Thread by @jaygirimd

1/ THREAD Is @ABIMFoundation Interventional Cardiology (IC) board certification associated with improved outcomes after PCI? Latest study from all-star cardiovascular and @PennLDI fellow Sameed Khatana in @CircIntv investigates:
ahajournals.org/doi/10.1161/CI…

2/ Sameed used data from the NY state PCI registry crosslinked with info from @doximity and other publicly available sources to see if the following MD characteristics were associated with better outcomes after PCI...

3/ Years since Fellowship Graduation
International vs US Medical Graduate
Graduation from US News Top 25 Med School
IC Board Certification
Participation In @ABIMFoundation Maintenance of Certification Program

4/ Which of these MD characteristics do you think will positively impact PCI mortality?

5/ Sameed initially discovered a very modest statistically positive association for IC board certification with IC boarded MDs having 1.06 deaths per 100 cases and non-IC having 1.14 death per 100 cases.

However, 1 key adjustment factor was missing...

6/ We had not yet adjusted for operator case volume. After factoring this in, the board certification association vanished.

7/ Caseload ended up being the only factor independently associated with mortality after accounting for clinical/presentation factors already included in the NY PCI mortality model.

8/ For those up on this literature, you may recall this @CircAHA paper from our co-author, Paul Fiorilli, looking at some of these issues using NCDR data a few years ago: ahajournals.org/doi/abs/10.116…

9/ The conclusion from that study was a very small positive association with IC board certification and PCI mortality.

So why the discrepancy with our current work?

A couple of explanations...

10/ First, the association they found was VERY small, a difference in in-hospital mortality of 0.08% corresponding to a NNT for board certification of 1250.

Whenever you see such small associations that are statistically significant with large datasets, view with skepticism.

11/ Please see this great article on the “e-value” that explains how to quantify the vulnerability of your observational data to an unmeasured confounder:

annals.org/aim/article-ab…

12/ As it turns out, Paul had accounted for operator volume but modeled it dichotomously (high volume was greater than 50 procedures in year 2010).

13/ Interestingly, a secondary analysis in the paper modeled in interaction terms for operator and hospital volume and this caused the statistical association with board certification to vanish.

14/ The authors understood their finding was not a “positive study” for board certification and were measured in their writing.

This has not stopped @ABIMFoundation from using this, along with plenty of other suspect observational data, to push a board cert and MOC agenda

15/ It remains disappointing that despite so much 💰 flowing in, there appear to be no attempts on the horizon to carry out an independent RCT that tests the value of these ABIM products.

Thanks very much to @rwyeh and @rkwadhera for penning the associated editorial covering this issue and the broader context of how ICs should best me measured and reported on:

ahajournals.org/doi/10.1161/CI…

And if you believe that a moratorium needs to be put on the non-evidence based ever expanding ABIM MOC net, please support important effort of @doctorwes:

gofundme.com/practicing-phy…

Sorry, for those interested, link to full study in first tweet is broken. Here is a working link:

ahajournals.org/doi/full/10.11…