Abstract
The Kidney Allocation System fundamentally altered kidney allocation, causing a substantial increase in regional and national sharing that we hypothesized might impact geographic disparities. We measured geographic disparity in deceased donor kidney transplant (DDKT) rate under KAS (6/1/2015–12/1/2016), and compared that with pre-KAS (6/1/2013–12/3/2014). We modeled DSA-level DDKT rates with multilevel Poisson regression, adjusting for allocation factors under KAS. Using the model we calculated a novel, improved metric of geographic disparity: the median incidence rate ratio (MIRR) of transplant rate, a measure of DSA-level variation that accounts for patient casemix and is robust to outlier values. Under KAS, MIRR was 1.751.811.86 for adults, meaning that similar candidates across different DSAs have a median 1.81-fold difference in DDKT rate. The impact of geography was greater than the impact of factors emphasized by KAS: having an EPTS score ≤20% was associated with a 1.40-fold increase (IRR=1.351.401.45, p<0.01) and a 3-year dialysis vintage was associated with a 1.57-fold increase (IRR=1.561.571.59, p<0.001) in transplant rate. For pediatric candidates, MIRR was even more pronounced, at 1.661.922.27. There was no change in geographic disparities with KAS (p=0.3). Despite extensive changes to kidney allocation under KAS, geography remains a primary determinant of access to DDKT.
INTRODUCTION
Considerable geographic inequity in access to deceased donor kidney transplantation (DDKT) in the United States has persisted since the creation of the national organ allocation system in the 1980s 1–6. Unadjusted DDKT rates in each DSA ranged from 11.5 to 49.8 per 100 active waitlist years in 2012 and 2013 7. Unadjusted DDKT rate in each state ranged from 0.7 to 27.5 per 100 dialysis patient years in 2015 8. These patterns are observed with casemix adjustment: DDKT rates ranged from 60% lower to 150% higher than the national average from 1996 to 2005 4. Moreover, these disparities have worsened over time; the range of wait times for DDKT among DSAs increased from 0.41–3.67 years in 2000 to 0.50–5.22 years in 2009 6.
The Kidney Allocation System (KAS) was implemented on December 4, 2014, and is the most dramatic revision of kidney allocation in more than 20 years 9. Major changes of the new allocation system include improving longevity matching using the Estimated Post Transplant Survival (EPTS) score and Kidney Donor Profile Index (KDPI), and providing greater access to blood type B candidates 10. These changes have been accompanied by broader sharing for sensitized candidates, broader sharing for KDPI over 85, and elimination of local policy variances 11. Specifically, regional imports increased from 8.8% pre-KAS to 12.5% under KAS, and national imports increased from 12.7% pre-KAS to 19.1% under KAS 12. Under KAS, kidney allocation points are influenced by waiting time, candidate age, Calculated Panel Reactive Antibody (CPRA) score, EPTS score, blood type, and whether the candidate is a prior living donor 11. Though KAS was not designed to decrease geographic disparities, these multiple changes in organ allocation and expansions in sharing might have affected geographic disparity in kidney allocation 10.
Measuring geographic disparity in access to transplantation is both contentious and methodologically challenging. Various metrics to quantify disparity have been attempted, yet all of these have flaws that limit their accuracy in capturing the impact of geography on access to DDKT. Unadjusted DDKT rates and waiting times to transplant fail to account for patient casemix 1,13,14. Metrics based on time to DDKT among transplanted patients fail to account for the experience of patients who do not get transplanted 14. The range of waiting times for DDKT per DSA is sensitive to outliers 6. The Institute of Medicine, examining the Final Rule in liver transplantation, recognized the importance of selecting an appropriate disparity metric and concluded: “overall median waiting time, which has dominated the policy debate, is a poor measure of differences in access to transplantation. Status-specific rates of pretransplantation mortality and transplantation are more meaningful indicators of equitable access 3.” As such, the goal of this study was to calculate DDKT rates before and after KAS, adjusted for casemix and allocation factors that should impact DDKT rate (i.e. intentionally introduced by the allocation system), in accordance with the Institute of Medicine’s recommendation, incorporating the experience of both transplanted and censored patients, using a novel method that is robust to outlier values.
METHODS
Data Source
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 US, submitted by the members of the Organ Procurement and Transplantation Network (OPTN), and has been described elsewhere 15. The Health Resources and Services Administration (HRSA), U.S. Department of Health and Human Services provides oversight to the activities of the OPTN and SRTR contractors.
Study Population
Our study population consisted of 97862.87 active person-years on the kidney-only waitlist between June 1, 2015 and December 1, 2016 (“under KAS”). Inactive person-years were excluded because inactive candidates were not eligible to receive DDKT. That is to say, if a candidate was active, then inactive, then active again, only the active time on waitlist was included. We have also excluded person-years in medical urgency status. Since prior studies have demonstrated a “bolus effect” of newly-prioritized candidates receiving transplants during the first six months after KAS implementation, we excluded the first six months of time following policy implementation 16. Candidates were censored at waitlist removal for a reason besides receiving a transplant (died, too sick to be transplanted, refused transplant, et cetera).
Model Specification
We constructed an adjusted multilevel Poisson model with random intercepts at the DSA level to estimate a separate, casemix-adjusted DDKT rate for each DSA 17,18. The outcome measure was DDKT rate, defined as number of DDKT per active person-year. In a Poisson model, active person-years were specified as an offset. We adjusted for all other factors influencing allocation to estimate the impact of DSA on DDKT rate independent of casemix. Because allocation policies differ for adult (≥ 18 years) and pediatric (<18 years) candidates, we separated models for these two groups. Covariates included Calculated Panel Reactive Antibody (CPRA, categorized as 0–19%, 20–94%, 95–97%, 98%, 99%, 100%), waiting time (as defined by KAS), ABO blood type. Waiting time was calculated according to OPTN policy, accounting for both creatinine clearance, eGFR ≤20mL/min, and time on regularly administered dialysis 11. In the model for adult candidates, we also included Estimated Post Transplant Survival (EPTS) score (categorized as ≤20th percentile, >20th percentile) and whether the candidate was a prior living donor 11. CPRA, waiting time, and EPTS score were modeled as time-varying covariates that were re-calculated at least every 30 days. We performed a complete case analysis excluding person-years with missingness on EPTS score (1.14%) for adult candidates and waiting time (0.97%) for adult and pediatric candidates.
Incidence Rate Ratio (IRR) and Median Incidence Rate Ratio (MIRR)
Incidence rate ratio (IRR) is the ratio of DDKT rates between two groups with different values of a covariate. We analyzed median incidence rate ratio (MIRR) - a measurement of between-cluster variation - in a multilevel framework for DDKT rate 19–23. MIRR of DDKT rate is derived from comparing DSA-level random effects, and can be interpreted as the median of the DDKT rate IRRs for all pairs of DSAs when comparing two candidates with similar allocation points 24,25. In other words, if we consider two random candidates with similar allocation points from two different DSAs, and compare the candidate in the DSA with greater access to transplantation to the candidate in the DSA with less access, we get an incidence rate ratio for DDKT for this pair of DSAs. We then calculated the IRRs of all DSA pairs, comparing the higher DDKT rate to the lower DDKT rate. MIRR is the median ratio of DDKT rates over all pairs of DSAs. We showed distribution of IRRs and MIRR using a histogram. Estimated DDKT rate was obtained by empirical Bayes estimates for each DSA from the adjusted multilevel Poisson regressions. IRRs were calculated by comparing the higher DDKT rate to the lower DDKT rate for each pair of DSA.
MIRR can be interpreted as the median increase in DDKT rate that would occur if an individual moved from an DSA with less access to an DSA with greater access to DDKT. An adjusted MIRR represents the geographic disparity in DDKT rates for candidates with the same allocation priority. The 95% confidence intervals of MIRR were obtained using a bootstrap with 200 repetitions that resampled registrants in each DSA with replacement.
Comparing Geographic Disparities Pre-KAS and under KAS
In a separate analysis, we compared geographic disparity under KAS vs. pre-KAS (6/1/2013–12/3/2014). We calculated unadjusted (observed) DDKT rates for each DSA pre-KAS and under KAS and displayed them using a scatterplot and color-scaled maps. On color-scaled maps, darker colors represented higher DDKT rates, while lighter colors represented lower DDKT rates. We were forced to use unadjusted models to measure the change in geographic disparities under KAS, because the allocation factors adjusted for in the pre-KAS era vs. under KAS differ. Adjusting for different factors by time period would prevent direct comparisons of MIRR pre-KAS vs. under KAS. Because there was evidence of increased sharing, which is likely due to increased imports for highly sensitized candidates, we examined geographic disparities among this subgroup of candidates using the same methods stated above 12,16. In addition, we calculated the number of DDKT and the DDKT rate for each of the region pre-KAS and under KAS. The statistical significance of the change in MIRR from pre-KAS to under KAS was assessed using a bootstrap with 100 repetitions that resampled registrants in each DSA with replacement. The two-sided p-values were derived empirically from the distribution of bootstrapped results, as p-value = 2 - 2 * max [proportion of times that bootstrapped pre-KAS MIRR was higher, proportion of times that bootstrapped post-KAS MIRR was higher].
Sensitivity Analysis
In a sensitivity analysis, we compared geographic disparity under KAS that included all active person time after implementation of KAS (including the “bolus” time period) with that pre-KAS. In addition, we repeated analyses using a “bolus” period of 1 month and 3 months.
Statistical Analysis
Confidence intervals were reported as per the method of Louis and Zeger 26. All analyses were performed using Stata 14.1/MP for Windows (College Station, Texas). All maps were generated using R 3.3.3 GUI for Mac OS.
RESULTS
Study Population
In total, 120,557 registrants (118,926 adult and 1,631 pediatric candidates) were included in our analysis, comprising 97862.87 active person-years, with baseline characteristics shown in Table 1. Our comparison population included 96575.69 active person-years pre-KAS, and 98643.30 active person-years under KAS.
Table 1. Kidney Waitlist Candidate Characteristics at Baseline under KAS.
Baseline is the first time after June 1, 2015 candidates were observed active on the waitlist. Each candidate had multiple person-year records, each less than 30 person-days. Time-varying characteristics including age, CPRA, waiting time, EPTS score changed across person-year records.
| Adult | Pediatric | |
|---|---|---|
| N | 118,926 | 1,631 |
|
| ||
| Age in years | ||
| Median (IQR) | 54 (44–63) | 13 (7–16) |
| Range | 18–90 | 0–17 |
|
| ||
| Female, N (%) | 45,450 (38.2%) | 699 (42.9%) |
|
| ||
| Race/ethnicity, N (%) | ||
| White | 46,135 (38.8%) | 679 (41.6%) |
| African-American | 38,029 (32.0%) | 368 (22.6%) |
| Hispanic | 22,745 (19.1%) | 461 (28.3%) |
| Asian | 9,701 (8.2%) | 78 (4.8%) |
| Other | 2,316 (1.9%) | 45 (2.8%) |
|
| ||
| Blood type, N (%) | ||
| O | 61,310 (51.6%) | 824 (50.5%) |
| A | 35,068 (29.5%) | 491 (30.1%) |
| B | 19,028 (16.0%) | 264 (16.2%) |
| AB | 3,520 (3.0%) | 52 (3.2%) |
|
| ||
| CPRA, N (%) | ||
| 0–19 | 82,959 (69.8%) | 1,231 (75.5%) |
| 20–94 | 23,639 (19.9%) | 257 (15.8%) |
| 95–97 | 2,138 (1.8%) | 17 (1.0%) |
| 98 | 1,222 (1.0%) | 19 (1.2%) |
| 99 | 2,182 (1.8%) | 23 (1.4%) |
| 100 | 6,786 (5.7%) | 84 (5.2%) |
|
| ||
| Waiting Time in years, median (IQR) | 2 (1–4) | 1 (0–2) |
|
| ||
| EPTS score ≤20th percentile, N (%) | 28,885 (24.3%) | N/A |
|
| ||
| Prior living donor, N (%) | 57 (<1%) | 0 (0%) |
Note: For adults, EPTS score was mapped to a reference population of all adult kidney candidates on the waiting list on December 31, 2016. Waiting time included years from regularly administered dialysis, creatinine clearance or eGFR≤20mL/min on or after registration, whichever came earlier.
Adjusted Multilevel Poisson Models under KAS
For adult candidates, the DDKT rate was 57% higher with each 3-year increase in waiting time (IRR=1.561.571.59, p<0.001; Table 2). The DDKT rate was 40% higher for candidates in the top 20% of EPTS scores (IRR=1.351.401.45, p<0.001) vs. those in lower percentiles. Compared to candidates with CPRA 0–19%, the DDKT rate was 17% lower for those with CPRA 99% (IRR= 0.760.830.91, p<0.001), and 46% lower for those with CPRA 100% (IRR=0.510.540.57, p<0.001). By blood type, DDKT rates were highest for candidates with type AB and A and lowest for candidates with blood type B. Candidates who had previously been living donors had a DDKT rate over 10 times higher than non-donor candidates (IRR= 7.0810.1414.53, p<0.001).
Table 2. Adjusted Rate of DDKT and MIRR for Adult Kidney Transplant Candidates under KAS.
MIRR was larger than the IRR of 3-year increase in waiting time and the IRR of having EPTS score ≤ 20%. This means for adult candidates, moving to a DSA with a higher transplant rate is associated with a median increase in the transplant rate that is higher than 3-year difference in waiting time and having EPTS score ≤ 20%.
| aIRR (95% CI) | p-value | |
|---|---|---|
| Waiting time, 3 years | 1.561.571.59 | <0.001 |
|
| ||
| EPTS score | ||
| >20th percentile | ref | |
| ≤20th percentile | 1.351.401.45 | <0.001 |
|
| ||
| CPRA (%) | ||
| 0–19 | ref | |
| 20–94 | 0.930.971.01 | 0.10 |
| 95–97 | 0.760.840.93 | <0.01 |
| 98 | 0.490.560.65 | <0.001 |
| 99 | 0.760.830.91 | <0.001 |
| 100 | 0.510.540.57 | <0.001 |
|
| ||
| Blood type | ||
| O | ref | |
| A | 1.611.671.72 | <0.001 |
| B | 0.870.910.95 | <0.001 |
| AB | 2.432.602.79 | <0.001 |
|
| ||
| Prior living donor | 7.0810.1414.53 | <0.001 |
|
| ||
| MIRR | 1.751.811.86 | |
Note: DDKT: deceased donor kidney transplantation, aIRR: adjusted incidence rate ratio, MIRR: median incidence rate ratio. EPTS score was mapped to a reference population of all adult kidney candidates on the waiting list on December 31, 2016. Waiting time included years from regularly administered dialysis, creatinine clearance or eGFR≤20mL/min on or after registration, whichever came earlier.
For pediatric candidates, the DDKT rate was 30% higher with each 3-year increase in waiting time (IRR=1.151.301.47, p<0.001; Table 3). The DDKT rate decreased dramatically as CPRA increased. Comparing to candidates with CPRA 0–19%, the DDKT rate was 56% lower for CPRA 20–94% (IRR=0.350.440.54, p<0.001), but was 97% lower for CPRA 99% (IRR=0.000.030.20, p<0.01), and 88% lower for CPRA 100% (IRR=0.070.120.18, p<0.001; Table 3). Similarly as adult candidates, by blood type, DDKT rates were higher for candidates with type AB, and lower for candidates with type B.
Table 3. Adjusted Rate of DDKT and MIRR for Pediatric Kidney Transplant Candidates under KAS.
MIRR was larger than the IRR of 3-year increase in waiting time. This means for pediatric candidates, moving to a DSA with a higher transplant rate is associated with a median increase in the transplant rate that is higher than 3-year difference in waiting time.
| aIRR (95% CI) | p-value | |
|---|---|---|
| Waiting time, 3 years | 1.151.301.47 | <0.001 |
|
| ||
| CPRA (%) | ||
| 0–19 | ref | |
| 20–94 | 0.350.440.54 | <0.001 |
| 95–97 | 0.030.090.29 | <0.001 |
| 98 | 0.070.170.43 | <0.001 |
| 99 | 0.000.030.20 | <0.001 |
| 100 | 0.070.120.18 | <0.001 |
|
| ||
| Blood type | ||
| O | ref | |
| A | 1.021.201.42 | 0.03 |
| B | 0.480.600.76 | <0.001 |
| AB | 1.211.782.63 | <0.01 |
|
| ||
| MIRR | 1.661.922.27 | |
Note: DDKT: deceased donor kidney transplantation, aIRR: adjusted incidence rate ratio, MIRR: median incidence rate ratio. Waiting time included years from regularly administered dialysis or registration on the waiting list (regardless of creatinine or eGFR levels), whichever came earlier.
Geographic Disparity under KAS
After adjusting for all the non-geographic factors included in the allocation algorithms, distribution of IRRs and MIRR were shown in Figure 1. MIRR was 1.751.811.86 for adult candidates and 1.661.922.27 for pediatric candidates (Table 2–3). In other words, if two candidates with similar allocation points were selected from two DSAs, the DDKT rate of the candidate with higher access to transplant would be a median of 81% higher than the DDKT rate of the other candidate if they were adults, and a median of 92% higher if they were pediatric candidates. The distribution of IRRs in Figure 1 shows that for some pairs of DSAs, a candidate could increase their DDKT rate twelve-fold by moving from the DSA with the lower rate to the DSA with the higher rate.
Figure 1. Distribution of IRRs and MIRR under KAS between June 1, 2015 and December 1, 2016.
DDKT rates were estimated for each DSA using empirical Bayes estimation. For each pair of DSAs, IRRs were the ratio of the higher DDKT rate to the lower DDKT rate and were always larger than 1. MIRR was the median of all these IRRs.
Comparing Geographic Disparities Pre-KAS and under KAS
The unadjusted DDKT rates for each DSA pre-KAS and under KAS are shown in Figure 2 and Figure 3. DDKT rates were heterogeneous both pre-KAS and under KAS (Figure 3). Areas with high DDKT rates pre-KAS continued to have high DDKT rates under KAS, and areas with low DDKT rates pre-KAS continued to have low DDKT rates under KAS (Figure 2). DDKT rates pre-KAS and under KAS had a correlation of 0.70. DDKT rate for each region was shown in Table 4. DDKT rate increased from pre-KAS to under KAS in 9 regions, and decreased in the other 2 regions.
Figure 2. Unadjusted (Observed) DDKT rate of DSAs Pre-KAS (6/1/2013–12/3/2014) and under KAS (6/1/2015–12/1/2016).
Each point represented an DSA. The diagonal line indicated identical DDKT rate pre-KAS and under KAS. Correlation between pre-KAS DDKT and under KAS DDKT was 0.70. DDKT rate was defined as number of DDKT per active person-years.
Figure 3. Unadjusted (Observed) DDKT Rate in the United States Pre-KAS (6/1/2013–12/3/2014) and under KAS (6/1/2015–12/1/2016), by DSA.
We calculated unadjusted DDKT rates for each DSA pre-KAS (A) and under KAS (B) and displayed them using color-scaled maps. Darker colors represented higher DDKT rates, while lighter colors represented lower DDKT rates. DDKT rates were heterogeneous both pre-KAS and under KAS.
Table 4. Number of DDKT and DDKT rate by region pre-KAS and under KAS.
DDKT rate was the number of DDKT per 100 active person-years. DDKT rate increased in 9 regions and decreased in 2 regions from pre-KAS to under KAS.
| Region | Pre-KAS | Under KAS | ||
|---|---|---|---|---|
| Number of DDKT | DDKT rate (per 100 PYs) | Number of DDKT | DDKT rate (per 100 PYs) | |
| 1 | 655 | 19.76 | 713 | 20.06 |
| 2 | 2157 | 15.87 | 2534 | 18.64 |
| 3 | 2343 | 17.85 | 2747 | 18.23 |
| 4 | 1622 | 14.22 | 1924 | 17.38 |
| 5 | 2993 | 15.13 | 3475 | 17.01 |
| 6 | 752 | 34.86 | 706 | 32.04 |
| 7 | 1295 | 15.26 | 1407 | 17.90 |
| 8 | 1226 | 30.76 | 1198 | 29.74 |
| 9 | 1050 | 15.35 | 1388 | 19.60 |
| 10 | 1373 | 20.60 | 1484 | 24.42 |
| 11 | 1900 | 26.01 | 2033 | 26.32 |
Unadjusted MIRR was 1.62 pre-KAS and 1.64 under KAS. There was no evidence of change in MIRR from pre-KAS to under KAS (p=0.3). For candidates with CPRA of 99 and 100%, the unadjusted DDKT rate increased for all DSAs. There was no evidence of change in geographic disparities for candidates with CPRA of 99 and 100% from pre-KAS to under KAS (p=0.9, Figure 4).
Figure 4. Unadjusted (Observed) DDKT Rate among Person-Years with CPRA ≥ 99% Pre-KAS (6/1/2013–12/3/2014) and under KAS (6/1/2015–12/1/2016), by DSA.
Because there was evidence of increased sharing, which is likely due to increased imports for highly sensitized candidates (CPRA ≥99%), we examined unadjusted DDKT rates for each DSA pre-KAS (A) and under KAS (B) among this subgroup of candidates. For candidates with CPRA of 99 and 100%, the unadjusted DDKT rate increased for all DSAs from pre-KAS to under KAS.
In the sensitivity analysis including all person-years after the implementation of KAS, we found no difference in MIRR pre-KAS (1.62) and under KAS (1.61, p=0.5). In addition, we found no difference in MIRR pre-KAS and under KAS in repeated analyses excluding “bolus” period of 1 month (p=0.5) or 3 months (p=0.8).
DISCUSSION
In this national registry study of kidney transplant candidates, we found that under KAS, geographic disparity in DDKT rate among DSAs remained large both for adult (MIRR=1.751.811.86) and pediatric candidates (MIRR= 1.661.922.27). For adults, while geographic disparity impacted DDKT rate by a median factor of 1.81, having EPTS score at or below 20th percentile only impacted DDKT rate by a factor of 1.40 (IRR=1.351.401.45, p<0.001) and an extra 3 years of waiting time impacted DDKT rate by a factor of 1.57 (IRR=1.561.571.59, p<0.001). For pediatric candidates, geographic disparity impacted DDKT rate more than any other allocation factors including waiting time, CPRA, and blood type. DDKT rates were highly correlated pre-KAS and under KAS (correlation=0.70). We found no change in geographic disparities from pre-KAS to under KAS.
We adopted a new measure of variation - MIRR - to investigate the geographic disparity in DDKT rate because it has several distinct advantages over other metrics of disparity 25. MIRR has a straightforward interpretation as the median factor by which DDKT rate would change if a candidate listed in a different DSA. Particularly, since the MIRR is interpretable as an incidence rate ratio, it is directly comparable to covariate effects - the IRRs of other factors influencing allocation 19,25. This allows us to compare disparities caused by geographic location to the differences in DDKT rate that are by design within the allocation system, e.g., moving to a DSA with a higher transplant rate is associated with a median increase in the transplant rate that is higher than 3 year difference in waiting time, or being in the top 20th percentile of EPTS score. No other known metric of disparity allows us to do such comparisons. Moreover, MIRR can be adjusted for covariates influencing allocation and MIRR summarizes geographic disparity as a single quantity, which facilitates hypothesis testing of whether disparity is impacted by a new allocation policy.
MIRR is informative for various stakeholders including patients, policy makers, healthcare professionals, and transplant researchers. Knowing the degree of geographic disparity in access to transplant helps patients and physicians assess the tradeoff between listing in another DSA and being on the waitlist for a longer time. The direct comparison between MIRR and the factors intended to influence allocation priority invites policy makers to consider whether allocation realities truly reflect the relative priority of various candidates for transplant. Researchers can extend this metric to study geographic disparity in the allocation of other organs, and to summarize trends in geographic disparity over time. MIRR can be calculated from the output of simulated allocation models to understand the effects of proposed allocation policy changes on geographic disparity. We urge the transplant community to consider MIRR as a standard measure of geographic disparity.
Geographic disparities in access to transplant are persistent and unacceptably high. In 2012, the United Network for Organ Sharing (UNOS) board called for “optimized systems” to reduce these disparities. In liver transplantation, broader sharing and redistricting have been proposed to address geographic disparity 27,28. In kidney transplantation, efforts to improve geographic equity are less developed 29, and the usual concerns about increased transport costs and possibly increased cold ischemia times with broader sharing have been expressed 16,30. Mitigating geographic disparity in kidney transplant was explicitly excluded as a goal of KAS 10,16. Accordingly, we found that geographic disparity remains severe overall following the implementation of KAS. Geographic location has a larger impact on DDKT rate than EPTS scores at or below the 20th percentile, or 3-year waiting time. This seems to contradict the “Final Rule”, which states that neither place of residence nor place of listing should be a major determinant of access to a transplant. 31.
We acknowledge some limitations of our study. The DDKT rate is influenced by variation in access to the transplant waiting list. Geographic disparity in access to the transplant waiting list should be studied, but is outside the scope of our analysis. Our study might be confounded by factors not captured in national registry data. However, we explored all the covariates collected by OPTN. Additionally, DSA-level geographic disparity might be partly mediated by center-level variation in practice; centers that are more aggressive in accepting marginal organs might have higher DDKT rates without greater organ availability 32. However, greater aggressiveness is usually seen at centers with longer waiting times and worse organ shortage 32. This suggests that disparities in DDKT rates between DSAs with aggressive versus non-aggressive centers are likely underestimated, because aggressive behavior bolsters DDKT rates. Finally, we did not adjust for allocation factors when comparing DDKT rates pre-KAS and under KAS. We were forced to use an unadjusted model because the allocation system changed greatly under KAS, and we cannot compare MIRR between different models adjusted for different sets of allocation factors. Not adjusting for allocation factors likely underestimates geographic disparity, so again our findings comparing pre-KAS to under KAS are conservative, but still reveal persistent and striking geographic disparity.
Our study quantified the substantial geographic disparities in DDKT rate in the KAS era. Using a novel metric - MIRR, we found that DSA of listing remains a major determinant of transplant rate under KAS. The impact of geography on transplant rate was larger than the impact of factors emphasized by KAS. Targeted policy changes would be needed to reduce geographic disparity.
Supplementary Material
Acknowledgments
This work was supported by grant number 1R01DK111233-01 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). The analyses described here are the responsibility of the authors alone and do not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the U.S. Government. The data reported here have been supplied by the Minneapolis Medical Research Foundation (MMRF) as the contractor for the Scientific Registry of Transplant Recipients (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 the SRTR or the U.S. Government.
Abbreviations
- KAS
Kidney Allocation System
- DDKT
Deceased Donor Kidney Transplantation
- IRR
Incidence Rate Ratio
- MIRR
Median Incidence Rate Ratio
- OPTN
Organ Procurement and Transplant Network
- SRTR
Scientific Registry of Transplant Recipient
- UNOS
United Network for Organ Sharing
- DSA
Donation Service Areas
- CPRA
Calculated Panel Reactive Antibody
- EPTS
Estimated Post Transplant Survival
- KDPI
Kidney Donor Profile Index
Footnotes
DISCLOSURE
The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.
Additional Supporting Information may be found in the online version of this article.
References
- 1.Davis AE, Mehrotra S, Ladner DP, Kilambi V, Friedewald JJ. Changes in geographic disparity in kidney transplantation since the final rule. Transplantation. 2014;98(9):931–936. doi: 10.1097/TP.0000000000000446. [DOI] [PubMed] [Google Scholar]
- 2.Mathur AK, Ashby VB, Sands RL, Wolfe RA. Geographic Variation in End-Stage Renal Disease Incidence and Access to Deceased Donor Kidney Transplantation. American Journal of Transplantation. 2010;10(4p2):1069–1080. doi: 10.1111/j.1600-6143.2010.03043.x. [DOI] [PubMed] [Google Scholar]
- 3.Organ Procurement and Transplantation: Assessing Current Policies and the Potential Impact of the DHHS Final Rule. Washington (DC): 1999. [PubMed] [Google Scholar]
- 4.Ashby VB, Kalbfleisch JD, Wolfe RA, Lin MJ, Port FK, Leichtman AB. Geographic variability in access to primary kidney transplantation in the United States, 1996–2005. Am J Transplant. 2007;7(5 Pt 2):1412–1423. doi: 10.1111/j.1600-6143.2007.01785.x. [DOI] [PubMed] [Google Scholar]
- 5.Axelrod DA, McCullough KP, Brewer ED, Becker BN, Segev DL, Rao PS. Kidney and pancreas transplantation in the United States, 1999–2008: the changing face of living donation. Am J Transplant. 2010;10(4 Pt 2):987–1002. doi: 10.1111/j.1600-6143.2010.03022.x. [DOI] [PubMed] [Google Scholar]
- 6.Davis AE, Mehrotra S, McElroy LM, et al. The extent and predictors of waiting time geographic disparity in kidney transplantation in the United States. Transplantation. 2014;97(10):1049–1057. doi: 10.1097/01.tp.0000438623.89310.dc. [DOI] [PubMed] [Google Scholar]
- 7.Matas AJ, Smith JM, Skeans MA, et al. OPTN/SRTR 2013 Annual Data Report: kidney. Am J Transplant. 2015;15(Suppl 2):1–34. doi: 10.1111/ajt.13195. [DOI] [PubMed] [Google Scholar]
- 8.USRDS Annual Data Report. 2017 https://www.usrds.org/adr.aspx2017.
- 9.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. 2014;25(8):1842–1848. doi: 10.1681/ASN.2013070784. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. [Accessed 3/1/2017];The New Kidney Allocation System (KAS) Frequently Asked Questions. 2014 https://optn.transplant.hrsa.gov/media/1235/kas_faqs.pdf.
- 11. [Accessed 6/29/2017, 2017];Organ Procurement and Transplantatio Network (OPTN) Policies. 2017 https://optn.transplant.hrsa.gov/media/1200/optn_policies.pdf#nameddest=Policy_08.
- 12.Massie AB, Luo X, Lonze BE, et al. Early Changes in Kidney Distribution under the New Allocation System. J Am Soc Nephrol. 2016;27(8):2495–2501. doi: 10.1681/ASN.2015080934. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Couchoud C, Stengel B, Landais P, et al. The renal epidemiology and information network (REIN): a new registry for end-stage renal disease in France. Nephrol Dial Transplant. 2006;21(2):411–418. doi: 10.1093/ndt/gfi198. [DOI] [PubMed] [Google Scholar]
- 14.Davis AE, Mehrotra S, Kilambi V, et al. The effect of the Statewide Sharing variance on geographic disparity in kidney transplantation in the United States. Clin J Am Soc Nephrol. 2014;9(8):1449–1460. doi: 10.2215/CJN.05350513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Massie AB, Kucirka LM, Segev DL. Big data in organ transplantation: registries and administrative claims. Am J Transplant. 2014;14(8):1723–1730. doi: 10.1111/ajt.12777. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Stewart DE, Kucheryavaya AY, Klassen DK, Turgeon NA, Formica RN, Aeder MI. Changes in Deceased Donor Kidney Transplantation One Year After KAS Implementation. Am J Transplant. 2016;16(6):1834–1847. doi: 10.1111/ajt.13770. [DOI] [PubMed] [Google Scholar]
- 17.Rabe-Hesketh SSA. Multilevel and Longitudinal Modeling using Stata. 2. College Station, TX: Stata Press Publication; 2008. [Google Scholar]
- 18.Carlin BP, Louis TA. Bayes and empirical Bayes methods for data analysis. 2000 [Google Scholar]
- 19.Larsen K, Merlo J. Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol. 2005;161(1):81–88. doi: 10.1093/aje/kwi017. [DOI] [PubMed] [Google Scholar]
- 20.Merlo J, Chaix B, Ohlsson H, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60(4):290–297. doi: 10.1136/jech.2004.029454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Pruitt SL, Schootman M. Geographic disparity, area poverty, and human papillomavirus vaccination. Am J Prev Med. 2010;38(5):525–533. doi: 10.1016/j.amepre.2010.01.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Austin PC, Merlo J. Intermediate and advanced topics in multilevel logistic regression analysis. Stat Med. 2017;36(20):3257–3277. doi: 10.1002/sim.7336. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rosano A, Ricciardi W, van der Zee J. Analysis of the role of General Practice in preventing Avoidable Hospitalisation through a multilevel approach. Epidemiol Biostat Public Health. 2016;13(2):10. [Google Scholar]
- 24.Ravanan R, Udayaraj U, Ansell D, et al. Variation between centres in access to renal transplantation in UK: longitudinal cohort study. BMJ. 2010;341:c3451. doi: 10.1136/bmj.c3451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Austin PC, Wagner P, Merlo J. The median hazard ratio: a useful measure of variance and general contextual effects in multilevel survival analysis. Stat Med. 2016 doi: 10.1002/sim.7188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Louis TA, Zeger SL. Effective communication of standard errors and confidence intervals. Biostatistics. 2009;(1):1–2. doi: 10.1093/biostatistics/kxn014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gentry SE, Massie AB, Cheek SW, et al. Addressing geographic disparities in liver transplantation through redistricting. Am J Transplant. 2013;13(8):2052–2058. doi: 10.1111/ajt.12301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Gentry SE, Chow EK, Wickliffe CE, Massie AB, Leighton T, Segev DL. Impact of broader sharing on the transport time for deceased donor livers. Liver Transpl. 2014;20(10):1237–1243. doi: 10.1002/lt.23942. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Davis AE, Mehrotra S, Friedewald JJ, et al. Improving Geographic Equity in Kidney Transplantation Using Alternative Kidney Sharing and Optimization Modeling. Med Decis Making. 2015;35(6):797–807. doi: 10.1177/0272989X14557696. [DOI] [PubMed] [Google Scholar]
- 30.Bray RA, Gebel HM. The new kidney allocation system (KAS) and the highly sensitized patient: expect the unexpected. Am J Transplant. 2014;14(12):2917. doi: 10.1111/ajt.12974. [DOI] [PubMed] [Google Scholar]
- 31.Final Rule 2000. 2016 https://optn.transplant.hrsa.gov/governance/about-the-optn/final-rule/
- 32.Garonzik-Wang JM, James NT, Weatherspoon KC, et al. The aggressive phenotype: center-level patterns in the utilization of suboptimal kidneys. Am J Transplant. 2012;12(2):400–408. doi: 10.1111/j.1600-6143.2011.03789.x. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.