O.R. ApplicationsThe development and evaluation of a fuzzy logic expert system for renal transplantation assignment: Is this a useful tool?
Introduction
Renal transplantation is recognized as being of significant benefit to patients in renal failure, and the demand for donor allografts continues to rise. The allocation of scarce health-care resources has been the subject of much debate (Brich and Gafni, 1992; Lawrence et al., 1993). As noted by Childress (1989), common law in the USA (and elsewhere) has recognized organs as “quasi-property” and in most jurisdictions public and social policies have been invoked to resolve the issues surrounding disbursement of solid organs for transplantation. Underlying this debate has been the implicit and explicit understanding that respect for persons, utility (to maximize good) and justice (equity) are the principles underlying any allocation system. These principles can be undermined by bias occurring prior to as well as subsequent to recipients being accepted onto the transplant waiting list (Alexander and Sehgal, 1998). The debate over the “rules of allocation” has been protracted and frequently fractious (McNamee, 1996; Starzl and Fung, 1996; Briggs, 1996; Chang, 1996). Indeed, as noted by Terasaki et al. (1996), the term “allocation” is probably a misnomer, as distribution of kidneys for transplantation more closely resembles a “lottery”, albeit one with winnable odds. Some authorities have emphasized equity as the major criterion for allocation, while others have argued that optimal outcome of the renal allograft should be a critical factor in allocation, as graft failure simply increases the pressure on this scarce resource by returning recipients to the waiting pool. It is important to investigate the possible balances between these two criteria (Yuan et al., 1994).
In the absence of any agreement about the best balance between “equity” and “medical utility” in any given individual renal allocation situation, most organ procurement organizations (OPOs) have opted for a rigid hierarchical set of rules to distribute cadaver donor kidneys, with physicians being given the right to over-ride these decisions at the local level, subject to auditing. In general, marginal or less well-defined predictors of improved outcome have been sacrificed in order to allow physicians and OPOs to arrive at mutually acceptable allocation algorithm compromises (Starzl and Fung, 1996; Zachary et al., 1997).
“Experts” in transplantation have been most active in clarifying the issues surrounding “medical utility”. While rigid allocation algorithms have solved some problems, the potential benefits of complex interactions have been correspondingly sacrificed. A decision-making system that could incorporate such complexities but minimize physician or OPO bias at the individual allocation event would be very helpful, particularly if such a system could be tested and evaluated prospectively in a “real-life” simulation.
Vagueness and ambiguity are often involved in human reasoning. A physician often faces a dilemma such as: “Should a cadaver kidney from a middle aged male donor be given to a poorly matched, highly sensitized young female who has been waiting for a long time in preference to a non-sensitized, very well matched older recipient who has been on the waiting list for a moderate period of time?” Individuals can make these choices, but their decisions are subject to bias and non-reproducibility. Existing allocation algorithms can generate a solution but do not directly represent physicians' intuitive thinking. We believe that fuzzy set theory and fuzzy logic introduced by Zadeh (1965) are inherently well suited to solving this particular problem. A fuzzy logic expert system can be used to represent physicians' intuitive thinking well, and at the same time avoid inconsistency and emotional biases.
The objective of our research was to investigate whether or not a fuzzy logic expert system for kidney transplant allocation could perform better in comparison with existing allocation algorithms. What follows are the explanations of kidney allocation criteria and existing allocation algorithms; the introduction and implementation of a fuzzy logic approach for kidney allocation; the simulation and the analysis of testing results; and finally, the conclusion and further research directions.
Section snippets
Kidney transplant allocation criteria
Kidney allocation is a complex, multiple criteria decision-making problem. Here we list several major criteria under consideration. Our focus is to show the nature of these criteria. It is not our intention to make the list complete, since the decisions regarding which criteria should be included are themselves very complex.
Current allocation schemes
It is clear that all the above criteria are stated in natural language and by nature they are fuzzy. The interaction and tradeoffs among them are also not well defined. Many algorithms have been proposed to convert these criteria into a simple, more or less rigid set of guidelines in order to eliminate vagueness. Here we discuss two important allocation algorithms. The first algorithm was used by the multiple organ retrieval and exchange (MORE) program of Ontario in Canada. The second was used
A fuzzy logic approach for kidney allocation
Fuzzy set theory and fuzzy logic were established in 1965 by Zadeh in order to deal with the vagueness and ambiguity associated with human thinking, reasoning, cognition, and perception processes (Zadeh, 1965, Zadeh, 1973; Zimmermann, 1987). Fuzzy logic has been successfully used in solving many decision making and industrial control problems (Hruschka, 1988) and has yielded results superior to those obtained from traditional numerical methods (Munakata and Jani, 1994). It has also been used in
Simulation objectives and hypotheses
One main objective of the simulation was to determine whether or not the fuzzy system could accurately approximate Dr. Expert's thinking. Therefore, it was necessary to check how closely Dr. Expert's choices of suitable patients matched with those of the system. As a form of control group, it would also be interesting to see how his choices compared to those made by both the UNOS and MORE algorithms. However, the interesting aspect of the simulation lay in the fact that Dr. Expert was being
Analysis of testing results
The simulation test results consist of the ranking orders of the possible recipients generated by different methods for each donor kidney against the waiting list. We want to compare the ranking results between two methods, the experts' manual choices versus an algorithm's choices. Three types of measurements will be used to measure the differences: the “overlapping rate”, the “rank consistency index”, and “the first choice breakdown”.
The overlapping rate is the percentage of candidate
Conclusion and further research
We have successfully developed a pilot fuzzy logic expert system for kidney allocation based on a kidney transplant expert's knowledge. The fuzzy logic system can represent the expert's thinking well in handling complex tradeoffs. A simulated kidney allocation experiment based on real donor and patient data has shown that overall, the fuzzy logic derived recommendations were more acceptable to the expert than those from the MORE and UNOS algorithms. However, in order to derive a more general
Acknowledgements
We wish to thank the staff at MORE Ontario for their help with the case data. Thanks also go to anonymous referees who provide constructive comments for the improvement of the paper.
References (44)
- et al.
Fuzzy sets in life sciences
Fuzzy Sets and Systems
(1997) - et al.
Rating and ranking of multiple-aspect alternatives using fuzzy sets
Automatica
(1977) Allocation of cadaver kidneys the UKTSSA users kidney advisory group audit subcommittee
Lancet
(1996)- et al.
Fuzzy multiple criteria decision making: Recent developments
Fuzzy Sets and Systems
(1996) How should cadaver kidneys be allocated?
Lancet
(1996)- et al.
A model of consensus in group decision making under linguistic assessments
Fuzzy Sets and Systems
(1996) Use of fuzzy relations in rule-based decision support system for business planning problems
European Journal of Operational Research
(1988)- et al.
Preoperative HIV testing: Is it less expensive than universal precautions?
Journal of Clinical Epidemiology
(1993) Deciding about kidney-graft allocation
Lancet
(1996)- et al.
Medical diagnosis from patient records – a method using fuzzy discrimination and connectivity analysis
Fuzzy Sets and Systems
(1987)
Task complexity and contingent processing in decision making: An information search and protocol analysis
Organizational Behavior and Human Performance
Fuzzy modelling of the expert's knowledge in ECG-based ischaemia detection
Fuzzy Sets and Systems
Fuzzy multiple attribute decision making: A review and new preference elicitation techniques
Fuzzy Sets and Systems
Exchanging kidney transplants – is it worth it?
Lancet
The politics of grafting cadaver kidneys
Lancet
Allocation of cadaver kidneys
Lancet
Induction of fuzzy decision trees
Fuzzy Sets and Systems
A genetic algorithm for generating fuzzy classification rules
Fuzzy Sets and Systems
Providers' incentives to participate and contribute: A neglected aspect in the development of a central matching system for the allocation of cadaveric kidneys
Socio-Economic Planning Sciences
Fuzzy sets
Information Control
Barriers to cadaveric renal transplantation among blacks, women and the poor
JAMA
Cost effectiveness/utility analysis do current decision rules lead us to where we want to be?
Journal of Health Economics
Cited by (24)
An optimal learning model for training expert system to detect uterine cancer
2021, Procedia Computer ScienceOrgan transportation and allocation problem under medical uncertainty: A real case study of liver transplantation
2020, Transportation Research Part E: Logistics and Transportation ReviewAn integer programming approach for fuzzy rule-based classification systems
2017, European Journal of Operational ResearchCitation Excerpt :Fuzzy rule-based classification systems (FRBCSs) have two advantages as a data mining technique. First, their capability in encompassing nonlinear and complex relations in a data set enables them to establish precise classification models, especially when a system deals with ambiguous and imprecise data (Al-Ebbini, Oztekin, & Chen, 2016; de Andrés, Landajo, & Lorca, 2005; Nakandala, Samaranayake, & Lau, 2013; Yuan, Feldhamer, Gafni, Fyfe, & Ludwin, 2002). Second, the rule learning procedure used to extract the fuzzy rules for a FRBCS can be used to discover the latent knowledge in large data sets in the form of a set of interpretable fuzzy rules compatible with human logic.
Deciding kidney-offer admissibility dependent on patients' lifetime failure rate
2016, European Journal of Operational ResearchCitation Excerpt :Such directions are heuristics - but still more analytically sound than the extant point system. For example, Yuan et al. (2002) suggest a fuzzy logic approach. The authors show, by way of example, that the fuzzy logic based policy is closer to an expert’s (a medical practitioner) opinion than the policy attained by the UNOS point system.
FLAS: Fuzzy lung allocation system for US-based transplantations
2016, European Journal of Operational ResearchCitation Excerpt :Determining essential variables for decision making: This step refers to the fact that certain critical variables are extracted from the database for decision making with consultation to the medical experts in the field; reflecting their thought process within this problem on hand. Moreover, these variables are fairly common in credible literature (Yuan et al., 2002). The input variables are: distance, recipient's age, donor's age, blood matching, HLA matching, and number of days on the waiting list.
A bi-level belief rule based decision support system for diagnosis of lymph node metastasis in gastric cancer
2013, Knowledge-Based Systems
- 1
Present address: Canadian Imperial Bank of Commerce World Markets, New York, US 10017. Tel.: (212) 856-3584; fax: (212) 885-4999.
- 2
Tel.: (905) 525-9140x22681; fax: (905) 546-5211.
- 3
Tel.: (905) 521-6049; fax: (905) 521-6088.
- 4
Tel.: (905) 521-6049; fax: (905) 521-6088.