An autism test used by GPs to flag adults with the condition in adults is ‘unreliable’, academics have said.
The Autism-Spectrum Quotient (AQ10) screens adults who are suspected of having the condition but do not have learning disabilities.
They are asked 10 questions including ‘do you notice small sounds others don’t?’ and ‘do you find it hard to read people’s emotions from their face?’.
British psychologists examined the effectiveness of AQ10 in measuring autism using data from over 6,500 people in the general population.
They said the test had ‘poor reliability’ in spotting true autistic traits, which could be damaging for patient care.
The results of scientific research which has long used the test to get a better understanding of autism may also be invalid, the team said.
An autism test used by GPs to flag adults with the condition in adults is ‘unreliable’, academics have said. They said the test had ‘poor reliability’ in spotting true autistic traits (stock)
Autism is a lifelong developmental disability that affects how people perceive the world and interact with others. It is measured on a spectrum.
There are around 700,000 autistic people in the UK – more than 1 in 100, according to the National Autistic Society.
Although it is normally diagnosed in children, it’s quite common for people to have gone through life without an autism diagnosis, feeling that somehow they don’t quite fit in and learning to cope with their differences.
The National Institute for Health and Care Excellence (NICE), the government body who makes guidelines for the NHS, recommends the AQ10 as a screening tool for autism in adults.
If adults who suspect they have autism score between six and ten, they may be referred for more tests.
The score is then linked to their performance on other tasks, in order to inform if autism is the cause of their social difficulties.
Results from questionnaires are also often used in large-scale studies to measure autistic traits in the general population.
The research published today in Cambridge University Press’ journal Experimental Results, University of Bath, King’s College London, and Cardiff researchers used data from over 6,500 participants from the general population – the largest sample to date.
Their results found that the AQ10 had poor reliability across several statistical techniques.
Dr Punit Shah, senior author of the study and expert on cognitive processing at the University of Bath, told MailOnline: ‘The tool isn’t for diagnosis but a screening tool used in the general population to “flag” potential cases of autism for GPs to investigate further, potentially referring people to specialist autism clinics.
‘Given that the tool lacks reliability, it is possible that it may not be doing the screening/flagging process as well as we’d like, with knock-on consequences for autism diagnosis and research.’
Dr Shah said the test was not incorrect in any specific part, rather the questions do not ‘hang together very well’.
It may measure several psychological processes and not just autism alone.
‘It is worrying, though it is difficult to know the extent to which is impacts patient care,’ Dr Shah said.
For research purposes, results from the questionnaire are used in large-scale studies to measure autistic traits in the general population. This shapes how autism is understood, and therefore treated.
Dr Shah said: ‘A lot of what we know about autism – and how to help people with autism – has come from studies where these tools are used.
‘However, if the measure of autism is unreliable, as we suggest, so too are the findings and conclusions.
‘Without reliable measures, it is unclear if the findings from these studies are valid, and may be hampering the support we provide for people with autistic personalities or diagnosed autism in society.
‘Our findings add further evidence to a growing body of literature indicating that the measures of autism and autistic traits currently used in research are inadequate.
‘Our research is just a first step in flagging potential problems and this all now requires further research and a review by NICE.
‘This issue is not just limited to autism tools. These issues are widespread in (poor) psychological screening tools used in mental health research and clinical practice.’
The team now suggest the reliance on AQ10 as a measure of autistic traits needs to be questioned.
