Knowing what we don’t know: Can statistics support autistic people in living their best life?

Father giving young son a piggyback looking happy

Gordon Forbes is a researcher in the department of Biostatistics and Health Informatics. He is currently undertaking a NIHR doctoral fellowship to develop models that predict what the future may hold for an autistic child. In this blog he talks about the importance of describing the level of uncertainty in any predictions. 

Most of us struggle with uncertainty. When a child is formally diagnosed with autism, some certainty is provided to their parents, but they may still feel overwhelmed by questions, not least about their child’s future. Will they be able to find work and how independently will they be able to live? Will they be happy?

In a recent study, we explored whether we could somehow predict the challenges an autistic child will go on to experience in adulthood. Where these could be anticipated, health professionals and families could plan the support and care needed to minimise their impact. Crucially, recognising what we can’t predict in the future of an autistic child, as well as what we can, is an important step towards providing the best support to autistic people, their carers and health professionals.

Predicting Uncertainty?

The future holds countless possibilities for any child, and in fact, some research suggests that in many ways, an autistic person’s health and wellbeing is no different to those of their neurotypical peers. Even so, it has been shown that autistic people are more likely to face certain challenges, including mental health issues, struggles in finding employment and social isolation.

Predicting what will happen in someone’s adult life based on who they are as a child is very difficult, though. Some children may find school overwhelming but thrive in adulthood when they have more control over their environment. Children also develop at different speeds. A loss of language in early childhood, for example, may resolve naturally and have no impact on wellbeing in later life.

Complicating matters further, the tests clinicians use to measure behaviour and personal characteristics are not perfect – they only capture a snapshot of who someone is. And life itself is inherently unpredictable – major events and their ramifications, from relocations and bereavements to a change in financial circumstances, are typically impossible to foresee in any meaningful way.

Predictions about an autistic child’s future can also have serious consequences - if incorrect, someone could be denied opportunities or have unnecessary or unhelpful support put in place for them. Consequently, predictions must be handled with care, with an understanding of how much confidence we can have in them. A more pertinent question, then, is can we identify which kinds of challenges are more likely to eventuate and which are more uncertain as an autistic child grows up?

Working with uncertainty: a statistician’s bread and butter

As statisticians, measuring and describing uncertainty is our raison d’etre. Using statistics in this way to understand the paths that autistic children take through life has been a focus of my supervisor Professor Andrew Pickles’ research for the last 30 years. It’s also the subject of my PhD in the Life Course Epidemiology group, which Andrew leads, in the Department of Biostatistics and Health Informatics.

In our most recent study, we wanted to see how well we could predict the aspects of adult life of individual autistic children where they would face less challenges, and those in which they might go on to face more serious problems. These included their mental health and wellbeing, the ability to manage tasks in everyday life and work, and their social life. In this work, identifying what we couldn’t predict with any certainty was as important as identifying what we could predict about an autistic person’s adult life.

To investigate, we used data from the Autism Early Diagnosis Cohort, a group of 123 participants in the US who were referred for assessments for autism as toddlers or at the age of nine in the 1990s. The assessments included tests of behaviour, communication and cognitive abilities, for example, the ability to spot similarities in a pair of pictures. Crucially, for our work, participants also took part in further detailed assessments up to the age of 27, making the cohort one of the richest resources of its kind.

Our approach was a personalised one: we focused on how well we could make individual predictions for a specific child, rather than generalised predictions for the study cohort as a whole. We used statistical models to make the predictions and describe how certain or uncertain any prediction would be for an individual.

The positives of uncertainty

A key result of our study was that we couldn’t predict someone’s mental health or wellbeing in adulthood with any certainty. Such uncertainty could be daunting, but there is an upside. In these areas of life, an autistic child’s future is, in effect, still unwritten. If a negative outcome is no more certain than a positive one, this can give hope and motivation to families, and children can be encouraged to explore life, knowing that anything is possible - much like neurotypical children.

Knowing that mental health issues and a lack of wellbeing cannot be predicted with confidence, all autistic people should receive support in these areas from health professionals, families and carers. That healthcare professionals avoid negative generalisations about a child’s future and communicate uncertainty effectively in conversations with families, given the potentially harmful consequences, is also crucial. The importance of good communication of uncertainty was backed up in conversations we had with parents and clinicians that was part of our research.

Using what we do know

Using data collected when the children in the cohort were nine, we did identify some aspects that could be predicted with some certainty: the children’s IQ – both verbal and non-verbal – and everyday living skills in adult life. In turn, we could also make reasonably good predictions about the potential for employment and independent living, and whether a child may form friendships.

Knowing who is likely to face these kinds of challenges in adulthood means support can be put in place, enabling autistic people to live their best life. For example, if we know at age nine that a child is likely to face difficulties with everyday tasks in adulthood, families and carers have time to plan help such as supported housing.

An important caveat is that predictions are, nevertheless, fluid; a multitude of factors and their timings may - or may not - influence a child’s development. Consequently, predictions should be revisited through an individual’s childhood and adolescence. This way, a child’s future is not unnecessarily pigeon-holed.

Our next steps are to bring together all existing datasets that follow autistic people from childhood to adulthood. Using this, we plan to develop a tool that can indicate what the future may hold for an individual, tailored to a child’s characteristics and experiences, and incorporating their personal priorities in life. Predictions need not be one-off; ideally, they will evolve as the child grows, taking into account new assessments and changes in priorities.

Thanks to the IAMHealth PPI panel who contributed to the ideas presented and provided valuable feedback.

This blog was written by Gordon Forbes with support from Jude Dineley, science communicator and researcher on the RADAR-CNS project. 

Tags: Informatics -

By NIHR Maudsley BRC at 29 Jul 2021, 10:16 AM

Back to Blog List