In a previous post, I discussed the roles assigned by a decision analysis. The patient provides the utilities, the provider gives the probabilities of outcomes, and optimization combines utilities and probabilities to give the best decision. In a follow-up post, I described how things often work differently in the real world, and that sometimes it is best that way.

However, I will argue below that, in theory, the roles assigned to the patient and provider by decision analysis are kinder than would be the roles they would otherwise assume when making a decision more informally.

A decision analysis is kinder to the patient. With a decision analysis, the patient just needs to know their utilities: for example, how they will feel if a certain event, such as a side effect, occurs. Admittedly, this can be challenging (and the provider sometimes helps, as discussed in this post), but focusing only on utility frees the patient from having to think about the probability of different events, which can be difficult without formal medical training. Knowing one’s utilities is ultimately easier than the Sisyphean task of figuring out how to put probabilities and utilities together in the end (which should, according to decision analysis, be offloaded to mathematical optimization). Also, the patient’s focus on utility clearly defines it as their territory, which protects the incorporation of their preferences.

A decision analysis is kinder to the provider. With a decision analysis, if the provider gives probabilities of outcomes that are valid (e.g., based on the current medical literature or their clinical experience), they have done their job. This is easier than making a full decision for a patient (i.e., easier than the task of putting probabilities and utilities together, which, according to decision analysis, again, should be offloaded to mathematical optimization). A decision analysis also allows the provider to offload some cognitive burden to the literature, if the probabilities for which the provider is responsible are available there. This can be particularly helpful with complex, continuous outcomes like survival and toxicity burden, as in the scenario described here. Sometimes, the necessary probabilities might not yet be present in the literature, and, in such a case, providers can steer the research community in ways that will make these probabilities more easily available. Also, a decision analysis allows the provider to avoid the need to assert true understanding of another person’s utilities, which can be fraught with ethical quandaries.

In general, as mentioned, decision analysis spares both the provider and the patient from figuring out how to put probabilities and utilities together (i.e., from having to make a final medical decision). This often falls onto the shoulders of one or the other, otherwise. With a decision analysis, given the probabilities from the provider and the utilities from the patient, a calculator can be used to combine probabilities with utilities and to then find the “optimal” decision.

One might be averse to the presence of a calculator, or mathematics in general, in something so deeply personal as a medical decision—it can feel almost as if one is relinquishing control. However, a decision analysis clearly defines where the non-mathematical, human component of the decision belongs (the utility). This empowers patients to focus on this human aspect of the decision. Further, a decision analysis clearly defines where the provider’s expertise belongs (the probabilities), empowering them to focus on providing this in the best way possible.

In a sense, a decision analysis suggests that the mathematics of probability and optimization lurk behind every medical decision. By acknowledging this, we free ourselves from the mathematics, and—as discussed here—from one another’s responsibilities, giving ourselves kinder roles in the process.