At a time when the UK has recently appointed a minister for suicide and every local area is expected to a have a suicide prevention plan, there is a great need for us to understand what works when it comes to suicide prevention. In the course of my work, I regularly talk to commissioners who are keen to invest in interventions that are safe and effective, but they often don’t know where to start. This is particularly true when it comes to working with young people as, where it does exist, the evidence base in terms of suicide prevention largely draws on adult cohorts. So I was pretty excited when I saw Robinson et al’s paper appear in press.
Given its title: ‘What Works in Youth Suicide Prevention? A Systematic Review and Meta-Analysis’ you could be forgiven for assuming this paper would tell the reader what works in youth suicide prevention. It doesn’t.
It does, however, provide a comprehensive review of the data to date and raise some important questions and suggestions for moving the field forwards. However, at a time when suicide prevention is a hot topic and commissioners are keen to buy into interventions that are safe and effective, some more concrete answers would have been welcomed. I’ve no doubt the paper authors would echo this sentiment; they conducted a very thorough review and found the evidence base wanting; they get ten out of ten for effort, but points deducted for the click-bait title.
Robinson et al conducted a systematic review and, where possible a meta-analysis of studies designed to lead to a reduction in suicide related behaviours in young people.
They searched Medline, PsycINFO, and EMBASE from January 1 1990 to September 21, 2017 using keywords relevant to suicide-related behaviour, intervention type and youth.
Studies were included if they:
- Were written in English
- Assessed a suicide-related outcome
- Targeted 12 to 25 year olds
- Were published in or referenced by a peer reviewed journal.
Studies were excluded if they:
- Were not primarily concerned with effecting suicide related outcomes
- Did not measure or report on a suicide related outcome
- Data related to the target age group could not be disaggregated from a larger sample
- Employed a non-experimental design.
Data were extracted from the resulting studies and classified and synthesised by two researchers to enable comparison of a diverse range of studies.
A whopping 34,463 articles were retrieved via the initial search, of which 105 articles corresponding to 99 unique research studies met the inclusion criteria. Here’s a snapshot of the makeup of the studies:
- 52% were conducted in clinical settings
- 31% were conducted in educational or workplace settings
- 16% were conducted in community settings
- 67% tested indicated interventions
- 17% were universal
- 49% were randomised controlled trials (RCTs)
- The RCTs included 64% of the clinical setting studies, 49% of the education/workplace studies and none of the community setting studies
- Half of the studies were conducted in the US
- 12% were UK based and 11% Australian based
- 2% of studies were conducted in low to middle income countries
- The number of studies doubled in 2005-2017 compared to 1990-2004.
Where possible, the data from the studies was synthesised to allow for meta-analysis so the researchers could draw conclusions about what works in youth suicide prevention across a range of interventions. Due to the diverse nature of the studies, the results are very granular and can be viewed in full in the original paper, but here are some headlines. Please note that the original paper (PDF) contains a LOT of detail, so please refer back to it to see beyond the headlines. I have not shared any data on individual interventions, which is all in the original paper.
Clinical studies: randomised controlled trials (RCTs)
Studies examined the impact of a range of interventions, including individual and group cognitive behavioural therapy (CBT), dialectical behavioural therapy (DBT), family therapy, and brief contact interventions on self-harm, suicidal ideation and suicide.
Across 32 studies included in this meta-analysis; compared to controls, there was:
- No evidence of any intervention effect on self-harm post intervention
- No evidence that the type of intervention modified the size of the treatment effect post-intervention
- At follow up, a reduction in the proportion of people who had received an intervention who went on to have a repeat self-harm episode (k = 16, RR = 0·83, 95% CI 0·70 to 0·99, I2 = 40·9%)
- When measured dichotomously, no evidence of any effect of intervention on the proportion of people who experienced suicidal ideation post-intervention or at follow up
- When measured continuously, strong evidence of a small effect of the intervention on suicidal ideation post-intervention (k = 15, SMD = −0·28, 95% CI −0·48 to −0·08, I2 = 76·3%). The effect was smaller at follow-up (k = 11, SMD = −0·18, 95% CI −0·34 to −0·02, I2 = 41·1%)
- 43% of studies had a dropout rate of 15% or higher.
Clinical studies: other study designs
All nineteen studies in this category tested therapeutic interventions:
- 5 studies tested CBT:
- two of these reported reductions in suicide-related behaviour
- Six studies tested DBT:
- five of these reported reductions in suicide-related behaviour
- four reported reductions in suicidal ideation
- Three studies tested family-based interventions:
- two reported reductions in suicidal ideation
- one reported a reduction in suicide attempts
- 63% of the nineteen studies had a dropout rate of more than 15% and all but one study was underpowered
- Only one study reported that outcome assessors were blind to treatment allocation.
Studies conducted in Educational and Workplace settings (RCTs)
Fifteen of the 31 studies conducted in educational and workplace settings were RCTs. They were all either educational or therapeutic in nature, including four which tested internet-based interventions. Eleven of the studies reported data amenable to meta-analysis. Across these, compared to controls, there was:
- Evidence of an intervention effect on self-harm when measured dichotomously:
- at post-intervention (k = 3, RR = 0·31, 95% CI 0·15 to 0·61, I2 = 0%)
- at follow-up (k = 3, RR = 0·63, 95% CI 0·42 to 0·96, I2 = 0%)
- Compared to control, the effect on suicide ideation when measured dichotomously was not significant:
- at post-intervention (k = 1, RR = 0·76, 95% CI 0·50 to 1·16)
- at follow-up (k = 2 (4 intervention arms), RR = 0·72, 95% CI 0·51 to 1·03, I2 = 0%)
- There was stronger evidence of an effect of the intervention on suicidal ideation when measured continuously:
- at post-intervention (k = 7, SMD = −0·41, 95% CI −0·57 to −0·24, I2 = 15·2%)
- but at follow-up, the effect was no longer significant.
The current paper is an important summary of the field of youth suicide prevention interventions to date. The key thing it tells us is that when it comes to suicide prevention in young people, we don’t really know what works yet and that there is a clear need for us to improve the quality of research being conducted.
Strengths and limitations
This is a huge piece of work that drew on data from ninety-nine studies. However, due to a lack of homogeneity and poor data quality, it was difficult to come to clear answers to the overarching research question of ‘what works in youth suicide prevention?’
A great summary
The current study does a fantastic job of summarising the field to date and will be an important starting point for future researchers who should look to build upon the studies herein. It is clear that there is a lack of coordination between researchers resulting in many piecemeal studies. It would be fantastic to see better national and international coordination between researchers to see some of these important questions asked and answered more collaboratively and coherently (a girl can dream..).
Are we measuring the right thing?
I can’t review a paper about the impact of suicide prevention work without raising this old chestnut. The stated aim of this meta-analysis is to review what works in suicide prevention, but the main outcomes measured were self-harm and suicidal ideation (with a few notable exceptions which measured suicide attempts). What, if anything, do quasi measures tell us about changes in suicides or suicide attempts? Both the self-harm to suicide pathway and the suicidal ideation to action pathways are young areas of research with much for us to learn.
It is difficult to measure suicide behaviours; but the research tells us that asking people about suicide is safe. Surely, if the aim of a programme is to reduce suicides and suicide attempts, we must come to a consensus about how to measure this appropriately and move away from quasi measures that may not fit the bill. The current study showed suicide prevention is a fast-growing area of research; as researchers, we need to ensure we are asking the right questions.
Significant gaps and interesting decisions were highlighted
Pooled together, the data about youth suicide prevention showed that whilst young people are considered a higher risk group when it comes to suicide, many interventions trialled with young people have not been developed specifically with them in mind. Is it appropriate to adapt adult interventions and expect positive results, or do we need to recalibrate our thinking?
The study also showed an absence of high-quality interventions aimed at GPs (general practitioners), although they are often the first port of call for young people at risk and they (GPs) have highlighted a need for training.
The potential impact of online interventions was also highlighted by Robinson et al who suggested “we are likely missing crucial opportunities for intervention, such as delivery via online platforms. Future research should adapt known effective interventions for young people, and for delivery online.”
Important questions raised about sample size
Pooling the data in this way highlighted an important question for researchers; many studies were too small and under-powered, whereas very large studies could have a small but statistically significant impact due perhaps to the statistical power afforded by a large sample size. It is worth noting too that when pooled for meta-analysis, large studies may skew the overall data; in this instance the Hassanian-Moghaddam (2011) trial with 2,133 participants was implicated as driving effect sizes, though no steps were taken to recalculate effect sizes minus the implicated trial. So what is the optimal trial size? Is there a balance to be struck between too much and too little statistical power?
Despite the increase in quantity of research, Robinson et al point to the paucity of high-quality youth suicide prevention intervention studies. The current meta-analysis should act as a wakeup call to researchers and commissioners (the field of youth suicide prevention research needs a shake-up) lives depend on it.
Conflicts of interest
Robinson J, Bailey E, Witt K, Stefanac N, Milner A, Currier D, Pirkis J, Condron P, Hetrick S. (2018) What Works in Youth Suicide Prevention? A Systematic Review and Meta-Analysis (PDF). EClinicalMedicine Vol 4, p 52-91, Oct 01, 2018 https://doi.org/10.1016/j.eclinm.2018.10.004
Hassanian-Moghaddam H, Sarjami S, Kolahi AA, Carter GL. Postcards in Persia: randomised controlled trial to reduce suicidal behaviours 12 months after hospital-treated self-poisoning. Br J Psychiatry 2011;198(4):309–16.