Entomologists got it right!
What have scientists working to eliminate mosquitoes understood that epidemiologists have not?
It has been two weeks since I published the first entry in this newsletter, in which I showed that Mindel Sheps’ idea from 1958 has been independently rediscovered several times in several very different academic fields. Since then, I have been made aware of several more very interesting instances of this, some of them actually preceding Sheps (though applied to a very different setting):
In a very short paper from 1925, W.S. Abbott, an entomologist working at the US Department of Agriculture, proposed measuring the effect of insect sprays using what is now known as Abbott’s Formula. Abbott’s formula, which is still used by entomologists today, is mathematically equivalent to Sheps’ suggestion for the case where the intervention increases risk of the outcome. Abbott does not consider the situation where exposure reduces the risk of the outcome.
In 1939, another entomologist, C.I. Bliss from the Institute for Plant Protection in Leningrad, extended Abbott’s formula to the setting where exposure reduces incidence. To do so, he developed the Joint Independent Action model, which has become central to how toxicologists think about interaction between poisons. Toxicologists have made attempts to convince epidemiologists about the utility of this framework for interaction, which according to Howard and Webster has “firm biological foundations” in contrast with epidemiological models that consider interaction in terms of departures from risk additivity.
(I will note that while Bliss’ works appears to precede Sheps’ , Sheps was certainly the first to recognize the implications for medical statistics and epidemiology, and appears to have done so independently of the earlier work).
So if these models are good enough to be standard tools of the trade for toxicologists and entomologists, why are they still not being used by epidemiologists, medical statisticians or clinicians? Every time I find a new researcher who has rediscovered part of this idea, I get more and more puzzled over this.
It is not because people haven’t tried. In 1986, Clarice Weinberg, suggested using the Joint Independent Action model in epidemiological research, leading to recommendations that are identical to Sheps’. Weinberg is a leading epidemiologist publishing in a leading journal, and the journal did not publish any convincing counterarguments.
Yet Sheps recommendations are still not used in practice. There seems to be a deep resistance to these ideas, and it is unclear where it is coming from. Hopefully, our manuscript can contribute to clarifying the scope and limitations of this line of reasoning, and bring together closely related insights that have been scattered across the literature in fields as diverse as entomology, psychology, philosophy and computer science!
Tell us Anders, where did you find out about Abbott's paper and why did you obtain and read it? After all,at the Datamethods blog on July 4 you repeatedly dismissed my comments that mechanistic models and risk results go back to the 1920s, replying
"It therefore isn’t at all obvious to me that reading the old literature is generally the best use of time" and "I wouldn’t try to learn calculus from Newton or Leibniz’ original writings, and likewise, I wouldn’t try to learn causal inference from Robins (1986)",
to which I replied at length, ending with:
"For those with a historical bent, in a paper invited by an engineering toxicology journal (“Elementary models for biological interaction”, Journal of Hazardous Materials 1985;10:449-454) I attempted to provide a connection between bioassay and epidemiologic models for interaction." Among other things this 1985 paper cites W.S. Abbott, A method of computing the effectiveness of an insecticide, J. Econ. Entomol., 18 (1925) 265--267.
- See https://discourse.datamethods.org/t/should-one-derive-risk-difference-from-the-odds-ratio/4403/213 and the posts leading up to that one.
Also, you never responded to my deduction of risk additivity from the assumptions of no interaction response types and of no confounding of either single or joint factor effects:
https://discourse.datamethods.org/t/should-one-derive-risk-difference-from-the-odds-ratio/4403/221