Thread by @AnupamBJena

Our piece in @HarvardBiz today on the 'Art of Evidence-Based Medicine,' where @ChrisWorsham & I argue that EBM, while impt (b/c knowing something is clearly better than knowing nothing), is something we should be cautious to apply in rote fashion. THREAD.

hbr.org/2019/01/the-ar…

2/x There is a sentiment that docs who depart from EBM - perhaps they prescribe a drug that's inferior in RCTs - either are clueless as to the evidence or worse, they think their anecdotal evidence outweighs systematic investigations. But, this isn't what RCT evidence tells us.

3/x RCTs tell us what works on average, not for individual patients, which is well recognized. But consider this though experiment

The best doctor you know suggests a treatment to some patients that *isn't* consistent w/ EBM-guidelines. Trust the doc or guideline?

I'd go w/ doc

4/x Why? B/c I assume that doctor incorporates the scientific evidence into decisionmaking? If he/she doesn't, then that's different. But RCTs tell us if we have NO information on predicted treatment effect, better to treat all patients with drug A than all patients with drug B

5/x This means that of 3 choices:
(1) Docs who always adhere to guidelines
(2) Docs who never adhere
(3) Docs who mostly but not always adhere

We may expect patient outcomes to be better for (3) > (1) > (2)

This idea could, btw, be tested in actual data, but it hasn't been

6/6
Summary: when docs practice in ways that deviate from EBM, it doesn't mean that they are 'clueless' in some way. To argue that's true, one needs to show that's true by studying doc-level outcomes as a function of % deviation from EBM. The same logic appears to quality metrics