[pillowkusis]

I'm pretty skeptical of this study.

The breakdowns by age and gender make me suspicious of p-hacking. It seems this journal does not require pre-registration of studies so it's hard to know if the study was pre-registered. It very likely was not.

Note that the only statistically significant finding was for boys, ages 10-17. No other category had a significant increase. That sounds like a negative result across most dimensions of analysis to me.

>There was a 28.9% increase in suicide among Americans ages 10-17

>The study found that boys were far more likely than girls to kill themselves after the show debuted. Suicide rates for females did increase, but it was not statistically significant

This is an enormous effect size. What exactly is the supposed method of action here? So only boys, who watch netflix, only ages 10-17, who decided to watch a show about a girl committing suicide (teen boys do not usually watch female-led dark serial dramas), account for the majority of the 30% increase in suicide across their entire age group?

>The spokesperson noted that the study conflicts with research published last week out of the University of Pennsylvania. That study found that young adults, ages 18-29, who watched the entire second season of the show "reported declines in suicide ideation

Now we have two seemingly "significant" studies with opposite conclusions. It beggars belief (if either of them are even representative of a real life causative factor at all).

Girls, older teens, and adults are completely immune from this effect. What possible method of action would affect such a specific demographic so disproportionately, with absolutely no spillover to any other demographic? How can it be that teenage girls, who watch more dramas and presumably empathize with a female lead more, show absolutely no effect? Or could it be that the researchers had an expected result, broke down the data until they found a dimension that had a large enough increase by chance, and submitted it to a journal full of peer reviewers who would find the result plausible and prestigious to their field?

[learc83]

>How can it be that teenage girls, who watch more dramas and presumably empathize with a female lead more, show absolutely no effect?

In general women are more likely to attempt suicide, but men are much more likely to succeed because of the methods they tend to choose. The data for attempted suicides is much less granular and reliable than the data for successful suicides.

[adjkant]

I don't see how an age group of 10-17 which is at a much more formative time in their lives with some not having fully developed brains points to p-hacking. Gender difference is interesting but doesn't mean p-hacking because the reasoning for it is non-obvious.

It's also possible that the 10-17 audience of 13 Reasons Why was significantly more male, but Netflix does not release that data. But there are plenty of plausible reasons for the gender divide.

> Now we have two seemingly "significant" studies with opposite conclusions

But not conflicting results. There's nothing that says one age group, with vastly more life experience, would be affected less or even differently than another with significantly less life experience.

> Or could it be that the researchers had an expected result, broke down the data until they found a dimension that had a large enough increase by chance, and submitted it to a journal full of peer reviewers who would find the result plausible and prestigious to their field?

A result in either direction would have been interesting for the field, so I don't think this is the motivation. The positive does get a bit more media attention, but I think this view is overly skeptical.

[crazygringo]

The paper's own discussion section explicitly acknowledges potential weaknesses:

This study has several important limitations. First, the quasi-experimental design of our study limits our ability to draw any causal conclusions between the release of 13 Reasons Why and increased suicide rates in young people in the U.S. Nevertheless, the time series and forecasting approaches employed in this study allow us to make credible inferences about this association. The initial increase in youth suicide rates in the month immediately following the series release is concordant with a prior report showing a spike in Internet searches about suicide in the month following release,46 and a small single-hospital study showing an increase in suicide attempt admissions after the series’ premiere. Second, we were unable to assess whether the observed increase in youth suicide rates was attributable to the portrayal of suicide in the series, a lack of adherence to media guidelines (e.g., failure to provide national suicide prevention resources until later months), or other factors. The observation that the series was first released on March 31, 2017 and suicide rates increased that month also raises questions about effects of pre-release media promotion of the series premiere. Third, we did not examine the impact of 13 Reasons Why on specific methods of suicide (e.g., suicide by cutting) due to small cell sizes, which would result in unstable estimates. Fourth, there may have been other events or unmeasured factors that occurred during the study period that might be associated with increased suicide rates. Fifth, our study may have lacked sufficient statistical power to detect a significant association in 10- to 17-year-old girls. Finally, as with most studies looking at possible contagion, we have little understanding of “dose” or context, including who specifically watched the series, when they watched, whether they binge-watched, if it was further discussed in peer-groups, how secondary discussions may have influenced vulnerable individuals, and whether the subsequent focus on suicide prevention may have actually mitigated some of the pronounced contagion effects.

[pillowkusis]

Thanks, I couldn't find the paper anywhere.

"Second, we were unable to assess whether the observed increase in youth suicide rates was attributable to the portrayal of suicide in the series."

This is so intellectually dishonest. It feels like a Motte and Bailey[0] argument. Scientists go around parading a correlation as causation. The title of the paper implies causality. They speculate idly about the causation in public. They make no effort to correct media sources suggesting a cause. And then, when challenged, they fall back on "well, nobody can establish causation here..."

I'm not even sure there's a correlation, let alone causation!

0: https://rationalwiki.org/wiki/Motte_and_bailey

[tssva]

"credible inferences" means assumptions which fit the narrative we want to tell.

That disclaimer is basically a long-winded way of saying our study provides no evidence of anything.

[robocat]

However there is good reason to suspect a series about suicide would cause suicides.

"The average increase in motor vehicle fatalities is 9.12 percent in the week after a suicide story" (I love the extra 0.12 points!): https://www.ncbi.nlm.nih.gov/m/pubmed/867044/

So the implication isn't so much about copycats, but about contagion.

[DanBC]

Young women are the group that has lowest risk for death by suicide, although much higher risk for self harm.

Young people in general have lower risk of death by suicide, and any increase in this age group is worrying.

These are mostly preventable deaths. The reward - a not very good tv show - doesn't in my mind justify the risk - more young people dying preventable deaths.

[leephillips]

Thank you for taking the time to write this comment; it's a good example of the kind of critical thinking that everyone should apply to research reports.

[saas_sam]

Great observations. At first glance I'd wonder if Fortnite wasn't the true culprit here. (Not really but you get what I mean.) Very shady study.

[hajile]

I think these issues are grossly overstated.

Let's look at the statistics (the newest from the US government are 2017). Around 79% of suicides are men. That edges up to around 82% in the 14-25 age group and down to 75% in the 10-14 age group (though the rate there is much lower).

Suicide is the second biggest cause of death in the 15-24 age group. Furthermore, there is a youth suicide epidemic with the rate for the 15-24 age group going from 9.7 in 2007 up to 14.46 in 2017 (per 100K with most of that jump occurring in just the last 4 years). That's an almost 50% increase over the past decade generally speaking. Between 2016 and 2017 there's an abrupt jump of 1.3 per 100K.

Most interesting to the study in question is the 10-14 age group. despite rises in other rates, their rate is basically unchanged until the year that show was released. At that point, we see a doubling of the suicide rate (from 0.73 to 1.34) for the 10-14 age from 2016 to 2017. That is definitely statistically significant.

Suicide Contagion is a rather well-known effect. Furthermore, it is known that adolescents and young adults are particularly susceptible to this effect.

Season 1 and 2 are very different (also reflected in their critic ratings). The first season has been out much longer than the second, so statistics there are probably more accurate. It may also be true that those who dealt with the first season and moved on to the second were less at risk to begin with. In addition, the show is about a bunch of high-school kids, so I'd guess they would find it more attractive than most dramas while lots of older people would find it less appealing.

TL;DR: Males are generally the ones committing suicide. The show targets the age groups in question. Those age groups saw a big jump in suicide rates the year the show came out (including a group that had previously been all but unaffected by other changes). This all seems inline with the study in question.

https://afsp.org/about-suicide/suicide-statistics/

https://www.nimh.nih.gov/health/statistics/suicide.shtml#par...

https://www.hhs.gov/answers/mental-health-and-substance-abus...