Abstract
Individuals with autism spectrum disorders (ASD) experience a significant number of co-occurring medical conditions, yet little is known about these conditions beyond their prevalence. This study hypothesized that individuals with ASD face an increased burden of co-occurring conditions as measured by the presence, frequency, and duration of medical visits. Using a novel tool (pyPheWAS) to analyze medical record data from 3,097 individuals with ASD and 3,097 matched controls, the researchers identified distinct medical profiles based on the age of ASD diagnosis. Early diagnosis (before age 7) was associated with neurological conditions like epilepsy, while later diagnosis (after age 7) was strongly linked to psychiatric conditions and suicidal ideation. These findings suggest that the timing of an ASD diagnosis is a critical marker for clinical management and long-term health outcomes.
Introduction
Autism Spectrum Disorder (ASD) is characterized by significant clinical heterogeneity. While it is well-established that individuals with ASD have higher rates of co-occurring conditions compared to the general population, traditional research has often relied on prevalence—a binary measure of whether a condition exists. This approach fails to capture the "medical burden," or the intensity and persistence of these conditions over time.
Furthermore, the timing of an ASD diagnosis varies widely. Some children are identified in early childhood, while others are not diagnosed until adolescence or adulthood. This study explores whether the age at diagnosis serves as a proxy for different clinical subtypes, characterized by unique patterns of co-occurring neurological and psychiatric conditions.
Methodology
Data Source and Cohort Selection
The researchers utilized the Vanderbilt University Medical Center’s BioVU, a massive de-identified electronic medical record (EMR) database. The primary cohort consisted of 3,097 individuals with ASD matched with 3,097 typically developing (TD) controls based on age, sex, and race. To ensure fair comparisons of visit frequency and duration, a "Lifespan Cohort" subset was created, matching 1,420 ASD patients with 2,210 TD controls based on the total length of their medical records.
The pyPheWAS Tool
The study employed a "Phenome-Disease Association Study" (PheDAS) approach using pyPheWAS, a Python-based tool. This method maps standard ICD-9 billing codes to "phecodes"—groups of related symptoms that define distinct clinical conditions. Unlike standard studies, the researchers analyzed three dimensions of medical burden:
- Presence: Whether a specific phecode appeared in the record.
- Frequency: The total count of incidences of a code.
- Duration: The time span between the first and last appearance of a code.
Subtyping by Age
The ASD cohort was divided into two groups:
- Early Diagnosis: First ASD code recorded before age 7.
- Late Diagnosis: First ASD code recorded at or after age 7.
Key Insights and Results
The Multi-Dimensional Nature of Burden
The study demonstrated that prevalence alone is insufficient to describe the ASD experience. For example, conditions like ADHD were only significantly associated with ASD when examining the duration of the condition rather than just its presence. This suggests that while many people may experience ADHD symptoms, individuals with ASD require more prolonged clinical management for them.
The Late-Diagnosis Signature
A striking divergence emerged between the early and late diagnosis groups. The late-diagnosis group was more likely to be female (24.6% vs. 16.1% in the early group). This supports the "camouflaging" hypothesis, which suggests that females may mask ASD symptoms until the social and biological pressures of adolescence trigger a "second hit," leading to psychiatric crises.
Key disparities included:
- Psychiatric Prevalence: The late-diagnosis group had a 54.4% prevalence of psychiatric conditions, compared to just 14.7% in the early-diagnosis group ($p < 2.2 \times 10^{-16}$).
- Suicidal Ideation: This was significantly enriched in the late-diagnosis group, highlighting the severe mental health risks associated with delayed identification.
- Neurological Profiles: Early diagnosis was more strongly linked to hearing loss and neurological conditions such as encephalopathy and seizures.
Temporal Trajectories
The study also mapped when co-occurring conditions appeared relative to the ASD diagnosis. In both groups, epilepsy-related codes typically appeared before the ASD diagnosis, suggesting that neurological symptoms often serve as early red flags. However, psychiatric issues followed a different path: in the early group, they usually appeared after the ASD diagnosis, whereas in the late group, they often appeared before or concurrently with the ASD diagnosis. This suggests that many individuals enter the medical system for depression or anxiety before their underlying autism is recognized.
Discussion and Clinical Implications
The findings confirm that individuals with ASD carry a significant medical burden characterized by increased visit frequency and duration across various conditions, including sleep disorders, mood disorders, and developmental delays.
The distinction between early and late diagnosis profiles suggests that age at diagnosis is not merely a matter of timing but may reflect different clinical phenotypes. For clinicians, these profiles offer a roadmap for personalized medicine. By identifying "onset peaks" for specific comorbidities, providers can introduce targeted risk assessments and interventions at the appropriate stages of a patient's life.
Limitations
The study relies on ICD-9 billing codes, which are designed for administrative purposes rather than clinical research and may contain inaccuracies. Additionally, the data is sourced from a single academic medical center in the Mid-South United States, which may limit its generalizability to other geographic or socioeconomic populations. Future work should integrate socioeconomic status and genetic data to further refine these clinical profiles.
Summary updated