Montreal, Canada – A recent study conducted in Quebec, Canada, suggests that the rising rates of autism may be attributed to over diagnosis and the need for reevaluation of diagnostic criteria. Researchers utilized an advanced AI algorithm to analyze over 4,000 clinical reports of children being assessed for autism, focusing on the criteria commonly used for diagnosis as outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5).
The DSM-5 criteria for autism encompass a range of behaviors, including avoiding eye contact, displaying highly limited interests, engaging in repetitive movements, and experiencing challenges in forming friendships and engaging in conversations. Surprisingly, researchers found that behaviors related to social interactions, such as nonverbal communication and relationship-building, were not significantly more prevalent in individuals diagnosed with autism compared to those without a diagnosis.
Instead, behaviors like repetitive movements, known as ‘stimming,’ and intense fixations were strongly associated with an autism diagnosis. The study suggests that healthcare providers may be over diagnosing autism by placing undue emphasis on social-related factors, neglecting to thoroughly examine behaviors that are more closely linked to the condition, such as stimming.
Experts propose streamlining the evaluation process for autism by focusing on non-social behaviors and utilizing AI technology to assess language skills, which could lead to more efficient and accurate diagnoses. This approach may facilitate quicker access to appropriate therapies and treatments for individuals with autism, even though the condition currently has no cure.
Dr. Danilo Bzdok, a neuroscientist from the Montreal Neurological Institute-Hospital and Quebec Artificial Intelligence Institute, believes that the advancement of large language model technologies could revolutionize the understanding and classification of autism in the future. The study, published in the journal Cell, scrutinized over 4,200 clinical reports of 1,080 children undergoing autism evaluations, with AI algorithms predicting diagnosis outcomes.
By inputting the seven criteria descriptions from the DSM-5 into the AI model, researchers identified that individuals diagnosed with autism often exhibited non-social behaviors such as repetitive actions, echolalia, rigid interests, and sensory sensitivities. They argue that evaluating non-social behaviors may offer a more effective diagnostic approach than focusing solely on social indicators, potentially reducing the risk of over diagnosis in autism cases.
It is important to note that the study acknowledges limitations, including a lack of data on older children who may present different symptoms. The findings come amidst a surge in autism diagnoses in the United States, with current CDC statistics indicating that one in 36 children in the U.S. are diagnosed with autism, representing a significant increase from previous years.
While experts attribute part of this surge to heightened awareness and improved detection by healthcare providers, advocacy groups emphasize that the underlying causes of autism remain largely misunderstood. The research underscores the need for ongoing evaluation and refinement of diagnostic criteria to ensure accurate and timely identification of individuals with autism spectrum disorders.