The acceptance-refusal asymmetry in competence requirements

In 2016, this Journal published an article by Rob Lawlor1 on what we might call the acceptance-refusal asymmetry in competence requirements. This is the view that there can be cases in which a patient is sufficiently competent to accept a treatment (viz., to give consent to it), but not sufficiently competent to refuse it (viz., to withhold consent to it). Though the main purpose of Lawlor’s paper was to distinguish this asymmetry from various other asymmetries with which it has sometimes been confused,¹ Lawlor also presented a brief case in favour of it. Developing an earlier argument of Ian Wilks’,2 3 Lawlor argued that, when the risks associated with refusing a treatment are graver than those associated with accepting it, a higher level of competence may be required to refuse a treatment than to accept it.

This claim could have important implications for the law, since determinations of competence often play a central role in determining the lawfulness of refusing or imposing a treatment (eg, in England and Wales under the Mental Capacity Act 2005). Indeed, the acceptance-refusal asymmetry in competence requirements, or something close to it, has played an important role in court judgments.² However, Lawlor himself suggests that his conclusion will have practical implications only in a narrow range of cases.³

The challenge

In this issue, Pickering, Giles Newton-Howes and Simon Walker (henceforth ‘the authors’) respond to Lawlor’s piece.4 They deny that competence requirements should depend on the level of risk associated with a decision, and thus that there is any basis for the acceptance-refusal asymmetry in competence requirements.

Part of the authors’ argument involves contesting the way in which Lawlor uses cases to support his view. In one case cited by Lawlor—and drawn from Wilks—we are invited to consider a whether an acrobat is sufficiently competent to walk a tightrope. According to Wilks and Lawlor, it is intuitively plausible that an acrobat might be competent enough to walk the tightrope with a safety net, but not without one. This, they suggest, supports the view that whether one is sufficiently competent to do something can depend on the riskiness of doing it. The authors’ worry, however, that, insofar as we have the intuitions about this case that Lawlor and Wilks expect us to have, this may be due to our conflating competence and safety. Once we take care to distinguish the two, we might be more inclined to say that that the presence of a safety net affects the acrobat’s safety, but not whether he is sufficiently competent.

The authors’ main strategy, however, is to deny that mere appeals to case-based intuitions are sufficient to establish that competence requirements should depend on risk. What is also required, they claim, is a convincing explanation of why risk is relevant. The authors doubt that any such explanation can be furnished: they claim that ‘there is no logical connection that shows why riskier choices should require a higher degree of capacity’. While they concede (perhaps only for the sake of argument) that the difficulty of a decision to accept or refuse a treatment might be relevant to the competence requirements that apply to it, they deny that the difficulty of a decision will track the risks associated with taking it. As they rightly note, we often face difficult to decisions between low-risk options, and easy decisions between high-risk options.

Two types of competence requirement

I agree that Lawlor owes us an explanation for why competence requirements should depend on risk, but I want to suggest that this demand for an explanation can be at least partiality satisfied by distinguishing two different ways of understanding ‘competence requirement’. The distinction is very similar—perhaps identical—to one drawn by Lawlor himself.⁴ In what follows, I first introduce the distinction, and then invoke it in response to the author’s challenge.

We can approach the distinction I have in mind by noting, to begin with, that there will be some level of competence that in fact suffices for a person to be autonomous with respect to a decision—including a decision to accept or refuse a medical treatment. Call this level of competence the true standard of competence.⁵

We will sometimes not be able unable to determine, with a high degree of certainty, whether a patient meets the true standard of competence. There are two reasons for this. First, there is empirical uncertainty. We will sometimes not be able to empirically ascertain, with complete certainty, which autonomy-relevant capacities a person has and to what degree. Second, there is normative uncertainty. We do not know for sure what capacities a person needs to have, and to what degree, in order for a decision to be autonomous. Different theories of autonomy will give different answers to this question, and we do not know for sure which theory of autonomy is correct.

Still, we need to get by. We need to decide whether to administer treatments. And, since a person’s own autonomous decisions to accept or refuse a treatment can affect the moral permissibility of administering that treatment in a way that her non-autonomous decisions would not, we need to make judgements about autonomy. Thus, we need some operationalisable measure of competence. More specifically, we need to decide, on the basis of the capacities that a person evidences, whether to treat that person’s decision to accept or refuse a treatment as if it were autonomous. Let us call the level of capacity that a person must evidence if we are to treat her decision as if it were autonomous the test for competence. The test for competence might be a particular score that the patient needs to receive on some standardised scale, such as the MacCAT-T.

False positives and false negatives

The true standard of competence is not something that we can choose. It just depends on what theory of autonomy is correct and what verdict that theory gives in a particular case. But the test for competence is something we can choose. We can decide how demanding we want this test to be—for example, how well an individual has to perform on the MacCAT-T. And we could decide to set this threshold at a different level for treatment acceptance than for treatment refusal.

In setting this threshold, we ought to attend to two possible negative outcomes—what we might call ‘false positives’, and ‘false negatives’.

False positives are cases in which a person passes the test for competence—so we treat her decision to accept or refuse a treatment as though it were autonomous—though in fact, since the person does not meet the true standard of competence, her decision is not autonomous. False positives are problematic because, in these cases, we will typically respect the patient’s decision (withholding treatment if the patient refuses it, and providing the treatment if she accepts it) though in fact, since the patient’s decisions are nonautonomous, we should decide whether to provide the treatment on the basis of the patient’s best interests.

False negatives are cases in which a person does not pass the test for competence—so we treat her decision as though it were nonautonomous—though in fact the person meets the true standard and the decision is autonomous. False negatives are problematic because, in these cases, we may disrespect the patient’s autonomy. We may provide or withhold the treatment on the basis of a patient’s best interests, since we assume that the patient’s decision is nonautonomous, when really the patient’s decision is autonomous, and we ought to be respecting this autonomous decision.

How to set the test

When setting the test for competence, we will face a trade-off between false positives and false negatives. The less demanding the test, the more false positives we will get. But the more demanding the test, the more false negatives we will get.

It is plausible that, in seeking to resolve this trade-off, we should have an eye both to the number of false positives and false negatives we will get, and to the moral cost associated with each. But these moral costs will differ depending on whether the patient’s decision is to accept the treatment or to refuse it. False positives will typically come with greater moral costs when the patient rejects the treatment than when the patient accepts it. This is because medical treatments are typically only offered when they are thought to be in line with the patient’s best interests. Healthcare professionals are typically under an obligation to offer only treatments that they deem to be in their patients’ best interests, and it is reasonable to suppose both that they typically comply with this obligation, and that their assessments of best interests are somewhat reliable. We can thus assume that treatments offered in medical contexts at least tend to be in line with the patient’s best interests.

Thus, when we decline to provide a treatment on the basis that the patient has refused it, we will likely be withholding a treatment that was in the patient’s best interests. And when the case is a ‘false positive’—when the patient’s decision to refuse was in fact nonautonomous—we will thus be failing to do what we ought to do—viz., act in line with the patient’s best interests.

On the other hand, when we provide a treatment on the basis that the patient has accepted it, we will likely be providing a beneficial treatment. Thus, even if the case is a false positive—the patient’s decision to accept was nonautonomous—we will, fortuitously, likely to be acting as we should act—viz., in line with her best interests.

All this suggests that, when setting a test for competence in relation to treatment refusal we should be more concerned about false positives than when setting a test in relation to treatment acceptance. Making the test too easy—and thus generating a lot of false positives—is more problematic in relation to refusal, where it will typically result in a failure to provide beneficial treatments, than it is in in relation to acceptance, where it will typically result in the provision of beneficial treatments.

Final thoughts

What I have just given is a kind of pragmatic argument for adopting a test for competence that is sensitive to risk, and thus to whether a treatment is accepted or refused. I interpret Lawlor himself as claiming also that the true standard of competence is sensitive to risk. However, my argument entirely consistent with thinking that the true standard of competence is invariant to risk—perhaps for precisely the kinds of reasons that the authors give.⁶ Drawing a distinction between the true standard of competence and the test for competence may thus allow us to maintain that both the authors and Lawlor are partly correct.