|
|||
 |
|||
|
|
|
|
|
|||
 |
|||
|
|
|
|
 | ||||
 | ||||
| Example of (medical) Decision Analysis Example below is not necessarily factual, accurate, or correct; but is adapted loosely from p. 286 of User's Guide to the Medical Literature, edited by Gordan Guyatt, MD and Drummond Rennie, MD, AMA Press, 2002 for the purpose of providing an example of the strength and weakness inherent in conventional, consequentialist reasoning. --------------------------------------------------------------- Impetus for Making a Decision Patient w/Atrial Fibrillation [something which can cause circulating clots (embolisms) leading to a stroke] --------------------------------------------------------------- Numbered Rival Strategies (along with their 4 "lettered" outcomes) --------------------------------------------------------------- 1. No Prophylaxis [assumption of a 10% chance of Stroke and a 1% chance of Bleed; a 90% chance of No Stroke and a 99% chance of No Bleed] --------------------------------------------------------------- a) No Stroke, No Bleed probability = 0.891, utility = 1.0 outcome value = [probability] x [utility] = 0.891 x 1.0 = 0.891 b) Stroke, No Bleed probability = 0.099, utility = 0.5 outcome value = 0.0495 c) No Stroke, Bleed probability = 0.009, utility = 0.8 outcome value = 0.0072 d) Stroke and Bleed probability = 0.001, utility = 0.4 outcome value = 0.0004 --------------------------------------------------------------- Total [a+b+c+d] value of "No Prophylaxis" strategy [to be contrasted against rival strategies] = 0.948 --------------------------------------------------------------- 2. Aspirin [assumption of a 8% chance of Stroke and a 2% chance of Bleed; a 92% chance of No Stroke and a 98% chance of No Bleed] --------------------------------------------------------------- a) No Stroke, No Bleed probability = 0.9016, utility = 1.0 outcome value = 0.9016 b) Stroke, No Bleed probability = 0.0784, utility = 0.5 outcome value = 0.0392 c) No Stroke, Bleed probability = 0.0184, utility = 0.8 outcome value = 0.0147 d) Stroke and Bleed probability = 0.0016, utility = 0.4 outcome value = 0.0006 --------------------------------------------------------------- Total [a+b+c+d] value of "Aspirin" strategy [to be contrasted against rival strategies] = 0.9561 --------------------------------------------------------------- 3. Warfarin [assumption of a 4% chance of stroke and an 8% chance of bleed] a) No Stroke, No Bleed probability = 0.8832 utility = 1.0 outcome value = 0.8832 b) Stroke, No Bleed probability = 0.0368, utility = 0.5 outcome value = 0.0184 c) No Stroke, Bleed probability = 0.0768, utility = 0.8 outcome value = 0.0614 d) Stroke and Bleed probability = 0.0032, utility = 0.4 outcome value = 0.0013 --------------------------------------------------------------- Total [a+b+c+d] value of "Warfarin" strategy [to be contrasted against rival strategies] = --------------------------------------------------------------- Summary of relative values No Prophylaxis = 0.948 Aspirin = 0.9561 Warfarin = 0.9643 --------------------------------------------------------------- As a talking point, a preference for the use of Warfarin in such patients arises from the decision analysis above, as it has the highest relative value of the 3 rival strategies. But along the way there were several assumptions, such as a discounting or overlooking of other negative outcomes besides bleeding and stroke (such as treatment costs and patient/doctor inconveniences), validation of the probability estimates for the 4 paired outcomes, and evaluation of the actual, true, and objective utility of these paired outcomes to the patients in question. One can imagine such decision analyses applied to mundane tasks such as the decision to get into your car and drive to work (even though, if you do that, you run the risk of being killed in a car accident) or to weightier issues, such as the moral and political collectivisation of mankind. Perhaps the toughest task is to ascribe utilities to expected outcomes. It may be impossible to escape subjectivity when attempting to do something like that. Ed | ||||
| ||||
 |
 |
|
 |
|
