“The failure to train doctors about clinical uncertainty has been called “the greatest deficiency of medical education throughout the twentieth century.”
Djulbegovic (2004), https://www.bmj.com/content/329/7480/1419
Life is full of uncertainties. In decision making uncertainty means that it is impossible to define future outcomes. Amount of medical information continues to grow exponentially, but uncertainty is still inevitable in medical decision making. Uncertainty is everywhere in medicine. We do not know whether this pill will reduce patient´s blood pressure, we do not know whether patient will have side effects after starting the medication, we can´t be 100% confident that the diagnosis which was made, is truly right, we can´t be 100% confident that all life-threatening conditions possibly causing patient´s symptoms were ruled out, we can´t be 100% confident that patient will get better after an operation. Decisions for individual patients are like navigating though a sea of uncertainty (Djulbegovic, 2004).
Dealing with uncertainty is easier when one understands the framework of medical decision making and some qualitative characteristics within this framework. Below is a summary, which describes some properties of the medical decision making framework.
The most important is the knowledge situation. Decision can be made under certainty, risk, uncertainty or ignorance. Annotations for each knowledge situation is described in the graph. These knowledge situations have at least three dimensions, which give more insight and help to identify which situation is at hands.
Subject matter research availability and quality. If we wish to have a knowledge situation in which we can make a decision under risk, we need extensive amount of literature which is also of high quality. Decision under risk means that we have exact probabilistic knowledge which can be used and applied in the decision making. Example of such information is exact probability estimate whether a new drug will, say lower patient´s systolic blood pressure below a certain target value. Or we have an exact estimate of the probability whether our patient in the need of the drug will suffer from certain side effect. This sort of research data is extremely scarce. The lesser the quality of the research data and the scarcer it is, we have to scale down our knowledge situation. For numerous surgical interventions only retrospective and descriptive data is available meaning that data is scarce and poor quality. At best we can estimate some sort of range of probabilities (ie. less than 10%, or 1-5%) for certain complication to happen. This naturally means what the knowledge situation in which we are workin is decision under uncertainty. The poorer the data the more we are working under ignorance.
Likelihood of heuristics in decision making. As said, medical decision making is full uncertainties. Since decisions must be usually done during the course of treatment of patients, decision maker must be able to deal with uncertainties. Dealing and facing uncertainty can increase cognitive load. Cognitive load can be eased by using heuristics. Heuristics are simple strategies to make a decision. Use of heuristics will only lessen the cognitive load to make a decision and they are not always optimal, perfect, logical, or rational. Numerous heuristics have been proposed. Those outlined by Tversky and Kahneman are probably the best known and they are also very applicable in medical decision making.
Level of treatment effect heterogeneity. This aspect is not as robust as the other two. Treatment effect heterogeneity indicates how different patients respond differently to same treatment given. In short, treatment effect heterogeneity is a very multifaceted topic. In medical decision framework I propose that the less there is treatment effect heterogeneity, the closer we are working under risk. By this I mean that if average treatment effect is very similar and with low variability across different patients, it is easier to establish some sort of probabilistic knowledge. On contrary, if treatment effect heterogeneity is large meaning that average treatment effect is associated with large variability, it is harder to establish probabilistic knowledge and hence we are working closer to ignorance than risk.
All this represents very hypothetical decision framework. As shared decision making continues to evolve and gain more wider foundation, it is probable that we will see new perspectives regarding decision theory in medicine.