This article is better than its headline (can we just stop the antiquated tradition of letting late-night copyeditors write the titles that may be all anyone reads of our long articles that we labor over?), "Models vs. Evidence". bostonreview.net/science-nature…
I appreciate the effort to figure out what is going on with the debates over the pandemic response. But I think this article gets a few things right and a number of key points wrong.
It gets right that "Public health epidemiology uses theory... to model infection and understand patterns and causes of disease. Public health epidemiology also relies on a diversity of data." Thank you for that description of our field at its best.
I'd call that bayesian reasoning, which is awfully close to my working definition of rationality. Of course randomized evidence is in some ways more precise and convincing than nonrandomized evidence. But ignoring nonrandomized evidence is a recipe for a short miserable life.
For a lighthearted example see bmj.com/content/363/bm…
Clinical epidemiology, and specifically the evidence-based medicine tradition within clinical epidemiology, does privilege randomized evidence over anything else, as do some economists and sects within other fields.
But from here the article just misunderstands our fields. First, to clarify (pace the headline), as the author notes, public health epidemiologists (as he calls us) privilege models and evidence, not models over evidence. That's just the headline, not a fault of the article.
Second, clinical epidemiologists (the good ones) are open to all forms of evidence, appropriately weighted. For example a leader of the Clinical Epi program at @HarvardEpi is @_MiguelHernan. He's made his career on rigorous evaluation of observational data hsph.harvard.edu/miguel-hernan/…
Using John Ioannidis as the sole representative of the clinical epidemiology tradition is like using @JerryFalwellJr as the representative of the Christian tradition. He represents one part of it but not the whole thing.
Clinical epidemiologists who are rational about decision making under uncertainty understand that with noncommunicable diseases, inaction in the face of uncertainty is rational (it is unlikely to get much worse or better if we do nothing), but not so for infectious disease.
This point was beautifully articulated by two of our former department chairs, Hans Olov Adami and the late Dimitrios Trichopoulos nejm.org/doi/full/10.10… 
Rationality (roughly, Bayesian decision theory) takes into account the costs of inaction and of action when deciding. Fanatical devotion to tenets of "evidence based" decisions would recommend staying on the traintracks until an RCT showed benefits of evading an oncoming train.
This is what Adami/Trichopoulos, two giants of clinical epidemiology, meant. This is also noted by ID modelers such as Ed Kaplan @YaleSPH in this classic paper pnas.org/content/99/16/…. It's been criticized but gets the decision problem exactly right (fig 3 & accompanying text).
Finally a couple of other notes: first, don't blame ID epidemiologists (otherwise almost coextensive with "public health epi") for @IHME_UW model. Everyone agrees it is unfit for purpose. annals.org/aim/fullarticl… for example, but really everyone. Even Bill Gates funds some flops.
Last, the conclusion (like the title) is just misleading. The body said that public health epi as defined here combines theory with evidence. The conclusion is that we need the best of both worlds. Surely, clinical epi has much to offer, as does the public health tradition.
But what they have to offer -- the good stuff -- is that both already value theory and evidence. Emphasis may be different, frameworks for thinking may be different, but rational thinking is rational thinking, and blind adherence to arbitrary principles isn't it.
Serious scientists in both traditions use theory for hypotheses & guidance when evidence is scarce, use data of all kinds to test all theories & slay wrong ones, & realize that decisionmaking under uncertainty rationally weights type 1 and 2 errors differently according to costs.
As some ad hominem examples, @_MiguelHernan , clinical epidemiologist extraordinaire, is playing a major role in the Spanish pandemic response, while it was modelers (@nataliexdean @CT_Bergstrom and me among others) who noted the problems with the Santa Clara study.
Good science is good science, and good scientists are flexible thinkers not wedded to arbitrary rules. Finis.
Here is a link to my proof from Bruce’s birthday. The fist slide is aphorisms I learned from him. @RELenski was there. As was @CT_Bergstrom who did an interpretive dance as I recall. dropbox.com/s/o4fcxa9p7036…
Correction: @BostonReview editor Matt Lord tells me @jonathanjfuller approved the headline, sorry for the confusion and criticism of copyeditor (perhaps hypocritical criticizing "late night" when tweeting 11pm).
Several suggest the article and I may not disagree that much. Will take a fresh look and perhaps write something more conventional than a tweet thread.
It may be a normative/descriptive thing -- narrow ways of thinking exist in science but I hate to see them legitimized as ways of doing science (normative-ish), rather than as professional rules of thumb/habits (descriptive-ish).
