What protects a teenager from self-harm depends on how old they were

Research on childhood self-harm and suicidality has focused overwhelmingly on risk factors — what makes an outcome more likely. A cohort study published in The Lancet Psychiatry on September 8 inverts the question and asks what protects, and when [s1].

Its most useful result is a negative one. No variable showed consistent associations across ages and across the four outcomes examined [s1]. Protection, in this data, is not a stable property of a child's circumstances. It moves.

What was done

The analysis used four predictor waves and one outcome wave of the Longitudinal Study of Australian Children, a cohort born in 2003-04 and surveyed every two years [s1]. Predictors were measured across ages 6 to 13; outcomes were measured at ages 14 to 15 [s1].

The analytical sample was 3,044 children: 1,570 (51.6%) male and 1,474 (48.4%) female [s1]. Almost all — 3,034 (99.7%) — were born in Australia; 858 (28.2%) had at least one parent born overseas, and 76 (2.5%) were Aboriginal or Torres Strait Islander [s1].

Predictors were drawn from the Nest framework, an existing multidimensional participatory child wellbeing framework with domains covering valued, loved, and safe; material basics; healthy; learning; and participating [s1]. Each domain was operationalised through multiple subdomains, which were the actual predictors [s1].

The four outcomes were self-harm ideation, self-harm behaviour, suicidal ideation, and suicide attempt in the past 12 months [s1]. People with related lived experience were involved in the study design and implementation [s1].

The findings

Across all ages, most children had most of their wellbeing needs met on individual subdomains [s1]. Against that baseline, the strongest associations were:

• Self-harm ideation: a good state of mind (odds ratio 0.55, 95% CI 0.42 to 0.71) and sense of belonging (0.60, 0.46 to 0.77), both measured at age 12-13 [s1].
• Self-harm behaviour: relationships with parents (0.45, 0.21 to 0.97) and involvement in the community (0.53, 0.32 to 0.89), both at age 10-11 [s1].
• Suicidal ideation: sense of belonging at 12-13 (0.57, 0.42 to 0.78) and school satisfaction at 10-11 (0.32, 0.12 to 0.89) [s1].
• Suicide attempts: school safety at age 6-7 (0.59, 0.39 to 0.89) and school satisfaction at 10-11 (0.47, 0.26 to 0.85) [s1].

The authors' reading is that protective factors change as children develop, and that the timing of policy interventions therefore needs careful consideration — offering school safety interventions in primary school years as an example of likely long-term value [s1].

Reading the intervals

Several of the headline estimates have confidence intervals that reach close to 1. Relationships with parents for self-harm behaviour runs 0.21 to 0.97; school satisfaction for suicidal ideation runs 0.12 to 0.89 [s1]. Those are statistically significant and imprecise, which is what happens when relatively rare outcomes are modelled in a cohort of 3,044.

More importantly, the study examined many subdomains against four outcomes at four timepoints. That is a large number of comparisons, and the ones that emerge strongest are selected from among them. The authors' own framing — that no variable was consistent across ages and outcomes [s1] — is the honest summary, and it also means the specific age-outcome pairings above should be read as hypotheses rather than established windows.

Where it agrees with the wider literature

A systematic review and meta-analysis published in the same journal in August pooled prospective longitudinal studies of family dynamics and self-harm and suicidality in children and adolescents under 20, searching from database inception to May 20, 2025 without language restrictions [s2]. It was designed to test whether negative family dynamics confer prospective risk and positive family dynamics confer protection — a question the field had not resolved [s2].

The cohort study's parent-relationship finding sits inside that question. One cohort at one timepoint does not settle it, but the two papers together indicate that the protective side of family dynamics is now being studied on its own terms rather than as the absence of risk.

Limits

The cohort is Australian, overwhelmingly Australian-born, and includes 76 Aboriginal and Torres Strait Islander children [s1] — too few to say anything about a population with markedly different self-harm and suicide rates. The findings should not be read as applying to that group.

Outcomes were self-reported at ages 14 to 15 over a 12-month recall window [s1]. Predictors were measured before outcomes, which establishes temporal order, but observational association across that gap does not establish that changing a predictor would change the outcome.

The children were born in 2003-04, so the outcome wave predates the most recent period of change in adolescent social environments. Whether the same protective factors operate for cohorts growing up later is untested here.

What to watch

Replication in cohorts from other countries and later birth years; whether interventions that target sense of belonging or school satisfaction at the ages this study flags produce measurable changes in self-harm outcomes; and whether the age-specificity survives in samples large enough to estimate it precisely.

This article describes observational cohort research and a meta-analysis, and is informational only. It is not medical advice.

If you or someone you know is struggling, contact a local crisis line or health service.

Sources

• [s1] Campbell PD, Proulx JDE, Sollis K, Cruwys T, Calear AL, Rathbone J, Sheanoda V, Reynolds KJ, Early-life protective factors for adolescent self-harm and suicidality: a longitudinal cohort study in Australia, The Lancet Psychiatry, 2025;12:746-757, published online 2025-09-08.
• [s2] Hammond NG, Semchishen SN, Geoffroy MC, et al., Family dynamics and self-harm and suicidality in children and adolescents: a systematic review and meta-analysis, The Lancet Psychiatry, 2025;12:660-672, published online 2025-08-12.