Most doctors make decisions on behalf of patients on the basis of an impression of what they would have decided if they had been in the patient’s place. Patients can do the same of course by predicting how they would feel about each outcome or adverse effect of treatment. However, if doctors are expected to provide a transparent rationale then they might have to use a technique called ‘Decision Analysis’, which involves a considerable amount of arithmetic. Instead of listing each possible outcome for each treatment, it is necessary to list each possible combination of outcomes (see table 2). There can be a large number of such possible combinations. Each of these combinations represents the possible results of a theoretical study on huge number of patients with identical features and personal opinions to the patient. Ideally the data should depend on direct experience of the outcomes of a randomised control trial conducted on patients with a similar age, gender, symptoms, examination findings and test results to the patient on whom the decision is to be made. However, it is impossible to conduct such detailed studies and the figures would have to be estimated from the results of similar trials.

For each treatment option (represented by four treatments in the example shown in table 2), the doctor has to estimate the frequency of each combination of outcomes. (These are written as a percentage e.g. 0.1% in table 2.) The doctor or the patient also has to estimate the proportion of times the same patient would feel satisfied with that outcome if the study was done repeatedly at different times. (These proportions are written as decimals in table 2 e.g. 0.9.) By multiplying the percentage experiencing that combination of outcomes with the proportion being satisfied, one gets the proportion of times an identical patient would be satisfied with that combination of outcomes for that treatment. By adding up these proportions of an identical theoretical patient being satisfied for each combination, we get the total proportion of times that identical theoretical patients would be satisfied for that treatment (e.g. diet alone). By doing this for each treatment, we get a comparison of the proportion of times identical theoretical patients would be satisfied with each treatment. The patient should choose the treatment option which provides identical patients with the most frequent satisfaction.

Table 2

For each treatment, % with each outcome multiplied by proportion with satisfaction

Outcome combinations Diet alone Diet + Metformin Diet + Gliclazide Diet + Insulin

Outcome combination 01:

e.g. fatigue improved and 0.1%x0.9 0.08%x0.8 0.7%x0.7 0.7%x0.6

weight reduced and etc =0.09% =0.064 % =0.49% =0.42%.

Outcome combination 02: 0.1%x0.8 0.1%x0.7 0.1%x0.8 0.1%x0.5

=0.08 =0.07 =0.08 =0.05

Outcome combination 03: 0.09%x0.7 0.1%x0.6 0.08%x0.5 0.09%x0.4

=0.063% =0.060% =0.040% =0.036%

Etc

What often happens in practice is that the patient and doctor simplify the decision by focusing on the most important benefit from the patient’s point of view (e.g. resolution of persistent hitherto uncontrollable severe pain) and the worse adverse effect from the patient’s point of view (e.g. death during surgery). Other less important outcomes may be assumed to cancel each other out. If the probability of greatest perceived benefit is greater than the probability of the perceived worst loss, then the decision may be made to go ahead. If there is uncertainty then the patient (and doctor) will opt for an action where severe adverse effect (e.g. death) is unlikely.

No comparison has been made of the best method of deciding the best course of action in medicine or any other field (such a study would be very difficult to design). In practice, the majority of choices in day to day medical care are clear. Many patients find comfort in letting someone else with more experience of the situation decide for them. When patients are expected to make a decision in the face of great uncertainty, they usually opt for no treatment or the treatment with least dangerous adverse events. It might also help to ask for a second opinion and showing a doctor experienced in the relevant speciality a self-explanatory summary of the kind described in this webside.

© Huw Llewelyn 2016