Epic Research is not viewable using Internet Explorer. Please try accessing it with an alternate browser.
Cosmos Study

Patients Diagnosed with Autism Are Less Likely to Receive On-Time Cancer Screening, Except for Colorectal Cancer

July 28, 2026
Dual-Team Study
Team A:Kersten Bartelt, RNNicholas Volker
Team B:Louis Kazaglis, MDGrant Keane

Key Findings

  • Among patients newly eligible for cancer screening, those diagnosed with autism had a 35.6% lower likelihood of on-time screening overall compared to patients without autism. The gap was largest for cervical (50.7% lower likelihood) and breast (28.7% lower likelihood) cancer screening, and smaller but still significant for prostate cancer screening (14.9% lower likelihood).
  • Colorectal cancer screening ran opposite to this pattern: patients with autism had 12.1% higher likelihood of on-time colorectal cancer screening than patients without autism. This coincided with greater use of at-home, stool-based tests, which made up 46.2% of colorectal cancer screenings among patients with autism versus 37.6% among those without.

Patients with autism and other developmental disabilities are diagnosed with cancer at roughly the same rate as the general population, but they are far more likely to die from it, a disparity widely attributed to gaps in preventive care, delayed diagnosis, and barriers to treatment rather than to the disease itself.¹ Routine screening is one of the most effective tools for catching cancers early, yet prior research suggests it reaches this population unevenly. In one survey of adult women with autism, it was found that intellectual disability, communication difficulty, and non-independent living were all linked to going without having a Pap smear performed.² Understanding where screening falls short and where it does not can help health systems target outreach and accommodations to the patients most likely to be missed.

We studied more than 6 million U.S. adults who newly became eligible for breast, cervical, colorectal, or prostate cancer screening between 2022 and 2024, each followed for at least one year after becoming eligible. Patients were included if they had outpatient care before, during, and after the eligibility window and met age and sex criteria for the relevant screening type; those with a prior history of the corresponding cancer or related conditions were excluded. We compared the share of patients screened on time (within one year of first becoming eligible) between patients with autism and patients without autism. We accounted for demographics, social vulnerability and rurality based on recent address, census region, and eligibility year.

Patients with autism were less likely to be screened on time for most cancer types. Overall, patients with autism had a 35.6% lower likelihood of on-time screening than patients without autism, as seen in Figure 1. The largest gaps were for cervical cancer screening, where patients with autism had a 50.7% lower likelihood, and breast cancer screening, where they had a 28.7% lower likelihood; prostate cancer screening showed a smaller but still significant 14.9% lower likelihood.

Figure 1
Likelihood of On-Time Cancer Screening by Presence of Autism Diagnosis
Likelihood of On-Time Cancer Screening by Presence of Autism Diagnosis
Figure 1. The likelihood of on-time cancer screening among patients diagnosed with autism compared to patients without, by screening type.

Colorectal cancer screening was the exception to this pattern. Unlike the other screening types, patients with autism had a 12.1% higher likelihood of on-time colorectal cancer screening than patients without autism. This appears to be explained by how the screening was done: at-home, stool-based tests (FIT/gFOBT and sDNA-FIT) accounted for 46.2% of colorectal cancer screenings among patients with autism, compared with 37.6% among patients without autism, while in-clinic colonoscopy made up a correspondingly smaller share (52.5% versus 61.4%). Stool-based options can be completed at home without sedation or a procedural visit, which might lower barriers that depress screening rates elsewhere. Other screening methods (including sigmoidoscopy and colonography) were uncommon for both populations.

Figure 2
Colorectal Cancer Screening Method by Presence of Autism Diagnosis
Colorectal Cancer Screening Method by Presence of Autism Diagnosis
Figure 2. The distribution of colorectal cancer screening methods among patients with and without diagnosed autism.

Because patients with autism tend to have more frequent healthcare encounters, we conducted a sensitivity analysis additionally adjusting for healthcare utilization to test whether differences in visit frequency accounted for the observed screening gaps. The overall pattern was unchanged, indicating that the lower screening rates among patients with autism are not explained by differences in how often these patients access care.

These data come from Cosmos, a dataset created in collaboration with a community of Epic health systems representing more than 307 million patient records from 2,000 hospitals and more than 49,000 clinics from all 50 U.S. states, Canada, Lebanon, and Saudi Arabia. This study was completed by two teams that worked independently, each composed of a clinician and research scientist. The two teams came to similar conclusions. Graphics by Brian Olson.

  1. Vuattoux D, Colomer-Lahiguera S, Fernandez PA, et al. Cancer Care of Children, Adolescents and Adults With Autism Spectrum Disorders: Key Information and Strategies for Oncology Teams. Front Oncol. 2021;10:595734. Published 2021 Jan 19. doi:10.3389/fonc.2020.595734
  2. Chen Y, Powers J, McDougle CJ, Zürcher NR, Thom RP. Cervical Cancer Screening and Prevention Uptake in Females with Autism Spectrum Disorder. J Autism Dev Disord. 2026;56(1):1-12. doi:10.1007/s10803-024-06565-2

Data Definitions

Study period
Study population: inclusion
Study population: exclusion
Index date
Exposures
Outcomes
Breast cancer screening
Cervical cancer screening
Colorectal cancer screening
Prostate cancer screening
Lung cancer screening
Confounders
Race and ethnicity
Model specifications
Limitations