June 13, 2019

What Causes AIDS Deaths?

By Austin Frakt

Here’s a puzzle: To what would you attribute deaths from AIDS today?

  • Genetics?
  • Behavior?
  • Social or environmental factors?
  • The health system (or its failure)?
  • Something else?

Once you have your answer, how would you know it was right? How would you test it? What evidence would you need? What studies would you do?

For example, suppose you thought AIDS deaths are predominantly due to behavior — unprotected sex and sharing of needles are the cause, you think. How would you prove that?

It’s not enough to show that people who don’t engage in those behaviors don’t die of AIDS. Lots of people who do engage in those behaviors also don’t die of AIDS. And why is that?

One reason is that we have life-sustaining treatments for AIDS today. This suggests that the health system has something to do with AIDS deaths. So maybe it’s not behavior, or not entirely.

But, what if someone with AIDS lacks good access to the health system? Maybe they’re uninsured for financial reasons. Maybe they have low health literacy and have difficulty navigating the complex US health care system? Maybe they can’t get timely care or have inadequate transportation to receive it? And, what if as a result, they do not receive lifesaving care?

These all fall into the “social or environmental factors” category. So maybe it’s not entirely behavior and it’s also not entirely the health system.

It’s hard to avoid the conclusion that it’s impossible to assign AIDS deaths to one cause.

This brings us to the temptation to assign percentages to the various factors. So you could assign 40% of the responsibility for the death to behavior, 30% to social or environmental factors, and 30% to the health system, as an example.

How would you test this? How would you know what the “right” proportional attributions are?

That’s a hard question. I don’t believe it has a complete and convincing answer. I don’t think we know how to attribute AIDS deaths across these categories in an unassailable way.

One thing we can do, however, is to test interventions that act on these different dimensions. For example, we could create programs to provide better access to health insurance or health care and evaluate them and their impact on AIDS deaths.

We could create programs designed to reduce risky behaviors and evaluate those.

We could do this for lots of programs that nudge people and systems in all manner of ways and look for changes in AIDS deaths. If well designed, we could get some convincing answers and build up an inventory of interventions’ effects on AIDS deaths.

This would be hard work, and there would be lots of limitations. For instance, would a program that reduces AIDS deaths by, say, 10% in one city do the same in another? How generalizable are the findings?

Nevertheless, we would have some numbers and could compare the relative value of intervention A vs. B for AIDS deaths (again if the interventions were well designed for evaluation). Though it’s exceedingly hard, perhaps impossible, to attribute AIDS deaths to causes, it’s much easier (though still hard) to see how specific interventions affect them.