The Skeptic

I usually get to lab around 7-something in the morning, which is earlier than almost everyone else. Truth be told, I've been trying to work in a 630am arrival schedule but my sleeping times have been much less resilient. In any case, I get some reading done in the morning - mostly stuff on my Google reader feeds.

This morning, I found this entry - Desiree Jennings on 20/20 - and had an epiphany:  (you can see a summary of my epiphany in a comment at the bottom). Most, if not almost all of the problems associated with the perception of science by the general public can be attributed to one word - proof.

With all the stories I'm hearing about people changing biology curriculuae, the anti-vaccination movement, I would be very happy if more people were scientifically literate, which is different from really knowing science in itself. I don't expect everyone to become a scientist, or major in a hard science; it's up to you how much you like flipping burgers, but I do expect people to be scientifically literate, or in the least, be amenable to be taught it.

I won't talk about scientific literacy now.  However, I can give my understand of proof from a medical perspective. That too, I'm going to talk about a very specific (but incredibly important) aspect of proof - causation.

MEDICAL CAUSATION, as I will call it. Actually, it's a summary of Hill's Criteria of Causation, a very eloquently put together set of tenets which I think are pretty easy to understand (for the non-scientific audience). It's important to note that there are rigorous numerical methods behind each one of these, except maybe #4. Let's begin.

Say you want to comment on ... vaccines (or the mercury they have). How would you know what effects they have? How would you prove that vaccines actually caused this effect? How would you prove that a vaccine causes autism? I'm not going to go further with this question; I don't have the data nor do I wish to really turn anyone away. Let's talk about apples, and how apples may keep the doctor away.

How to prove one thing causes another: 
The Tenets

With a short commentary on each.

1. Temporal relationship - The cause must come before the effect. You must eat the apples before your number of doctor visits start to decrease. Simple enough, right?


2. Dose-response gradient - The more of the 'cause' you have the more of the 'effect' you have. The more apples you eat, the more doctor visits are avoided. Still pretty simple, so far.

3. Consistent association across studies - In similar and/or different situations, the cause still produces the same effect. The more different situations you have, the stronger this tenet becomes. E.g. Your cousin finds that when he eats apples, he avoids doctor visits. Your aunt/uncle/neighbor/sister/brother/sick lady across the street/cancer patients/people in third world countries all find this same effect. Thus the association is consistent across studies (different population groups). 

4. Face validity - The association makes sense, applying common sense. E.g. Eating apples makes you healthier, so somehow you avoid doctor visits as you get sick less often? If this was something like, every morning you wake up and snap your fingers twice and that decreases your # of doctor visits, well, that doesn't make too much sense. It lacks face validity. Apples --> better health --> less doctor visits? Might make sense, thus, has face validity. 

5. Strength of association - This one is the hardest one to throw out there without statistics. Basically, how strong is it? Probably best to look at the example for this one. E.g. One apple avoids 20 doctor visits vs 100 apples avoid one doctor visit. The former is a much stronger association than the latter. Let me take a stab at explaining it. In the latter, the fact that you're eating more apples probably means you're eating healthier in general, which would certainly decrease the # of doctor visits. In the former, the implication is strong that there might be something in the apple itself that does the trick, not any personal habits you may have picked up.


There are more tenets to Hill's criteria, but I feel like if you understand these 5, you're well on your way to understanding medical causation. Particularly important are 2, 3, and 5. #1 is easy to understand and easy to refute if it doesn't exist, #4 is murky and dependent on how far research has advanced. 2, 3, and 5 however, will hold true regardless of how much research has advanced.

More on this, and proving things, and scientific literacy later.