Applications of Operations Research in Solid Organ Transplantation and Random Utility Choice Models

Vikram Kilambi

doi:https://doi.org/10.21985/N2DT9X

Work

Applications of Operations Research in Solid Organ Transplantation and Random Utility Choice Models

Public Deposited

Download PDF

Download All Files (.zip)

This work is a collection of articles featuring applications of operations research primarily on solid organ transplantation. At the time of writing, 111,434 Americans were waiting for a liver or kidney transplant. Only 26,901 transplants were performed last year – a consequence of the scarcity of organ donors and the lack of technologies that confer the same survival and quality of life as transplantation. 5,541 individuals died waiting for a transplant and 6,059 became too sick to receive one; and perhaps the most unfair hardship borne by many is that they must wait years more for a transplant than an equally sick patient somewhere else. The national organ procurement and transplantation network is the complex logistical system responsible for allocating organs obtained from deceased donors to potential recipients. Surprising to me also is that this system discarded 4,372 of the organs obtained last year. Most of these organs were of lesser quality but would have otherwise provided lifesaving benefits to patients. The first two chapters propose restructuring the national system for liver allocation with the aims of reducing geographic disparity in access to liver transplantation and annual mortality. The structures are based on principles from manufacturing and systems engineering and have graph-theoretical and topological motivations. Using heuristics or stochastic, non-convex integer optimization, we obtain several new designs and test their performances with large-scale discrete-event simulations of the entire system. The appendices include additional technical information. These designs significantly reduced geographic disparity, total mortality, and sometimes average transportation cost. The next two chapters investigate the decision-making of kidney transplant candidates. The first of these develops a multi-state Semi-Markov process model of the patient’s overall experience as a candidate for transplantation. The model calculates the average survival time for a newly listed patient that can then be used for delivering prognoses and benchmarking performance. The following chapter responds to the discards of lesser quality organs by conducting individualized decision analyses that determine when it would be beneficial for a patient to accept such organs for transplantation. A comprehensive and realistic computation engine based on decision trees is constructed and demonstrated. The last chapter is unrelated to the others and presents a method based on robust optimization for dealing with well-known nuisance parameters in the conditional logit discrete-choice model used in applied microeconomics and marketing.

Last modified