Flawed analysis of Florida Medicaid data used to promote false claim that vaccines cause autism

In January 2025, a claim that vaccines cause autism circulated online, using a study of children enrolled in Medicaid as supporting evidence for the claim. The claim was propagated through a Substack article by entrepreneur Steve Kirsch, which he also shared on X. His tweet sharing his article received more than 12,000 likes and more than 5,000 retweets. Kirsch emerged as a vocal opponent of vaccination during the COVID-19 pandemic; Science Feedback debunked several of his vaccine-related claims in previous reviews.

In his tweet, Kirsch claimed that “RFK and Andy Wakefield were RIGHT” and that “vaccines cause autism”. This is in spite of the fact that Wakefield’s study, which alleged a link between the MMR vaccine and autism, was eventually shown to be fraudulent.

Kirsch’s claim is based on an analysis that appeared in the publication Science, Public Health Policy and the Law. Ostensibly a scientific journal, it was created by James Lyons-Weiler, a scientist who previously spread vaccine and COVID-19 misinformation. It also republished a retracted study that baselessly alleged the COVID-19 vaccines killed hundreds of thousands of people.

The publication’s editorial board includes several figures who spread health misinformation, including Peter McCullough, Brian Hooker, and Russell Blaylock.

The first author of the analysis is Anthony Mawson, who has published flawed studies claiming that unvaccinated children are healthier than vaccinated children. His most recent analysis also made the same claim. But the analysis, like his earlier studies, contains significant flaws that render its conclusion meaningless. We explain below.

What did the study do and what did it find?

The study claimed: “Although vaccination is said to be safe and effective for the great majority of children, a legitimate question is whether the expanded schedule could be contributing in some way to rising rates of NDDs [neurodevelopmental disorders]”.

This premise falsely implies that whether vaccines cause autism remains an unresolved scientific question. Previous Science Feedback reviews shared evidence from numerous large-scale studies investigating a potential association between childhood vaccination and autism. None of these studies found an association between the two. The absence of this association provides strong evidence against the hypothesis that childhood vaccination is causally linked to autism.

Nevertheless, in its quest to address this “legitimate question”, the analysis drew on a database of Florida Medicaid records for children born between 1 January 1999 and 31 December 2002. Data collected spanned the time of a child’s birth up until they were nine years old.

The analysis determined the vaccination status of the children using Medicaid billing codes. Children that fell within the vaccinated group were defined as those whose records contained at least one vaccination-related billing code. Conversely, children whose medical records didn’t contain a vaccination-related billing code fell within the unvaccinated group.

To identify children with neurodevelopmental disorders (NDD), such as autism spectrum disorder, the analysis also used billing codes related to the diagnosis of NDDs. However, the analysis counted diagnoses of NDDs only if the diagnosis took place between the ages of five and nine.

The authors then sought to determine if there was an association between vaccination status and an NDD diagnosis based on this data. And they did find that children in the vaccinated group were more likely to be diagnosed with an NDD. They also found that the greater the number of medical visits with at least one billing code related to vaccination, the greater the likelihood of an NDD diagnosis.

Finding these associations led the authors to conclude that their analysis raised “concerns about the safety of the current vaccination schedule and its possible contribution to rising rates of neurodevelopmental disorders”.

Multiple red flags call into question the reliability of the analysis

Bertha Hidalgo, an epidemiologist and associate professor at the University of Alabama at Birmingham, spoke to the Unbiased Science Podcast about several issues with the analysis. This review sums up a few of the key issues highlighted by Hidalgo; readers who wish for a detailed and comprehensive breakdown can read the podcast’s article on Substack.

Firstly, the analysis didn’t account for important confounding factors, one of which is healthcare utilization bias:

“Children who receive vaccinations typically have more frequent medical visits, naturally increasing their likelihood of receiving developmental screenings and subsequent diagnoses. Conversely, unvaccinated children often have less healthcare engagement, potentially leading to underdiagnosis.”

The same flaw was also highlighted by pediatric cardiologist Frank Han, who noted that NDDs often require multiple clinic visits to be diagnosed correctly. Experts note that one of the difficulties in diagnosing ASD stems from the heterogeneity of the condition, that is, diversity in the way the condition manifests between individuals and the severity of the symptoms^[1]. Another obstacle to diagnosis is that symptoms of ASD can overlap with symptoms of other conditions.

This means that children who get vaccinated are more likely to go to the doctor. At the same time, children who go to the doctor more often are also more likely to be diagnosed with an NDD, if one is present. Thus the analysis’ finding that vaccinated children are more likely to have an NDD is potentially due to the fact that the two groups—vaccinated children and children with NDDs—have in common a greater frequency of healthcare utilization, rather than because vaccines are causally linked to NDDs.

Secondly, the analysis’ reliance solely on Medicaid billing codes for determining the vaccination status of a child introduces additional room for error. Medicaid is a program that provides medical and health-related services to low-income groups in the U.S. Children can receive vaccinations through Medicaid, but they can also be vaccinated through other programs outside of Medicaid, such as the Vaccines for Children (VFC) program. Per the U.S. Centers for Disease Control and Prevention:

“The goal of the VFC Program is to ensure that a VFC Program-eligible child does not contract a vaccine-preventable disease because of their parent or guardian’s inability to pay for the vaccine.”

However, vaccinations outside of the Medicaid program aren’t billed to Medicaid. This means that children who were vaccinated through another program like VFC could very well be incorrectly categorized as unvaccinated in the analysis, since their records wouldn’t contain a vaccination-related Medicaid billing code.

The analysis’ Discussion section indicates that the authors were aware of this drawback. They also noted that “the rate of unvaccinated children in this sample is high compared to other groups and national averages”. One explanation for this observation is that many children who were classified as unvaccinated in the analysis were actually vaccinated. However, the authors didn’t mention additional steps they took to mitigate the potential bias posed by incorrect classification, such as verifying vaccination status through more comprehensive medical records.

Thirdly, the authors’ decision to only include NDD diagnoses between the ages of five and nine years would lead the authors to miss early diagnoses, as “[a]utism is typically diagnosed between ages 2-4”, Hidalgo said. This raises the question of whether including early diagnoses may have changed the associations that the authors observed between vaccination status and NDD diagnosis.

Overall, the analysis contains several methodological weaknesses that call into question the reliability of its findings.

Finally, the analysis was funded by the National Vaccine Information Center (NVIC) and its publication costs were also partially funded by NVIC and the Institute for Pure and Applied Knowledge (known as IPAK). Both groups are known for their opposition to vaccines and IPAK was founded by Lyons-Weiler, the same individual who created the publication that Mawson’s analysis was published in. As such, the analysis may involve a potential conflict of interest that readers should be aware of.

Conclusion

There have been several analyses claiming to find that childhood vaccines are causally associated with neurodevelopmental disorders and poorer health. But these haven’t stood up to scrutiny, as a closer examination of these analyses turn up multiple methodological flaws. As this review shows, the latest analysis by Mawson is no different.

The claim that childhood vaccines cause ASD was first propagated thanks to a study that eventually proved to be fraudulent. Numerous epidemiological studies that looked at large cohorts found no association between childhood vaccination and ASD, thereby refuting claims of a causal link between the two.

REFERENCES

• 1 – Zwaigenbaum et al. (2015) Early Identification of Autism Spectrum Disorder: Recommendations for Practice and Research. Pediatrics.