[rflrob]

While I agree that you want to be wary of any particular claims, and more study is warranted to establish the effectiveness of any particular intervention based on data, there’s a Bayesian justification to looking at otherwise unexplained deaths. “When you have eliminated the impossible, all that remains, no matter how improbable, must be the truth.”

[phyzome]

And what about when "all that remains" is 47 different possibilities? That's not a very satisfying truth.

[ziddoap]

You have to start the process of elimination somewhere. Sometimes that process involves multiple steps.