Share this comment
I agree with you about the biasing effect of certain stats, particularly relative numbers and risks.
I appreciate the work done at the Patient Preferences Project around scenario planning: patientpreferences.org
The problem is, humans struggle to weigh numbers associated with incommensurate values. What does it mean to weigh x% chance of s…
I agree with you about the biasing effect of certain stats, particularly relative numbers and risks.
I appreciate the work done at the Patient Preferences Project around scenario planning: https://patientpreferences.org
The problem is, humans struggle to weigh numbers associated with incommensurate values. What does it mean to weigh x% chance of success (but with some downside) against y% risk of some burden against z% risk of a different burden. Furthermore, folks often end up transforming those numbers in their own minds into decision weights: this will happen, this won't happen, this probably will/won't happen.
Instead, you could tell a story, incorporating your knowledge of the evidence as well as your experience as a clinician, about what you think the best, worst, and most likely case scenarios would be with and without the intervention. Human minds are made for stories and much more easily compare and contrast them. You can do this after you discover what values are most important to this person (e.g., spending time with grandkids, vs avoiding healthcare interactions, vs living as long as possible no matter the trade-off).
This storytelling, of course, requires that the patient trust you: they need to trust that your account of things is accurate. Maybe they don't trust you, and they want the numbers so they can trust those instead. But they still need to trust that your account of the numbers is trustworthy (and you haven't fallen prey to believing relative number bias yourself).
Hi Joshua: I'm thinking there might be 2 different issues here - the examples on patientpreferences.org are all around conditions which are "symptomatic" dialysis, ICU, colectomy etc however when it comes to treating BP, lipids and glucose these are almost always taking asymptomatic people and lowering an estimated risk of getting a symptomatic thing. I can't think of how the patientpreferences.org approach would work for risk factors. Would love to hear how you think it would work for say a person with an "elevated" cholesterol where we might reduce the risk from 10% over 10 years down to 8% over 10 years and show this without using the type of tools I'm showing. The best case scenario is the 1 in 50 people like in the example will benefit but the other 49 won't.