Mental Health and Diseases of Older Age Among People with Autism or Autistic Traits: AI and Medical Record Linkage in UK Biobank and SPARK

Background

Later-diagnosed autism is on the rise as awareness of this condition increases. Of people living with autism, 96% of people over 60 and 89% of individuals between 40-59 years are undiagnosed¹. Little information is available about the ageing process in autism¹. Autistic people have a shorter life expectancy due to multiple factors. Individuals with autism are at higher risk of mental illness and diseases of older age such as diabetes, dyslipidemia, heart disease, obesity², and dementia³, but frequently face barriers to health care, resulting in health care inequality. The factors that lead to this are not fully understood. 

UK Biobank (www.ukbiobank.ac.uk) has 161,269 participants over the age of 40 who have results (both total and item-specific) for the Autism Spectrum Quotient (ASQ), a screening tool for autism, which contains questions about social interactions, theory of mind, and interests. In addition, Dr. Tammimies will provide access to the Simons Foundation Powering Autism Research (SPARK) cohort which has 425,000 autistic participants as well as family members. These large data resources will be used with lifestyle, genetic, medical and prescription records to examine comorbidities, trends in health care utilisation and diagnoses, focusing on diseases of older age in these populations. Using item-level questions (e.g., “I don’t know how to keep a conversation going”) may provide a snapshot of individuals who may struggle to obtain adequate health care.

Aims

• To create subtypes of individuals who share clusters of autistic traits. For example, use unsupervised machine learning methods to create subtypes of individuals based on the ASQ in UK Biobank or a diagnostic instrument such as the ADOS in SPARK 
• Examine these subtypes using XAI and deep learning to relate those subtypes and other risk factors to mental health and ageing-related disease outcomes
• To create a fuller understanding of individuals who may not be receiving the health care that they require by analysing medical and prescribing records of these subtypes using XAI; utilising patterns of prescribing, frequency of GP and hospital visits, and diagnoses. This work will be informative for future research on tailored support for autistic individuals. 

Training outcomes

• Training in biostatistics, epidemiology, statistical genetics and ML/AI applied to biomedical data
• Training in working with “big data”
• Programming experience in bash scripting, R, Python and available software for ML/XAI
• Mastery of autism domain-specific knowledge
• Deep understanding of diseases of ageing
• Working with medical record linkage
• Biomedical writing experience
• Supervision experience of co-supervising MSc and/or Honours students
• At minimum 3 published manuscripts
• Input from 4 Supervisors, to be maintained via frequent meetings together with the student
• Encouragement and discussions about future directions, be it in academia or industry

References

1. Stewart GR, Happé F. Aging Across the Autism Spectrum. Ann Rev Devel Psychology 2025;7:461-484
2. Dhanasekara CS, Ancona D, Cortes L, Hu A, Rimu AH, Robohm-Leavitt C, Payne D, Wakefield SM, Mastergeorge AM, Kahathuduwa CN. Association Between Autism Spectrum Disorders and Cardiometabolic Diseases. JAMA Pediatr 2023;177:248-257
3. 3. Vivanti G, Tao S, Lyall K, Robins DL, Shea LL. The prevalence and incidence of early-onset dementia among adults with Autism Spectrum Disorder. Autism Res 2021;14:2189-2199