Despite advancements in early autism detection, approximately 25% of children still receive their diagnosis after the age of six. Since early intervention is crucial for improving outcomes, delayed diagnosis prevents many individuals from accessing timely and effective support. A recent study published in JAMA Pediatrics leverages big data and machine learning to explore factors contributing to delayed autism diagnosis and identifies two distinct subgroups.

Key Findings

Researchers analyzed data from 23,632 individuals with autism using the Simons Foundation Powering Autism Research for Knowledge (SPARK) dataset. Machine learning techniques, including k-means clustering and random forest classification, helped uncover two major groups among those diagnosed after age six:

1. D1 Group — Individuals with lower support needs, displaying fewer autistic traits and co-occurring conditions compared to those diagnosed earlier.

2. D2 Group — Individuals with higher support needs, characterized by greater challenges, particularly in repetitive and restrictive behaviors. They also had more co-occurring conditions, suggesting that these complexities may have overshadowed their autism diagnosis.

The study achieved a high level of accuracy in distinguishing these groups, shedding light on the factors influencing late diagnosis.

Implications for Diagnosis and Treatment

These findings highlight the need for more nuanced diagnostic approaches. The study suggests that individuals with fewer apparent challenges (D1) may be overlooked, while those with multiple co-occurring conditions (D2) may receive a delayed autism diagnosis due to diagnostic overshadowing. By leveraging data-driven tools, clinicians can enhance early detection strategies, ensuring individuals receive the right support at the right time.

Future research should focus on longitudinal studies and more refined clinical measures to further improve diagnostic accuracy and intervention planning.