitted work. R. Patzer reports ownership interest in Vital Software (spouse has ownership); and is a scientific advisor or reports membership: Editorial Board of American Journal of Transplantation, CJASN Editorial Board, and Chair of United Network for Organ Sharing Data Advisory Board. P. Reese reports investigator-initiated grants from CVS Caremark and Merck to support research on medication adherence; investigator-initiated grants from Massachusetts General (with AbbVie) and Merck; personal fees from Associate Editor, American Journal of Kidney Diseases; and personal fees from Collaborative Healthcare Research & Data Analytics, all outside the submitted work. All remaining authors have nothing to disclose.

Funding

This work was supported by a National Kidney FoundationYoung Investigator Grant (to S. Husain). S. Husain is also supported National Center for Advancing Translational Sciences grant KL2 TR001874. S. Mohan is supported by National Institute of Diabetes and Digestive and Kidney Diseases grantsU01-DK116066 and R01 DK114893 and U01DK and National Institute of Minority Health and Health Disparities grant R01 MD014161.

Data Sharing Statement

The data reported here have been supplied by the Hennepin Healthcare Research Institute as the contractor for SRTR. The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy of or interpretation by SRTR or the US Government.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020030335/-/DCSupplemental.

Supplemental Figure 1. Nationwide distribution of adjusted center-level probability of deceased donor kidney transplant within 3 years of wait listing in 2011.

Supplemental Figure 2. Nationwide distribution of adjusted center-level probability of deceased donor kidney transplant within 3 years of wait listing for highly sensitized candidates (PRA≥98%) in the pre-KAS and post-KAS eras and by dialysis vintage.

Supplemental Figure 3. Distribution of unadjusted center-level probability of deceased donor kidney transplant within 3 years of wait listing in 2011 or 2015 by OPTN region.

Supplemental Figure 4. Differences across centers in adjusted probability of deceased donor kidney transplant within 3 years of wait listing in 2015 for subgroups of candidates by OPTN region.

Supplemental Figure 5. Additional correlation scatterplots between center-level or donation service area–level characteristics and the adjusted probability of deceased donor kidney transplant within 3 years of wait listing in 2015.

Supplemental Table 1. Candidate characteristics and 3-year waiting list outcomes of the post-KAS cohort (wait listed in 2015) by OPTN region.

Supplemental Table 2. Range in center-level adjusted probability of deceased donor kidney transplant within 3 years of wait listing across all centers and within OPTN regions for groups of patients targeted in the Kidney Allocation System policy change.

Supplemental Table 3. Correlations between center-level or donation service area–level characteristics and the adjusted probability of deceased donor kidney transplant within 3 years of wait listing in 2015 by OPTN region.