Researchers say an overdiagnosis of autism has lead to a tripling in cases in US children.

The way autism is diagnosed in the United States has led to an apparent tripling in cases in recent years that does not reflect reality, researchers say.

Rather, more youths with intellectual or developmental disabilities are being reclassified as autistic, said the study in the American Journal of Medical Genetics, released on Wednesday.

The prevalence of autism in the United States was just one in 5000 in 1975.

It leapt to one in 150 in 2002, and reached one in 68 in 2012, according to the US Centers for Disease Control and Prevention.

"This new research provides the first direct evidence that much of the increase may be attributable merely to a reclassification of individuals with related neurological disorders rather than to an actual increase in the rate of new cases of autism," said the study in the American Journal of Medical Genetics.

Scientists at Penn State University analysed 11 years of special-education enrolment data on an average of 6.2 million children per year.

They found "no overall increase in the number of students enrolled in special education," said the study.

"They also found that the increase in students diagnosed with autism was offset by a nearly equal decrease in students diagnosed with other intellectual disabilities that often co-occur with autism."

Therefore, what may appear to be an epidemic of autism is more likely the result of shifting patterns of diagnosis over time.

In addition, autism is a complicated condition with many degrees of severity, and can overlap with other related disorders.

"The high rate of co-occurrence of other intellectual disabilities with autism, which leads to diagnostic reclassification, is likely due to shared genetic factors in many neurodevelopmental disorders," said lead researcher Santhosh Girirajan, assistant professor of biochemistry and molecular biology and of anthropology at Penn State.

"Every patient is different and must be treated as such. Standardised diagnostic measures incorporating detailed genetic analysis and periodic follow up should be taken into account in future studies of autism prevalence."