To understand autism, you need to understand incidence and prevalence

In 2012, the CDC put out a prevalence study of autism in the United States. It changed the prevalence number from “1 in 110” to “1 in 88”. There were many who were alarmed at this statistic. They thought that the chances of a child being born with autism increased from 1 in 110 to 1 in 88. Well, they didn’t. This was only a rise in prevalence, not a rise in incidence. While the two are related, they are not necessarily tied to each other. That is, if one rises, the other one doesn’t have to. One can rise and the other can fall. Why?

It’s a little complicated, but I’ll try to explain it.

Let’s look at the definition of incidence. Incidence is the number of new cases in a population, per unit of time (usually a year), divided by the number of people at risk in that population. So, if your population is 100,000 and 100 people get the disease in one year, then your incidence is 0.001 or 0.1%. But what if we’re talking about cervical cancer? In a normal population of 100,000, only half of the people in that population — the women — would get cervical cancer. Men don’t have the right equipment for that. In that case, 100 cases in an at risk population of 50,000 is an incidence of 0.002 or 0.2%.

If you fully recover from the disease, then you move over to the “at risk” population again. If you don’t — because it stays with you forever or because it kills you — then you stay out of the at risk population. You could have 100 cases each year, no more and no less, and the incidence would continue to rise if no one recovers or your at risk population is not replenished by new births fast enough. In the example I just gave you, the population at risk for year two is 49,900. If you get another 100 cases, then your incidence is 0.00200401 or 0.2004%. It’s a small increase, but it’s an increase nonetheless.

So, remember this: If the disease is not curable (because it is chronic, pervasive, incurable, or deadly), then the population at risk dwindles if it is not replenished by births or immigration. Lower the denominator in incidence, and you will get a higher number. To decrease incidence, you either increase the number at risk or you decrease the number of new cases.

Now, let’s move on to prevalence. Prevalence is the number of existing cases in a population, per unit of time, divided by the total number of people in that population. That’s total population, regardless of whether or not they have the disease. So, if you have 100 cases of cervical cancer on year one, your prevalence will be 100 divided by 50,000, which is 0.002 or 0.2%. Year two, you get another 100 cases, and you will now have 200 existing cases divided by the same population of 50,000, which is 0.004 or 0.4%. Your prevalence doubled!

This assumes, of course, that no one died of the disease or that the total population stayed static through some means. In real life, population levels change.

In year three of the above scenario, you get another 100 cases, making it 300 existing cases in a population of 50,000, for an overall prevalence of 0.006 or 0.6%.

So, remember this: If a disease is not curable, then the prevalence will increase as long as there are new cases. Prevalence will decrease if the increase in population outpaces the new number of cases or the number of existing case decreases because of death or recovery.

Now, onto autism.

As far as science and medicine can tell us, autism is not curable. It is treatable. With the right interventions and depending on the level of severity of the autism signs and symptoms, autism is treatable. Plenty of people with autism go on to live happy and fulfilling lives. Again, it is not curable. Not at this time. So any new cases of autism will pile-on to existing cases and… Prevalence will increase.

Not only that, but the number of new cases per year can go down, but there will still be all those previously-diagnosed cases of autism which are still being added on to even if the incidence falls. Incidence would have to reach zero, the number of new births would have to continue (some countries have a negative birth rate), and people with autism would have to start passing away before the prevalence of the condition decreases.

Here are some theoretical numbers, as an example:

Note that there was a successful intervention in this example.

The column headings are self-explanatory, but let’s just go over them again for clarity.

  • New cases – Number of newly diagnosed cases that year.
  • Existing cases – The number of new cases for the year plus the number of existing cases the previous years. (Let’s pretend that there were no existing cases in 1999.)
  • Incidence – The number of new cases for the year, divided by the population at risk.
  • Prevalence – The number of existing cases (new cases plus existing cases) for the year, divided by the total population.
  • Population – The total population.
  • Population at risk – The total population minus the number of new and existing cases.

As you can see, we had a steady increase in the number of new cases from 2000 to 2009. From 2009 to 2015, the number of new cases declined. Appropriately, the number of existing cases continued to increase throughout because the condition is not deadly. (Again, this is theoretical. People with autism die from other causes, like the rest of us.) As you can see, incidence climbed along with the number of new cases until 2009/2010, then it began it’s decline. On the other hand, prevalence started its increase in 2000 and continued increasing to 2015. Also note that I increased the population every five years or so in our theoretical place (city, county, state) because that’s what populations in the United States have been doing. We don’t have a negative birth rate.

So, as cases dropped and population increased, incidence dropped. Because cases didn’t die, and the new number of cases outpaced the population increase, prevalence continued to increase.

On a side note, one of the criticism someone mentioned about HIV/AIDS treatment is that prevalence continues to increase. In their mind, HIV/AIDS treatment is not working if that particular measure of disease continued to increase. Can you see now why they were wrong?

Can you see why the prevalence of autism increasing from 1 in 110 (0.9%) to 1 in 88 (1.1%) is not a clear indicator that the number of new cases each year is rising? It’s only an indicator that people with autism are living and that the number of existing cases each year is outpacing population growth. For prevalence to decline, you would have to drop the new number of cases per year to zero and wait for the number of existing cases to drop on their own as people with autism get old and die.

That right there is what puzzles me when certain groups say they want the prevalence of autism to plummet. And their ranting and raving about an increase in autism signaling an “epidemic” of autism is also puzzling. Hopefully, it will be puzzling to you as well, now that you have seen how incidence and prevalence work.

Of course, this assumes that all things are equal when it comes to autism surveillance. But they are not. But that’s for another blog post at a later time.

By the way, here is the graph of the information in the table above, for those of you who are more visual:

Even with a theoretical, successful intervention in 2009, prevalence continued to increase. Why?


2 thoughts on “To understand autism, you need to understand incidence and prevalence

  1. I have it on good authority that he has embarrassed himself and his professors on more than one occasion. And doesn't include his harassment of pro-vax folks, nor does it include his lack of understanding of libel versus slander.

  2. I would love to see this question posed to upcoming MPH graduate, Jake Crosby at his thesis defence. He seems to have a wee problem with this concept.

Comments are closed.