select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
Research, published in JAMA Pediatrics, reports on a possible link between screentime for infants and later autism specrtum disorder symptoms.
Dr Peter Etchells, Reader in Psychology and Science Communication, Bath Spa University, said:
“Unfortunately, this study appears to add little more than noise to an already messy and controversial area of research. The research question appears to focus on screen use/screen time, but this is poorly operationalised – focusing on TV/DVD watching using a binary yes/no response for 12 month olds. This is followed up with a question about how much time is spent watching these media, but at 18 months. Many researchers have been repeatedly saying for a long time that such generalised measures don’t tell us anything useful – we need a much more granular approach that considers both the content children are consuming, as well as the context within which they are consuming it. Moreover, the outcome measures used relate to ASD-like symptoms and ASD risk, but as the authors themselves note, no actual ASD diagnosis measure was used. As such, I would urge caution in taking anything significant from these findings.
“Given the nature of this sort of research, it is essential that studies like this one are pre-registered, and the data and analysis scripts made openly available. As far as I can tell, this has not happened here; nor is the funding source declared. Coupled with the conflation between general screen time and watching TV, it is difficult to see how the results from this study can lead to the authors recommending adherence to AAP guidelines suggesting no screen time before 18 months of age. This is especially frustrating given the current pandemic situation, where screen-based video calls are the only way that families can maintain social contact with each other.”
Prof Dorothy Bishop, Professor of Developmental Neuropsychology, Department of Experimental Psychology, University of Oxford, said:
“It is remarkable that JAMA Pediatrics have published this paper. In an accompanying editorial, numerous flaws with the paper are noted. It is hugely surprising to me they have decided to publish it despite what the editorial says: “It is most unfortunate, then, that for this key and dramatically changed environmental exposure, media, the measurement procedure was not sufficient. It was flawed because it was not collected with sufficient granularity to assess direct effects. After 1 year of age, we know how much media children in the NCS watched, but not what they watched, and it was flawed because the data were collected initially too late (at 1 year of age, when most children are already regularly watching).”
“This is not the only thing that is unfortunate about the study. The title might lead one to expect that the focus is on digital media more broadly, but the only medium studied is watching of TV and DVDs. The presentation of the results leaves much to be desired. The results are presented solely as odds ratios with confidence intervals; distributions of scores are not provided, but for most of the groupings shown in tables, the median score on the outcome scale is zero. In other words, children have been assessed on a scale measuring autistic features that potentially ranges from 0 to 20, on which most children score zero. Screen exposure at 12 months, the only screen-related variable that was significantly associated with outcomes (not of ‘ASD-risk’ but of ‘ASD-like symptoms’) was evaluated by asking ‘Does your child watch TV/DVDs?’. We are not told how many parents responded ‘yes’, but we are told that the ‘interquartile range’, i.e. range of scores on the autism scale (potentially ranging 0 to 20) went from 0 to 1, both for those who said ‘yes’ and for those who said ‘no’. This suggests that the analysis conducted on the data was highly misleading, because it implies some general effect of TV-watching on the population, which is simply not present.
“Furthermore, the analyses do not appear to have been pre-registered, and the funding for the work is not stated.
“I regard it as unfortunate that the journal decided to publish this study, and it is in my view irresponsible of the editorial to use it as a vehicle to promote the American Academy of Pediatrics Media and Young Minds statement recommending no digital media before 18 months of age – when the study gives no support to such a recommendation.”
Dr David Nunan, Senior Researcher at the Centre for Evidence Based Medicine, University of Oxford, said:
“The effect size here is small and there is no mention of its clinical relevance (which is more likely to be irrelevant). “Autism-like symptoms” is not the same as autism, the tools used are not useful for measuring autism and there was no diagnosis of autism. The study did not distinguish between the type of content being watched nor parent-child interaction before, during or after screen watching. The main issue is that of reverse causality – children who are not comfortable in ordinary social situations are offered more screen time as a common activity and that may actually be supportive and helpful for them. Indeed, a recent systematic review confirms increased screen-time in children with ASD and discusses these very reasons as potential explanations for their findings.
“If there are worrying news headlines from this study there may well be many anxious parents as a result. This would be no help whatsoever to children or their families. It is ridiculous to publish this study in its current format when almost half the world are in lockdown, including millions of families and teachers already worried about the impact of all of this on children’s physical and mental wellbeing.
“We are seeing a lot of opportunism on the back of COVID-19. It would be sad if this paper, which provides no answer as to whether more screen time results in autism, leads to worry among families, teachers and the children in their care. No one should worry about the risk of autism due to screen time based on this study.”
Prof Andrew Przybylski, Associate Professor and Director of Research at the Oxford Internet Institute, University of Oxford, said:
“In this paper the authors analyse data collected from parents in 2009 in which a diverse range of questions about the lives of infants and toddlers were asked of parents. The authors do something we have seen many times before and go through such datasets for technology-related questions and then publish implausible conclusions which are not backed up by evidence. Unvalidated measures of screen time were used in this analysis and we are already up to our eyeballs with low quality studies like these.
“Like jazz, studies like these are all about absent details that matter the most. In this paper there are two key details missing.
“First, there was no registration of the authors’ analysis plan before the data was collected or analysed. The richness of datasets like these make them ideal candidates for ‘cherry-picked’ results if not all of the relevant data are analysed. A second screen time question was asked: “How often does child watch TV and/or DVDs?” but this was not analysed. The authors did not report that that this was asked as a follow-up question to parents when their children were 12 months old. Its absence is worrying and I don’t know why the authors have not reported on that.
“Second, there was no mechanism proposed or tested in the data linking screens to autism. If a scientist is going to make a bold sweeping claim that answering yes to the question “Does your child watch TV/DVDs?” at 12 months is associated with greater Autism-like symptoms 6 months later it should incumbent upon them to do the research required to explore before publishing. This was not done.
“It’s difficult to understand how the publication of this paper is not an example a failed peer review process.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“Overall I think that this study does not provide any very clear message about video screen time and autism-like symptoms. Drawing conclusions from the study is not easy, largely because of unavoidable limitations in the data that were available. The data came from a previous US study that was not specifically designed to look at the effects of screen media exposure on autism spectrum disorder (ASD), but was much broader in scope – it did collect some data relevant to that question, but not in much detail. There are some measures of whether the children watched TV or DVD, and (in a limited way) of how long they spent doing this at age 18 months, but no measure of what they were watching on their screens. According to the linked editorial by Christakis, these measurements were not made at the most appropriate age anyway. The research report makes it clear that the cohort of people who provided the data were not intended to be representative of the US population, and furthermore that results were obtained only for children whose caregivers volunteered to complete the M-CHAT questionnaire. This would be likely to move the results further away from being typical of the US (or any other) population.
“This is an observational study, and as with all such studies, it is pretty well impossible to draw conclusions about what causes what from just one study. For example, in looking at a possible association between whether the children used screens at age 12 months and whether their caregivers reported autism-like symptoms later, there will inevitably be many differences between the children (and their families) who used screens and those who did not, apart from the screen use. Any combination of those other differences might be, partly or wholly, the cause of any differences in the prevalence of ASD-like symptoms, rather than the screen use itself. The researchers made some adjustments in their statistical models to try to allow for some of these possible other factors, but that can be done only for factors on which they have data. A positive point about the research is that it’s prospective – that is, the data on screen use and on other factors that might be related to risks of ASD were collected before the data on autism-like symptoms were collected. In principle that might at least rule out some possible patterns of cause, because an effect can’t come before its cause. But in this study that’s arguably not much of an advantage anyway. All the data came from what the caregivers reported to the original researchers. So it’s possible that what the caregivers told researchers when their children were 12 or 18 months old was affected in some way by what the caregivers had observed of their children’s behaviour up to that age, and in turn that could, maybe, have affected their responses, later, on the M-CHAT questionnaire that was used to measure both ASD-like symptoms and the risk of future ASD-like symptoms. And it’s important to note, as is made clear in the research paper and the editorial, that M-CHAT is not a tool for diagnosing ASD or for defining symptoms as ASD-like – I understand that it’s a screening questionnaire designed for early identification of children who might benefit from early interventions in relation to possible ASD. So we (and the researchers) just can’t tell whether in fact these children would demonstrate diagnoses of ASD, or of a wide range of ASD-like symptoms, after a proper diagnosis. The researchers, in the Limitations section near the end of their report, and Christakis in the linked editorial, list and explain the limitations that I have described (and several others).
“In relation to video screen use, to whether the caregiver plays with toys with their children daily, and to how often the caregiver reads to the child, the researchers found only two factors where there was acceptable statistical evidence of an association with one of the two different measures of ASD-like symptoms that they derived from M-CHAT. On one of those measures, the ASD risk, no statistically significant associations with screen use, play or reading were found. On the other, which measured the actual ASD-like symptoms M-CHAT asks about, there was some evidence of an association with whether the child used screens at 12 months old, and with whether their caregiver played with them daily, but not with the amount of screen use at 18 months or with being read to. However, the evidence for these two associations is far from strong. (The relevant confidence intervals come close to including zero effect.) You might think that this weakness of evidence arose because of the issues with the data that I’ve already mentioned, and that stronger evidence would therefore be found in better data. Well, that’s a possibility, but it’s also possible that the evidence would entirely disappear with better data. We just can’t tell. A technical statistical point is that, in looking for so many different associations between ASD-like symptoms and a big range of factors about the children and their families, in just two complicated statistical models, arguably the evidence for the associations that were found could well be weaker than it looks.”
Dr Punit Shah, Psychology lecturer, University of Bath, said:
“This is an exciting and important new study, which will inevitably be controversial given the reluctance of autism researchers and wider society to consider the role of parenting in the developmental trajectory of autism. Although there are several obvious limitations of the study – such as the reliance on suboptimal measures of autism-like symptoms and crude ‘yes/no’ measures of media exposure – it raises important new questions about the contribution of parenting behaviors to autism. While we should be cautious not to read too much into this individual study and be mindful of its potential to cause unwarranted blaming of parents, it suggests that the role of the environment in autism warrants much more attention than it currently receives.”
Prof Uta Frith, Emeritus Professor of Cognitive Development, Institute of Cognitive Neuroscience, UCL, said:
“From this study we cannot tell which is cause and which is effect. A case can be made for increased screen time being a consequence, not a cause of “autism-like symptoms”. Not all children are equally comfortable in ordinary social interactions and parents/carers can sense social discomfort. This may lead to an increase in screen time as a common default activity. I worry that the paper will do harm, if it feeds the ‘blame the parents’ meme.”
Dr James Cusack, Director of Science, Autistica, said:
“It is absurd to claim, based on these results, that screen time has any association with autism or how autistic someone is. Families deserve better science than this – particularly right now.
“First of all the effect of screen time observed in the study is small. Secondly, the tool used to observe the effect is not particularly effective at detecting autism. Thirdly, the measure is used at two years old, an age where children develop at different rates and where we know it is hard to accurately diagnose autism.
“In the very unlikely event that this study was accurately detecting autism or the presence of traits, this study would still be fatally flawed. This is because a result like this could be simply down to an increased preference for screen time in children who show increased traits associated with autism.
“In the editorial which accompanies this article, the author suggests that screen time must be limited because we must start with the Hippocratic oath – “to do no harm”. I would argue, on the contrary, there is nothing more harmful than poor science.
“Clearly, as most families understand, there is a balance to be struck in terms of screen time, but science like this is completely unhelpful. Scientists and journal editors must do a better job of serving families than this. Families should not be concerned that allowing their young children to watch screens will cause autism – it would be absurd to reach this conclusion based on this evidence.”
Mr Hugo Pedder, Senior Research Associate in Statistical Modelling at the Bristol Medical School, University of Bristol and Statistical Ambassador to the Royal Statistical Society, said:
“I am not an expert on ASD and so cannot comment on the difference between ASD-like symptoms and autism, and I also do not have a strong knowledge of the existing evidence. However, as a statistician I can comment on the methodology of the study.
“This observational study shows an association between autism spectrum disorder (ASD)-like symptoms and screen viewing in the first 18 months of life. As such, it does not imply a causal link between screen time and ASD-like symptoms.
“The authors have accounted for some confounding variables, though it is unclear how these were selected. They also assessed the sensitivity of their models to unmeasured confounders and found that quite small changes in these could have led to different results, particularly for the model of ASD-like symptoms in which the association with screen time was found.
“There are also several other quite substantial limitations, many of which were identified by the authors. The sample was self-selected by parents who were keen to assess their child using the ASD screening tool, M-CHAT, and so may not be representative of the general population, and the amount of screen time was reported by parents, which may not be an accurate measure of true screen time use and could introduce bias.
“Despite being a prospective study, it is not possible to identify which effect came first – whether screen time leads to ASD-like symptoms or whether ASD-like symptoms at a very young age (12-18 months) leads to increased screen time. Given these issues, and that the study can only examine correlations rather than causes, the real-world implications of this finding are limited.”
Dr Lucy Livingston, Lecturer in Psychology, Cardiff University, said:
“This is an interesting study that addresses a much under-researched area; the relationship between ASD and screen time. Although the researchers find a positive link between screen time and later ASD symptoms in infancy, we cannot conclude that screen time has necessarily caused this increase in ASD symptoms. Those infants that went on to develop higher ASD symptoms may have already been showing early autistic-like characteristics earlier on in development, regardless of screen time. It will be really interesting to see further research prompted by this study.”
‘Association of early-life social and digital media experiences with development of autism spectrum disorder–like symptoms’ by Karen Frankel Heffler et al. was published in JAMA Pediatrics at 16:00 UK time on Monday 20 April 2020, which is also when the embargo will lift.
DOI: 10.1001/jamapediatrics.2020.0230
Declared interests
Dr Peter Etchells: “PE is the author of Lost in a Good Game: Why We Play Games and What They Can Do For Us.”
Prof Andrew Przybylski: “In the past five years my work has been financially supported by UK taxpayers, the UK’s Economic and Social Research Council, the British Academy, the Diana Award, the John Fell Fund, the Leverhulme Trust, Barnardo’s UK, and the Huo Family Foundation. As part of my science communication and policy outreach, I have served in an unpaid advisory capacity to; the Organization for Economic Co-operation and Development, Facebook Inc., Google Inc., and the ParentZone. For information on where the Oxford Internet Institute gets its funding from, see:
https://www.oii.ox.ac.uk/about/our-founding-donor/
https://www.oii.ox.ac.uk/about/transparency/
https://www.oii.ox.ac.uk/about/giving/”
Prof Kevin McConway: “Prof McConway is a member of the SMC Advisory Committee, but his quote above is in his capacity as a professional statistician.”
Hugo Pedder: “I have no conflicting interests to declare.”
None others received.