Wednesday, July 25, 2012

Medicine as the trolley problem

Are you familiar with the trolley problem? It is an ethics dilemma first formulated by the great Philippa Foot as a part of a series of such dilemmas. Her formulation goes roughly like this. Imagine there is a tram hurtling down a track. If it keeps going straight, it will hit and kill 5 people who are working on that track. The conductor is able to throw a switch and divert the train to another part of the track, where 1 single worker will be killed by the trolley. The question is what should the conductor do? Most people when asked respond that yes, he should throw the switch and sacrifice 1 life to save 5. After all, the net benefit is n=4.

There are literally thousands of alternative formulations of this problem, but one of them from the philosopher Judith Jarvis Thomson merits special consideration. The problem starts out similarly, with 5 lives on a track in potential peril. The vantage point and the solution are quite different, though. Now there is a bridge over the rail track, and a very large man is looking at the tracks from the bridge. One way to stop the train is to throw a heavy object in its path, like this large man, for example. You are on the bridge standing behind the man. Would you be justified in pushing him off the bridge in front of the tram to meet his death in order to spare the 5 workers down the tracks? Most people when faced with this formulation say an emphatic "no." This is somehow puzzling, since the net benefit is the same, n=4, as in the original Foot formulation.

Philosophy professors have puzzled over this difference for decades, and there are several potential explanations for why we respond differently to the two scenarios. One explanation has to do with the proximity of the operator (conductor in the first case and the person doing the pushing in the second) to the sacrificial lamb -- in the first case one is enough removed from the action of killing by merely redirecting the tram, whereas in the second the action is, well, more active, and the operator is actually pushing an innocent person to his death.

Though in some ways the scenarios seem to bear no practical distinction from one another, we see the morals and ethics of each differently. This difference in the view point is instructive to the field of medicine, where it has implications to how policy relates to the individual patient encounter. Here is what I mean.

Suppose you are a policy maker, and you recommend that every woman at age 40 start to receive an annual screening mammogram to reduce deaths from breast cancer. At the population level, if we screen 1,000 women for about 30 years, we will save approximately 8 of them from a breast cancer death. (Yes, it's 8, not 80, and not 800). At the same time, among these 1,000 women, there will be over 2,000 false alarms, and over 150 of these will result in an unnecessary biopsy. Some of these biopsies will incur further complications, though currently we  do not seem to have the data to quantify this risk. But what if even one of these biopsies were to lead to death of or another dire lasting complication in a woman who turned out not to have cancer? And by the way the accounting is not all that different when applying the new USPSTF mammography screening recommendations. Well, then we have the trolley problem, don't we? We are potentially sacrificing 1 individual to save 8. And who does the sacrificing is where the variations of the trolley problem come in.

Payers levy financial penalties on primary care physicians when they fail to comply with screening recommendations in their patient panels. The payer certainly sees this issue as the original formulation of the problem: Why not throw this financial switch to achieve net life savings? But for a clinician who deals with the individual patient this may be akin to pushing her over the bridge toward a potentially fatal event. Because we don't have a crystal ball, we cannot say which woman will die or incur a terrible complication. But the same population data that tell us about benefits must also give us pause when reflecting on the risks. Add the ubiquitous uncertainty (and lack of data) into this equation, and the implications are even more shocking. So, while making policy recommendations based on population data is sensible, policing uniform application of these recommendations to individual patients is fraught: of course, clinicians and patients need to be cautious about making individual decisions even when in population data benefits outweigh risks.

On the surface risk-benefit equations for many interventions may appear favorable, leading to blanket policy recommendations to employ them on everyone who qualifies. In the office, the clinician, caught in a tug of war between mountains of new literature and the ever-shrinking appointment times, is hard-pressed to take the time to consider these recommendations in the context of the individual patient. And furthermore, financial incentives from payers act as a short-hand justification, a "nudge," for doing as recommended rather than for giving it thought. So, who must look out for the patient's interest? The patient, that's who. Who understands the patient's attitude toward the risks and the benefits? The patient, that's who. Who now has to be responsible for making the ultimate informed decision about which track to stand on? The patient, that's who.

For me the trolley problem gives clarity to the reservations that I walk around with every day. I have done a lot of soul searching about why it is that, even if the benefits seem to outweigh the risks, I am still more often than not skeptical about whether a particular intervention is right for me. And since every intervention in medicine has a real risk, though mostly quite low, of going terribly awry, my skepticism is justified. This is my approach to evaluating these risks and benefits, based on my values and my understanding of the data as it is today.

What's the answer to this ethical conundrum in medicine? I cannot see that policy makers will stop throwing the switch in the near future, and so as a society we will be forced to accept the tram's collateral damage. And while this may make sense in an area such as vaccination, where thousands of lives can be saved by sacrificing a very few by throwing the switch, in most everyday less clear-cut medical decisions the answer is less clear-cut. Will doctors rebel against being forced to throw some patients on the tracks in order to save some marginally larger number of others? I don't think that they have the time or the energy or the incentive to do this, since the framing of the switch-throwing is through the rhetoric of "evidence." Right or wrong, doctors are shackled by the stigma of ignorance that comes with not following evidence-based guidelines, and this may act to perpetuate blind compliance. This leaves the patients, for some of whom the right thing will be just to get themselves off the tracks altogether, far away from the hurtling trolley until its brakes are fixed.                        

If you like Healthcare, etc., please consider a donation (button in the right margin) to support development of this content. But just to be clear, it is not tax-deductible, as we do not have a non-profit status. Thank you for your support!


  1. Great post! There's another parallel to be drawn between the trolley problem and healthcare decision making. Just as empirical philosophy is now being done - to figure out what people actually *think* about the trolley problem when you ask them, the solution to healthcare decisional dilemmas, as you point out, is often to ask the patient.

    The difference from the trolley problem is however also instructive. And it's the same difference that always applies between philosophical test cases and the real world. "Other things being equal..." but things are NEVER equal. There's a balancing which must be done of a kind that doesn't apply in the trolley case. The whole point of the trolley is that the same number of people die, assumed for the sake of argument. That never happens in clinical decision making.

  2. Zack, thanks for your insightful comment. I agree that there are many instructive lessons that we can learn if we reflect back on some of the arguments that already exist. And agree that the uncertainty inherent in life and medicine makes for much messier decision making. However, I find some of these thought exercises really useful in clarifying the tensions in our decisions.

  3. I agree with Zackary - great post. The two trolley car scenarios pit utilitarianism vs deontology and as you point out philosophers and ethicists have not resolved the conflict. The breast cancer screening example is terrific (or tragic) since it so nicely demonstrates the conflict. Some of the critics of the USPSTF 2009 MMG recommendations cited that detecting the 40 y/os breast CA was more valuable than detecting it in a 55 y/o (unless you happen to be the 55 y/o). This can be seen as a utilitarian position - the 55 y/o has less utility than the 40 y/o - yet to make utility estimates you have to assume some definition of human worth and this sort of thinking wanders into deontology. So why are humans valuable? Is it the net worth they bring to society? None of this necessarily conflicts with EBM but EBM is not equipped to consider these sorts of questions. Baconian science will not address the questions of 'why' or of the 'first cause' and in this way to uphold science as the solution to everything is disingenuous. Yes better understanding of EBM would render better, more efficient and less injurious care, but I think it is the societal expectation of science to answer everything that prevents physicians and patients to see the bigger picture.

    Keep up the great work!


  4. Thank you, UtilityKnife, for coming and commenting. I have struggled with these ideas myself for a very long time, an the persistent conclusion that I invariably get to is exactly what you have said: to expect science to answer ALL of these questions is ridiculous. It is, in fact, setting science up for failure. This why a considered dialogue is so needed at this time (and at all times, really), before we do the usual and rush forth with our sexy but pointless propensity to innovate.

  5. thanks for sharing.

  6. Whatever the case, it seems like the patients are the ones left in the scenarios. Experiments are done but we haven't got the answers yet. I'm not playing safe, it's just that it is apparently the issue we can't get over for years already.

  7. Great Post! If the patient ultimately is responsible to decide if screening/test are individually appropriate.. who bears the responsibility to ensure informed decision making per the patient? Can't be delivered in the short office visit or the all inclusive world wide web without research interpretation proficiency. Then there are the different levels of cognitive capability. Seems like the doctor is not left with a greater burden than the patient?

    New to the your post. Thanks for sharing!