Joint Over- and Underdiagnosis

Today I had several more terrible lectures on ADHD.

In one of them, I was informed that America is medicalizing normal childhood mischief and loading anyone who gets worse than a B+ up with Ritalin or amphetamines as part of the pathologization of everyday life.

In another, I was informed that ADHD is shamefully underdiagnosed and most of the children who need stimulants most are going without them and failing school unnecessarily, so we need better screening programs and more efforts to seek out potential sufferers of the condition.

So I asked one of my attendings, Dr. L, which one it was. Are we overdosing ADHD? Or underdiagnosing it?

He answered that we are both overdiagnosing and underdiagnosing ADHD, the same as every other psychiatric disease, and then explained this so it made perfect sense and I was embarassed for not realizing it before.

Suppose that 3% of the population has ADHD.

Suppose that of people with ADHD, 50% of them realize they have ADHD like symptoms and go to a psychiatrist to get checked out.

Suppose that of people without ADHD, 10% of them falsely believe they have ADHD and also go to a psychiatrist to get checked out.

The Conners Continuous Performance Test is a commonly used test that evaluates children for ADHD. It is found to have a sensitivity of 75% and a specificity of 73%. In theory our system is based on faith that a trained psychiatrist can do better than a neuropsychological test; in practice they probably do much worse. Let’s give them the benefit of the doubt and say this is an excellent psychiatrist who outperforms the test handily and has both a sensitivity and specificity of 85%.

We can see that of every 100 people, 3 will have ADHD and 97 won’t. 1.5 true patients and 9.7 false patients will show up for psychiatric evaluation. The psychiatrist will diagnose 1.275 true patients and 1.455 false patients with the condition, and prescribes stimulants according to the diagnosis.

So we have three things that, surprisingly, all happen at once:

1. We have an excellent psychiatrist who outperforms the tests and is right 85% of the time.
2. The majority of people who are on Ritalin, shouldn’t be.
3. The majority of people who should be on Ritalin, aren’t.

Number two sounds a lot like what we mean by “overdiagnosis”, and number three sounds a lot like what we mean by “underdiagnosis”. So even with a pretty good psychiatrist acting honestly, we expect ADHD to be both overdiagnosed and underdiagnosed at the same time.

Even in conditions that do not quite satisfy the “majority” part of (2) and (3), we might still expect it to be true at the same time that a sizeable chunk of people diagnosed with the disease don’t have it and a sizeable chunk of people with the disease aren’t diagnosed.

If this seems counterintuitive, it is just another example of the annoying world of medical sensitivity and specificity statistics, which are constantly tripping up even the most experienced doctors. See also the infamous mammogram problem.

Once I understood this joint-overdiagnosis-and-underdiagnosis problem, several other candidate situations immediately leapt to mind. Antidepressants are almost certainly both overprescribed and underprescribed. So are opiate pain medications.