The point is, I think that is has to be understood "race" in this context is used as proxy for a set of (often unknown) features that are correlated with a condition. On top of it, the effect has to be known. I.e. if e.g. a certain group has a 5% increase in some rare condition, it is likely a meaningless proxy. That being said, in the medical community folks are not afraid of doing so. Rather the data is often (not always) problematic as the association with race categories are not strong enough to make useful prediction just based on that (for the already stated reasons). For example certain drugs are generally not proscribed to black Americans (or only), as they seem to be less efficient (which shines light on the fact that most medical data we have are in fact from white males, for a variety of reasons). However, some newer researcher puts the validity of some of those assessments in question.
Even worse, there is evidence that race differences in diagnosis and treatments have been, in fact, misused. For example, there was the (unfounded) notion that African Americans are more prone to drug abuse, and were prescribed less pain medication than their white counterparts for the same indications. Perhaps unknowingly that was not a bad thing, but effectively it resulted in race-based differences that were ultimately not based on medical or biological indications.Of course, there are also other disparities that fall across racial lines, such as lower rates of referral, even when corrected for social-economic status, which indicates bias in medical staff. This has far-reaching consequences for epidemiological studies.
For example, if we just look at the raw numbers, it appears that African American are more susceptible to a range of coronary artery disease (CAD). It has been suggested that allels that are more prevalent in African Americans and which might be associated with salt-related hypertension could be the culprit.
So, intuitively one might think that they should be informed about that and more measures have to be take to address that. On the negative side, this could impact other aspects, such as health and life insurance cost, which could be a problem. Still, one might think that is justified. However, digging deeper into data we encounter a problem. African Americans are less likely to receive proper care and treatment in CAD, including major procedures such as revascularization procedures, or even common drugs, such as beta blockers or blood thinners. Even when we look among African Americans, it appears that certain social factors play a role. Strangely, hypertension in African Americans was highest for those who were actually of higher socioeconomic status, and highest for those who had overall high achievement (i.e. were climbing the social ladder). So perhaps it is not just in the genes or not just being black. Rather, the observed outcome is the result of complex biological as well as social issues, which require very different approaches to address properly. If one simplifies the model too much without understanding the mechanisms, one risks in doing more harm than good by e.g. assigning likelihoods of disease to certain groups . It can result in over- or false treatment, added stress and impede future research.
In that context it makes more sense to me to one has to have a clear assessment of the usefulness of a given diagnosis. I think too many folks are thinking about that in terms of being offended or that is is based on PC. But those working in the area realize that the situation is complex and by given simplifications to the public, one may unintentionally do harm. It is the same with the whole overarching discussion. Is race as often defined a social construct? Certainly. But it does not mean that there are no difference that can be used in certain contexts. The issue is that most folks will get it wrong (as most won't study the subject in sufficient depth) with can result in quite harmful practices (and even laws). It should be noted that even among specialists simple errors such as bad statistical practices or simply not being aware of confounding factors it has happened quite frequently that effects are overestimated. It would be quite problematic if then physicians (who are for the most part woefully undertrained in the interpretation of epidemiologial data would try to apply initial findings without knowing how firm the literature is on a given association.