What COVID Can Teach Us About Autism Epidemiology

Success with COVID-19 and the rates of new cases

Alexander MacInnis

Daily new cases of COVID-19

Source: Alexander MacInnis

The COVID-19 pandemic was a worldwide disaster, and it’s not over yet. But, it’s also an incredible success story. In just a matter of weeks, doctors and scientists first observed a disease with a novel set of symptoms, identified and sequenced the virus that causes it, developed high-quality tests to determine who has the disease, and designed multiple safe and effective vaccines to prevent it. One basic yet essential part of the success was focusing on the rate of occurrence of new cases, called incidence in epidemiology. The incidence rate is fundamentally useful for exploring causes. It also indicates the level of risk from a disease or disorder and the urgency of addressing it. Today, daily news stories and science reports on COVID continue to highlight the daily rates of new cases.

Try to imagine what would have happened if those same doctors and scientists had ignored the incidence and instead focused on the counts of currently active cases, which is prevalence. That would have made it much harder to understand the cause of COVID, its spread, and the effectiveness of preventive measures.

The terms incidence and prevalence are frequently confused and misused in both the popular press and scientific papers. You can think of them like this: prevalence is like the amount of water in a pond, while incidence is like the rate that water flows into the pond.

Slow progress in understanding autism

Contrast the experience of COVID-19 with that of autism. In 1943, Dr. Kanner reported on eleven children with novel, severe symptoms, and he gave the disorder a name. Today, nearly 70 years later, we have made insufficient progress in understanding autism’s causes and the biological mechanisms behind the symptoms. Doctors don’t have approved medical treatments to treat autism itself, and they don’t have methods of prevention to recommend. There are roughly 6,000 scientific journal papers mentioning autism published every year. Most describe various aspects of autism, and few of them address treatments or potentially preventable causes. The progress towards treatment and prevention is still very slow.

Valid reasons explain some of this slow progress. Autism has substantial differences from COVID-19 that make it harder to identify the causes. Since it is defined in terms of behavioral symptoms, we don’t know how many different biological mechanisms might be causing the symptoms in different people. It does not spread like an infectious disease, and apparently, it is not caused by an infectious agent. Most likely, each case is caused by a combination of multiple factors, and that combination may differ between individuals. While we should expect progress to be slower than with COVID, we should have been able to do much better over nearly 70 years. Unfortunately, there is a continuing lack of urgency to figure out the causes and find treatments.

Focus on the rate of occurrence of new cases: birth year prevalence

One important lesson from COVID-19 is to focus on the rate that new cases occur, which is fundamental to epidemiology. Unfortunately, most autism epidemiology papers describe prevalence estimates. Some use the word “incidence,” but from context, they mean prevalence. Some papers use “incidence” to mean the rate of diagnoses rather than new cases. If autism epidemiology focused on the incidence of the disorder or an equivalent, we could make quicker progress in understanding autism’s causes.

Autism is generally considered a disorder of early childhood, whose existence is either present or predetermined at birth or perhaps shortly after that. We generally cannot observe the time of occurrence for such disorders and diseases, which may be prenatal. That is particularly true when symptoms develop over time and diagnosis comes even later. For early childhood disorders where we cannot observe the occurrence of the disorder, epidemiology uses something called birth year cohort prevalence to serve the function of incidence. For each birth year, it’s the proportion of the people born that year who have or develop autism. We can use abbreviations such as birth year prevalence to mean the same thing. If autism epidemiology focused on birth year prevalence rather than generic prevalence, which is quite different, we could make much quicker progress. Also, using birth year prevalence, we could predict with confidence how many older people will have autism in the future, based only on people who are already born.

In principle, birth year prevalence estimates are simple: for each birth year, count the number of cases and divide by the population born that year. Some high-quality studies, including the biannual CDC reports, already produce birth year prevalence estimates, but they don’t call them that, which is confusing. Producing accurate estimates is more complex than the simple description above; a future post will address that.

Alternative hypotheses and not trusting the data

The greatest impediment to producing solid, widely accepted birth year prevalence estimates is that many people don’t trust the data. That objection is based on two hypotheses. One idea is that we undercounted cases in previous years, and the observed increase is due to “better diagnosis.” Specifically, an increase in factors such as awareness might explain much of the increase. Another hypothesis is that broadening diagnostic criteria over time has greatly increased the number of people who qualify as having autism, explaining the increase. When using generic prevalence, it is very hard to determine whether either hypothesis is true. However, when studying birth year prevalence, there is much less ambiguity. It is still a challenge to evaluate these hypotheses correctly, but it is possible.

It is one thing to consider hypotheses, but the literature indicates these two hypotheses are widely believed to be true, sometimes to the point of being stated as facts. However, I have not found any scientific papers that support either hypothesis with proper analysis. Nevertheless, it is appropriate to give these hypotheses the full benefit of the doubt. Later installments in this series will explore all this and more.