 |
 |
Presumably the degree of precision needed in a probability estimate depends on the relevant decision threshold. But decision thresholds in turn depend on the data used to calculate them, which may be just as problematic as clinicians' probablity judgments. IMHO, quantitative decision theory applied to health care makes logical sense. However, really trying to apply it to real health care situations is extremely difficult in practice. John Epling wrote: Responding to Roy Poses et al. about pre-test probabilities... (Hi Roy...long time, no email!) How much precision do we need in this? Acknowledging that there is a lot of cognitive bias work that we're ignoring here, I teach something akin to what Huw was getting at - a principal utility of this knowledge is to decide whether a test will change your management of the patient. And for that, maybe we just need broad ranges of pretest probability..."pretty sure disease present", "completely unsure" and "pretty sure disease absent". (Corresponding PTP numbers would be 80-90%, 50%, and 10-20%). Then if the test results would move you across those ranges, you should consider ordering it. I'm not sure clinicians (or patients) need much more precision than that. I know that's not terribly intellectually satisfying for a bunch of people that like statistics, but an honest guess like this would be better than misapplied attempts at precision. The question then remains whether physicians are any good at these broad brushes of prediction. We probably aren't... John Epling, MD, MSEd Professor and Chair, Department of Family Medicine SUNY Upstate Medical University Syracuse, NY -- Roy M. Poses MD FACP President Foundation for Integrity and Responsibility in Medicine (FIRM) [log in to unmask] Clinical Associate Professor of Medicine Alpert Medical School, Brown University [log in to unmask] "He knew right then he was too far from home." - Bob Seger