Has anyone seen "channeling bias" in the context of epidemiological research? I'm working on a manuscript describing analysis of a workplace-based, non-medical occupational health intervention using observational survey data. The intervention is treated as an exposure and is not randomly allocated. I recently discovered the pharmacological research term "channeling bias", where individuals most in need of treatment are more likely to receive it, usually in a trial setting. This could result in the treatment's observed effects being smaller than they actually are, since sicker individuals have a lower baseline. Participants may also be channeled if researchers think they will receive greater benefit, and thus inflate observed effects. With a workplace-based intervention, I suspect employers may be frugal and, the healthier the worker is, the less likely they are to provide the intervention (despite the intervention being mandatory), thus biasing the exposure group to be in worse shape than those without the exposure.
I think the term channeling bias would apply well to epidemiological research, but have not seen it used in this context. Are there any reasons why it would be inappropriate? Note, I only discovered the term when reviewing a doctoral student's thesis in pharmacology, in which the student used "healthy worker bias" to describe this phenomenon, which sounded the opposite of what they meant, and looked into what the term would actually be. Could its use only/primarily in pharmacological research just be the result of siloing across disciplines?