I help train surgery residents in the US and agree completely with your observations and commentary. Are there specific resources (books, articles, videos etc …) you might recommend to improve my analytical skills while doing the same for my residents.
The incentives are awful and the thing we could do tomorrow in the UK to help with that would be to remove research/audits etc from any part of compulsory training. It's a race to the bottom, stewed in cynicism. Some people do not care, and just want to do what they are told, and that's OK. We're asking to be lied to and people are hearing our call and lying to us
It strikes me that we should consider most observational studies as hypothesis generating not hypothesis testing.
That would mean any finding in an observational cohort should be confirmed with an RCT or an a priori formed hypothesis tested in another observational cohort.
It also very important to realize that our knowledge only advances when the data is inconsistent with a hypothesis, not when data supports a hypothesis.
For medical therapies I think this is absolutely true. There are of course instances where RCT's are not feasible/ethical, in which case we are confined to observational studies. These alone are hypothesis testing, but could rarely ever provide enough certainty on their own to be conclusive. They would need to be assessed as part of a body of evidence which takes into account their inherent risk of bias or unaccounted for confounding.
What I have seen is that most major medical journals won't accept a biomarker paper based upon retrospective analysis of existing data. They at least want to see the result confirmed in a validation cohort, which could also be existing data.
The challenge we have is that with increasingly engineered and therefore potentially highly personalised therapies the large scale RCTs will not be possible. This is also the case for adaptive AI algorithms.
I help train surgery residents in the US and agree completely with your observations and commentary. Are there specific resources (books, articles, videos etc …) you might recommend to improve my analytical skills while doing the same for my residents.
Many Thanks
Thank you! I think it is challenging to do, but I did compile some of my preferred resources to get started in this previous post - I hope it helps!
https://open.substack.com/pub/alasdairmunro/p/so-you-want-to-be-a-clinical-academic?r=1fhhmw&utm_campaign=post&utm_medium=web
Thank you! Thoroughly enjoy your posts.
The incentives are awful and the thing we could do tomorrow in the UK to help with that would be to remove research/audits etc from any part of compulsory training. It's a race to the bottom, stewed in cynicism. Some people do not care, and just want to do what they are told, and that's OK. We're asking to be lied to and people are hearing our call and lying to us
I completely agree.
Much more important is training people to interpret research properly, and to be able to support research activities under appropriate supervision.
All these mandatory output requirements do nothing but hurt the cause, especially as they only breed resentment in the process.
It strikes me that we should consider most observational studies as hypothesis generating not hypothesis testing.
That would mean any finding in an observational cohort should be confirmed with an RCT or an a priori formed hypothesis tested in another observational cohort.
It also very important to realize that our knowledge only advances when the data is inconsistent with a hypothesis, not when data supports a hypothesis.
For medical therapies I think this is absolutely true. There are of course instances where RCT's are not feasible/ethical, in which case we are confined to observational studies. These alone are hypothesis testing, but could rarely ever provide enough certainty on their own to be conclusive. They would need to be assessed as part of a body of evidence which takes into account their inherent risk of bias or unaccounted for confounding.
What I have seen is that most major medical journals won't accept a biomarker paper based upon retrospective analysis of existing data. They at least want to see the result confirmed in a validation cohort, which could also be existing data.
The challenge we have is that with increasingly engineered and therefore potentially highly personalised therapies the large scale RCTs will not be possible. This is also the case for adaptive AI algorithms.
hicpac@cdc.gov
TELL THIS COMMITTEE TO STOP RIDICULOUS, OUTDATED COVID RESTRICTIONS LIKE FORCED NASAL RAPES (IE TESTING) REGARDLESS OF SYMPTOMS, MASKING AND ISOLATION