9621 Agnes Crossing, Lake Suzanneview, New Mexico Island 84604-9295.
Photo: Mohamed Hassan on Pixabay Consider the following scenario: Two patients with chronic liver disease (severe liver disease) are waiting for a transplant. The condition of both patients
Consider the following scenario: Two patients with chronic liver disease (severe liver disease) are waiting for a transplant. The condition of both patients is so serious that if they do not receive a healthy organ in a short time, their chances of survival will gradually disappear.
Fortunately, a liver arrived at the transplant unit. This organ is compatible with both patients and both can benefit significantly from liver transplantation. Whoever gets it, in principle, a long survival with a good quality of life can be predicted.
Unfortunately, this liver has only one transplantable lobe. This means that one of the candidates will not be transplanted. In bioethics and public health, the term ration it refers precisely to the fact that while allocating a resource to a beneficiary, at the same time not giving a resource to another person who needs it.
In such a situation, apportionment becomes a tragic dilemma, an ethically contradictory situation where people’s lives are at stake. To which of the two patients should the liver be assigned? And according to what features?
In most countries, the distribution of transplantable livers is made mainly by considering the short-term mortality estimation in the waiting list. Priority is given to people most likely to die while waiting for a transplant.
The measurement of mortality risk is based on a linear regression with three variables (a mathematical formula commonly used in statistics): creatinine, bilirubin, and International normalized rate (standardizes prothrombin times). These three biomarkers are indicators that are normally used in various health analyzes.
The interaction of these three values creates a scoring system called the MELD scale, which is called the end-stage liver disease model (End Stage Liver Disease Model). This scale measures the severity of chronic liver disease.
However, some difficulties have led to the need to rethink this system. An important factor is the increased availability of suboptimal organs. These organs are of lower quality as they come from donors who are in worse health. If we transplant a damaged organ into a healthy patient, that person may need another transplant later on. On the other hand, transplanting a liver to a patient with a short life expectancy may mean losing the greater benefits that organ could have in another person.
To overcome these inefficiencies, the characteristics of not only the receiver but also the transmitter must be taken into account, as was done in the death estimation in the previous system. Thus, the UK Transfer Benefit Points (TBS) – transplant benefit score – taking into account 7 variables of the donor and 21 of the organ recipient. The goal is to maximize the lifespan of the transplanted liver.
The interaction of these 28 variants of TBS is more complex than the three of MELD. This spurred research to develop artificial intelligence (AI) models applicable to this system. For example, neural networks have been used for this purpose. Neural networks are advanced AI algorithms based on machine learning that train on interconnected layers of data.
The use of neural networks was able to predict the risk of death after transplantation at 90 days, with greater precision than previous statistical models. AI tools are also more effective at accurately predicting graft (i.e. transplanted organ) survival in each recipient. All this ensures that the scarce resource is channeled to the person who will apparently benefit most from it.
However, there is a problem with these AI-based models that still require further research. Because neural networks work autonomously, they do not allow the weight of each variable in the configuration of the final result to be known. In other words, these AI systems generate suggestions developed in an algorithmic process, the causal mechanisms of which even experts cannot explain. This lack of opacity and transparency is often referred to as the “black box” of artificial intelligence.
Given its multiple repercussions, the inexplicability of AI is a hot topic of discussion in disciplines such as computer science, medicine, business, law, and philosophy. How should the use of unexplained algorithms in medical artificial intelligence be evaluated?
There are those who argue that precision, not explainability, should be the core value of AI in medicine. They argue that medical knowledge is often incomplete and grappling with multiple uncertainties. In fact, this ambiguity already exists in some effective medical treatments, such as the prescription of lithium for psychiatric illness. In other words, we shouldn’t worry about not being able to explain the underlying reasons for these benefits as long as treatments are successful.
However, it becomes particularly problematic when one wants to use AI to support the prioritization of scarce health resources, where AI is as accurate as possible and whenever that leads to inexplicability.
In recent research I’ve enjoyed pioneering, we’ve looked at how inexplicable falls short of important ethical ideals.
Unexplainability prevents attribution of responsibility for error, makes it difficult to examine bias (which is problematic at both the clinical and ethical levels), and controls the transparency and trust necessary for the public acceptance of these technologies.
Since these criteria are fundamental to the fair distribution of scarce health resources, we should support the implementation of AI algorithms whose operation is explainable. Prediction accuracy is also absolutely important. But we must encourage the development and implementation of explainable precise algorithms.
Returning to the first case, let’s assume that the distribution of the liver is decided by the advice of artificial intelligence. What would you do if you were injured? If it were my situation, I wouldn’t hesitate to ask for clarification.
This article won second prize in the 2nd edition of the youth promotion competition organized by the Lilly Foundation and The Conversation Spain.
Jon Rueda Etxebarria, La Caixa INPhINIT Fellow and PhD candidate in philosophy, University of Granada
This article was originally published on The Conversation. Read the original.
Source: El Nacional
Alice Smith is a seasoned journalist and writer for Div Bracket. She has a keen sense of what’s important and is always on top of the latest trends. Alice provides in-depth coverage of the most talked-about news stories, delivering insightful and thought-provoking articles that keep her readers informed and engaged.
