Inefficiency in cadaveric organ procurement.

1. Introduction

As most people are now aware, there is a severe and longstanding shortage of cadaveric human organs made available for transplantation in both the United States and abroad. Specifically, the supply of organs retrieved from cadavers--hearts, livers, kidneys, etc.--has failed to meet the demand by patients suffering organ failure for at least the past two decades. (1) As a result, waiting lists of potential transplant recipients now total over 80,000 patients in the United States alone. Of these, more than 6000 have died in each of the last three years because of the lack of a suitable donor organ, since still more patients were added to the list. (2) At current organ collection rates, it is expected that approximately half of the patients on that list will die before the needed organs become available.

At the most fundamental level, there appear to be at least two important causes of this shortage and the tragic loss of life associated with it. Notably, both of these causes are directly attributable to provisions of the National Organ Transplant Act of 1984 (NOTA). (3) First, NOTA proscribes any payment to organ donors or their surviving family members in order to encourage an increased rate of consent for organ removal. (4) Thus, the acquisition price of organs is legally set at zero.

Second, NOTA institutionalizes a set of firms known as Organ Procurement Organizations (OPOs) that are solely responsible for acquiring cadaveric organs in the United States. (5) Importantly, under NOTA, these firms are (i) set up as nonprofit organizations that are reimbursed at average cost and (ii) assigned exclusive geographic territories. Thus, the organ procurement system established by NOTA consists of a set of nonprofit monopsonists that are constrained to pay a zero price for this essential input. Clearly, this is not a system that is likely to maximize the procurement of cadaveric organs for transplantation. Indeed, it would be difficult to devise a system that would better ensure a shortage. (6)

Fortunately, there now appears to be some emerging hope for significant reform of this system. In June 2002, the American Medical Association (AMA) passed a resolution allowing trials in which organ donors' surviving families would be offered a financial incentive (i.e., be paid a positive price) in exchange for their consent to allow the organs of the deceased to be removed for transplantation purposes. (7) The AMA action has been followed by several similar resolutions passed by other organizations within the medical community. (8) This growing support for experimentation with donor payments has sparked several bills in Congress that would fund trials and provide the necessary exemption from the legal ban on payments contained in NOTA. (9) Congress, however, has not yet acted on such legislation. As a result, the zero price policy remains in effect.

Moreover, while the medical community appears to be relaxing its longstanding opposition to paying organ donors, it remains steadfast in its support for maintaining the existing market structure in organ procurement. That is, under all of the current proposals, the OPOs would maintain their nonprofit monopsony status. The only change would be removal of the zero price constraint.

While that change is likely to have a significant positive impact on cadaveric organ collection rates, it is unlikely to fully resolve the shortage (or it will resolve it but at unnecessarily high prices) if that shortage is caused in part by inefficiencies inherent in the current, legislatively mandated structure of the procurement system. Such inefficiencies appear likely under that system, because the OPOs are presented neither the incentive (carrot) of profits nor the pressure (stick) of competition to improve their collection rates. As a result, their performance, even with removal of the zero price constraint, is likely to fall far short of that achievable with for-profit companies facing effective competition.

The inefficiencies of the current system have recently assumed heightened policy relevance. Specifically, some opponents of proposals to pay organ donors have argued that such payments are not necessary to resolve the shortage. Noting the wide variation in relative performance across existing OPOs, these parties allege that the organ shortage can be eliminated by bringing the performance of all OPOs up to the observed level of those with the highest collection rates. (10) Such a proposal, however, rings hollow in the absence of a practical and effective method to bring about the necessary increase in performance. (11)

In addition, it is questionable whether such an approach would succeed. That is, given the observed collection rates of the most successful OPOs under the current market structure, it is not clear whether, after controlling for demographic and other exogenous factors, bringing all procurement organizations up to that level would provide a sufficient number of organs to fully resolve the shortage. Consequently, the potential success of the proposed approach has an important empirical dimension.

We address that dimension here. To do so, we employ a frontier estimation methodology to gauge the relative degrees of inefficiency exhibited by existing OPOs while controlling for exogenous determinants of observed collection rates. Our results show that, even if all of the relative inefficiencies could (somehow) be eliminated, the organ shortage would still be substantial. (12) This finding suggests that efficiency improvements that occur within the current organ procurement framework will be insufficient to address the dire shortage faced by patients.

Our paper is organized as follows. Section 2 provides some background and institutional detail regarding the cadaveric organ procurement system in the United States. Section 3 then presents our empirical model of observed organ collection rates across our sample of OPOs. Section 4 describes our data and estimation results. Finally, section 5 concludes the paper.

2. Background/Institutional Setting (13)

The current organ shortage has grown out of a public policy that was devised and implemented more by historical accident than conscious design. The earliest transplant operations were confined to kidneys. (14) Moreover, these operations were uniformly performed with living related donors. At that time, transplantation technology--in particular, the state of knowledge regarding immunosuppressive drugs--effectively precluded the use of cadaveric organ donors. Consequently, at that time, organ transplant patients brought the necessary donor with them when they checked into the hospital for the transplant operation. If there was no acceptable living donor, no transplant operation was possible. As a result, there were no waiting lists and no apparent shortage.

Moreover, under the living related donor system, no obvious need existed for an explicit payment to encourage donor cooperation. The affection between relatives was generally thought to be sufficient to motivate the requisite organ supply. And, where it was not, any necessary payment (or coercion) between family members could easily be arranged without the sort of middlemen generally required for market exchange. Such intrafamily cajoling by emotional pressure or outright payments or both also remained out of sight of the transplant centers and attending physicians. Therefore, a system of altruistic supply seemed to make sense in such a setting, and reliance on that system did not seriously impede the use of this emerging medical technology.

The situation gradually changed, however, as new drugs, improved tissue matching, and advanced surgical procedures led to both transplantation of cadaveric organs and greatly improved transplant success rates. Significantly, this newfound ability to make use of cadaveric organs expanded the application of transplant technology to vital organs other than kidneys. Although we have not been able to locate reliable data to pinpoint a precise time, probably sometime during the mid-to-late 1970s organ waiting lists began to appear as transplant candidates began to form queues for cadaveric organs. Those queues were generally managed by the transplant physicians located at the center where the operation was to be performed. In addition, the centers themselves generally coordinated the necessary organ procurement efforts.

These emerging shortages began to increase rapidly during the 1980s as a result of at least three developments that greatly stimulated transplant demands. First, the above-noted advancements in immunosuppressive therapy improved transplant success rates significantly. (15) Second, third-party payment was substantially increased as the federal government initiated coverage for all kidney transplants and began to provide coverage, along with private insurance companies, for other organ transplants as well. And third, the federal government also began to provide coverage of dialysis treatments for all patients suffering renal failure, thereby keeping a larger number of potential kidney transplant patients alive for much longer periods. (16) Together, these developments have caused transplant demands to grow steadily at about 10% per year over the past two decades, while cadaveric organ donations have increased only marginally over this period. The upshot has been substantial annual shortages that, each year, have added to the stock of patients who are placed on the waiting lists.

It was at the time these shortages were beginning to develop that NOTA was passed. (17) This legislation contains a number of substantive provisions affecting the transplant industry. Two in particular stand out. First, as noted earlier, NOTA codified into law the existing industry practice, inherited from the earlier era of living related donors only, of not paying organ donors or their surviving family members. That is, it set the legal price of donor organs--both living and cadaveric--at zero. And second, it created a complex system of largely private, completely nonprofit organizations to manage both organ procurement and allocation. The former function was assigned to the OPOs, while the latter was given to the United Network for Organ Sharing (UNOS).

The OPOs fall into two groups--independent and hospital based. The latter are generally associated with a specific transplant center, while the former operate separately and can serve multiple centers within their collection region. The OPOs' primary function is to acquire cadaveric organs by working closely with the hospitals in their region and approaching the families of recently deceased individuals who qualify as potential organ donors to request permission to remove the organs. Once the organs are acquired, they are placed into the UNOS system for allocation to recipients. That allocation, in turn, operates through a computerized point system that assigns weights to various patient characteristics (e.g., health status, length of time on the waiting list, and so on).

From a practical standpoint, procurement of organs from cadavers involves several steps. First, suitable cadaveric donors must be identified in a timely fashion. Such donors must have healthy, well-functioning organs and be free of infection at the time of their death. In addition, the donor organs must be free of cancer, because the use of posttransplant immunosuppressive therapy tends to accelerate progression of the disease. In the past, an age limit on donors was generally applied. That limit was increased from 50 to 55 during the mid-1990s to expand the potential organ supply. More recently, an explicit age limit has been dropped, and a more subjective case-by-case evaluation is used to screen for acceptable donor organs. (18)

The above conditions require that the majority of cadaveric organs come from either trauma or stroke victims who have been declared brain dead. Because of these rather stringent requirements, it has been estimated that only about 1% to 2% of all deaths occur under circumstances that allow the organs of the deceased to be used in transplantation. (19) As a result, identification of potential donors in a timely fashion is important in the procurement process, and failure to do so undoubtedly results in a large number of potential donations going uncollected. (20)

Once a potential organ donor has been identified, permission of surviving family members is generally sought. Notably, under the Uniform Anatomical Gift Act of 1968 (UAGA), such permission is not required in cases where the decedent has signed an organ donor card or executed some other valid premortem statement indicating a desire to donate upon death. In fact, under that legislation, the hospital where the death occurs has an affirmative legal obligation to honor the donor's wishes and remove the organs if they are medically suitable for transplantation. In practice, however, permission of surviving family members is usually sought even in cases where a valid donor card is present; and, in clear violation of the UAGA, the organs are generally not collected if the family denies consent. (21)

Thus, as a practical--though not legal--matter, the property rights to the organs of deceased individuals rest with the surviving family members. Any premortem decision of the deceased to donate organs upon death can be overridden by the family's wishes. In such an environment, then, donor cards, where they exist and are located, serve the more limited role of informing the family of their deceased relative's desire to have the organs donated. (22) Those cards are not, however, controlling in the ultimate decision.

Given the need to obtain the family's consent to remove the organs of the deceased, the methods used in seeking that consent become extremely important in determining the outcome of the procurement effort. Current regulations require that all families of potential organ donors be approached after death for permission to remove the organs. Because that request is determinative, how it is made is crucial to the outcome. Several studies have shown that both the timing of the request and how the request is presented can have significant effects on observed consent rates. (23) Also, the party making the request--physician, nurse, or organ procurement professional--tends to influence outcomes as well. In particular, a number of cases have been reported in which physicians or others have fulfilled the legal requirement that a request be made in a fashion that is either inept or intentionally designed to elicit a negative response. (24) Thus, while government can legally require that a request be made, it cannot ensure that the request is sincere and presented in a competent, or even humane, fashion.

The above process suggests that there are several reasons that cadaveric organ donation does not occur with all (or nearly all) qualifying deaths. The potential organ donor may not be identified in time, the family of the deceased may not be asked to donate, the request for donation may not be made in a competent manner, or the family may simply refuse. (25) A breakdown at any point in the process results in a failure to collect the organs. It is little wonder, then, that only between one-fourth and one-half of all potential donor deaths result in organ donations. (26)

3. A Model of Procurement Rates

It seems reasonable to regard OPO harvesting of cadaveric organs for transplant as a production activity not unlike other harvesting operations, for example, catching fish or picking mushrooms, although cadaveric organ donation does not exhibit the dynamic population stock effects of the former. Thus, as in any case of harvesting, the yield reflects (i) effort; (ii) the fecundity of the targeted resource-bearing area; and (iii) luck or random factors.

Given the above considerations, we adopt the following specification for our empirical model of the observed variation in organ procurement rates across OPOs. We begin by observing that cadaveric organ donation is feasible only in the event of a death. Moreover, owing to the rather stringent conditions required for the decedent's organs to be transplanted, it is also necessary, as a practical matter, that the death occur in a hospital setting. Therefore, we assume that

E([D.sub.i]) = [[lambda].sub.i][HD.sub.i], (1)

where [D.sub.i] is the number of cadaveric organ donors obtained by the ith OPO; E is the expectation operator; [HD.sub.i] is the number of hospital deaths within the ith OPO's collection region; and [[lambda].sub.i] is a proportionality factor, where 0 [less than or equal to] [[lambda].sub.i] [less than or equal to] 1. Specification of the model then reduces to a search for the determinants of [[lambda].sub.i]. Three sets of factors appear relevant.

First, OPOs engage in certain observable activities intended to encourage cadaveric organ donation. Primary among these are their educational programs. Two specific types of these programs are used. First, professional education involves the provision of training and educational materials to physicians and other health care workers within the OPO's collection region, especially those involved in emergency room and intensive care activities. These programs are intended to help identify potential organ donors and prepare surviving family members for the donation request. And second, public education involves the now familiar general advertisements promoting organ donation, initiatives aimed at particular ethnic groups, and so on. These programs attempt to convince the general population and subgroups within the population that organ donation is a worthwhile and rewarding choice.

Together, these two types of educational activities--which we denote [PR.sub.i] and [PU.sub.i], respectively--form the basic tools used by these organizations to obtain organs for transplantation. With price fixed at zero, persuasion becomes the only instrument available to encourage families of the deceased to donate. Because of the paramount role played by advertising and persuasive speech in generating donations, the effectiveness of these programs and the efficiency of the OPOs in carrying them out are matters of significant public importance.

The second set of variables expected to influence [[lambda].sub.i] involve demographic/cultural factors that may influence the receptivity and suitability, or fertility, of the population within each OPO's collection region. The effectiveness of the OPO's collection efforts in any given case, [[lambda].sub.i], appears likely to reflect both the extent and efficiency of its recruitment efforts and the predisposition and suitability of the targeted population to effectively consent to organ donation. The latter factor presumably depends upon demographic variables beyond the OPO's control. For example, the failure of an OPO to obtain donations at a higher rate should not be attributed to inefficiency if that OPO faces a less fertile pool of potential donors.

Several studies, including the recent work of Siminoff et al. (2001), have investigated factors that influence the donation decisions of surviving family members. (27) To account for these results, we include a set of demographic variables constructed for each OPO's collection region using U.S. census data. No demographic data are currently reported by OPO collection region. Nonetheless, our data allow us to reconstruct such variables by using state-level census data weighted by OPO region populations by state. (28) Thus, we are able to include (i) the percentage of the population in the OPO region i that reports its racial demographic as white, [PW.sub.i]; (ii) OPO collection region i median family income, [INC.sub.i]; (iii) the percentage of the OPO population over the age of 65 in 1995, [OLD.sub.i]; and (iv) the percentage of OPO population with a college degree, [EDUC.sub.i]. Based upon previous studies of the donation decision, we expect these factors to affect cadaveric organ donations. In particular, based on previous research and medical criteria, it is probable that (i) nonminorities are more likely to donate; (ii) higher income families may be more amendable to donation; (iii) older residents, regardless of their predisposition to donate, are less likely to meet all medical criteria for donation; and (iv) more educated populations may be more likely to donate. In addition, we include the OPO region's total population, [POP.sub.i], to control for possible scale effects in OPO operations. Also, we define a set of regional dummy variables, [R.sub.j], in an attempt to capture other, unmeasured demographic or cultural factors.

Finally, OPOs' observed performance, [[lambda].sub.i], is likely to be influenced by a number of unobservable, or random, factors. For example, the motivation and skill of the procurement personnel in approaching grieving family members to obtain permission to remove the organs of the deceased are likely to affect observed collection rates. Also, owing to the general lack of incentives to perform efficiently that result from the market structure imposed on this industry, we expect substantial inefficiencies to exist. Moreover, such inefficiencies and random effects are not directly observable and, therefore, cannot be directly reflected in the specification of the regression model. Thus, we include a one-sided error, which we denote as [v.sub.i], to account for the relative inefficiency of the OPOs in our sample. (29) In addition, a traditional two-sided error, [u.sub.i], is included to account for purely random variations in the OPOs' performance.

The stochastic frontier methodology has been widely applied for efficiency analyses. Extensive discussion of these models, and an explanation of their estimation by maximum likelihood, can be found in Greene (1990).

Thus, suppressing the subscript i, [lambda] is specified as

[lambda] = exp([b.sub.0] + [b.sub.1]PR + [b.sub.2]PU + [b.sub.3]PW + [b.sub.4]INC + [b.sub.5]POP + [b.sub.6]OLD + [b.sub.7]EDUC + [summation] [b.sub.j][R.sub.j] + v + u), (2)

where the bs are the coefficients to be estimated. Econometric implementation is achieved by substituting Equation 2 into Equation 1 and rewriting to obtain the transformed expression

ln(D/HD) = [b.sub.0] + [b.sub.1]PR + [b.sub.2]PU + [b.sub.3]PW + [b.sub.4]INC + [b.sub.5]POP + [b.sub.6]OLD + [b.sub.7]EDUC +[summation] [b.sub.j][R.sub.j] + v + u, (3)

where u is specified as U ~ N(0, [[sigma].sup.2.sub.u]), and v is the absolute value of a normal variate with mean 0 and variance [[sigma].sup.2.sub.v]. Thus, E(v) = ([square root of 2]/[pi])[[sigma].sub.v] and V(v) = (([pi] - 2)/[pi]) [[sigma].sup.2.sub.v]. Letting e = u + v, the marginal density of e is f(e) = (2/[sigma])[f.sup.*](e/o)[[F.sup.*](qe/[sigma])], where [f.sup.*], [F.sup.*] are the standard normal probability density function (pdf) and cumulative density function (cdf), respectively, [[sigma].sup.2] = [[sigma].sup.2.sub.w] + [[sigma].sup.2.sub.v] and q = [[sigma].sub.v]/[[sigma].sub.w]. The model specified above is estimable by maximum likelihood.

The form of Equation 3 has both advantages and disadvantages. Use of the dependent variable In (D/HD) addresses, in a conventional way, heteroskedasticity concerns (see Caudill, Ford, and Gropper 1995). On the other hand, using the transformation In (D/HD) instead of D itself presents difficulties in providing any useful interpretation of firm-specific measures of inefficiency based on statistics of the form E(v|e). We return to this second point below.

Particular interest attaches here to a basic issue. What is the extent of relative inefficiency among OPOs, and to what degree would some hypothetical elimination of such inefficiency resolve the organ shortage? We emphasize that there are presumably a number of potential institutional sources of OPO inefficiency, including their nonprofit status, the monopsony market structure, and the lack of a market (price) mechanism to encourage donations. Within this environment, however, it seems likely that observed performance will vary considerably from one OPO to another. Because all OPOs are not-for-profit monopsonies operating in an environment in which payments for organs are prohibited, any findings of inefficiency here will be purely relative. That is, they must refer to the inefficiency exhibited given these rather severe constraints. Yet, despite all this, the possibility of reducing the shortage by fine-tuning the existing system should at least be examined. Lacking any market mechanism to discipline inefficient OPOs, society is left only with studies of the present sort and any political or regulatory actions such studies might motivate.

4. Data and Estimation Results

Our data include counts of the number of cadaveric organ donations per 1000 hospital deaths by OPO for the period 1994-1996, in addition to OPO spending on both public and professional education for the years 1995 and 1996. These data were obtained by the authors through a Freedom of Information Act request. Mean donations per 1000 deaths are about 15.4 in our sample, with variation between 6.4 and 29.8 as the system extremes. Thus, owing to a number of factors, less than 2% of all hospital deaths result in organ donations, and, as expected, there is considerable variation in observed OPO collection rates. Of course, many of these deaths fail to satisfy minimal requirements for donation for medical reasons.

Because donations may fluctuate over short intervals as a result of chance and because any effects of educational spending are likely to reflect long-lived, capital, or goodwill effects such as those typically associated with advertising, we aggregate our educational spending variables and collection rates across several years using a three-year window (1994-1996) for collections, and two years (1995-1996) for educational expenditures, where the latter is due to data availability. The three-year aggregation over donations is intended to reduce the (relative) importance of the two-sided disturbances, although such a course reduces the sample size.

Table 1 presents definitions and descriptive statistics for the variables in our model. The final sample contains 42 observations, which represents all independent OPOs in the United States except for those with incomplete data. The OPOs in our sample represent about 85% of the U.S. population.

Table 2 gives our estimation results. We note first that the regression is highly significant (Wald [chi square](10) = 1894). Also, the frontier formulation appears to be coherent (the conventional test for this is a likelihood ratio test that [[sigma].sub.u] is zero, which here yields [chi square](1) = 1.40, with probability > [chi square] equal to 0.11). Thus, we do appear to have a frontier with both two-sided and one-sided disturbances, although inefficiency appears to be more important than luck in explaining donation variability, presumably a welcome consequence of temporal aggregation of donations.

We note that the coefficients in Table 2 define the efficient levels of (the log of) donations per 1000 hospital deaths and must be interpreted in this light. For example, an increase in any variable with a positive coefficient implies that an OPO must increase its donation yield to be efficient, but such a change does not imply such an increase would or would not occur for any actual OPO.

We turn next to an examination of the estimated model coefficients. First, and rather importantly, we find no significant effects for public (PU) educational spending and a marginal effect for professional educational spending (PR). Although the frontier specification makes the policy relevance of this result somewhat opaque, it is apparent that both sorts of spending have weak capacities to increase the number of organs collected by efficient organizations. (30) Using the coefficients uncritically as they appear, however, we find that, for a typical OPO with about 300 donations in the sample period, additional spending of almost $167,000 on public education is necessary to solicit a single additional donation. (31)

Results for the demographic variables are consistent with previous research on the donation decision. First, the (self-identified) white racial demographic is significantly and positively related to the efficient, in other words frontier, levels of donation. In other words, it is easier, in some sense, to induce donations from individuals who identify themselves as white. For a variety of reasons, many minority communities have historically placed little trust in officials--medical or otherwise. Unfortunately, the chronic shortage of minority donors has created relatively disproportionate waiting times for minority transplant candidates, a circumstance that may itself reduce trust by minorities in the current system.

Our other results vary in significance. Somewhat surprisingly, an older donor base appears to have no significant effect on donation rates, despite the probable medical effects of aging on donor suitability. Education (EDUC), somewhat crudely measured as the ratio of college graduates to the total population, plausibly has a mild positive effect, but significance is low (t = 1.32). Income, which is quite collinear with education and other demographics, has no noticeable effect. (32) Some regional variation in donation rates is observed, with residents of the Western region appearing to be the most generous.

Of potential importance, we find that OPO population (POP) size reduces donation rates per 1000 hospital deaths, although death counts are almost proportional to population in most OPOs. In other words, OPOs with larger populations, for whatever reasons, are less able to successfully recruit acceptable donors. This effect is small but statistically significant: A large, 500-donor OPO that experienced an increase in population of 20% at the mean (about 1,000,000 additional residents) would lose about 13 donors (almost 3% of their total). Apparently, given the constraints under which donations must be obtained (e.g., moral suasion), larger OPOs may face an inherent disadvantage. If this effect is real, its cause, whether managerial difficulties, a more impersonal environment, geographic size, or something else, is an issue worthy of further investigation. (33)

We turn next to an evaluation of our results on the relative efficiencies of sample OPOs. We note first that there are two sources of deviations in OPO performance in the context of our model: random, two-sided variation and inefficiency, given here as a one-sided disturbance that never increases donation rates. A variety of suggestions have appeared in the frontier estimation literature for measuring the inefficiency component, the most popular being that of Jondrow et al. (1982), who suggest calculating the conditional expectation E(v|e) = [[sigma].sup.*][(e[lambda]/[sigma]) + [f.sup.*](e[lambda]/[sigma])/[F.sup.*](e[lambda]/ [sigma])] where [[sigma].sup.*] = [([[sigma].sup.v][[sigma].sup.u]/[sigma]).sup.2] and [f.sup.*], [F.sup.*] are as defined above. We take a somewhat different approach here because our question--whether the elimination of all inefficiency among OPOs could solve the shortage--does not require that any particular emphasis be placed on which OPOs are efficient and which are not. Additionally, our dependent variable y, having the form y = ln (D/HD) instead of merely y = D, does not allow any simple or exact interpretation of the Jondrow statistic in terms to inefficiency. Thus, since our interest focuses on the performance of the whole industry, we instead compare estimated efficient levels of donation to actual (observed) levels of donation. Unlike Jondrow's statistic, this approach implies that some firms (the lucky ones) will be more efficient than the frontier.

We begin by noting that 8 of 42 OPOs exhibit actual performance beyond the frontier, while 34 OPOs perform below it. See Table 3. Efficiency varies widely among these organizations: A few OPOs exhibit donation shortfalls of more than 35%. The total shortfall is about 2669 donations over three years (in calculating this, negative losses are included, otherwise the figure would be higher). This result is equivalent to 890 donors per year, most of whom, of course, would provide multiple organs for transplantation. (34)

In the case of kidney transplantation, UNOS reported there were 27,498 patients, 31,045 patients, and 34,550 patients on waiting lists in 1994, 1995, and 1996, respectively. Following the estimation methodology of Kaserman and Barnett (2002), we obtain annual shortages of 3890 (1994), 5050 (1995), and 5307 (1996). (35) Assuming an optimistic 1.5 successful kidney transplants per donor, the complete and costless (and miraculous) elimination of this estimated inefficiency will fail to eliminate these shortages: The average annual increases in transplants would be 1335, arising from 890 additional donors, yielding shortages of 2555 (1994), 3715 (1995), and 3972 (1996). Thus, while eliminating inefficiency helps, it will only cause the waiting list to grow more slowly.

A final point is worth making regarding our findings and methodology. Our calculations of inefficiency are related to, but differ from, more basic techniques familiar in the epidemiological literature (for example, Stogis et al. 2002). To illustrate this, we ordered the sample from lowest to highest by the ratio of actual donations to predicted donations, and also ordered the sample by the ratios (don/deaths) and (don/pop), where don is the number of donors obtained, 1994-1996; pop is OPO region 1995 population; and deaths is the number of OPO region hospital deaths. We then calculated Spearman rank correlation coefficients [lambda] for these various orderings. We obtain the following Spearman statistics:

(don/E(don)) vs. (don/deaths))[lambda] = 0.607

(don/E(don)) vs. (don/Pop))[lambda] = 0.552

(don/deaths) vs. (don/Pop))[lambda] = 0.521

where [lambda] = Spearman statistic = 1 - 6[summation][([R.sub.i] - [R.sup.-.sub.i]) .sup.2]/([n.sup.3] - n) and [R.sub.i], [R.sup.-.sub.i] are the ordinal ranks of [OPO.sub.i] under the two ordering schemes being tested. (36) While it is unsurprising that all three measures are (significantly) correlated, the correlations between our derived efficiency rankings and those based on normalized donor ratios are far from perfect. A Spearman correlation of less than 0.67 indicates that the uncritical use of simple donor ratios may well produce a conclusion that a given OPO is quite inefficient when, in fact, it exhibits average efficiency but is located in a region with demographic impediments to donation. For any incentive scheme intended to encourage efficiency to succeed, these effects would have to be evaluated in detail.

5. Conclusion

Competitive markets and for-profit enterprises are generally acknowledged to be relatively adept at devising efficient and innovative methods of production. In industry after industry, the incentive mechanisms faced by profit-seeking firms under competitive market conditions push firms to minimize costs by maximizing the output obtained from a given set of inputs. Organ markets are likely to be no exception.

The evidence we have presented here suggests that the current organ shortage cannot be addressed solely by improving OPO efficiency, even if it were known how such a result could be obtained. Indeed, our estimates indicate that even a complete elimination of all relative inefficiencies in the current system would not yield a sufficient number of donors to end the shortage. Therefore, as policy makers seek to resolve the shortage, it is imperative that the resulting reforms address all causes of the current shortages. Not only should NOTA be amended to allow the payment of positive market-clearing prices to organ donors, it should also allow entry and other measures to improve the efficiencies of poorly performing OPOs, or else replace them with more efficient organizations.

The authors thank, without implicating, two anonymous referees and the editors of this journal. All errors are the authors' alone.

Received May 2004; accepted November 2005.

References

Abadie, Alberto, and Sebastien Gay. 2004. The impact of presumed consent legislation on cadaveric organ donation: A cross country study. NBER Working Paper No. 10604.

Cate, Fred H. 1994. Human organ transplantation: The role of law. Journal of Corporation Law 20:69-90.

Caudill, S., J. Ford, and D. Gropper. 1995. Frontier estimation and firm-specific inefficiency measures in the presence of heteroskedasticity. Journal of Business and Economic Statistics 13:105-11.

Comarow, Avery. 2003. "Transplant trauma: New tactics fight the growing organ donor deficit." U.S. News and World Report, 13 January. Accessed 5 June 2005. Available www.usnews.com/usnews/health/articles/030113/13donor_2.htm.

Denise, Susan H. 1985. Regulating the sale of human organs. Virginia Law Review 71:1015-38.

Engstrom, Paul. 2001. Damaged goods. Washington Post, 3 May, p. HE08.

Evans, Roger W., Carlyn E. Orians, and Nancy L. Ascher. 1992. The potential supply of organ donors: An assessment of the efficiency of organ procurement efforts in the United States. Journal of the American Medical Association 267:239-46.

Gortmaker, Steven L., Carol L. Beasley, Lorie E. Brigham, et al. 1996. Organ donor potential and performance: Size and nature of the organ donor shortfall. Critical Care Medicine 24:432-9.

Greene, William H. 1990. Econometric analysis. New York, NY: MacMillan Publishing Company.

Howard, David H. 2002. Why do transplant surgeons turn down organs? A model of the accept]reject decision. Journal of Health Economics 21:957-69.

Jondrow, J., K. Lovell, I. S. Materov, and P. Schmidt. 1982. On the estimation of technical inefficiency in the stochastic frontier model. Journal of Econometrics 23:267-74.

Kaserman, David L., and A. H. Barnett. 2002. The U.S. organ procurement system: A prescription for reform. Washington, DC: The American Enterprise Institute Press.

Mendeloff, J., K. Ko, M. S. Roberts, M. Byrne, and M. A. Dew. 2004. Procuring organ donors as a health investment: How much should we be willing to spend? Transplantation 78(12):1704-10.

National Organ Transplant Act. 1984. Public Law 98-507, U.S. Statutes at Large 98 (1984): 2339 (Title 42, [subsection] 273 et seq.).

Pessemier, Edgar A., Albert C. Bemmaor, and Dominique M. Hansens. 1977. Willingness to supply human body parts: Some empirical results. Journal of Consumer Research 4:131-40.

Peters, Thomas G., Dilip S. Kittur, L. J. McGraw, M. Roy First, and Edward W. Nelson. 1996. Organ donors and nondonors: An American dilemma. Archives of Internal Medicine 156:2419-24.

Saranow, Jennifer. 2003. What is your body worth? Putting prices on the pieces. Wall Street Journal Online. 6 May. Accessed 5 June 2005. Available at http://online.wsj.com/article/0,,SB105217044930202200,00.html.

Sheehy, Ellen, Suzanne L. Conrad, Lorie E. Brigham, Richard Luskin, Phyllis Weber, Mark Eakin, Lawrence Schkade, and Lawrence Hunsicker. 2003. Estimating the number of potential organ donors in the United States. New England Journal of Medicine 349:667-74.

Siminoff, Laura A., Nahida Gordon, Joan Hewlett, and Robert M. Arnold. 2001. Factors influencing families' consent for donation of solid organs for transplantation. Journal of the American Medical Association 286:71-7.

Stogis, Sheryl, Richard A. Hirth, Robert L. Strawderman, Jane Banaszak-Holl, and Dean G. Smith. 2002. Using a standardized donor ratio to assess the performance of organ procurement organizations. Health Services Research 35:1329-44.

U.S. Department of Health and Human Services. 2002. The organ donation breakthrough collaborative: Best practices final report. Washington, DC: USDHHS, contract: 240-94-0037.

Warren, Jim. 2003. Legislation authorizing HHS to fund organ donation financial incentive trials introduced in Congress. Transplant News 13:1-3.

(1) For a brief history of the organ shortage and a description of the U.S. organ procurement system, see Kaserman and Barnett (2002), Chapter 2.

(2) The United Network for Organ Shortage reports over 40,000 deaths of patients on waiting lists since 1995. See http:// www.optn.org/latestData/rptData.asp.

(3) National Organ Transplant Act (1984).

(4) Specifically, the Act makes it a felony to "knowingly acquire, receive, or otherwise transfer any human organ for valuable consideration for use in human transplantation." Penalties include fines of up to $50,000 and prison sentences up to five years, or both. See 42 U.S.C. 274(a) (1994).

(5) Kaserman and Barnett (2002).

(6) A third factor that may contribute to the problem of cadaveric organ procurement is the widespread practice in the United States and many other countries of routinely following surviving family members' wishes, even in the presence of an affirmative premortem commitment by the deceased to donate. Recent empirical evidence suggests that adoption of a system of presumed consent (such as that used in Belgium and Spain) in which such consultation with family members is not required would reduce, but not eliminate, the organ shortage. See Abadie and Gay (2004).

(7) See Comarow (2003).

(8) At the same time, several other organizations (e.g., the National Kidney Foundation) and individuals have expressed adamant opposition to the use of payments to encourage increased donations.

(9) See Warren (2003).

(10) See U.S. Department of Health and Human Services (2003).

(11) This proposal is a bit like arguing that we can solve the problems of our educational system by bringing all students' performance up to that displayed by the very best and brightest individuals. It avoids addressing the question of exactly how we go about doing that.

(12) We refer to our estimated inefficiencies as relative because they can only reflect departures from the observed performance of the most efficient OPOs under the existing organ procurement system. Given the available data, they cannot reveal the magnitude of inefficiencies in comparison to performance under a different system (for example, a for-profit competitive organ collection industry).

(13) This section draws heavily from Kaserman and Barnett (2002).

(14) The first successful human organ transplant in the United States was performed on December 23, 1954. On that date, a kidney was transplanted from a living donor who was an identical twin of the recipient.

(15) The most important new drug discovery at this time was cyclosporine, which was introduced in the United States in 1983. Following the introduction of this drug, one-year patient survival rates improved from approximately 70% to 85% for kidneys, from 58% to 80% for hearts, and from 25% to 50-70% for livers. See Kaserman and Barnett (2002).

(16) The End Stage Renal Disease Program was established under Medicare in 1972. It provides federal funding for both kidney transplants and dialysis. See Kaserman and Barnett (2002).

(17) The impetus for this legislation appears to have come from a Virginia physician's attempt to alleviate the shortage by brokering living donor kidneys. The medical community's outrage at this development then led to political pressure that resulted in the passage of this Act. See Denise (1985).

(18) Elimination of the age restriction is but one indication of the willingness of transplant centers to make use of increasingly marginal organs. Organs that would have been discarded 10 years ago on quality grounds are now being transplanted. See Saranow (2003), Peters et al. (1996), and Engstrom (2001). Also, Howard (2002) provides an economic analysis of surgeons' decision-making process regarding whether to transplant an organ given the quality of the organ and the patient's health status.

(19) See Evans et al. (1992); Gortmaker et al. (1996); and Sheehy et al. (2003).

(20) In some cases, a death that could have yielded transplantable organs is not recorded as a potential donor because the attending physician fails to take the actions required to preserve the organs adequately. For example, if the blood pressure of the deceased is allowed to drop below a critical level, the organs will be damaged and will be unsuitable for transplantation. It is not known how many potential donors are lost from this cause.

(21) A recent exception is provided by the OPO based in Pittsburgh, Pennsylvania, which we have been informed no longer requests the family's permission when a valid donor card is located. Nonetheless, the medical community's widespread insistence that, despite legislation to the contrary, family members' permission be obtained appears to stem from several potentially valid considerations. See Kaserman and Barnett (2002). It also persists, no doubt, because of a lack of enforcement of the UAGA's provisions. To our knowledge, no one has ever been prosecuted under this Act for failing to remove the organs of a deceased individual.

(22) Where a donor card has been executed and is located, however, surviving family members generally tend to honor the decedent's wishes. It is also the case, however, that in many instances donor cards are not located. See Cate (1994).

(23) See, for example, Siminoff et al. (2001). It has been found that a brief period of time (around 20 minutes) needs to separate the act of informing the family of their loss and requesting that the organs be donated. This separation between the notification of death and the request for donation is apparently needed for family members to accept the fact that their relative has died.

(24) In one case, the physician, having just informed the family that their relative had died, asked, "You don't want to give away any of his parts, do you?"

(25) Also, for various reasons, a number of organs are collected but do not get transplanted. These organs may exhibit physical flaws that are only discovered when they are removed from the cadaver, they may be damaged during removal or transport, and so on. See Howard (2002).

(26) There is currently a great deal of speculation regarding the actual aggregate collection rate. The problem is that the denominator--the number of potential organ donors is not directly observable. See, for example, Evans et al. (1992); Gortmaker et al. (1996); and Sheehy et al. (2003). The range given here appears to be a consensus figure.

(27) See, also, pessemier, Bemmaor, and Hansens (1977).

(28) In particular, state-level census data for 1995 are averaged using state population shares in OPO regions to obtain the relevant variables.

(29) Such composed error or frontier models are commonly used in cost and production function estimation. See, for example, Greene (1990) and Jondrow et al. (1982).

(30) A respecification of the model, in which spending is normalized by OPO population (for example, PU/POP replaces PU) produces qualitatively identical conclusions.

(31) This would still be a bargain, however. Mendeloff et al. (2004) provide evidence that a donor is worth over $1,000,000 in welfare terms.

(32) We duplicate the estimation while dropping out the education variable. Other results show no significant change. These results are available from the authors on request.

(33) Alternately, the result may reflect a statistical artifact arising from the near proportionality of HD and population. Runs omitting population, however, show no significant change in the other results.

(34) UNOS data for 2001 indicate that, on average, cadaveric donors yield 1.35 kidneys, 0.36 hearts, and 0.77 livers.

(35) By this methodology, a shortage is measured as the increase in patients on the waiting list, plus deaths of such patients who could wait no longer.

(36) The ordinal approach is used for purposes of robustness.

T. Randolph Beard, * David L. Kaserman, ([dagger]) and Richard P. Saba ([double dagger])

* Department of Economics, College of Business Building, Auburn University, AL 36849, USA; E-mail rbeard@ business.auburn.edu; corresponding author.

([dagger]) Department of Economics, College of Business Building, Auburn University, AL 36849, USA; E-mail kaserman@ business.auburn.edu.

([double dagger]) Department of Economics, College of Business Building, Auburn University, AL 36849, USA; E-mail sabaric@ auburn.edu.

Table 1. Variables and Summary Statistics

                                                              Standard
Variable    Definition                                Mean    Deviation

ln (D/HD)   Log of donations per 1000 hospital
              deaths, 1994-1996                       2.68         0.31
PR          Professional education spending,
              1995, 1996, in 1000s of 1996 $        370.83       654.65
PU          Public education spending, 1995,
              1996, in 1000s of 1996 $              328.14       381.90
POP         OPO collection region population in
              millions, 1996                          5.09         3.49
PW          % white in OPO collection region,
              1996 census                             0.74         0.12
INC         Median family income in OPO region,
              1996, in 1000s of 1996 $               36.208        4.29
OLD         % over 65 years in OPO collection
              region, 1996 census                     0.13         0.02
EDUC        % with college degree in OPO
              collection region, 1996 census          0.12         0.02
[R.sub.1]   Dummy = 1 for Southeastern and
              Atlantic states region                  0.31         0.47
[R.sub.2]   Dummy = 1 for Northeast and Midwest
              states region                           0.41         0.49
[R.sub.3]   Dummy = 1 for Western region              0.28         0.45

n = 42. OPO region demographics compiled by weighting reported OPO
state populations by state census demographic values for those OPOs
exhibiting multistate operations. All dollar figures in 1996 dollars.

Table 2. Composed Error Regression Results (Dependent Variable =
ln (D/HD))

Variable                                  Coefficient   t score

PR                                            0.00009      1.55
PU                                            0.00002      0.21
POP                                          -0.02671     -2.31
PW                                            1.91443      5.85
INC                                           0.00047      0.04
OLD                                           1.85075      0.96
EDUC                                          2.97552      1.32
[R.sub.1] (SE, ATL)                           0.85775      1.56
[R.sub.2] (MW, NE)                            0.83973      1.44
[R.sub.3] (W)                                 1.15063      2.09
ln [[sigma].sup.2.sub.v] (inefficiency)       -3.8335     -6.15
ln [[sigma].sup.2.sub.u] (two-sided)          -2.7320     -3.99

n = 42. Wald [chi square] (10) = 1894. Log likelihood = 5.987.
SE = Southeastern U.S., ATL = Atlantic Coast U.S.,
MW = Midwestern U.S., NE = New England, W = Western U.S.

Table 3. Losses Due to Inefficiency 1994-1996

OPO   Actual Donations   Predicted Donations   Losses

 1          311                  323             12
 2          112                  167             55
 3          451                  579            128
 4          113                  130             17
 5          591                  529            -62
 6          253                  314             61
 7          266                  249            -17
 8          349                  440             91
 9           83                   93             10
10          319                  299            -20
11           37                   41              4
12          736                  785             49
13          281                  363             82
14          158                  196             38
15          253                  328             75
16          308                  299             -9
17          201                  244             43
18          537                  895            358
19          509                  661            152
20          449                  371            -78
21           92                  171             79
22          294                  355             61
23          102                  130             28
24           65                  161             97
25          340                  504            164
26          118                  114             -4
27           89                  121             32
28          578                  925            347
29          114                  127             13
30          146                  178             32
31          257                  336             79
32          215                  182            -33
33          691                  766             76
34          422                  491             69
35          179                  230             51
36          248                  306             58
37           97                  128             31
38          498                  548             50
39          498                  761            263
40          190                  290            100
41          154                  135            -19
42          240                  341            101