Blog: Decision making in healthcare: from doctor-level to system-level care
The world we know is being reshaped by all rapid technological advancements that are taking place and in their rise these technologies change structures and levels at which decisions are taken. This will not be different for healthcare. One of the changes there will probably be a shift from doctor-level care to system-level care.
The human body is an impressive system of complex interacting processes. Doctors need years (if not lifetimes) to learn to understand how it works. For everybody else it is often simple: “When you’re sick, you go to see your doctor.”, putting a great responsibility on their doctors’ shoulders as a result. Yet, with this responsibility, also comes great power. Doctors have a large autonomy in deciding whom receives which care at what moment, effectively making them the gatekeepers to healthcare and giving them a monopoly on (the access to) its resources. As a result, healthcare can be said to be driven from the level of doctors. Their impartiality and patients-first attitude (among others sworn in their Hipocrates oaths) make them the logical place to decide on access to care, often even not even needing to take consequential costs into account. With external factors out of play, doctors are able to decide solely on what is best for the individual in front of them.
But, this is about to change. The discretionary power to decide what patient receives which care might start to be reduced soon, changing the place of doctors in healthcare.
The high speed at which new technologies are coming up and improving might soon enable them to take over parts of decision-making processes of doctors. Artificial Intelligence, in its infancy still, is already able to assist clinicians in calculating their patients’ life expectancy based on their risk factors. Helping doctors and their patients in making treatment choices. The uPrevent initiative of the UMC Utrecht in the Netherlands for example, brings a dashboard that can predict cardiovascular risks for patients and shows the possible effects of receiving treatment. Although these recommendations are often basic still, they sometimes already outperform clinicians on some areas since they can take way more factors into account. Currently, researchers are exploring possibilities in taking the next step by recommending treatments for (subgroups of) patients instead of risk scores. Decision support systems like these might soon start replacing parts of current clinical practice guidelines and will probably develop to slowly take over (simple) care decisions that are now still made by human doctors. Even if AI would not become ‘smarter’ it will be able to take over an increasing amount of care decisions by breaking these decisions into ever smaller decisions until they are feasible to answer based on data.
Although, chances are slim that algorithms will take over the roles of doctors anytime soon it does mean that doctors might have to say less in care decisions in the future. Already, doctors are ought to follow guidelines most of the time. Yet until now it was never verifiable whether they did. With AI supporting decisions it will become harder for doctors to deviate from the programmed standards. Decisions made by these algorithms will be driven by how they are programmed by their designers and programmers. This will shift the decision making powers currently held by doctors more to a system-level. Doctors might keep a say in an indirect way on these decisions since they have the substantive knowledge, but their say will more limited in examination rooms. It will not make doctors superfluous, but will change their profession and position in the healthcare system. Their current position based on the dependence of their knowledge and decisions, will shift to a more caring one with less decision power.
Increased compliance to clinical practice guidelines will probably be good for healthcare outcomes on an aggregate level and might increase efficiency. However, it also contains risks. When the margin of appreciation in decision making of doctors is limited they will also lose the ability to personalize care for their patients. Even if algorithms are designed to give personalized decisions it will be impossible to tailor them exactly to every patient. This might prove to be a problem that will emerge mainly in policy recommendations deciding who should get care. These decisions are often based on factors that are hard to capture in standard algorithms, currently solved by letting doctors tailor them to the patients in front of them. When algorithms would decide on these decisions it could become harder to deliver the right care to the right patient, especially when these algorithms contain biases resulting from the data on which they were trained.
These shifts in decision making levels from a doctor to system level and the related risks of implement algorithms should not stop us from implementing them in healthcare. However, it is important to think about their consequences and how to handle these.
Disclaimer: I derrived the ideas for this blog (shifts in decision making levels) from a paper of a similar phenomenon in bureaucracies by Mark Bovens and Stavros Zouridis, enthusiasts can find it here.