We don't? If a diabetic person crashes a car because his blood sugar is low and he is not coherent, it will likely be listed on the crash report and maybe the death certificate. If a person who is having a heart attack dies after the ambulance taking him to the hospital crashes, is his death because of the crash or the heart attack? There are differences in criminal matters. If a person shoots and kills someone with leukemia that wouldn't have died if they didn't have leukemia, the person who shot them is still held liable for the immediate cause of death. However, should the fact that someone shot them cause their leukemia to not be counted in statistics that show how extensive and dangerous leukemia is? There are differences in situations that make immediate cause of death important. I would say that in epidemiology and a pandemic, it isn't as important as in a murder trial.
The bigger question is what difference does it make. If the numbers are inflated by a factor of 3, or under reported by a factor of 3, then that would be a big deal and would matter to epidemiologists. The current numbers are about 14,800 in the US. What difference would it make if it is actually 14,400 or 15,200? I am sure that epidemiologists are going to use uncertainty in their models and predictions. That uncertainty along with many other uncertainties are the reason that the IHME model shows a range of 939 to 4,023 deaths for today. I am not an epidemiologist, but I am certain that they understand there are not 100% accurate ways to measure any of the numbers they use.
Again I ask, if the actual number of deaths in the use is 14,672 instead of 14,814 what would you be doing differently? What should the government be doing differently? Why does the difference between 14,672 and 14,818 matter?