I was reading around the web on the Supreme Court’s recent ruling on California’s law banning the sale or rental of violent video games to minors. Then, I was watching a news clip from the CBC with Peter Mansbridge. Near the end of the clip was a mention of a recent study and this showed up.
That image about children’s evening media use and sleep problems raised my bullshit alarm and I promptly dropped everything. The study in question was published in Pediatrics and authored by Michelle Garrison (Seattle Children’s Research Institute) and colleagues.
Background: Media use has been shown to negatively affect a child’s sleep, especially in the context of evening use or with a television in the child’s bedroom. However, little is known about how content choices and adult co-use affect this relationship.
Objective: To describe the impact of media content, timing, and use behaviors on child sleep.
Methods: These data were collected in the baseline survey and media diary of a randomized controlled trial on media use in children aged 3 to 5 years. Sleep measures were derived from the Children’s Sleep Habits Questionnaire. Media diaries captured time, content title, and co-use of television, video-game, and computer usage; titles were coded for ratings, violence, scariness, and pacing. Nested linear regression models were built to examine the impact of timing, content, and co-use on the sleep problem score.
Results: On average, children consumed 72.9 minutes of media screen time daily, with 14.1 minutes occurring after 7:00 pm. Eighteen percent of parents reported at least 1 sleep problem; children with a bedroom television consumed more media and were more likely to have a sleep problem. In regression models, each additional hour of evening media use was associated with a significant increase in the sleep problem score (0.743 [95% confidence interval: 0.373–1.114]), as was daytime use with violent content (0.398 [95% confidence interval: 0.121–0.676]). There was a trend toward greater impact of daytime violent use in the context of a bedroom television (P = .098) and in low-income children (P = .07).
Conclusions: Violent content and evening media use were associated with increased sleep problems. However, no such effects were observed with nonviolent daytime media use.
I did not believed those numbers I saw and I wondered whether they were made up, so I tracked down what I believe is the source of these statistics floating in the sea of news, the source of all this is from the Associated Press:
Some findings for the children studied:
– Daily screen time averaged about 73 minutes, with 14 minutes after 7 p.m.
– Children with bedroom TVs watched about 40 minutes more TV daily
– About 60 kids averaged an hour or more daily of violent TV; 37 percent had frequent sleep problems vs. 19 percent who saw little or no violence.
– Almost 100 kids averaged more than half an hour of nighttime TV; 28 percent had frequent sleep problems vs. 19 percent who watched little or no nighttime TV.
My problem is where do these numbers come from? I have contacted the lead author, Dr.Michelle Garrison, who provided the numbers to the journalists and she acknowledged that they oversimplify a complex issue.
When I started getting phone calls from journalists, though, it became quickly apparent that trying to describe to a lay audience the marginal effect of each additional hour of media per category in terms of the adjusted mean effect on sleep scale just wasn’t working. As a result, there were some journalists who were trying to calculate their own dichotomous statistics from the tables and there were some issues there. So while also attempting to explain the journal article findings to the press, I also calculated dichotomous statistics (percentages and chi-squares) using the same data and same variables as the paper, and provided those numbers to journalists who asked for a clearer way to present the findings to the public. While I think that the dichotomous statistics oversimplify a complex issue to a degree, they are indeed valid numbers that address the original hypotheses.
I wholly agree that it is quite difficult to translate statistical analyses into a lower level discourse. Although, I am worried how those numbers were framed on various news sources, some might seem to bloat the findings as was indicated earlier in my post.
Most the articles that contained these statistics seemed to have derived from the AP release in one form or another. Analyzing the original source (AP), it seemed to follow the gist of the study. However, “violence” was not detailed as Medpage Today’s where the toddlers were “not watching super-gory adult fare or playing shoot-‘em-up computer games. Instead they were seeing cartoons and other material aimed at slightly older children.” It’s sad to see how news sources are playing telephone and we will have a portion of the public misunderstanding the study. What’s even sadder is that I noticed some news adding their flavour or removing some stuff and contacting their own expert’s opinions on something that might not even relate to the study itself.
Time’s Healthland, Mail Online, and redorbit had the following sentences: “Those who watched TV after 7 p.m. had the most difficulties, with 28 percent suffering sleep disturbances compared to 19 percent of those who watched little or no TV at night.” The manner readers will interpret this is that watching TV after 7 p.m. seem to increase sleep problems. My problem is 28% of what? I know you are saying children who watched TV after 7 p.m., but how many are there? Looking back at the AP source, it suggested that about 100 children out of 612 or 16.3% of the sample and out of that sub-sample, 28 parents reported sleep problems with their child or 28%. But without the context, a person is likely to think that a child who watches after 7 pm will have a 9% increase of sleep problems.
Moving on to the 37% of those 60 children who reported who averaged an hour or more of daily violent TV. The absolute number came down to 22 parents who reported sleep problems versus the 19% of those who watched little or none. Again, 19% of how many? 557? The more inquisitive person is likely to ask what are the characteristics of those 60 children and parents. Another interpretation is how the sample is likely to alarm parents, discounting because of the relative size or over-reaction from well-to-do parents who will immediately send their children to bed (for a few weeks before going back to their usual routine).
“Children with bedroom TVs watched about 40 minutes more TV daily”. 59 children have television in their bedrooms or about 9.6% of the sample. A peculiar observation is that in the study indicated that “children with a bedroom television were more likely to be low-income or single-adult households”. This changes my reading of the study that other potential factors come into play as well. It does so when their regressional model (see table 4) found that along with nighttime TV time, low-income and single-adult household do have a significant relationship with sleep problems. Evening media use is still significant and taken those factors into account (and Dr. Garrison replied back my questions about them which reminded me about how regressions are made.).
The way I see it now is that not only having a TV in your bedroom increases TV use and therefore poses a risk for sleep problems, but can it be possible for low-income or single-parents to remove them? It is possible that this may be due to parenting style in regards to children’s media use, they may use it to distract their children while resting because maybe they are exhausted in working in their stressful low-paying jobs. Or perhaps they lack the energy to parent their children’s media use. Contrast those with more financial resources or two-parent households, most of them may already have made good media decisions for their children. The problem is that socioeconomic status and parental status was not mentioned at all in the news, the focus is rather on the nighttime TV hours. In my interest for brevity on interventions as there will be an upcoming paper, Dr. Garrison said there are possible mediational differences.
My mind is turning into mush, but I now understand that translating knowledge from one level to another is a difficult task. Overall, the news give a pretty good description of the study. Except for that one : ” children who watched violent television, including cartoons with violence during the day showed as much as an 18% increase in sleep problems“. This is why I started this blog, full of facepalming and hand wringing of bad journalism.
Garrison, M. M., Liekweg, K., & Christakis, D. A. (2011). Media use and child sleep: The impact of content, timing, and environment. Pediatrics . doi: 10.1542/peds.2010-3304