You are out for a drive, cruising along and enjoying the sights, when you notice a flashing blue light in your rear view mirror – a police cruiser is on your rear bumper. You pull over to the shoulder of the road and roll down your window. The police officer steps up to your car, pokes his head inside, glances at your dashboard, then declares “I see that you have a full tank of gas. I’m giving you a ticket for speeding!” This story is preposterous, of course. Not even the greenest rookie would issue a speeding ticket based on your gas gauge. And yet, lots of folks go around making the same error when they quote prevalence data to support their claim we’re in an epidemic of autism.
Prevalence = the percent of the population that is affected by a condition. Incidence = the rate of occurrence of new cases. An epidemic is defined as a spike in the rate of occurrence of new cases (i.e., an increase in incidence). Incidence and prevalence measure different things, just like your gas gauge and your speedometer: You can be creeping along in the slow lane with a full tank of gas, or speeding along in the passing lane with your gas gauge on empty. Likewise, incidence can go up or down independently of prevalence. Up until the 1950s, there were waves of polio epidemics every few years. Then polio vaccine was introduced. Once mass immunization took hold, the rate of occurrence of new cases of polio (incidence) dropped dramatically. But the number of people in the population affected by polio (prevalence) remained unchanged – at least initially. Over the past 50 years, the people who acquired polio prior to vaccines slowly have died off, and the generations of infants who were immunized early in life have taken their place, so the prevalence of polio in the general population has gotten diluted down. But that took a long time, compared to the nearly instantaneous decline in incidence. At the other extreme, an outbreak of a highly contagious, rapidly fatal disease (like Ebola) can cause a spike in incidence, but – because people who are infected die almost immediately – little or no increase in prevalence. And of course there are all sorts of other scenarios in between these extremes. But the bottom line is that you cannot substitute prevalence for incidence – any more than our hypothetical police officer can ticket you for a full tank of gas.
It is certainly possible that there has been a change in incidence of ASD. But we can’t prove it by citing the prevalence data, any more than our hapless policeman can prove you were speeding by checking your gas gauge, and for the same reason: It won’t stand up in court.
Unless you have taken a course in medical statistics or epidemiology, you can be forgiven for not knowing the difference between incidence and prevalence - up until now. And yet, lots of people who should know better persist in fudging the difference – either out of willful ignorance, or because it suits their purposes (remember Foxy Loxy?) An epidemic – just like a misplaced fear that the sky is falling – brings in customers (for questionable therapies) and donors (often, for legitimate charitable or research organizations). But treatment and research are important in and of themselves, without hyping a questionable "explosion."
And to top it off: We can’t even be sure that the prevalence of ASD has really changed over the past 50 years. Yes, I know: the number of children and adults who have the diagnosis (the so-called "administrative prevalence") has increased, but most of those people have been there all along, and up to now have simply been overlooked. The lion’s share of this increase is most likely attributable to changes in the definition of ASD, and changes in case-finding and reporting methods. From the DSM-III through the DSM-IVR (the Diagnostic and Statistical Manual of the American Psychiatric Association, the diagnostic standard in the US), the criteria for ASD got broader and broader, and as a result, more children (and adults) fit within the definition of a "case." Imagine that you are doing a survey to determine the prevalence of “tall stature” among the readers of Psychology Today. If you define “tall stature” as anything over 7 feet, you’re going to get an extremely low prevalence. Let’s say, one in 10,000. Then we decide to revise the definition of “tall” to anything over 6’ 10.” Now the prevalence is 1 in 1000. Change the criterion to 6’6”, and now the prevalence is 1 in 500. A twenty-fold increase in prevalence! But nobody is one inch taller than before, and the rate at which tall people are being born hasn’t changed. All we did was change the criteria for a “case.” And the same thing has happened with the DSM definitions of ASD over the past 40 years. (In my darker moments, I sometimes wonder if one motivation behind the more restrictive criteria in DSM5 was to make the numbers go down again. The authors of DSM5 resolutely insist that an adult must be "severely impaired" to be eligible for an ASD diagnosis - the subject of a future blog.) Regardless of the authors’ motivations for the successive expansions and the most recent contraction of the definition in DSM, we must not confuse changes in administrative prevalence (driven by changes in diagnostic criteria) for a change in actual prevalence – or incidence.
Another fact contributing to the "explosion" of cases has been change in Federal education law. Prior to 1990, Federal regulations did not even recognize ASD as a reimbursable disability until 1990. Prior to that date, children enrolled in special ed with ASD were usually labelled “Emotionally Disturbed,” “Mentally Retarded,” “Other Health Impaired,” or something else – but not autism. When the law changed in 1990, not only were newly diagnosed children correctly labeled for the first time; a huge backlog of mis-labelled children were recategorized. (In fact, the rise in reported number of students with ASD has been mirrored by an equal decrease in the number of students with a label of Intellectual Disability - strongly suggesting that a big portion of the boom in cases of ASD is due to reclassification. This graph says it all.)
Interestingly, a door to door study in England, using modern criteria for ASD, found that the prevalence among today’s senior citizens is about the same as among today’s school children. In other words, they were there 50 or 60 years ago, but not recognized as having ASD.
Here’s a link to one of my presentations that covers the foregoing issues in detail. Still more information is available here.
This still leaves the big questions unanswered: How autistic is “autistic enough” to warrant a diagnosis of ASD? Is ASD a disorder with sharp boundaries, or a fuzzy cloud of behaviors, widely distributed across the entire population – perhaps not even a “disorder” at all? More on these questions the next time. In the meanwhile, if you want to read ahead a little bit, go here