RCT shows flaws in less rigorous methods

FAIL

JAMA - Feb 26 2019
PACT-HF trial

Van Spall et al.
Effect of Patient-Centered Transitional Care Services on Clinical Outcomes in Patients Hospitalized for Heart Failure
The PACT-HF Randomized Clinical Trial
https://jamanetwork.com/journals/jama/fullarticle/2725688

FAIL
Health care hot spotting  NEJM 

https://www.nejm.org/doi/pdf/10.1056/NEJMsa1906848?articleTools=true

Health Care Hotspotting — A Randomized,
Controlled Trial
Amy Finkelstein, Ph.D., Annetta Zhou, Ph.D., Sar


OP ED NEJM

https://www.nejm.org/doi/full/10.1056/NEJMsb1901642

The Magic of Randomization versus the Myth of Real-World Evidence
List of authors.
Rory Collins, F.R.S., Louise Bowman, M.D., F.R.C.P., Martin Landray, Ph.D., F.R.C.P., and Richard Peto, F.R.S.

2020 NEJM 382_674 Collins Need RCT_Other Designs FAIL Op ed


___
Note that both articles above find flaws in the amount of value created by home health care (such as for intensive patients) - but this is often taken for granted, as in Harvard Business Review 10/2019 here.   You shouldn't take it for granted, as noted by Finkelstein.

See also "Why Nobody Believes the Numbers," about disease management flaws, by healthcare guru Al Lewis (here).

See also a too-little-noticed example where machine learning worked very well in retina diagnosis studies, but fizzled when used in real life (too many real world variables) - picked up by TechCrunch here.  Secondary article also here - commendably, by Google itself.   And original real world fizzle article is here as a conference proceeding abstract by Beede et al. 

These appeared about the same time as a favorable article on (controlled condition) machine learning retinal testing by Milea in NEJM, 4/30/2020, here.   (Milea pairs with an Op Ed by Ghandi et al., here.)