March 11, 2022

By Tom Insel, MD, and Matthew Hirschtritt, MD, MPH

Good morning, Research Roundup readers. Our package this week includes a longer than usual summary of a paper that investigated the accuracy and the uptake of TikTok videos about ADHD. Spoiler alert: the accuracy of the videos was low, and the uptake much, much higher than high quality, academic, peer-reviewed journals. Ouch. We also look at the use of AI-based linguistic tools to measure thought disorder, the social determinants for response to depression treatment, and – who knew? – the relationship of birth spacing to oppositional defiant disorder (ODD).

(If you like what you’re reading here, please forward this to a friend or colleague. You – or they – can sign up for our newsletters here.) 


TikTok and #ADHD – Viral but Fake News

Anyone who’s searched online for medical information is familiar with the motley assortment of results and sources that are available, ranging from anecdotal to academic. Without subject expertise – or the knowledge that as a consumer, you’re turning to a credible, trusted source – it’s nearly impossible to separate the wheat from the chaff. Yet increasingly, consumers are turning to social media to answer their health and medical questions. Recent reports have shown that the quality of medical information on platforms like YouTube and TikTok – including videos on acne, kidney stone disease, and alcohol use – is mixed at best and frequently poor. Now, a new study dives deep into the fast-paced, kaleidoscopic world of TikTok to assess the accuracy of information about ADHD available on the app.

Anthony Yeung of the University of British Columbia and the Centre for Addiction and Mental Health (Toronto) and colleagues identified the 100 most viewed and liked TikTok videos with the hashtag #adhd in July 2021. Two psychiatrists with ADHD expertise classified each video as misleading or useful. They also rated the understandability and actionability of each video using a standardized tool, as well as the overall quality of the medical information using a four-point scale.

Among their findings:

  • Among the selected videos, 52% were misleading, 27% personal experience, and 21% useful.
  • The top 100 ADHD-related videos had over 283 million total views; the average number of views per video was 2.8 million, the average number of shares was over 31,000, and the average number of likes was over 631,000.
  • Only 11 of the 100 videos were uploaded by healthcare professionals; the remainder were created by laypeople.
  • While the vast majority of the videos scored high on understandability, only 14% were actionable (that is, provided clear guidance to the viewer about how to apply the information).
  • On a scale from 1-4 (with 4 indicating excellent quality), the mean medical-information quality was a meager 1.3; the average quality score for “useful” videos was only slightly higher than the overall average at about 1.7.
  • Though videos uploaded by healthcare professionals tended to have higher quality, over a quarter were classified as misleading (compared with just over half of the videos uploaded by non-healthcare professionals).

Most of the misleading videos erroneously stated that symptoms such as anxiety and depression that are common among many psychiatric disorders are diagnostic of ADHD. Other examples of misleading statements included statements that vague symptoms like “anxiety shivers,” “random noise making,” and “being competitive” are signs of ADHD. Still others claimed that ADHD is “equally common among girls and boys” (in reality, the ratio of ADHD diagnoses in boys to girls is about 4-to-1).

In contrast, useful videos tended to be more nuanced. For instance, one useful video counseled viewers that people with ADHD may need extra time to find items, may feel anxiety about creating and managing to-do lists, and that appropriate treatment can improve their quality of life. Others accurately described common ADHD symptoms like daydreaming, disturbed sleep patterns, and academic challenges – and encouraged viewers to seek professional advice if they suspected they (or their child) have ADHD.

Perhaps the most stunning result of this study was the number of views. For comparison, the two most-viewed ADHD-focused reports published in JAMA Psychiatry in 2021 have been viewed about 5,000 and 7,000 times. Academic journals will never reach the mass audiences that social media platforms attract – nor should they, given the specialized content. But there is a clear need to enhance the appeal of high-quality, medically accurate, and useable content on popular platforms like TikTok. If you can’t beat ‘em, join ‘em.

TikTok and Attention-Deficit/Hyperactivity Disorder: A Cross-Sectional Study of Social Media Content Quality. Canadian Journal of Psychiatry, February 23, 2022.

–Matthew Hirschtritt


Birth Order and Disorder: an ODD Story

Image: Shutterstock

There’s increasing evidence that a short interpregnancy interval – the time between the birth of one child and the conception of a subsequent child – may be associated with spontaneous preterm delivery, autism, and schizophrenia, among other poor outcomes, in the subsequent child. Now a new study suggests that children conceived less than six months after the birth of an older sibling have a higher risk of being diagnosed with oppositional defiant disorder, a diagnosis made when children are uncooperative, argumentative, and frequently hostile to their caregivers and peers.

Berihun Assefa Dachew of Curtin University in Perth, Australia, and colleagues used data from the Avon Longitudinal Study of Parents and Children, an ongoing study in Bristol, England that tracked some 14,000 births and then assessed and interviewed children at various times after they turned 7. Dachew and colleagues reviewed the data to look at rates of oppositional defiant disorder at ages 7 and 10 years among over 3200 non-twin children.

The researchers found children conceived less than 6 months after the birth of a sibling were about 2.4 times as likely to be diagnosed with oppositional defiant disorder at 7 and 10 years old than those conceived 18 to 23 months after a sibling’s birth. For children conceived at any point beyond 6 months, the odds of being diagnosed with ODD were essentially the same. This association held true even when accounting for maternal and paternal age, maternal education, marital status, how many children the mother had already given birth to, maternal smoking or alcohol use during pregnancy, and maternal medical and psychiatric conditions. In general, oppositional defiant disorder occurs much more often in boys and this study was no different: More than three-fourths of the children diagnosed with ODD were boys.

These data reinforce results from previous studies that an interpregnancy interval shorter than six months is associated with specific psychiatric and neurologic disorders. One leading explanation for this relationship holds that maternal folate supply is depleted during pregnancy and not fully replenished within six months. Folate is essential for healthy fetal neural system development. However, before clinicians counsel would-be mothers to leave more time between children, there is also contrasting evidence linking pregnancy intervals longer than 23 months with poor childhood psychiatric and neurologic outcomes. Clearly more work is needed to untangle this complex set of associations.

Interpregnancy interval and the risk of oppositional defiant disorder in offspring. Development and Psychopathology, March 2, 2022.

– Matthew Hirschtritt


A Thermometer for Thought Disorder?

Photo illustration: Shutterstock

A hallmark of psychosis is disorganized thinking, assessed by incoherent or illogical speech. The development of natural-language processing tools have revolutionized computer analysis of speech. While these tools could provide objective approaches to assessing psychosis, this revolution has been slow to map the landscape of thought disorder. Now some recent findings are beginning to accelerate this young field.

One approach has measured semantic coherence or the likelihood that any two words would be found together. In the course of normal speech, “dog” and “bone” might be commonly associated, but “dog” and “steeple” would be less frequent companions. A computer program called word2vec has combed through billions of pages of text to establish vectors for the association of any two words. Coherence vectors measure the expected associations of any two words. In transcriptions of speech, people with thought disorders make more frequent use of word combinations that are rarely associated.

In a new preprint paper (one not yet peer-reviewed), groups from the University of Cambridge and the Institute of Psychiatry suggest a different approach to mapping thought disorder. Arguing that the problem is in how speech is strung together over time, they created an algorithm to graphically map the relationship between words – a semantic network. The algorithm, called netts, is available as an open-source tool. The team first mapped semantic networks in transcripts of speech from a general population. Having a basic image of the size and complexity of these semantic networks, they then compare people being treated for first-episode psychosis, those rated as having a clinical risk for psychosis, or controls with no history of or risk of psychosis. To provide some standardization to the speech tasks, subjects were asked to describe a picture or recall a story. When people in these different groups were asked to describe a picture, people with thought disorders used speech networks that were smaller and more fragmented than controls. This difference could not be explained by the more limited number of words used by people with thought disorder, or by their use of speech artifacts like “er” and “uh.” 

Why is it important to have objective tools for measuring thought disorder? If natural language processing tools can begin to precisely define different aspects of thought disorder, these algorithms could serve as a sensitive test for detecting the early phases of psychosis and help make diagnoses in settings where clinical expertise is not available. They might also be used to measure outcomes of treatment. Tools like word2vec, netts or algorithms yet to be developed could create the equivalent of a thermometer for thought disorder, bringing precision to the clinical assessment of psychosis.

Semantic speech networks linked to formal thought disorder in early psychosis  medRxiv, February 28, 2022.

Tom Insel


Social Determinants of Antidepressant Response

Photo illustration: Shutterstock

We hear a lot these days about social determinants of health. Where you live, how you live and who you live with, along with income and access to things like safe parks, healthy food and educational and social services, appear to be bigger factors for predicting health outcomes than your genes, your medication regimen or your clinical care. Now there is evidence that these same kinds of social determinants may influence your response to antidepressant treatment.

Joshua E. Buckman from the Centre for Outcomes Research and Effectiveness (CORE) at University College London and colleagues report on a re-analysis of several randomized clinical trials of people with depression treated in primary care. From nine studies they extracted data on nearly 5,000 patients to compare their baseline social factors to their outcomes after 3 to 4 months.

Ratings on scales that measure symptoms of depression were 28% higher for unemployed patients than for those with jobs and 18% lower for patients who were homeowners than for patients who were homeless or living with family or friends, or in hostels. These effects were independent of the kind of treatment received and factors such as educational attainment, severity of depression, and anxiety.

While these differences may not seem profound or surprising, they are roughly the effect sizes of medical or psychological treatments. Indeed, one could argue that if employment or housing were pills, they would be blockbusters. Yet despite their profound influence over health outcomes, neither are considered part of health care.

Socioeconomic Indicators of Treatment Prognosis for Adults With Depression: A Systematic Review and Individual Patient Data Meta-analysis  JAMA Psychiatry, March 9, 2022

–Tom Insel


If you or anyone you know is considering suicide, call the National Suicide Prevention Lifeline at 1-800-273-8255. And if you’re a veteran, press 1.



Tom Insel, MD, is a psychiatrist, neuroscientist, and former director of the National Institute of Mental Health (NIMH). He is a donor to MindSite News and chair of its Editorial Advisory Board. Dr. Insel’s financial conflict of interest statement, which includes equity and advisory roles in several early-stage mental health technology companies, can be found here.

Matthew Hirschtritt, MD, MPH, is a clinical psychiatrist with the Permanente Medical Group, Inc., a mental health services researcher with the Division of Research, Kaiser Permanente Northern California, and assistant program director of the Kaiser Permanente Oakland Adult Psychiatry Training Program. His current research focuses on identification and treatment of patients with first-episode psychosis, as well as implementation of a telehealth-based mental health evaluation and referral program.

What to Do When the World Is Ending

I am part of a generation that feels, constantly, and even in the most mundane moments, that the world is ending.

More Latino Men Dying By Suicide Even as National Rate Declines

U.S. suicide rates fell slightly in 2019 and again in 2020, the CDC reported last month. But there were outliers. Notably, suicides among Latino men increased by nearly 6 percent.

When Positivity Becomes Toxic

We are in the #GoodVibesOnly age, and it’s kind of a bummer. We’re buried in platitudes: “The difference between a good day and a bad day is your attitude.” The book Toxic Positivity points the way toward authenticity.