A General Approach to Broader Sharing in Organ Allocation

There have been political efforts to resolve controversy about, and improve, organ allocation. In addition to political considerations, the practical ingredients are of great importance. Each transplant program impacts upon other transplant programs, in a way that is different from any other area of medicine. This means that no political ruling altering the system can be ultimately successful without substantial agreement amongst the centers that it is worthwhile (1). The arguments regarding allocation center upon usefulness and fairness of diversion of an organ from a recipient at one program to a recipient at another program. It has not been possible to envision a system that will produce a major net benefit over the present balanced system (2). Attempts to alter the present organ distribution rules have foundered on this and the perception that these attempts are not driven with the best interests of the patients in mind. Allocation that encourages program-driven accumulation of patients asymmetric to donors available may not be desirable. However occasional major inequities exist, in which there would be a statistical advantage for broader sharing that would result in a different recipient receiving the organ than provided for by the present rules. An underlying conceptual problem at present is the categorical rule structure. This violates the statistical happenstance in possible donor-recipient pairings that is due to the relatively small numbers of donors and simultaneously critically ill recipients. On the average, the country’s patients are best served in the majority of instances by the local-regional-national hierarchy presently in place. However, occasions occur in which the best use of an organ for a critically ill recipient is not provided for due to these arbitrary boundaries. These instances seem to be unfair. In this article, I describe a process for cadaver donor liver allocation by which greater fairness could be achieved by broader sharing, allowing more opportunity for inclusion of statistical happenstance in optimization of selection of potential recipients and donors in the system. The principles guiding this approach are: (i) no fixed geographic boundary is used throughout the United States for allocation; (ii) provision is made for the practical, medical, and financial justification for local use, in the common settings in which there is no advantage due to happenstance of a more distant use; and (iii) urgency plays a quantified and critical, but not categorical, role, recognizing that all patients whose medical criteria justify listing for liver transplantation will die in a relatively short time they do not receive a transplant. Computer modeling (3) was used to predict allocation outcomes in the absence of categorical hierarchical sharing boundaries. Points in this plan were awarded for status, waiting time, and distance. The values for maximum waiting times and points and the results are given in Table 1. Historic data for donor and recipient streams and previous liver status codes were used for illustration purposes. Table 1: Comparisons of computer modeling results for an optimized allocation system using different maximum points for population distance (pop. points)Waiting time points were increased linearly up to the maximum time, at which time a maximum of 10 points accrued; thus a patient waiting 730 days in status 3, or 14 days in status 1, received the maximum of 10 points. Patients waiting for less than maximum time in status received a fraction of 10 points corresponding to the fraction of the maximum time they had waited; a status 1 patient who had waited 7 days received 5 points. Although patients could wait longer than the maximum time designated for the status, they did not receive more than 10 points for time waiting. For each donor, circles were drawn centered on the donor hospital, with radius equal to the distance between it and each program with potential recipients. The population within this circle according to a geoencoded referencing data base (4) was the basis for awarding distance points according to closeness. Thus, a recipient for a donor 0 miles away received the maximum, which decreased linearly to 0 points for recipients in programs outside a circle containing the maximum population value of 10 million. This allows weighting of each factor to be varied. For example, four values (5, 8, 15, or 20) were compared as the maximum points for minimum distance. As the maximum points for distance decreased, an increasing fraction of the transplants are allocated at a greater distance. Decreasing the maximum points for distance (thereby decreasing the effect of proximity of a potential recipient) in a graduated fashion allows increasing possibility for statistical happenstance to route an organ to a more distant center, when there do not happen to be as strong attractors (high status, long waiting) in recipients locally. In general, at an intermediate level (maximum distance points equal 15) a moderate increase in sharing at a distance occurs. This is achieved while providing good outcomes, compared with the rules currently in use. With this distance points maximum value of 15, the Organ Procurement Organizations (OPOs) in Region 5 and 6 were compared, because there has been concern about wider sharing in this part of the country. Table 2 shows that significant impacts would be felt in some programs (e.g., OPO 5 in Region 5 would go from 9 to 4 transplants), but over this 3-year modeling interval these would not be absolute in any location and would be relatively severe only uncommonly. Table 2: Comparison of computer modeling results for an optimized allocation system, using maximal population distance points=15, with “current” policy (1997)Refinements to this general approach would be desirable. The issue of distance is partly based on the population served, partly on the practical realities of transportation of organs. For this reason, it seems unreasonable to distinguish amongst centers within about an hour’s drive from the donor hospital. Furthermore, in many settings a trauma center serves the community, and arbitrary location differences of a few miles from that should not drive allocation. Therefore the maximum distance points should be given to all potential recipients in programs within, perhaps, 50 miles of the donor hospital. It would be useful to have a ceiling for the maximum number of points that a patient could receive for waiting time. This would decrease the advantage of a long waiting time in status, which could otherwise unreasonably overcome the effect of status points and distance points for long-waiting outliers. For this, a “wait time max” could be calculated from the longest time a patient would usually be expected to wait nationally. One way to do this would be to take the national average plus 1/2 of the 99% confidence interval in years (or fraction of a year) of waiting time to transplant per status per blood group and designate this the operational value for wait time max for that blood group. The points awarded for time waiting would not be equal to years of wait, but rather derived from this, using a “wait points max” that would be an absolute ceiling on the number of points awarded for the time waiting as in the example above. However, unlike the example, the wait time calculation could use a formula that generated a time:points curve that is asymptotic to the maximum. In this way, increasing waiting time, even beyond wait time max, would always provide increased points, but to a lesser degree as the wait points max were approached. A decimal fraction would be generated that, taken to a sufficient number of places, would be different for each potential recipient. This would serve to preserve time waiting as a fraction that is different for each recipient and therefore is a “tie-breaker” (as in the present system). Points could be increased linearly the first half of the maximum interval and then as a parabola segment, approaching the line asymptotically (Fig. 1). If there were a ceiling of e.g., 10 points for waiting time, 1/2 the maximum as defined above would correspond to 5 points. For a given blood group, the wait time max would be calculated as above, and 1/2 of this value would receive 5 points. Although actual wait for patients would occasionally exceed the calculated maximum time, the points received for waiting would only asymptotically approach 10 points. Figure 1: Proposed relationship between time waiting and waiting points for the proposed new system. Note that points continue to increase with time waiting, but as the maximum number of points is approached, the unit change lessens; the total number of points is never quite reached.Points might be awarded out to a circle of 25 or 50 million people, rather than the 10 million maximum used in the example given above. Perhaps geographic distance or a combination of this and population density would be more generally applicable for the distance variable. Waiting time and status are poor surrogates for need for a liver transplant. A better approach would be to use a continuous severity index such as the MELD (Mayo end-stage liver disease) system, presently under consideration by the United Network of Organ Sharing (UNOS) (5). Points could be calculated as in Figure 1 rather than being arithmetically related to MELD scores. Alternatively the severity index could be added as another category in addition to time waiting, if it is felt that this would provide the fairest comparison among patients. For this, the absolute MELD score points might be used, with waiting time calculated as in Figure 1. However the maximum points for waiting time might be lower than for the other factors. This would decrease the impact of waiting as compared with disease progression and the other considerations, but still recognize the human and practical burdens of waiting, plus providing a tie-breaker. It might be appropriate for the new allocation system to be phased in over a period of time, e.g., 3 years. This could be done by generating a set of points specified by the new algorithm and a set of points for comparison based on the present local-regional-national system, in parallel for each donor-recipient pair. Whichever points were highest would determine the offer, however, initially the points calculated by the new algorithm could be strongly discounted. Therefore initially, allocation would usually be by the present system. In subsequent years there would be a stepwise lessening of the discount, producing a relative increase in the points specified by the new system, and thereby an increased likelihood of sharing by the new system. When a donor became available, the new points would be calculated as described, for all patients. To compare the new system with the present system, recipients on the local list for that donor would all receive the maximum distance points (considered “0” distance), plus status and waiting points as in the new system. The new algorithm would apply to potential recipients in other regions. Nonlocal patients would receive, in addition to the same waiting time and status points, distance points calculated as described. The first offer would go to the recipient with the highest points. However initially during the first year these “new points” totals could be discounted, e.g., to 1/3 their value. During the second year, the discount could be to 2/3 the value, and third year, not discounted. Table 3 shows a possible set of ranges of points, with the status definitions now in use and the waiting time points at a maximum of 10 (the 1/2 max time would be different for each blood group). In the new points calculation all potential recipients listed at programs within 50 miles of the donor hospital would receive 10 distance points. To compare points for the present local-regional-national system all recipients in the present local area where the donor occurred would receive 10 points, and the points calculation would then decrease from that area. Table 3: Points in the new system, the discounted system, and the equivalent points for local recipients as designated by the old system, for comparisonIn general, during the first year this would lead to a recipient on the list local to the donor having more points by the local-regional-national conventional algorithm and receiving the organ as usual. However, on those occasions when statistical happenstance resulted in a much more urgent distant potential recipient with higher points, despite fewer distance points awarded, then the allocation would go to this more distant patient. In the first year of the phase-in this would probably result in less than 5% of organs being allocated (either locally or at a distance) by the new algorithm. During this phase in, initial real-world experience would be gained. In reviewing that experience, it would be important to remember that the full effect would only gradually become evident during the period it was being phased in because most offers would not be made by the new system. As the new system took effect in years 2 and 3 with resultant broader sharing, waiting time for patients in programs with shorter times would increase. In subsequent years they would be more likely to receive offers as times equalized. After the first 3 years, if review of the experience warranted, the new system could be instituted in place of the old, and the side by side calculation of points discontinued. Conviction that an effective practical review system can be devised to protect one’s patients from misunderstanding or impropriety on the part of other centers will be important in community support for a major change in the allocation algorithm. This could be done by creation of a national panel of reviewers to do prospective review of all status 1 and 2a listings, with a maximum 12-hr turnaround answer for each proposed listing. Active liver transplant surgeons and internists would be canvassed to ask if they would be willing to serve a month each year. Ideally each reviewer would have access to the worldwide web, using a password for a secure UNOS site. Alternatively fax or phone could be used. The requesting program would register a proposed listing according to criteria designated by the UNOS bylaws. The registration could be done by fax or by telephone or through a secure UNOS website. The Organ Center would transcribe it for review, coded under a UNOS candidate identification number. The Organ Center would then page 5 panelists selected at random (from 20 on call each month). The panelists would review the proposed listing and transmit a reply. The Organ Center would monitor the three information sites for returns every 2 hr. Once three “yes” or three “no” votes have been returned (i.e., a 3/5 majority vote for or against), the requesting program would be notified. If the vote was in favor, the patient would be listed. If, after 6 hr, a majority of three votes had not been achieved, two more panelists would be chosen at random. If, after 8 hr, there is no majority, two more panelists would be chosen; and after 10 hr, two more, so that there would be a total of 11 possible opinions sought, ensuring a sufficient number of responses by 12 hr. The first five to answer would constitute the review committee for that proposed listing. Patients with unambiguous fulminate failure, or with hepatic artery thrombosis, and medical status otherwise not questionable in the view of the proposing center, might be listed before conclusion of the review. On-site source validation should be done retrospectively with random selection of a small fraction of cases. Programs with a pattern of misrepresenting the patient data or violating the review decision would undergo UNOS membership review. In addition, Regional Review Committees would retrospectively review the information on all livers from their region that were transplanted into status 1 or 2a patients in any other region in the country. This review committee in the region of origin would review the listing data and certain further posttransplantation information on the recipient and circumstances, to be provided by the transplanting program when requested. The detailed information provided to the review committees would be kept confidential and be blinded as to center and patient identifiers. It would not be considered discoverable “data” that would be available long-term. Review panelists would need the traditional legal protection afforded to peer review. There could be concern regarding interstate commerce aspects of doctors collaborating across state lines to run this review system. Federal laws would need to be examined, and it may be that new legislation would be required to specify this protected status. This allocation algorithm provides for broader sharing by removing geographic boundaries, thus increasing fairness, in settings when better donor/recipient pairing can be done, while preserving the good relationships and results that occur in most cases in the present system in which local use is dominant. It is adjustable, allowing weighting of the factors (status, time waiting, distance, and potentially other factors) to be optimized; and it overcomes arbitrary geographic boundaries, while retaining the potential to maintain the good results provided by the present system. The principles are adaptable to any organ. A major objection made to this concept is that it is too complicated and difficult to understand. However it results from simple spreadsheet calculations that are easy to visualize and is essentially a logical extension of the present system, with enhancements. A system based on this has been in use for kidney transplantation in the heterogeneous multistate Region 1 of New England, with good results (6). Another objection is to the obliteration of geographic boundaries. It is feared that impairment of the public success of organ donor efforts might result from broader sharing. However a major impact of the concepts in the new algorithm is that organs will continue to be transplanted close to the donor on average. In fact, it could help in that way, because peoples’ relationships and loyalties are not necessarily related to the arbitrary geographic boundaries in place in the current system, and broader sharing from this sort of algorithm would commonly be to a relatively nearby patient. This proposal does not address the reasons for the dissociation between relative donor availability and number of patients on the list, which contributes to disagreements on allocation. Differential disease incidence seems to play a major role. In addition, however, success of organ donor efforts on the one hand, and center-directed accumulation of potential recipients for listing through marketing and contracts on the other, may not be aligned. Allocation is unalterably tied to these controversial issues, and in the future community consensus might allow incorporation of factors bearing on these in the allocation algorithm. For improved allocation, we need the following: (i) strong general agreement that a proposed system is clearly better and (ii) an appropriate oversight process (1). This approach to generating such an algorithm is proposed as a mechanism to fulfill these requirements. Acknowledgments. This article is dedicated to the memory of Alan Pritsker, PhD, whose technical accomplishments and humanitarian understanding benefited all transplant patients. I also thank the many clinical transplant professionals and UNOS staff members who have contributed to these concepts.