Population-Based Analysis and Projections of Liver Supply Under Redistricting : Transplantation

Journal Logo

Original Clinical Science—Liver

Population-Based Analysis and Projections of Liver Supply Under Redistricting

Parikh, Neehar D. MD, MS1; Marrero, Wesley J. BS2; Sonnenday, Christopher J. MD3,4; Lok, Anna S. MD1; Hutton, David W. PhD2,4; Lavieri, Mariel S. PhD2

Author Information
doi: 10.1097/TP.0000000000001785

Liver transplantation (LT) is a lifesaving therapy for patients with end-stage liver disease or hepatocellular carcinoma. Unfortunately, far more patients are in need of LT than there are organs available for transplantation.1 Thus, up to 20% of listed patients are removed from the waitlist due to clinical deterioration or death.2

The US Department of Health and Human Services, in its capacity governing the Organ Procurement and Transplantation Network (OPTN), has published the “final rule” in organ transplantation, a clause of which states that organ allocation should not be based on the potential recipient's place of residence or place of listing. The policy mandates expanding transplantation over as broad an area as feasible to maximize equity from a geographic perspective.3 Despite this, due to historical precedent in how donor service areas (DSAs) were established, the chances of receiving an LT vary widely depending on which region of the country a patient is listed in. The median Model of End-Stage Liver Disease (MELD) score at transplant can vary from 22 to 32 depending on which of the 11 regions a patient is listed in nationally.4 An analysis by Massie et al, showed that the 90-day likelihood of being transplanted in the intermediate MELD range (18-30 years) can vary from 15% to 70% between regions.5 This in turn translates to wide variations in risk of death while awaiting LT.5,6

This geographic inequity has led to recent policy changes designed to decrease geographic heterogeneity within the current regional allocation system. The “Share 35” policy was implemented in June 2013 to decrease the discrepancies in waitlist dropout of patients with the highest MELD scores between regions, with early results showing a decrease in mortality in this high MELD group (>35) and a slight decrease in adjusted overall waitlist mortality.7 However, significant geographic heterogeneity in LT remains due to imbalance in organ supply and demand throughout the country. There are limited projections about how population changes may change supply over the next decade,8 and how this may impact geographic inequity of supply between regions. The root cause of geographic inequity in donor availability is related to several factors—population demographic differences, regional differences in causes of death, organ procurement organization (OPO) activity, and variations in transplant center aggressiveness in organ utilization all likely play a role; however, there is uncertainty how much each factor is responsible for the inequity. A recently published study showed that increasing DSA authorization rates of eligible deaths in low-performing DSAs could improve geographic heterogeneity in donor availability.9 Increasing authorization rates requires systematic approaches to improve performance of suboptimally performing OPOs. Prior attempts at systematic improvement in application of best practices for improving organ supply from the Organ Donation Breakthrough collaborative resulted in improvement in donor conversion rates in low-performing hospitals and an overall increase in organs available for transplant.

To further decrease the geographic inequity in liver supply and demand nationally, there is a proposal to geographically redistrict the United Network of Organ Sharing (UNOS) regions from the current 11-region model into an 8-region model (Figure 1).10,11 A 5-year liver simulation allocation model of the regional proposal indicates that a net decrease in waitlist deaths might be achieved when compared with the current 11-region model.11 Although this analysis provides valuable insight into the impact of the redistricting proposal on waitlist mortality, it does not account for population demographic shifts in the United States, such as changes in age, obesity rates, and racial distribution, all of which vary regionally and have an impact on liver organ availability.8 In addition, demand for LT may fluctuate, given the aging US population and better therapies for hepatitis C, therefore the numbers of donors as a function of the total population may provide additional insight into geographic inequities in liver supply.12

F1-24
FIGURE 1:
Proposed 8-region allocation map from the UNOS.

We aimed to use a population-based approach to understand the impact of redistricting on the number of donors available per population and how this inequity will be impacted by the projected population and demographic changes in the United States. We also aimed to better understand the demographic and OPO-based variables that may contribute to geographic inequity.

MATERIALS AND METHODS

Liver Donor Projections by District Model

Historical Population

Data from the US Census Bureau was used to obtain 2002 to 2014 national population estimates. We used 2 data sets from the US Census Bureau for this purpose: (1) intercensal estimates of the resident population by single year of age, sex, race, and Hispanic origin: April 1, 2000, to July 1, 2010, and (2) monthly population estimates by age, sex, race, and Hispanic origin: April 1, 2010, to July 1, 2014. The national population estimates were used to calculate national donation rates.

Population by state and county were also derived from data from the US Census Bureau. Three data sets were used for this purpose: (1) intercensal estimates of the resident population by 5-year age groups, sex, race, and Hispanic origin: April 1, 2000, to July 1, 2010; (2) annual state resident population estimates for 6 race groups by age, sex, and Hispanic origin: April 1, 2010, to July 1, 2014; and (3) annual county resident population estimates by age, sex, race, and Hispanic origin: April 1, 2010, to July 1, 2014. These data sets were used to calculate donation rates by county and state, and subsequently, region.

The 8-region map was based on the proposed regions from UNOS/OPTN (Figure 1). Once divided by region and grouped by regional model (8 or 11), the data were stratified by race, sex, and age group (18-34; 35-49; 50-64; 65-84). The 18 to 34 years age group of our models overlapped between 2 different age groups in the US Census Bureau data set from 2002 to 2009. We accounted for this by estimating the 18 to 34 years age group by region based on the proportion of the remaining age groups by region (35-49; 50-64; 65-84) to the total population of the 35 to 84 years age group across all regions. Due to the lack of availability of county population data from 2002 to 2009, we assumed that the population of the states that extended along more than 1 region was divided per the average proportion of the population (relative to the total state population), which lived in those counties from 2010 to 2014.

Population Projections

We used data from the University of Virginia's Weldon Cooper Center for Public Service on state population projections from 2010 to 2030 (http://www.coopercenter.org/demographics/national-population-projections) to estimate the total population projection that corresponded to each OPTN region. The population projections were stratified by district model, race, sex, and age group. Because population projections by state only included information for years 2010, 2020, and 2030, the population projections from 2015 to 2019 and 2021 to 2025 were estimated assuming linear growth. The population projections of states that extend across more than 1 region in the different regional models were distributed per the average proportion of the 2010 to 2014 population that would have extended across the region. The proportion of the population projections by region was then applied to the population projections originally obtained using the US Census Bureau data to maintain consistency on the total population projected.

The national prevalence of obesity was obtained from National Health and Nutrition Examination Survey and was held constant across all regions and regional models, given recent evidence that obesity rates in the United States have stabilized.13 The details of the populations projection development is detailed in the Supplemental Material, SDC (https://links.lww.com/TP/B439).

Donor Availability

We performed a secondary analysis of the OPTN database from 2002 to 2014. We determined the utilization rates of whole and split livers for all donors with at least 1 organ transplanted. We calculated donation rate as the number of livers donated per total population stratified by age groups, body mass index (BMI) (<30 or > =30), sex, and race/ethnicity. Donors were regionally localized based on their registered zip code in the OPTN database. We used donor permanent address rather than site of procurement to maintain consistency in our projections. Within the OPTN database, only 6.4% of patients had a regional discrepancy between permanent residence and site of procurement.

Historical Analysis by OPO

To further the granularity of our analysis, we performed a historical analysis on the comparative performance of each OPO by identifying their service areas using a report from the Health Resources and Service Administration Data Warehouse.14 The details of this subanalysis are included in the Supplemental Material, SDC,https://links.lww.com/TP/B439.

We also analyzed the impact of donor demographic factors (race, age, BMI, sex) by OPO. We used transformed beta regressions to test the influence of each demographic predictor on donation and liver utilization rates. We also calculated R2 values to determine the ability of demographic factors to explain donation and utilization.15-17 We tested the statistical significance regression models using the likelihood ratio test.18

Exploratory Analysis of Geographic Inequity

To better understand the main contributors to geographic inequity, we performed an analysis where we standardized liver donation and utilization rates across the regions in both regional models by using the 2010 to 2014 mean national donation rates, mean national utilization, and both mean national donation and utilization rates. We then used the coefficient of variation to calculate geographic variation per D/100K in each regional allocation model.

RESULTS

Regional Historical Donor Availability

From 2002 to 2014, the absolute number of liver donors grew steadily from 4473 to 6376; however, the D/100K peaked in 2006 at 2.79 and decreased to 2.67 in 2014 (Table 1). On average, the highest donor availability was seen in region 3 during the period studied. In 2014, the regional D/100K ranged from 1.84 in region 9 to 3.26 in region 2. The magnitude of regional variation over the 14-year study period varied from 16.6% to 25.9% with significant heterogeneity on a year-to-year basis. The geographic variation fluctuated across the years equaling 16.6% in 2002, 18.1% in 2014, and reaching a peak of 25.9% in 2009.

T1-24
TABLE 1:
Historic population based deceased donor liver donation and geographic variation

Regional Projected Donor Availability

Figure 2 shows the projected donor availability using national projections of population growth and demographic changes. The D/100K is projected to decrease nationally over time from 2.53 in 2016 to 2.49 in 2025. Due to population growth, however, the overall number of liver donors is projected to increase from 6133 in 2016 to 6507 in 2025.

F2-24
FIGURE 2:
Projected national changes in donors/100 000 US population and Total liver donors available from 2016 to 2025.

Geographic Inequity Between Regions in the Allocation Models

The regional variation in D/100K is projected to decrease slightly from 20.3% in 2016 to 20.2% in 2025 for the 11- region model. In contrast, the proposed 8-region model is projected to slightly increase geographic heterogeneity from 13.2% in 2016 to 13.3% by 2025 (Figure 3).

F3-24
FIGURE 3:
Projected change in regional variance in donors/100 000 US population in each allocation model from 2016 to 2025.

OPO Analysis

Historical national coefficient of variation at the level the level of the OPOs (Table S1, SDC,https://links.lww.com/TP/B439) between 2002 and 2014 ranges from 21.3% to 30.8%. When compared with national rates in Table 1, the OPO level variation is higher. Table 2 depicts the within region coefficient of variation of donation by OPO. We see marked heterogeneity when comparing OPO performance within each region, with regions 1, 3, 4, 10, and 11 having between OPO variations of less than 20% and regions 2, 5, 6, 7, 8, and 9 having variations greater than 20% and up to 36%. Table S2, SDC (https://links.lww.com/TP/B439) shows the donation and utilization rates in the best- and worst-performing OPOs nationally by donation rates. Here, we again see wide disparities between OPOs that contribute to the observed region-based geographic disparities. There is wide fluctuation in utilization rates, even in the poor-performing OPOs; however, donation rates remain consistently low throughout the years.

T2-24
TABLE 2:
Coefficients of variation of donation by regional OPO

The results of our multivariate regression of the impact of demographic variables on donation and utilization rates, accounting for OPO, are shown in Table S3, SDC (https://links.lww.com/TP/B439). Male sex, white race, and younger donor age are associated with increased donation and utilization, whereas BMI of 30 or greater, nonwhite race, and older age are negatively associated with donation and utilization. The R2 values for the demographic variables explaining donation utilization were 25.7% and 31.7%, respectively, indicating they are poor of both donation and utilization.

Exploratory Analysis of Geographic Inequity

We sought to better understand the magnitude of regional disparity in donation and utilization rates. Figure 4A shows the heterogeneity in regional utilization rate from 2002 to 2014 (range, 69.7%-94.7%), and Figure 4B shows the heterogeneity in regional liver donation rate (range, 1.83-4.16 D/100K). The historical rates of utilization and donation are shown in Table 3. The lowest average donation rates for the period 2010 to 2014 were seen in regions 9, 6, and 1 (2.30, 2.36, and 2.47 D/100K, respectively), whereas the highest average donation rates were in regions 2 and 8 (3.82 and 3.60 D/100K, respectively). The lowest average utilization rates from 2010 to 2014 were seen in regions 1 and 6 (74.0% and 77.1%, respectively), whereas the highest average utilization rates were seen in regions 3 and 11 (88.7% and 84.2%, respectively). Notably, there was wide variation in utilization rates from year-to-year such that even in the regions with the lowest average utilization rates, there were years where utilization rates in these regions were near or above national averages. However, this was not seen with donation rates, where the lowest performing regions (1, 6, and 9) never reached national average donation rates even in their years with the highest donation rates.

F4-24
FIGURE 4:
A, Regional variation in liver utilization from 2002 to 2012. The vertical line represents the range and the horizontal line represents the mean for the period 2010 to 2014. B, Regional variation in liver donation from 2002 to 2012. The vertical line represents the range and the horizontal line represents the mean for the period 2010 to 2014.
T3-24
TABLE 3:
Regional average donation and utilization rates

Table 4 shows the effect of using the 2010 to 2014 national average rates of donation, utilization, and both utilization and utilization on geographic variance in D/100K in both regional models. Using national average utilization rates would reduce the geographic heterogeneity in 2016 by 3.0% to 17.3% in the current 11-region model and by 2.2% to 11.0% in the 8-region model. Standardizing national donation rates would reduce geographic heterogeneity by 11.5% to 8.8% in the current 11-region model and by 5.7% to 7.5% in the 8-region model. Using both national average donation and utilization rates would have the greatest impact in reducing geographic heterogeneity, reducing geographic variance to 4.6% in the current model and 4.9% in the 8-region model. Applying national average utilization and donation rates in our projection model have similar impact on regional heterogeneity in both regional models in 2025 as in 2016.

T4-24
TABLE 4:
Effect of standardization of liver utilization rates, liver donation rates, and both on geographic variation in donors/100 000 US population

DISCUSSION

Regional variation in LT leads to inequity in organ distribution and excess waitlist dropout. The 2 components of LT inequity included imbalance in supply of livers and demand for transplantation. In this analysis, we explored regional variations in donor supply and how potential redistricting may impact the disparity in regional supply. Using a population-based approach, we have shown that the geographic inequity in donor availability will be decreased with the 8-region model. This finding confirms the notion that larger regions will lead to less geographic inequity in liver graft supply. The geographic inequity in the next decade is projected to fall slightly in the current 11-region model with expected demographic changes in the US population, whereas, in contrast, geographic inequity is projected to increase slightly in the 8-region system. These changes are projected to be minor in both the 11 and 8-region allocation models, largely reflective of the fact that demographic changes play a minor role in geographic organ supply disparities. We have shown that OPO-based donation and utilization rates are the major drivers in disparity in organ supply. Although our modeling confirms the desired effects of redistricting on improvement in geographic inequity in supply, there are many barriers to implementing redistricting. These challenges must be weighed against the goal of decreasing waitlist mortality and the relatively urgent need to address geographic disparities in LT.

In our exploratory analysis to better understand the root cause of geographic inequity, we found that standardization of donation and utilization rates would lead to a greater reduction in geographic heterogeneity than redistricting. We found that variation in liver donation rates plays a bigger role in contributing to geographic inequity than utilization, suggesting that improving OPO activity to improve donation rates in lower-performing regions may be impactful to improve geographic inequity. Similarly, focused interventions to improve transplant center organ utilization in low-performing regions, would also aid in reducing inequity. However, improving donation and utilization rates are not easy tasks to accomplish. A previous attempt by the Organ Donation Breakthrough Collaborative to increase in donor conversion rates by instituting best practices in identifying and consenting potential organ donors; however, the sustainability of these efforts has been uneven.19,20 The Collaborative included 131 hospitals and before intervention conversion rates in participating sites was 51.5% (range, 21%-88%) which rose to 65% after the intervention was completed.21 Follow-up analysis of the long-term impact of the collaborative efforts have shown sustained improvement in consent rates; however, certain demographic subgroups of patients have seen little or no change in conversion rates.19 This indicates that although effective in improving overall donation rates, ongoing interventions to improve conversion rates could be effective in further improving OPO performance.

Based on our OPO-based analysis there is wide OPO-based variation in low-performing (6 and 9) and high-performing (2 and 8) regions. This suggests interventions should be targeted at the OPO level. OPO-directed efforts to increase donation should be initially directed toward those regions with the lowest overall donation rates, thus having the greatest impact on geographic variation. Improving donation rates in low-performing regions will likely require a concerted sustained effort, because these regions have consistently lower donation rates than national averages in our historical data. Systematic evaluation of donor recruitment practices in these OPOs may be an important first step in improving donor authorization rates. Such interventions, if successful, would also lead to an overall increase in number of donors and number of organs used—ultimate goals that would not be achieved with redistricting. In fact, there is some evidence from the LSAM analysis of redistricting, there may be a net decrease in the number of transplants conducted.22 Our historical data show that even low-performing regions have at some point been near or above national averages in liver utilization, indicating that every region can perform at a higher level and measures to consistently improve practices in these regions can be successful. There does appear to be fluctuation in historical utilization rates by region that appears to be somewhat random. Even in our analysis of the best- and worst-performing OPOs in the United States, the utilization rate in the worst performing one varied significantly from 91.8% to 79.7% from 2012 to 2013. Several converging factors, including transplant center personnel, DSA competition, and mix of eligible deaths in any given year may play a role in utilization. Thus, utilization may be difficult to systematically improve given this random fluctuation even in low-performing OPOs. By contrast, donation rates in the low- and high-performing OPOs seen in Table S2, SDC (https://links.lww.com/TP/B439) show relatively consistent performance over the years. Variations in regional demographics accounts for the remainder of the geographic inequity; however, our data show that demographic factors alone are poor predictors of donation and utilization (R2 = 25.7% and 31.7%, respectively).

Our study has many strengths and weaknesses that warrant attention. We compiled unique data elements from many sources to estimate historical and projected donor availability, and thus many of our input variables are dependent on single source data and thus we are not able to perform robust sensitivity analyses around our projections. In addition, given the methodology used, we are not able to perform statistical comparisons of each regional model in our projections. We also used the donor zip code to identify their region, which does not account for donor regional sharing.23 It is unclear to what extent sharing would occur in the 8-region system or change in coming years, thus we used the donor registration site as a proxy for donor availability. We also did not use eligible deaths as defined by the OPTN as a proxy for potential donor availability in our analysis, given the increasing proportion of donors who may fall out of that definition (ie, age > 70 years) because the population changes in the coming decade. Our population-based analysis and projections are based on historical donation trends, thus donor authorization rates (donors/eligible deaths) are implicit in this analysis. The projections of liver availability assume no major changes in donor availability or breakthroughs in procurement technologies that could dramatically increase liver utilization rates. We do not anticipate approval or widespread adoption of new technologies during the relatively short time horizon of this study. Similarly, we assumed that there would be little change in the factors that predict local organ availability (ie, obesity, proportion of deaths to cerebrovascular event or trauma, and so on). There has been a recent increase in donor availability due to the national epidemic opioid overdoses, a trend which bears monitoring and may impact our projections.24 Finally, our measure of equity, D/100K adult population, assumes that LT demand is proportional to the total population; however, there are wide variations in demand depending on several sociodemographic factors.25,26 Balancing donation rates would decrease geographic heterogeneity in donation, but would likely not completely balance geographic inequity in the ratio of eligible deaths to waitlist recipients. Differential regional demographics, disease prevalence, urban/rural population, insurance coverage, and density of transplant centers may impact demand that is not captured by our model. However, with the changing landscape of chronic liver disease in the United States, and changes in the availability in medical insurance with the adoption of the Affordable Care Act and Medicaid expansion, D/100K may be a complementary metric to consider in evaluating organ allocation policies.27

In conclusion, we have shown that redistricting regional allocation will result in decreased geographic inequity donor availability when measured from a population perspective. Geographic inequity is projected to increase in slightly in an 8-region model versus decrease slightly in an 11-region model when accounting for changes in US demographics over the next decade. However, improving performance of low-performing OPOs with respect to donation and utilization rates, especially in poor-performing regions, may have a more profound effect without the logistical challenges of redistricting. A hybrid approach to balancing LT supply and demand would likely yield the best results—that is, some form of redistricting coupled with systematic study and improvements in improvement in donation/utilization rates in low-performing DSAs. These factors are important considerations in the ongoing debate in how livers should be regionally allocated.

REFERENCES

1. Dienstag JL, Cosimi AB. Liver transplantation—a vision realized. N Engl J Med. 2012;367:1483–1485.
2. Lai JC, Feng S, Roberts JP. An examination of liver offers to candidates on the liver transplant wait-list. Gastroenterology. 2012;143:1261–1265.
3. U.S. Government Publishing Office. Organ procurement and transplantation network. http://www.ecfr.gov/cgi-bin/text-idx?SID=bb60e0a7222f4086a88c31211cac77d1&mc=true&node=pt42.1.121&rgn=div5. Updated April 3, 2017. Accessed April 15 2016.
4. Yeh H, Smoot E, Schoenfeld DA, et al. Geographic inequity in access to livers for transplantation. Transplantation. 2011;91:479–486.
5. Massie AB, Caffo B, Gentry SE, et al. MELD exceptions and rates of waiting list outcomes. Am J Transplant. 2011;11:2362–2371.
6. Rana A, Kaplan B, Riaz IB, et al. Geographic inequities in liver allograft supply and demand: does it affect patient outcomes? Transplantation. 2015;99:515–520.
7. Edwards EB, Harper AM, Hirose R, et al. The impact of broader regional sharing of livers: 2-year results of "Share 35". Liver Transpl. 2016;22:399–409.
8. Parikh ND, Hutton D, Marrero W, et al. Projections in donor organs available for liver transplantation in the United States: 2014–2025. Liver Transpl. 2015;21:855–863.
9. Goldberg DS, French B, Abt PL, et al. Increasing the number of organ transplants in the United States by optimizing donor authorization rates. Am J Transplant. 2015;15:2117–2125.
10. Pondrom S. The AJT report. Am J Transplant. 2014;14:2675–2676.
11. Gentry SE, Massie AB, Cheek SW, et al. Addressing geographic disparities in liver transplantation through redistricting. Am J Transplant. 2013;13:2052–2058.
12. Su F, Yu L, Berry K, et al. Aging of liver transplant registrants and recipients: trends and impact on waitlist outcomes, post-transplantation outcomes, and transplant-related survival benefit. Gastroenterology. 2016;150:441–453, e446.
13. Ogden CL, Carroll MD, Fryar CD, et al. Prevalence of obesity among adults and youth: United States, 2011–2014. NCHS Data Brief. 2015:1–8.
14. HRSA Data Warehouse. Organ donation and transplantation. https://datawarehouse.hrsa.gov/topics/organDonation.aspx. Published 2017. Accessed December 30 2016.
15. Ferrari SLP, Cribari-Neto F. Beta regression for modelling rates and proportions. J Appl Stat. 2004;31:799–815.
16. Cribari-Neto F, Zeileis A. Beta regression in R. J Stat Softw. 2010;34:1–24.
17. Smithson M, Verkuilen J. A better lemon squeezer? Maximum-likelihood regression with beta-distributed dependent variables. Psychol Methods. 2006;11:54–71.
18. Agresti A. Categorical data analysis. 3rd ed. Hoboken, NJ: Wiley; 2013.
19. Davis BD, Norton HJ, Jacobs DG. The organ donation breakthrough collaborative: has it made a difference? Am J Surg. 2013;205:381–386.
20. Howard DH, Siminoff LA, McBride V, et al. Does quality improvement work? Evaluation of the organ donation breakthrough collaborative. Health Serv Res. 2007;42(6 Pt 1):2160–2173; discussion 2294–2323.
21. Shafer TJ, Wagner D, Chessare J, et al. US organ donation breakthrough collaborative increases organ donation. Crit Care Nurs Q. 2008;31:190–210.
22. Goldberg DS, Abt PL, Gilroy RK. Reply to “Increasing the number of organs available to transplant is separate from ensuring equitable distribution of available organs: both are important goals”. Am J Transplant. 2016;16:730–731.
23. Massie AB, Chow EK, Wickliffe CE, et al. Early changes in liver distribution following implementation of Share 35. Am J Transplant. 2015;15:659–667.
24. Rudd RA, Aleshire N, Zibbell JE, et al. Increases in drug and opioid overdose deaths-United States, 2000–2014. MMWR Morb Mortal Wkly Rep. 2016;64:1378–1382.
25. Axelrod DA, Guidinger MK, Finlayson S, et al. Rates of solid-organ wait-listing, transplantation, and survival among residents of rural and urban areas. JAMA. 2008;299:202–207.
26. Hirose R, Gentry SE, Mulligan DC. Increasing the number of organs available to transplant is separate from ensuring equitable distribution of available organs: both are important goals. Am J Transplant. 2016;16:728–729.
27. Axelrod DA, Millman D, Abecassis MM. US health care reform and transplantation. part I: overview and impact on access and reimbursement in the private sector. Am J Transplant. 2010;10:2197–2202.

Supplemental Digital Content

Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.