Autism Spectrum Disorder (ASD) is a developmental disability that can lead to sustained challenges in social, communication and behavioral domains. It has also been observed that diagnoses are often delayed (mostly not until six years of age) due to the lack of providers or the primary care providers (PCPs) being unable to make confident ASD diagnoses. However, other information in the patient history may contribute to early detection and is currently under-used. Symptoms and associated conditions (either in the patient’s own medical history or their parents’) manifest much earlier and are well captured by the electronic health records (CPRD GOLD, CPRD AURUM).
This study will leverage medical histories using UK primary care data to develop early warning scores to empower PCPs to detect patients who are most likely to have ASD. It can help shorten the time to diagnosis, allow for early access to interventions, maximise developmental potentials of young patients and improve societal equity by reducing bias due to socio-economic or demographic factors. The early signals can also trigger advanced analyses, for example in genomics, that helps to move the research towards precision medicine and personalised treatment in ASD.
Early diagnosis of ASD would allow for early intervention so that children can fully leverage the developmental window to improve faster. Current ASD-specific screening tools have been ineffective, with poor sensitivity and low positive predictive value. This study will fully leverage the personal history of the evolution of conditions in the UK population to assess their potential to provide an early diagnosis. We will develop an early warning scoring system for the prediction of ASD based on a Cox proportional hazards model with time-varying covariates and possibly incorporating latent variables (also known as factors) that represent a low-dimensional summary of a child’s history of other relevant conditions. Both linear and non-linear specifications (e.g. interaction between risk factors) will be explored. We will demonstrate the performance of this system through a robust validation design and by comparing it with existing ASD screening methods (e.g. the modified checklist for Autism in toddlers) and recently developed algorithms in the literature.
Time to diagnosis of Autism spectrum disorders
Yajing Zhu - Chief Investigator - F. Hoffmann - La Roche Ltd
Yajing Zhu - Corresponding Applicant - F. Hoffmann - La Roche Ltd
Fiona Steele - Collaborator - London School Of Economics & Political Science
Sze Ming Lee - Collaborator - London School Of Economics & Political Science
YUNXIAO CHEN - Collaborator - London School Of Economics & Political Science
Christopher Chatham - Collaborator - Genentech, Inc.
Christopher Chatham - Collaborator - F. Hoffmann - La Roche Ltd
Fiona Steele - Collaborator - London School Of Economics & Political Science
Iori Namekawa - Collaborator - Roche
Sze Ming Lee - Collaborator - London School Of Economics & Political Science
Shu Wang - Collaborator - Genesis Research LLC
Shu Wang - Collaborator - Genentech, Inc.
YUNXIAO CHEN - Collaborator - London School Of Economics & Political Science
Kelly Zalocusky - Collaborator - Genentech, Inc.
CPRD Mother-Baby Link;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation