When to accept a living-donor liver

Organs are a scarce resource, and for some diseases, such as end-stage liver diseases, transplantation is the only viable therapy. The number of organ transplants has increased annually since the enactment of the National Organ Transplant Act of 1984. In 2002, nearly 25,000 Americans received an organ transplant. However, the number of donated organs has not kept pace with demand; approximately 6,500 patients died in 2001 while waiting for an organ. This research focuses on a deceptively simple decision problem that arises naturally within the existing allocation framework: If a patient is offered an organ for transplantation, should she (or her physician) accept it or wait for another organ? There are two reasons why a patient may reject an organ. First, depending on the quality of the offered organ, there is a chance that a future organ may be of higher quality. Second, one may view the components of life expectancy as pre-transplant and post-transplant. Accepting an organ ends the pre-transplant phase of a patient's life and begins the post-transplant phase. There is considerable evidence that post-transplant survival depends on the patient's health at the time of transplant. As long as the patient's health is not likely to decrease markedly, a patient may choose to prolong her pre-transplant life since it may not adversely affect post-transplant survival. In fact, an analysis of historical data suggests that livers are rejected approximately 60% of the time. Unlike global organ allocation schemes, the results of this research are immediately implementable by any patient, without the approval of any government agency. Living donors are an increasing source of livers for transplantation in America and Europe and are the primary source of livers in Japan. An advantage of considering organs from living donors is that the organ is always available, so the surgery can be scheduled at the best time for the patient. Livers are an appropriate organ for studying organ acceptance decisions, and many of the results may apply to other organs. Although there are more kidney than liver transplants, there is no alternative therapy for end-stage liver disease, such as dialysis for renal diseases. The Institute of Medicine noted that "much of the current debate has centered on the procurement and allocation of livers". This work uses real clinical data. We aggregated patient laboratory scores into a single measure, the widely used MELD (Model for End-Stage Liver Disease) score, a measure of pre-transplant survival which is currently used by UNOS (United Network for Organ Sharing) to rank patients according to medical urgency. MELD scores range from 6 to 40 in increasing order of sickness. The resulting policies gives a range of MELD scores where transplantation is advised, and suggests that the patient wait for other MELD score ranges. We are currently analyzing the case where the patient must accept or reject a cadaveric liver but can observe the waiting list. The resulting decision problems are large-scale Markov decision processes. One such MDP implicitly models the waiting list through the probability of organ arrivals. The other MDP explicitly models the waiting list, including the other patients as part of the state. In practice, the waiting list is not knowable, and thus the need for a decision model that reflects this fact. In future work we will model this decision as a partially observed Markov decision process (POMDP).