Does "Help" for Kids After Self-Harm Reduce Suicide Risk?

Spoiler alert: Yes.

My readers are well aware there is a youth mental health crisis. The rate of suicidal behavior in young people in 2023 was 13% for high school girls and 7% for high school-age boys. We've been at a loss for data that supports effective interventions, and this has led to a tremendous amount of fretting, hand waving, thumb twiddling, and talking about how mental health is so important. What's more important than talking about how important something is is having an intervention to meaningfully modify the bad outcome in a direction that is favorable.

Oral medications for depression? Although there is no good evidence that they increase the rate of death by suicide, poor data led to a black box warning from the FDA on increased suicidal behavior in youth across all oral medicines.

Similarly, I've written a great length about psychiatric hospitalization, which physicians often use to ensure “safety.” It doesn't work. I've looked, long and hard, to find any data that supports the use of psychiatric hospitalization for the reduction of suicide risk, and found none.

One of the everyday things that happens in a struggling child's life before there is a suicide attempt is self-injury. Self-injury is a predictor of subsequent suicide attempts and death by suicide.¹

As much as it pains me to say this, the New York Times, with its sometimes fetishistic coverage of wildly expensive and complicated interventions like dialectical behavioral therapy… was onto something.

New data out today demonstrates that therapeutic interventions, specifically engaging in self-harm behavior, decrease the long-term risk of suicide attempts and death by suicide. Still, unlike anti-psychotic medication, it has no long-term mortality risk.

Disclosures: I co-edited a treatment manual for one such Therapeutic Intervention, along with Laurel Williams, D.O., called Adolescent Suicide and Self Injury: Mentalizing Theory and Treatment I have also written a book on how oral medicines aren’t helpful called Inessential Pharmacology (Amazon affiliate links).

Today’s paper is titled:

Mortality in adolescents after therapeutic intervention for self-harm: A systematic review and meta-analysis

Let's dig into the data.

---

This paper, by Young et. al.,² reviewed all published data on therapeutic interventions (all of them) after a youth had engaged in self-harm. What is self-harm? This meta-analysis defined it thusly:

Self-harm, defined as self-injury or poisoning irrespective of motive

And then, using a structured search approach, Reviewed all published literature on:

RCTs comparing a [Therapeutic Interventions] defined as a replicable psychosocial or pharmacological intervention versus comparator condition (Ougrin et al., 2015); in adolescents up to 18 years that had engaged with treatment following at least one episode of self-harm.

So they're asking a broad question: does doing something intended to be helpful actually help a population of kids with an episode of self-harm?

How large a sample did this end up being? Big, but not huge:

The RCTs included 3470 randomised participants, of which 1729 were randomised to a TI.

It's delightful when things are published in British journals, and they spell the word randomized differently, right?

1729 is a large sample, and I hope those reading it who are familiar with papers I've co-authored will be very impressed that I have single publications on things like deep TMS with even larger sample sizes than that number³. This meta-analysis took 29 different papers to get to a Sample Size that large! Larger sample sizes are necessary if we're looking for small effects or rare outcomes. In the case of suicidal behavior, we are very explicitly talking about a rare outcome, which requires a large sample size. The rest of the medicine routinely gets to 10,000 patient samples, and in single trials, child psychiatry has miles to go.

I want my readers to have some sense of how large and undertaking this is…here is the work the authors had to do to sift through everything published to get to these results:

It's a lot of work. Moving on. What kind of interventions were included? I'm gonna save you some terrible, but it was all some kind of psychotherapy, with an app used here and there, But to be extremely clear: not a single biological intervention with a drug or device met the criteria to be included in this metanalysis. Everything you're gonna see subsequently is a bad psychotherapeutic or digitally assisted psychotherapeutic intervention. To say it one more time:

there are no drugs that have any evidence for the reduction of suicidal behavior or death by suicide in youth.

Additionally, unless these things that we are reviewing are integrated into psychiatric hospitalization,

there is no evidence for psychiatric hospitalization, focused on medication administration, as an evidence-based intervention to save the lives of youth after self-harm.

The good news is that we could do these psychotherapeutic things in hospitals, and they might be effective.

Child Psychiatry, as a medical discipline, still matters— just like physical medicine and rehab rehabilitation, it doesn't do all the physical therapy personally. What we're about to read provides further evidence that close partnerships with our psychotherapeutic colleagues— often trained in psychology and social work—create the minimum necessary teamwork to build systems of care that have evidence to reduce suicide in youth.

Since my readers have already learned how to interpret a meta-analysis through years of relentless articles in their inbox—maybe even reading my book, Inessential Pharmacology—I am, thusly, only going to do the very high-level introduction to interpreting these data.

What is a Meta-Analysis?

A meta-analysis combines several studies, does some math, and determines the best answer for the question.

The data is presented in a forest plot, which has a crucial feature: a vertical line in the middle, which is the line of equipoise. If you're on one side of the line, it's worse, and if you're on the other side of the line, it's better. If you cross the line, it's no different than chance.

How Do I Read A Forest Plot?

Every study has a confidence interval, which shouldn't cross that vertical line in a forest plot. You're about to see a bunch of forest plots! Each study is represented as a box; the larger the box, the larger the sample. A horizontal line shows how wide the confidence interval is; then, at the bottom, there's a diamond. The diamond is the important thing. If the diamond doesn't cross the vertical line, we have a difference between chance and the interventions! Now that we are prepared to examine this study's outcome—it is really the largest of all the available studies—let's get excited.

Do therapeutic interventions change the rate at which young people die—which is a rare event—after self-harm? The answer is no, at least not with this sample size.

An important note: I'm highly dubious that, absent a much larger data set, we would see a difference. Death is a rare outcome in young people. Suicide is an even more rare outcome. Although there is an enriched chance that investigators will capture a bad outcome in a sample of children who attempted self-harm, these are, at the same time, children who attempted self-harm but had their act together enough— and enrolled in a randomized control trial. Do we imagine there's something different about a family that has the ability to enroll in a clinical trial versus a family that does not? Exactly. They're probably different than the general population of kids, and regardless, the rate of death even in suicidal youth is relatively low, much less than the timeframe captured by any of these studies. Know how you can tell? Only a few of the studies even looked at mortality because they probably didn't have the funding to follow kids out for many, many, many years to figure that out.

Do therapeutic interventions reduce the rate of self-harm in kids who have already started self-harming? I'm very pleased to report the answer is yes!

Not all treatments work; some are even harmful, but psychotherapy for kids with self-harm reduces the rate of self-harm, broadly speaking. Can we do some analysis to figure out which approaches were the most helpful? And again, thanks to this large sample on these diligence authors, the answer is yes. Some features make it more likely that we're gonna be helpful to the kids who've had an episode of self-harm. But First, maybe just any treatment is helpful? Maybe you don't need specialized treatment for kids with self-harm in order to reduce self-harm.

NO.

Treatment, as usual, did not reduce self-harm behaviors.

The next time your insurance company says we'll pay for this non-specialized service, you can send them this article and say that the evidence does not support the use of just any treatment for children with a history of self-harm, it only supports specialized treatment interventions for youth with a history of self-harm. “Just any help” isn't helpful in this population and doesn't reduce the risk of self-harm prospectively.

The authors examined whether group therapy alone was successful, and the answer was no.

You can not just stick the kid in a group. What about families? Do you need to include the family in the care of suicidal youth for it to be effective?

The survey says yes.

And, of course, the most exciting part of this paper is the following finding: for kids with a history of self-harm, specialized interventions reduce not only self-harm, which is more frequent and easier to measure— but also attempts at suicide.

You'll get out that only a few studies actually looked at this outcome, which makes the statistically significant finding all the more meaningful; even in a relatively small sample compared to the overall cohort, we still got a difference.

Child psychiatry: as a discipline, we're doing something useful, which is providing specialized psychotherapies to kids that reduce their risk of subsequent suicide attempts.

These findings are important. We now know that group alone is not good enough. We know that we need to include families in the specialized therapies for self-harming youth, and if we do both of those things, we can reduce the rate of suicide attempts. We also know that these therapies don't increase the risk of death, but we don't know, over a long enough time scale, if they reduce mortality, and that's likely because we didn't have enough numbers to be able to tell.

Now, say it with me: We need to get insurance companies to actually pay for this sort of care. Of course, there is a corollary: They may not want to pay for care that doesn't work, and honestly, I think that's a good idea. The less we pay for demonstrably ineffective things, the better.

This kind of large-scale research is complicated and takes a lot of time. The initial studies took years of research, and the subsequent studies took years more. As somebody who has been the principal investigator of a clinical trial that just couldn't enroll enough kids, given the resources we had, I know how hard this work is.

We also need better interventions. The fact that there are no biological interventions in the above analysis that even looked at reducing suicidality among self-harming youth is an embarrassment. We should be profoundly ashamed that we have not created any brain-based interventions that met the bar for reducing self-harm and suicide, and we should diligently work to do so. This also will take funding. Of note, it's not that we haven't proved they don't work, per se; we just haven't attempted to demonstrate that they do. None of the biological or drug-based interventions they looked for met the inclusion criteria for the analysis.

We need to do better, but we know more now than we did in the past. We have interventions, like dialectical behavioral therapy, mentalization-based treatment, and the like, that, when they include family therapy and combined approaches with group and individual therapy, are more effective than treatment as usual in reducing the rate of suicide attempts and self-harm in young people who already hurt themselves on purpose before.

There is hope for kids who have engaged in self-harm.

And there is more work to do. Thanks for reading the newsletter. Please spread the word, and if you're asked to do a prior authorization with Insurance, educate your colleague on the other end of that peer review about this most recent data.

1. Griep, S. K., & MacKinnon, D. F. (2022). Does nonsuicidal self-injury predict later suicidal attempts? A review of studies. Archives of Suicide Research, 26(2), 428-446.

2. Mughal, F., Young, P., Stahl, D., Asarnow, J. R., & Ougrin, D. Mortality in adolescents after therapeutic intervention for self-harm: A systematic review and meta-analysis. JCPP Advances, e12302. https://doi.org/10.1002/jcv2.12302

3. Tendler, A., Goerigk, S., Zibman, S., Ouaknine, S., Harmelech, T., Pell, G. S., Zangen, A., Harvey, S. A., Grammer, G., Stehberg, J., Adefolarin, O., Muir, O., MacMillan, C., Ghelber, D., Duffy, W., Mania, I., Faruqui, Z., Munasifi, F., Antin, T., . . . Roth, Y. (2023). Deep TMS H1 Coil treatment for depression: Results from a large post marketing data analysis. Psychiatry Research, 324, 115179. https://doi.org/10.1016/j.psychres.2023.115179