Video Gamers…Fat and Sad?
A paper in the American Journal of Preventive Medicine just aged its way past the reporting embargo, and immediately hit the streets. It’s conclusions? Playing video games are bad for your physical and mental well-being. But how good is the paper and its findngs? Read on, for one random internet poster’s opinion.
For those interested, the title is Health-Risk Correlates of Video-Game Playing Among Adults by Weaver et al. It should be appearing in the October issue, but right now, its easiest to ferret out by going through their Press Releases section – something that does not impress me at all.
Authors: Never heard of any of them – not surprising, as preventive health is not my field. There’s alot of authors, mostly MDs and MPH/PhD types, from a number of affiliations – the CDC, the Rollins School of Public Health, and Andrews University. The first two, at least, are pretty major and generally fairly solid research universities (never heard of Andrews, which is not its fault). Should note however that there’s no epidemiologist or biostatistician on their authorship list. Given my experience with even very earnest, very well-educated and extremely good health behavior researchers, this makes me skittish. No offense to the health behavior readers out there, you guys do excellent, interesting work, but n times burned, n+1 times shy and all that.
And so we begin. Their data comes from an internet sampling scheme drawn from an e-Rewards marketing database, and response to an email solicitation. This sets alarm bells of in my mind: We’re drawing from heavy internet users already, and I’m a little skeptical about how representative a sample of people who respond to an email request to complete a survey are. There’s also no mention of any incentives (or lack thereof) for survey participation – their response rate was 17.8%, which is lowish for an epidemiology study, but I have no idea what’s common for behavior studies. But when only 1 in 5 people say yes, one wonders *why* that 1 said yes. They say that the sample’s demographic characteristics were similar to those in a random-digit-dialing sample of residents in the same area (Seattle-Tacoma), but RDD has its own substantial problems.
The original survey asked about video game use on a 10-point scale, ranging from 0 (I don’t use) to 9 (Very Important). The researchers then dichotomized this variable to either Non-Gamer (score of zero) or Gamer (1 to 9). Which means someone who plays Wii Tennis once a a week, and “I took time off my work to clear Ulduar when it was released” are in the same category. Why? – my suspicion is because doubly dichotomous variables are way, way easy to work with. Multiple categories of exposure are a pain. But a pain an epidemiologist or biostatistician could have dealt with – which would have probably given them both a much richer data set, and been able to parse out any effects that exist in say, very-very committed gamers verses casual players who like them some Peggle.
Personality, depression, and health-status are all self-assessed (it being an internet survey), using pretty standard metrics for these things.
First, does the sample match the population. Short answer: No. The study sample was way more educated, nearly twice the college and above graduates than the estimated population average, and 1/6th the number of High school or less types. Also considerably wealthier.
The study finds that gamers are more likely to be depressed, less extraverted and more psychotic than non-players. Additionally, they’re more likely to be unhealthy, have a higher BMI, a lower quality of life, and lower physical and mental health…
Except hardly any of these are significant findings. Look, I’m as quick to dismiss the worship of the p=0.05 as the end-all-be-all of whether a finding is meaningful or not. But these researchers *don’t* report the actual p-values, just whether or not its below 0.10 or 0.05, with the ever so annoying asterix-based notation scheme. None of the physical questions dip below 0.05, and only the extraversion and phychoticism questions do in the physical category. And I’m of the opinion that if you’re going to use out-of-the-box, borderline statistics statistics (dichotomized variables, not reporting your p-values) then by god I’m going to hold you to the 0.05 threshold. You cannot both have your cake and eat it too.
Incidentally, for this study, psychoticism means: Psychoticism is associated not only with the liability to have a psychotic episode (or break with reality), but also with aggression. Psychotic behavior is rooted in the characteristics of toughmindedness, non-conformity, inconsideration, recklessness, hostility, anger and impulsiveness. The physiological basis suggested by Eysenck for psychoticism is testosterone, with higher levels of psychoticism associated with higher levels of testosterone.
So, in the sample population, gamers were more likely to be male, and we find that gamers have higher levels of testosterone-correlated behavior. And no significant physical differences. And more likely to not be extraverted.
Alright authors, here’s where you might make it out okay. Yes, the mainstream media will publish your findings as incontrovertible, robust proof that gamers are fat lazy slobs who hate themselves, but what do *you* say.
“As hypothesized, health-risk factors – specifically, a higher BMI and a greater number of poor mental-health days = differentiated adult video-game players from non-players. Video-game players also reported lower extraversion, suggesting less sociability and less assertiveness”.
No, they don’t. They show less extraversion, certainly, but this is hardly a pathological condition. And the higher BMI? We’re talking about an average of 28.05 vs. 26.55 – and the players are more likely to be men, who often have a higher BMI than women. As far as I can tell, this was not adjusted for in the base analysis. Actual diminished physical health wasn’t different at all for the groups,
When they do look at men versus women generally, male players report higher BMI than non-players, and female players report higher depression than non-players – but again, not by all that much.
Basically, the authors found a mishmash of not terribly surprising, statistically dubious and largely fleeting findings, and have run with it in true style, asserting that this is clearly depressed women “self-medicating” for depression and claiming these are terrifically important findings. Except they’ve made no attempt to distinguish between heavy gamers and casual gamers (I’d really like to see the distribution of that 10-point scale). They authors do get some points for mentioning this is a cross-sectional study, so you can’t make any assertions about causative pathways, but it feels very much like they’ve put it in there because they have to – because they *are* making assertions of causation elsewhere in the paper.
Filed under: Epidemiology, Fat Acceptance | 3 Comments