Similarity and Probability
This week is school vacation week (because the kids are so exhausted from all the snow days they’ve had this month) and, as usual, many of my 6-year-old’s friends are away. Specifically, her best friend is in Florida. Since her best friend is the same age as my daughter and in the same grade and they like the same things, my daughter has correctly deduced that she and her friend are similar. And if they are similar, they should do similar things. Thus the question: “Why didn’t we go to Florida?”
Because we’re idiots, that’s why. No, I didn’t say that. I told her some version of We Can’t Afford It Whats Wrong With Snow Go Ask Your Father. It did occur to me, later, that the thought process behind the question, “Why didn’t we go to Florida?” is the same one patients use all the time to assess risk.
Assessments of similarity are natural for humans, being social creatures who like to find others like themselves. We do this automatically on issues of skin color, accent, socioeconomic status, educational level, etc. When something bad happens to someone, the people around that victim will tend to look at themselves and find ways to reassure themselves that they are not similar to the victim. Or, if asked to determine the risk that the same event will happen to them, they will inflate the risk if the number of similarities between themselves and the victim is large. For example, if I have an acquaintance who got breast cancer at age 38, and I’m 38, it might cause me to worry about my own risk of breast cancer. I might worry more if she was like me in more ways, like she was also a doctor or was also from California or had the same number of kids as me. It might not make much difference to me if some of my information about her were second-hand or gossip, or that being from California is not associated with breast cancer.
Another example might be Autism. A parent could see a child with Autism who has been vaccinated and compare that child with their own. The more the autistic child is similar, or appears to be similar, to the parent’s own child, the more worried about autism that parent might be. Having assessed that the two children are so similar, the parent decides he’d better make them a little less similar, so he’ll decide not to vaccinate his child. He decides not to vaccinate in part because other parents who share similarities with him have chosen not to. The fact that vaccinations and Autism are not linked or that the information he has about the autistic child might be false don’t impact his decision as much as you’d think.
The problem is that the trope “It could happen to anyone” is not really true, because judgements of similarity and risk don’t have the same rules.
Assessment of risk, or probability, must include information about average population rates and rates of the event in question in different demographic groups. Assessing probability also requires that the information you have is as accurate as possible. Assessment of similarity has no such restrictions. You can decide someone is similar to you regardless of how common the traits between you are in the population or how reliable your source of information is.
Nevertheless, patients equate similarity and probability all the time. This is why choices that patients make can seem irrational or crazy, and patients can get labeled as “worriers” or “malingerers”. Same thing with parents. Same thing with any question in which we are asked to make an impossible prediction. Comparing ourselves to others is natural and automatic, and forms the basis for more decisions than we would like to acknowledge. Yet another example of how medicine is not an exact science, because humans are involved.