Violence occurs at multiple ecological levels and can harm mental health. However, studies of adolescents’ experience of violence have often ignored the community context of violence, and vice versa. We examined how personal experience of severe physical violence and living in areas with high levels of neighbourhood disorder during adolescence combine to associate with mental health at the transition to adulthood and which factors mitigate this.
Data were from the Environmental Risk Longitudinal Twin Study, a nationally representative birth cohort of 2232 British twins. Participants’ experience of severe physical violence during adolescence and past-year symptoms of psychiatric disorder were assessed via interviews at age 18. Neighbourhood disorder was reported by residents when participants were aged 13–14. Potential protective factors of maternal warmth, sibling warmth, IQ, and family socio-economic status were assessed during childhood, and perceived social support at age 18.
Personal experience of severe physical violence during adolescence was associated with elevated odds of age-18 psychiatric disorder regardless of neighbourhood disorder exposure. Cumulative effects of exposure to both were evident for internalising and thought disorder, but not externalising disorder. For adolescents exposed to severe physical violence only, higher levels of perceived social support (including from family and friends) were associated with lower odds of psychiatric disorder. For those who also lived in areas with high neighbourhood disorder, only family support mitigated their risk.
Increasing support or boosting adolescents’ perceptions of their existing support network may be effective in promoting their mental health following violence exposure.
An estimated 1.6 million people worldwide lose their lives to violence each year, while many more suffer life-altering consequences including mental health problems . Violence is a complex phenomenon; it can take many different forms and occur at different—as well as multiple—ecological levels. At the inter-personal level, violence may be experienced personally (e.g., sexual and physical victimisation) through interactions with peers, family members, and others in the wider environment. At the community level, individuals may live in neighbourhoods where there is violence that is not directed at them or witnessed personally. This includes physical and social signs of violence, threat, and danger, collectively referred to as neighbourhood disorder .
Adolescence is a peak age for experiencing several types of violence including physical assault, sexual victimisation, and family violence . It is also when youth begin to spend more time unsupervised in the community, potentially exposing them to a wider range of violence and facilitating a greater awareness of neighbourhood disorder. At the same time, adolescence and the transition to adulthood is a high-risk period for the onset of a variety of common psychiatric disorders [4, 5] that for many will signal the start of recurrent mental illness throughout adulthood . Indeed, around 75% of adult mental health disorders will have onset by the age of 18 . A better understanding of how personal experiences of violence and high levels of neighbourhood disorder during adolescence are associated with the development of mental health diagnoses could inform early targeted intervention and prevention.
Evidence consistently shows that personal experiences of violence, especially during childhood and adolescence, elevate the risk for psychiatric disorders [8, 9] including internalising disorders such as depression and anxiety [10, 11], externalising disorders such as attention-deficit hyperactivity disorder (ADHD) , and psychosis [13, 14]. Separately, studies of community-level violence have also shown associations with residents’ mental health problems [15,16,17]. Although research specifically with adolescents is scarce, links between neighbourhood disorder and adolescent psychological distress  and psychotic experiences  have been evidenced.
However, different types of violence often converge, and therefore, it is also important to consider the combined impact of multiple forms of exposure. From the cumulative stress perspective, the greater the number of risk factors an individual experiences, the greater their risk of suffering from mental health problems [20, 21]. Alternatively, exposure to violence across multiple settings may normalise or desensitise individuals such that one exposure reduces the effects of the other . Individuals who experience one type of violence also commonly experience another type [23, 24] and this poly-victimisation confers even greater risk for mental health problems [9, 25,26,27]. Violence exposure may also converge across ecological levels; for example, people who live in neighbourhoods with high levels of disorder are more likely to have personal experience of crime [15, 28]. However, unlike poly-victimisation, there has been limited investigation of how multi-level violence exposure during adolescence combines to impact mental health. One study has shown cumulative effects of adverse living conditions (including neighbourhood disorder) and crime victimisation on adolescent psychotic experiences . However, studies of community-level violence and mental health have typically ignored  or been conflated  with personal experiences of violence, while studies of inter-personal violence often do not consider the social context of violence in which these may take place. Thus, understanding how these two levels of violence exposure operate in the context of one another in relation to mental health represents a significant gap in the literature.
Fortunately, not everyone who is exposed to violence develops mental health problems. For example, in a UK cohort, 40% of young people did not have a psychiatric disorder at age 18 despite experiencing severe childhood victimisation . Understanding what factors protect against poor mental health among adolescents exposed to violence is necessary to inform interventions at the individual, family, and community levels to mitigate its effects. Existing research has primarily focused on factors that are protective following violence that is personally experienced , but it remains unknown what factors are protective for adolescents who are also exposed to violence at the community level.
The present study addresses these knowledge gaps using data from the Environmental Risk (E-Risk) Longitudinal Twin Study. We investigate: (i) how the prevalence of psychiatric disorder at age-18 compares between adolescents with personal experience of severe physical violence, those who lived in neighbourhoods with high levels of disorder, and those with no such exposure during adolescence; (ii) whether there is a cumulative effect of having both personal experience of severe physical violence and living in a neighbourhood with high levels of disorder during adolescence on age-18 psychiatric disorder; and (iii) whether supportive relationships (maternal warmth, sibling warmth, and perceived social support), higher IQ, and higher family socio-economic status (SES) protect against the development of psychiatric disorder within those violence-exposed groups who are at elevated risk.
The putative protective factors that we investigate were identified during focus group discussions with a group of young people with lived experience of violence, abuse, and mental health problems (see Latham et al.  for details of the focus groups) and then matched to measures available in the E-Risk Study. These factors are also consistent with theoretical accounts of resilience that highlight physical, psychological, and social resources in the environment that can help individuals to sustain their wellbeing in the face of adverse circumstances  as well as empirical findings [32, 35, 36]. Involving individuals with lived experience in mental health research ensures that it is relevant, inclusive, and high quality [37, 38]. Accordingly, peer researchers also partnered with the academic research team to help interpret and present the study findings.
Participants were members of the Environmental Risk (E-Risk) Longitudinal Twin Study, which tracks the development of a nationally representative birth cohort of 2232 British children. The sample was drawn from a larger birth register of twins born in England and Wales in 1994–1995 . Full details about the sample are reported elsewhere , and in Supplementary Material. Briefly, the E-Risk sample was constructed in 1999–2000 when 1116 families (93% of those eligible) with same-sex 5-year-old twins participated in home-visit assessments. Sex was evenly distributed within zygosity (49% male).
Follow-up home-visits were conducted when the children were aged 7, 10, 12, and 18 years (participation rates were 98%, 96%, 96%, and 93%, respectively). There were 2066 E-Risk participants (47% male) who were assessed at age 18. The average age of the participants at the time of the assessment was 18.4 years (SD = 0.36); all interviews were conducted after the 18th birthday. There were no differences between those who did and did not take part at age 18 in terms of socio-economic status (SES) assessed when the cohort was initially defined (χ2 = 0.86, p = 0.65), age-5 IQ scores (t = 0.98, p = 0.33), age-5 behavioural (t = 0.40, p = 0.69) or emotional (t = 0.41, p = 0.68) problems, or childhood poly-victimisation (z = 0.51, p = 0.61).
The Joint South London and Maudsley and the Institute of Psychiatry Research Ethics Committee approved each phase of the study. Parents gave informed consent and participants gave assent between 5 and 12 years and then informed consent at age 18.
Personal experience of severe physical violence during adolescence
At age 18, participants were interviewed face-to-face about exposure to a range of adverse experiences between 12 and 18 years using the Juvenile Victimisation Questionnaire (JVQ) [41, 42] adapted as a clinical interview (see Supplementary Material and Fisher et al.  for full details). All information from the JVQ interview was compiled into victimisation dossiers. Using these dossiers, an expert in victimology (HLF) and three other members of the E-Risk team evaluated whether each participant was exposed to any physical violence, whether in the family, by peers, or by people in the wider environment. This “any physical violence” exposure variable was rated on a 6-point scale: 0 = not exposed, then 1–5 for increasing levels of severity (see Supplementary Table S1 for coding detail). The anchor points for these ratings were adapted from the coding system used for the Childhood Experience of Care and Abuse interview (CECA) [43, 44]. Consistent with previous studies using the CECA [43, 45], to index the most severe experiences of violence we dichotomised this variable such that those scoring at the upper end of the severity scale (4–5) were identified as having personal experience of severe violence (coded 1: 24.3% of participants, N = 502).
Neighbourhood disorder during adolescence
Neighbourhood disorder (14 items) was assessed via a postal survey sent to residents living alongside E-Risk families when participants were aged 13–14. Survey respondents, who were typically living on the same street or within the same apartment block as the Study participants, reported on various characteristics of their immediate neighbourhood, including levels of neighbourhood disorder. Surveys were returned by an average of 5.18 (SD = 2.73) respondents per neighbourhood, and there were at least 2 responses for 95% of neighbourhoods (N = 5 601 respondents). Residents were asked whether certain problems affected their neighbourhood, including muggings, assaults, vandalism, graffiti, and deliberate damage to property. Items were coded 0 (‘no, not a problem’), 1 (‘yes, somewhat of a problem’) or 2 (‘yes, a big problem’). Consistent with existing analyses of the E-Risk Study [28, 46,47,48,49], items were averaged to create a summary score. Scores for each E-Risk family were then created by averaging the summary scores of respondents within that family’s neighbourhood. The resulting variable approached normal distribution across the full potential range (M = 0.49, SD = 0.34, range = 0–1.93). We indexed high levels of neighbourhood disorder as those participants with above average neighbourhood disorder scores (coded 1: 42.2% of participants, N = 908).
Mental health problems at age 18
At 18 years of age, participants were privately interviewed about past-year symptoms of mental disorder. Ten disorder diagnoses were organised into three domains (externalising, internalising, and thought disorders) based on a reliable latent factor structure for psychopathology previously identified within the E-Risk Study . Full information on individual diagnoses is available in Supplementary Material. Participants were classified as having a research diagnosis of ‘externalising disorder’ when they met diagnostic criteria for ADHD, conduct disorder, alcohol dependence, cannabis dependence, or tobacco dependence. Participants were classified as having a research diagnosis of ‘internalising disorder’ when they met diagnostic criteria for general anxiety disorder, major depressive disorder, post-traumatic stress disorder, or presented at least 2 of 5 eating disorder symptoms from an established screening tool indicating a possible case of anorexia nervosa or bulimia nervosa . Finally, a ‘thought disorder’ classification was based on the definite presence of at least one of seven psychotic symptoms, centred on delusions and hallucinations. A total of 280 E-Risk participants met criteria for more than one classification. Thus, from these three domain-specific classifications, an overall binary outcome for ‘any psychiatric disorder’ was created, denoting the presence of any externalising, internalising, or thought disorder (coded 1), or the absence of all three (0). In the E-Risk study, 47.5% (N = 973) of participants met diagnostic criteria for any psychiatric disorder at age 18. Specifically, 31.9% (N = 656) met criteria for externalising disorder, 28.5% (N = 585) for internalising disorder, and 2.9% (N = 59) for thought disorder.
Maternal warmth during childhood
Maternal warmth during childhood was assessed using procedures adapted from the Five-Minute Speech Sample method [51, 52]. When children were aged 5 and 10, mothers were asked to speak for 5 minutes about each of the children separately. The speech samples were audiotaped and coded by two independent raters, who had good inter-rater reliability (r = 0.90). The warmth expressed by the mother in their interview about the child was coded on a six-point scale from no warmth (0; complete absence of warmth) to high warmth (5; definite warmth, enthusiasm, interest in, and enjoyment of the child). To capture high levels of maternal warmth during childhood, a binary variable was created that indexed the presence of high maternal warmth (i.e., a score of 4–5) at age 5 and/or age 10 (coded 1: 71.4% of participants, N = 1466).
Sibling warmth during childhood
Sibling warmth during childhood was assessed by asking mothers a series of questions about the quality of their twins’ relationship with one another when the children were aged 7 and 10 . Mothers responded on a three-point scale ranging from 0 ‘no’ to 2 ‘yes’ to six questions (e.g., “do your twins love each other”, “do both your twins do nice things for each other”). Internal consistency at age 7 was α = 0.77 and at age 10 was α = 0.80. As age-7 and age-10 scores were highly correlated (r = 0.57, p < 0.001), these were summed to create a single composite score (M = 19.92, SD = 3.35).
Perceived social support
Perceived Social Support was assessed at age 18 via self-reports using the multidimensional scale of perceived social support (MSPSS), which assesses individuals' access to supportive relationships with family, friends, and significant others . The 12 items comprise statements such as “There is a special person who is around when I am in need” and “I can count on my friends when things go wrong”. Participants rated these statements as “not true” (0), “somewhat true” (1), or “very true” (2). We summed scores to produce an overall social support scale with higher scores reflecting greater social support (M = 20.71, SD = 4.35). In addition, the family and friend sub-scales were utilised separately to examine whether social support from either family (M = 6.98, SD = 1.78) or friends (M = 6.71, SD = 2.01) was specifically protective.
Intelligence quotient (IQ)
Intelligence Quotient (IQ) was tested at age 12 using a short version of the Wechsler Intelligence Scale for Children-Revised (WISC-R)  which comprised three subtests (Matrix Reasoning, Information, and Digit Span). We converted the scores into an IQ score according to Sattler  and then standardised to a mean of 100 and standard deviation of 15.
Family socio-economic status (SES)
Family socio-economic status (SES) was measured at age 5 using a standardised composite of parental income (i.e., total household income), education (i.e., highest parent qualification), and occupation (i.e., highest parent occupation). These three SES indicators were highly correlated (r = 0.57–0.67) and loaded significantly onto one latent factor . The population-wide distribution of this latent factor was then divided into tertiles (i.e., low-, medium-, and high-SES).
Individual- and family-level covariates
Individual-level covariates included biological sex and participants’ history of emotional and behavioural problems during childhood, including attention-deficit hyperactivity disorder (ADHD) diagnosis, conduct disorder diagnosis, symptoms of depression and anxiety, self-harm and suicide attempts, and psychotic symptoms. Family-level covariates included family SES and family history of psychopathology. Measurement details for all covariates are provided in Supplementary Material.
Analyses were conducted using Stata 15. We accounted for the non-independence of our twin observations in all analyses using the Huber–White variance estimator . Analyses proceeded in three steps. First, we used a series of logistic regression models to examine the separate associations of (i) personal experience of severe physical violence and (ii) high levels of neighbourhood disorder during adolescence with (i) any psychiatric disorder, and then (ii) externalising disorder, (iii) internalising disorder, and (iv) thought disorder at age 18. We also conducted two sensitivity analyses: we examined the association of different types of severe physical violence (i.e., crime victimisation, maltreatment, sexual victimisation, and family violence) with the four age-18 mental health outcomes. We also tested associations using neighbourhood disorder categorised at different thresholds—above the median, above the 75th percentile, and using the full-scale (continuous) measure of neighbourhood disorder.
Second, to investigate the potential cumulative and interactive effects of personal experience of severe physical violence and high neighbourhood disorder, we created a 4-level categorical variable to reflect the four possible combinations of exposure: no exposure (coded 0); personal severe physical violence only (1); high neighbourhood disorder only (2); both personal severe physical violence and high neighbourhood disorder (3). We used Interaction Contrast Ratios (ICRs) to investigate whether personal experience of severe physical violence and high neighbourhood disorder during adolescence combined synergistically to increase the odds of psychiatric disorder at age 18 (indicated by a departure from additivity [59, 60]). This approach uses odds ratios (OR) derived from logistic regression models to estimate the relative excess risk as a result of synergy (i.e., ICR = ORexposure to both–ORpersonal severe physical violence only–ORhigh neighbourhood disorder only + 1).
Third, we used logistic regression to examine whether supportive relationships (maternal warmth, sibling warmth, and perceived social support including family and friend sub-scales), higher IQ, or higher family SES were associated with reduced odds of psychiatric disorder within those violence exposure groups found to be at risk in step 2. We also tested statistical interactions between significant protective factors and adolescent violence exposure in the whole E-Risk sample using logistic regression.
To test the robustness of the associations, all models were adjusted for sex, family history of psychopathology, and childhood emotional and behavioural problems. Family SES was also included as a covariate in models in step 1 and 2; however, because we investigated family SES as a potential protective factor, models in step 3 were not adjusted for this. Because we tested four mental health outcomes in steps 1 and 2, and seven potential protective factors in step 3, we controlled the false discovery rate (FDR) by applying the Benjamini–Hochberg method  to each collection of statistical tests. All p values are presented in their uncorrected form with an asterisk indicating that the p value remained significant after FDR correction. Missing data, which was minimal, were predominantly due to participants missing mental health and personal experience of physical violence data due to not participating in the E-Risk Study follow-up at age 18 (see “Sample” and “Measures” descriptions). This was not associated with living in areas with high levels of neighbourhood disorder (OR = 0.99, p = 0.967, 95% CI = 0.63–1.55) and we therefore analysed complete cases.
Is personal experience of severe physical violence during adolescence associated with mental health problems at age 18?
Table 1 shows the associations between personal experience of severe physical violence during adolescence and mental health problems at age 18. Having personal experience of severe physical violence was associated with significantly elevated odds of meeting criteria for any psychiatric disorder including externalising, internalising, and thought disorders at age 18. These odds remained significantly elevated after adjusting for covariates. Examination of different types of personal severe physical violence (i.e., crime victimisation, maltreatment, sexual victimisation, and family violence) also showed elevated odds for any psychiatric disorder, externalising disorder, internalising disorder, and thought disorder (see Supplementary Table S2; note these are similar to findings using dimensional measures of mental health in the E-Risk Study; see Schaefer et al. ).
Is living in a neighbourhood with high levels of disorder during adolescence associated with mental health problems at age 18?
Table 2 shows the associations between high (i.e., above mean) levels of neighbourhood disorder and mental health problems at age 18. Living in neighbourhoods with high levels of disorder during adolescence was associated with significantly elevated odds of meeting criteria for any psychiatric disorder at age 18. This association held after adjusting for covariates. The elevated odds of externalising, internalising, and thought disorders were no longer statistically significant after adjusting for covariates; however, the effect sizes for internalising and thought disorders were not attenuated. Sensitivity analyses using neighbourhood disorder as (i) a continuous variable or dichotomised at the (ii) median and (iii) 75th percentile revealed a similar pattern of associations (see Supplementary Table S3).
Is there a cumulative effect of having both personal experience of severe physical violence and living in a neighbourhood with high levels of disorder during adolescence on mental health at age 18?
Of the 502 E-Risk participants with personal experience of severe physical violence in adolescence, half (51.4%, N = 258) also lived in neighbourhoods with high levels of disorder. Table 3 shows the prevalence of age-18 psychiatric disorder according to adolescents’ exposure to personal severe physical violence and/or neighbourhood disorder. When both personal experience of severe physical violence and high levels of neighbourhood disorder were considered together, there was evidence that those exposed to both had the highest odds of meeting criteria for any psychiatric disorder at age 18 (Fig. 1, panel A). A similar pattern was evident for internalising and thought disorders—these outcomes were associated most strongly with exposure to both personal experience of severe physical violence and high neighbourhood disorder (Fig. 1, panels C, D). The higher odds were particularly notable for thought disorder. In contrast, the odds of externalising disorder were comparable for those adolescents with only personal experience of severe physical violence and those who also lived in high disorder neighbourhoods (Fig. 1, panel B). Adolescents who lived in neighbourhoods with high levels of disorder but did not have personal experience of severe physical violence were no more likely to meet criteria for a psychiatric disorder at age 18 than the non-exposed group.
Interaction contrast ratios for all mental health outcomes were non-significant showing that the combined effect of exposure to personal severe physical violence and high neighbourhood disorder was not significantly different to their summed effect.
Do supportive relationships, higher IQ, or higher family SES protect against mental health problems for those violence-exposed groups who are at risk?
Having established that adolescents’ personal experience of severe physical violence—with or without high levels of neighbourhood disorder—is associated with elevated odds of mental health problems across externalising, internalising, and thought disorders, we focus here just on the overarching ‘any psychiatric disorder’ outcome. Despite their elevated risk, in the E-Risk sample, just over 30% (N = 70) of adolescents who personally experienced only severe physical violence and 23% (N = 59) of those exposed to both severe physical violence and high neighbourhood disorder did not meet diagnostic criteria for a psychiatric disorder at age 18. We therefore examined whether supportive relationships, higher IQ, and higher family SES were operating as protective factors in these two violence-exposed subsamples (see Supplementary Tables S4 and S5 for descriptive statistics).
The results (Table 4) show that having family support at age 18 was associated with lower odds of any psychiatric disorder in adolescents exposed only to severe physical violence and those who also lived with high levels of neighbourhood disorder (though the latter did not remain significant after correction for FDR). For those with personal experience of only severe physical violence, higher overall levels of perceived social support and support from friends specifically were also associated with lower odds of any psychiatric disorder. Interestingly, the results showed that higher IQ, sibling and maternal warmth during childhood, and family SES were not significantly protective against age-18 psychiatric disorder in the violence-exposed subsamples.
Next, we tested for interactions between perceived social support, including family and friend sub-scales and adolescent violence exposure. None of these interactions were statistically significant (all p’s > 0.05, see Supplementary Table S6).
This study examined the association of adolescent violence exposure at the inter-personal and community level with mental health at the transition to adulthood. We found elevated odds for meeting diagnostic criteria for any psychiatric disorder (including externalising, internalising, and thought disorders) for adolescents with personal experience of severe physical violence. We also found evidence of a cumulative association with internalising and thought disorders for adolescents with personal experience of severe physical violence who also lived in neighbourhoods with high levels of disorder. Higher levels of perceived support (including from family and friends) at age 18 were associated with a reduced likelihood of psychiatric disorder following personal experiences of severe physical violence whereas only perceived support from family was related to reduced odds for those who were additionally exposed to neighbourhood disorder (though this association was not statistically significant). These results hint at a protective effect; however, perhaps, due to a lack of statistical power, interactions with violence exposure were not statistically significant.
Our finding that adolescents who personally experienced violence and lived in neighbourhoods with high levels of disorder had the greatest odds of internalising and thought disorders is consistent with the cumulative stress hypothesis and existing research on poly-victimisation . Interestingly, there was no cumulative association with externalising disorder—living in an area with high neighbourhood disorder during adolescence did not contribute any additional risk compared to having only personal experience of violence. This is in line with previous findings by Meltzer et al.  that adolescents’ feeling of safety in their neighbourhood was related to emotional disorders but not conduct disorder, suggesting that neighbourhood disorder may be differentially associated with internalising and externalising problems. We speculate that living in dangerous or threatening communities may promote maladaptive cognitive styles such as biased threat perception and paranoia that are implicated in disorders such as anxiety and psychosis [63, 64]. However, future studies that examine potential mechanisms and qualitative studies to better understand individuals’ experiences are needed to investigate this further.
We also revealed a stronger association between personal experiences of severe physical violence and mental health (compared to neighbourhood disorder). This may be understood in terms of the proximity of the violence to the adolescent; violence at the inter-personal level is a more proximal exposure than violence that occurs at the community level. Therefore, those who live in a neighbourhood with high levels of disorder may be able to ignore or more easily distance themselves from this whereas having personal experience of violence is likely very distressing and difficult to get respite from, especially if it is ongoing. Relatedly, because we used reports of neighbourhood disorder from near-by residents rather than the participants themselves, adolescents may not have perceived their neighbourhood in the same way . While our approach has the methodological advantage of avoiding same-source bias, which may inflate associations with mental health , a link between the two may depend on adolescents’ themselves perceiving there to be a high level of neighbourhood disorder where they live. Indeed, studies that have utilised both perceptions of violence in the neighbourhood and officially reported crime statistics suggest that it is people’s perceptions of their neighbourhood that are most relevant for their mental health [15, 18, 19].
Consistent with a wealth of existing research demonstrating the benefits of social support for health and wellbeing [66,67,68,69], we found evidence that perceived social support at age 18 helped reduce the likelihood that violence-exposed adolescents meet criteria for psychiatric disorder at the transition to adulthood. This suggests that having someone that adolescents can share their experiences and worries with and seek emotional support and advice from is important for maintaining good mental health. Given the prominence of peer friendships and individuals’ increasing independence from their family during adolescence, it is notable that perceived support from family also helped maintain mental health following violence exposure. In fact, for those exposed to both personal violence and neighbourhood disorder, it was higher levels of family support (not friend support) at age 18 that was associated (albeit non-significantly after FDR correction) with a reduced likelihood of psychiatric disorder. On the contrary, maternal warmth and sibling warmth assessed during childhood were not associated with a reduced likelihood of meeting criteria for psychiatric disorder. This may be because the level of sibling and maternal warmth is too low among those children who go on to experience violence in adolescence, or it may be that it is adolescents’ perception of supportive relationships that are currently available to them that is most valuable for maintaining mental health in the face of violence exposure.
Strengths and limitations
Study strengths include the use of a large nationally representative sample, longitudinal study design with excellent participant retention, and inclusion of a broad range of covariates to limit alternative interpretations. Nonetheless, we also acknowledge several limitations. First, participants’ mental health problems and personal experience of violence were both self-reported at age 18. Although adolescents are likely to be the most knowledgeable about their experiences, their current mental health may have impacted their reporting of violence exposure, resulting in reverse causation. For example, depressive disorder may bias recall of negative experiences  or improve the accuracy of reporting (so-called ‘depressive realism’ ). Cognitive avoidance strategies can also affect the retrieval of memory in individuals with post-traumatic stress . However, the prevalence of violent experiences during adolescence in the E-Risk study is comparable to other UK studies, suggesting that these were not significantly under- or over-reported . Similarly, perceived social support was reported at age 18 which has implications for interpreting its association with age-18 mental health. We did control for a range of earlier emotional and behavioural problems to try to rule out reverse causation, but it remains possible that the severity of adolescents’ mental health symptoms during the past year impacted their perceived level of support. Second, neighbourhood disorder was measured only once during adolescence when participants were approximately 13 years old. The majority of participants (71.4%, N = 1475) remained living at the same home address between the ages of 12 and 18, though levels of disorder within their neighbourhood may have changed over time. Third, our measure of neighbourhood disorder considers the immediate environment where E-Risk participants live. However, adolescents likely also spend time in other neighbourhoods (e.g., for education, work, and leisure) which may expose them to high levels of disorder that could impact their risk for later mental health problems. Fourth, items measuring neighbourhood disorder were averaged to create a total score. This approach considers less severe forms of disorder (such as graffiti) as being equal to more severe, potentially less common, forms (e.g., assault) which may underestimate the level of disorder in a neighbourhood. There are alternative ways of aggregating items to account for their differences in severity . Fifth, we focus on adolescents’ personal experience of severe physical violence; therefore, our findings may not generalise to other harmful experiences, such as non-physical bullying, cyber-bullying, and emotional abuse. Our findings also do not inform about associations with mental health beyond the age of 18; some of those who did not report symptoms may go on to develop psychiatric disorders in the future. Sixth, although we focused on disorders, there are other ways of conceptualising psychopathology (e.g., symptom continuum ) which our results do not necessarily inform about. Finally, our findings were based on a sample of twins, and these may differ from non-twins. However, the E-Risk sample is representative of UK families in terms of socio-economic distribution  and neighbourhood deprivation  and the prevalence of violence victimisation and mental health problems has been shown to be comparable between twins and non-twins .
Our findings reaffirm the need for early intervention to support adolescents who experience violence and highlights the vulnerability of those whose personal experience takes place in the context of community-level violence. Interventions focused on improving perceived social support by increasing the availability of supportive people or boosting adolescents’ perceptions of their existing support network may be effective in protecting their mental health.
Availability of data and materials
The dataset reported in the current article is not publicly available due to lack of informed consent and ethical approval for open access, but is available on request by qualified scientists. Requests require a concept paper describing the purpose of data access, ethical approval at the applicant’s institution, and provision for secure data access (for further details, see here: https://sites.duke.edu/moffittcaspiprojects/concept-paper-template/). We offer secure access on the Duke University and King's College London campuses. All data analysis scripts and results files are available for review on request from the corresponding author.
World Health Organisation (2002) World report on violence and health. https://www.who.int/violence_injury_prevention/violence/world_report/en/abstract_en.pdf. Accessed 3 November 2021
Sampson RJ, Raudenbush SW (1999) Systematic social observation of public spaces: a new look at disorder in urban neighborhoods. Am J Sociol 105:603–651. https://doi.org/10.1086/210356
Sickmund M, Puzzanchera C (2014) Juvenile offenders and victims: 2014 national report. Pittsburgh, PA: National Center for Juvenile Justice. https://www.ojjdp.gov/ojstatbb/nr2014/downloads/nr2014.pdf. Accessed 3 November 2021
Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE (2005) Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry 62:593–602. https://doi.org/10.1001/archpsyc.62.6.593
Solmi M, Radua J, Olivola M, Croce E, Soardo L, de Pablo GS, Shin JI, Kirkbride JB, Jones P, Kim JH, Kim JY, Carvalho AF, Seeman MV, Correll CU, Fusar-Poli P (2021) Age at onset of mental disorders worldwide: large-scale meta-analysis of 192 epidemiological studies. Mol Psychiatry. https://doi.org/10.1038/s41380-021-01161-7
Caspi A, Houts RM, Ambler A, Danese A, Elliott ML, Hariri A, Harrington HL, Hogan S, Poulton R, Ramrakha S, Hartman Rasmussen LJ, Reuben A, Richmond-Rakerd L, Sugden K, Wertz J, Williams BS, Moffitt TE (2020) Longitudinal assessment of mental health disorders and comorbidities across 4 decades among participants in the Dunedin birth cohort study. JAMA Netw Open 3:e203221–e203221. https://doi.org/10.1001/jamanetworkopen.2020.3221
Kim-Cohen J, Caspi A, Moffitt TE, Harrington H, Milne BJ, Poulton R (2003) Prior juvenile diagnoses in adults with mental disorder: developmental follow-back of a prospective-longitudinal cohort. Arch Gen Psychiatry 60:709–717. https://doi.org/10.1001/archpsyc.60.7.709
MacMillan HL, Fleming JE, Streiner DL, Lin E, Boyle MH, Jamieson E, Duku EK, Walsh CA, Wong MYY, Beardslee WR (2001) Childhood abuse and lifetime psychopathology in a community sample. Am J Psychiatry 158:1878–1883. https://doi.org/10.1176/appi.ajp.158.11.1878
Schaefer JD, Moffitt TE, Arseneault L, Danese A, Fisher HL, Houts R, Sheridan MA, Wertz J, Caspi A (2018) Adolescent victimization and early-adult psychopathology: approaching causal inference using a longitudinal twin study to rule out noncausal explanations. Clin Psychol Sci 6:352–371. https://doi.org/10.1177/2167702617741381
Lansford JE, Dodge KA, Pettit GS, Bates JE, Crozier J, Kaplow, J (2002) A 12-year prospective study of the long-term effects of early child physical maltreatment on psychological, behavioral, and academic problems in adolescence. Arch Pediatr Adolesc Med 156:824–830. https://doi.org/10.1001/archpedi.156.8.824
Nanni V, Uher R, Danese A (2012) Childhood maltreatment predicts unfavorable course of illness and treatment outcome in depression: a meta-analysis. Am J Psychiatry 169:141–151. https://doi.org/10.1176/appi.ajp.2011.11020335
Capusan AJ, Kuja-Halkola R, Bendtsen P, Viding E, McCrory E, Marteinsdottir I, Larsson H (2016) Childhood maltreatment and attention deficit hyperactivity disorder symptoms in adults: a large twin study. Psychol Med 46:2637–2646. https://doi.org/10.1017/S0033291716001021
Arseneault L, Cannon M, Fisher HL, Polanczyk G, Moffitt TE, Caspi A (2011) Childhood trauma and children’s emerging psychotic symptoms: a genetically sensitive longitudinal cohort study. Am J Psychiatry 168:65–72. https://doi.org/10.1176/appi.ajp.2010.10040567
Kelleher I, Keeley H, Corcoran P, Ramsay H, Wasserman C, Carli V, Sarchiapone M, Hoven C, Wasserman D, Cannon M (2013) Childhood trauma and psychosis in a prospective cohort study: cause, effect, and directionality. Am J Psychiatry 170:734–741. https://doi.org/10.1176/appi.ajp.2012.12091169
Polling C, Khondoker M, Hatch SL, Hotopf M, SELCoH Study Team (2014) Influence of perceived and actual neighbourhood disorder on common mental illness. Soc Psychiatry Psychiatr Epidemiol 49:889–901. https://doi.org/10.1007/s00127-013-0813-9
Ivert AK, Levander MT (2014) Adolescents’ perceptions of neighbourhood social characteristics—Is there a correlation with mental health? Child Indic Res 7:177–192. https://doi.org/10.1007/s12187-013-9210-x
O’Brien DT, Farrell C, Welsh BC (2019) Broken (windows) theory: a meta-analysis of the evidence for the pathways from neighborhood disorder to resident health outcomes and behaviors. Soc Sci Med 228:272–292. https://doi.org/10.1016/j.socscimed.2018.11.015
Goldman-Mellor S, Margerison-Zilko C, Allen K, Cerda M (2016) Perceived and objectively-measured neighborhood violence and adolescent psychological distress. J Urban Health 93:758–769. https://doi.org/10.1007/s11524-016-0079-0
Newbury JB, Arseneault L, Caspi A, Moffitt TE, Odgers CL, Baldwin JR, Zavos HMS, Fisher HL (2017) In the eye of the beholder: perceptions of neighborhood adversity and psychotic experiences in adolescence. Dev Psychopathol 29:1823–1837. https://doi.org/10.1017/S0954579417001420
Rutter M (1979) Protective factors in children’s responses to stress and disadvantage. In: Kent MW, Rolf JE (eds) Primary prevention of psychopathology, vol 3. Social competence in children. University of New England Press, New Hampshire, pp 49–74
Sameroff AJ (2000) Dialectical processes in developmental psychopathology. In: Sameroff A, Lewis A, Miller S (eds) Handbook of developmental psychopathology, 2nd edn. Kluwer Academic/Plenum Publishers, New York, pp 23–40
Mrug S, Windle M (2010) Prospective effects of violence exposure across multiple contexts on early adolescents’ internalizing and externalizing problems. J Child Psychol Psychiatry 51:953–961. https://doi.org/10.1111/j.1469-7610.2010.02222.x
Finkelhor D, Shattuck A, Turner HA, Ormrod R, Hamby SL (2011) Polyvictimization in developmental context. J Child Adolesc Trauma 4:291–300. https://doi.org/10.1080/19361521.2011.610432
Fisher HL, Caspi A, Moffitt TE, Wertz J, Gray R, Newbury J, Ambler A, Zavos H, Danese A, Mill J, Odgers CL, Pariante C, Wong, CCY, Arseneault, L (2015) Measuring adolescents' exposure to victimization: the environmental risk (E-Risk) longitudinal twin study. Dev Psychopathol, 27(4pt2):1399–1416. https://doi.org/10.1017/S0954579415000838
Ford JD, Elhai JD, Connor DF, Frueh BC (2010) Poly-victimization and risk of posttraumatic, depressive, and substance use disorders and involvement in delinquency in a national sample of adolescents. J Adolesc Health 46:545–552. https://doi.org/10.1016/j.jadohealth.2009.11.212
Turner HA, Finkelhor D, Ormrod R (2010) Poly-victimization in a national sample of children and youth. Am J Prev Med 38:323–330. https://doi.org/10.1016/j.amepre.2009.11.012
Crush E, Arseneault L, Moffitt TE, Danese A, Caspi A, Jaffee SR, Matthews T, Fisher HL (2018) Protective factors for psychotic experiences amongst adolescents exposed to multiple forms of victimization. J Psychiatr Res 104:32–38. https://doi.org/10.1016/j.jpsychires.2018.06.011
Newbury J, Arseneault L, Caspi A, Moffitt TE, Odgers CL, Fisher HL (2018) Cumulative effects of neighborhood social adversity and personal crime victimization on adolescent psychotic experiences. Schizophr Bull 44:348–358. https://doi.org/10.1093/schbul/sbx060
Baranyi G, Di Marco MH, Russ TC, Dibben C, Pearce J (2021) The impact of neighbourhood crime on mental health: a systematic review and meta-analysis. Soc Sci Med. https://doi.org/10.1016/j.socscimed.2021.114106
Lynch M, Cicchetti D (1998) An ecological-transactional analysis of children and contexts: the longitudinal interplay among child maltreatment, community violence, and children’s symptomatology. Dev Psychopathol 10:235–257. https://doi.org/10.1017/S095457949800159X
Meehan AJ, Latham RM, Arseneault L, Stahl D, Fisher HL, Danese A (2020) Developing an individualized risk calculator for psychopathology among young people victimized during childhood: a population-representative cohort study. J Affect Disord 262:90–98. https://doi.org/10.1016/j.jad.2019.10.034
Meng X, Fleury MJ, Xiang YT, Li M, D’arcy C (2018) Resilience and protective factors among people with a history of child maltreatment: a systematic review. Soc Psychiatry Psychiatr Epidemiol 53:453–475. https://doi.org/10.1007/s00127-018-1485-2
Latham RM, Temple R, Romeo C, Danese A, Fisher HL (2020) Understanding practitioners’ and young people’s views of a risk calculator for future psychopathology and poor functioning in young people victimised during childhood. J Ment Health 30:706–723. https://doi.org/10.1080/09638237.2020.1844869
Ungar M (2008) Resilience across cultures. Br J Soc Work 38:218–235
Afifi TO, MacMillan HL (2011) Resilience following child maltreatment: a review of protective factors. Can J Psychiatry 56:266–272. https://doi.org/10.1177/070674371105600505
Herrenkohl EC, Herrenkohl RC, Egolf B (1994) Resilient early school-age children from maltreating homes: outcomes in late adolescence. Am J Orthopsychiatry 64:301–309. https://doi.org/10.1037/h0079517
Goodare H, Lockwood S (1999) Involving patients in clinical research. Br Med J 319:724–725. https://doi.org/10.1136/bmj.319.7212.724
National Institute for Health Research (2015) Going the extra mile: improving the nation’s health and wellbeing through public involvement in research. https://www.nihr.ac.uk/documents/about-us/our-contribution-to-research/how-we-involve-patients-carers-and-the-public/Going-the-Extra-Mile.pdf. Accessed 3 November 2021
Trouton A, Spinath FM, Plomin R (2002) Twins early development study (TEDS): a multivariate, longitudinal genetic investigation of language, cognition and behavior problems in childhood. Twin Res 5:444–448. https://doi.org/10.1375/twin.5.5.444
Moffitt TE, E-Risk Study Team (2002) Teen-aged mothers in contemporary Britain. J Child Psychol Psychiatry 43:727–742. https://doi.org/10.1111/1469-7610.00082
Finkelhor D, Ormrod RK, Turner HA (2007) Polyvictimization and trauma in a national longitudinal cohort. Dev Psychopathol 19:149–166. https://doi.org/10.1017/S0954579407070083
Hamby S, Finkelhor D, Ormrod D, Turner H (2004) The comprehensive JVQ administration and scoring manual. University of New Hampshire, Crimes Against Children Research Centre, Durham, NH
Bifulco A, Brown GW, Harris TO (1994) Childhood experience of care and abuse (CECA): a retrospective interview measure. J Child Psychol Psychiatry 35:1419–1435. https://doi.org/10.1111/j.1469-7610.1994.tb01284.x
Bifulco A, Brown GW, Neubauer A, Moran P, Harris T (1994) Childhood experience of care and abuse (CECA) training manual. University of London, Royal Holloway College, London
Fisher HL, Bunn A, Jacobs C, Moran P, Bifulco A (2011) Concordance between mother and father offspring and retrospective reports of childhood adversity. Child Abuse Negl 35:117–122. https://doi.org/10.1016/j.chiabu.2010.10.003
Matthews T, Odgers CL, Danese A, Fisher HL, Newbury JB, Caspi A, Moffitt TE, Arseneault L (2019) Loneliness and neighborhood characteristics: a multi-informant, nationally representative study of young adults. Psychol Sci 30:765–775. https://doi.org/10.1177/0956797619836102
Newbury JB, Arseneault L, Caspi A, Moffitt TE, Odgers CL, Belsky DW, Sugden K, Williams B, Ambler AP, Matthews T, Fisher HL (2022) Association between genetic and socioenvironmental risk for schizophrenia during upbringing in a UK longitudinal cohort. Psychol Med 52(8):1527–1537. https://doi.org/10.1017/S0033291720003347
Odgers CL, Moffitt TE, Tach LM, Sampson RJ, Taylor A, Matthews CL, Caspi A (2009) The protective effects of neighborhood collective efficacy on British children growing up in deprivation: a developmental analysis. Dev Psychol 45:942–957. https://doi.org/10.1037/a0016162
Odgers CL, Caspi A, Bates CJ, Sampson RJ, Moffitt TE (2012) Systematic social observation of children’s neighborhoods using Google Street View: a reliable and cost-effective method. J Child Psychol Psychiatry 53:1009–1017. https://doi.org/10.1111/j.1469-7610.2012.02565.x
Morgan JF, Reid F, Lacey JH (1999) The SCOFF questionnaire: assessment of a new screening tool for eating disorders. BMJ 319:1467–1468. https://doi.org/10.1136/bmj.319.7223.1467
Magaña AB, Goldstein JM, Karno M, Miklowitz DJ, Jenkins J, Falloon IR (1986) A brief method for assessing expressed emotion in relatives of psychiatric patients. Psychiatry Res 17:203–212. https://doi.org/10.1016/0165-1781(86)90049-1
Caspi A, Moffitt TE, Morgan J, Rutter M, Taylor A, Arseneault L, Tully L, Jacobs C, Kim-Cohen J, Polo-Tomas M (2004) Maternal expressed emotion predicts children’s antisocial behavior problems: using monozygotic-twin differences to identify environmental effects on behavioral development. Dev Psychol 40:149–161. https://doi.org/10.1037/0012-1618.104.22.168
Jaffee SR, Caspi A, Moffitt TE, Polo-Tomas M, Taylor A (2007) Individual, family, and neighborhood factors distinguish resilient from non-resilient maltreated children: a cumulative stressors model. Child Abuse Negl 31:231–253. https://doi.org/10.1016/j.chiabu.2006.03.011
Zimet G, Powell S, Farley G, Werkman S, Berkoff K (1988) Psychometric characteristics of the multidimensional scale of perceived social support. J Pers Assess 55:610–617. https://doi.org/10.1080/00223891.1990.9674095
Wechsler D (1974) Manual for the Wechsler intelligence scale for children—revised. Psychological Corporation, New York
Sattler JM (2008) Assessment of children: cognitive foundations, 5th edn. JM Sattler, San Diego, CA
Trzesniewski KH, Donnellan MB, Moffitt TE, Robins RW, Poulton R, Caspi A (2006) Low self-esteem during adolescence predicts poor health, criminal behavior, and limited economic prospects during adulthood. Dev Psychol 42:381–390. https://doi.org/10.1037/0012-1622.214.171.1241
Rogers W (1994) Regression standard errors in clustered samples. Stata Tech Bull 3:19–23
Schwartz S, Susser E (2006) Relationships among causes. In: Susser E, Schwartz S, Morabia A, Bromet EL (eds) Psychiatric epidemiology: searching for the causes of mental disorders. Oxford University Press, Oxford, pp 62–74
Knol MJ, van der Tweel I, Grobbee DE, Numans ME, Geerlings MI (2007) Estimating interaction on an additive scale between continuous determinants in a logistic regression model. Int J Epidemiol 36:1111–1118. https://doi.org/10.1093/ije/dym157
Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Methodol 57:289–300
Meltzer H, Vostanis P, Goodman R, Ford T (2007) Children’s perceptions of neighbourhood trustworthiness and safety and their mental health. J Child Psychol Psychiatry 48:1208–1213. https://doi.org/10.1111/j.1469-7610.2007.01800.x
Okon-Singer H (2018) The role of attention bias to threat in anxiety: mechanisms, modulators and open questions. Curr Opin Behav Sci 19:26–30. https://doi.org/10.1016/j.cobeha.2017.09.008
Veling W, Pot-Kolder R, Counotte J, van Os J, van der Gaag M (2016) Environmental social stress, paranoia and psychosis liability: a virtual reality study. Schizophr Bull 42:1363–1371. https://doi.org/10.1093/schbul/sbw031
Chum A, O’Campo P, Lachaud J, Fink N, Kirst M, Nisenbaum R (2019) Evaluating same-source bias in the association between neighbourhood characteristics and depression in a community sample from Toronto, Canada. Soc Psychiatry Psychiatr Epidemiol 54:1177–1187. https://doi.org/10.1007/s00127-019-01718-6
Wang J, Mann F, Lloyd-Evans B, Ma R, Johnson S (2018) Associations between loneliness and perceived social support and outcomes of mental health problems: a systematic review. BMC Psychiatry 18:1–16. https://doi.org/10.1186/s12888-018-1736-5
Uchino BN (2009) Understanding the links between social support and physical health: a life-span perspective with emphasis on the separability of perceived and received support. Perspect Psychol Sci 4:236–255. https://doi.org/10.1111/j.1745-6924.2009.01122.x
Ozbay F, Johnson DC, Dimoulas E, Morgan CA III, Charney D, Southwick S (2007) Social support and resilience to stress: from neurobiology to clinical practice. Psychiatry (Edgmont) 4:35–40
Cohen S, Wills TA (1985) Stress, social support, and the buffering hypothesis. Psychol Bull 98:310–357. https://doi.org/10.1037/0033-2909.98.2.310
Dalgleish T, Werner-Seidler A (2014) Disruptions in autobiographical memory processing in depression and the emergence of memory therapeutics. Trends Cogn Sci 18:596–604. https://doi.org/10.1016/j.tics.2014.06.010
Ackermann R, DeRubeis RJ (1991) Is depressive realism real? Clin Psychol Rev 11:565–584. https://doi.org/10.1016/0272-7358(91)90004-E
Dalgleish T, Rolfe J, Golden AM, Dunn BD, Barnard PJ (2008) Reduced autobiographical memory specificity and posttraumatic stress: exploring the contributions of impaired executive control and affect regulation. J Abnorm Psychol 117:236–241. https://doi.org/10.1037/0021-843X.117.1.236
Barr PB (2018) Early neighborhood conditions and trajectories of depressive symptoms across adolescence and into adulthood. Adv Life Course Res 35:57–68. https://doi.org/10.1016/j.alcr.2018.01.005
Caspi A, Moffitt TE (2018) All for one and one for all: Mental disorders in one dimension. Am J Psychiatry 175:831–844. https://doi.org/10.1176/appi.ajp.2018.17121383
Odgers CL, Caspi A, Russell MA, Sampson RJ, Arsenault L, Moffitt TE (2012) Supportive parenting mediates widening neighborhood socioeconomic disparities in children’s antisocial behavior from ages 5 to 12. Dev Psychopathol 24:705–721. https://doi.org/10.1017/S0954579412000326
Caspi A, Taylor A, Moffitt TE, Plomin R (2000) Neighborhood deprivation affects children’s mental health: environmental risks identified in a genetic design. Psychol Sci 11:338–342. https://doi.org/10.1111/1467-9280.00267
Kendler KS, Martin NG, Heath AC, Eaves LJ (1995) Self-report psychiatric symptoms in twins and their nontwin relatives: are twins different? Am J Med Genet 60:588–591. https://doi.org/10.1002/ajmg.1320600622
We are grateful to the study mothers and fathers, the twins, and the twins' teachers and neighbours for their participation. Our thanks to Professor Avshalom Caspi, one of the founders of the E-Risk study, and to the E-Risk team for their dedication, hard work, and insights. We also thank Arif Hoque (Independent Researcher) for his thoughtful insights.
The E-Risk Study is funded by the Medical Research Council (UK MRC) [G1002190]. Additional support was provided by the U.S National Institute of Child Health and Human Development (NICHD) [HD077482]; the Jacobs Foundation; and a plus fund award from the UKRI Violence, Abuse and Mental Health Network (VAMHN) [ES/S004424/1]. Rachel M. Latham and Helen L. Fisher are supported by the Economic and Social Research Council (ESRC) Centre for Society and Mental Health at King’s College London [ES/S012567/1]. Louise Arseneault is the Mental Health Leadership Fellow for the UK ESRC. Joanne B. Newbury is supported by a Sir Henry Wellcome Postdoctoral Fellowship from the Wellcome Trust [218632/Z/19/Z]. The views expressed are those of the authors and not necessarily those of the VAMHN, ESRC, Wellcome Trust, or King’s College London. These funders played no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; nor in the decision to submit this article for publication.
Conflict of interest
The authors declare that they have no conflict of interest.
The Joint South London and Maudsley and the Institute of Psychiatry Research Ethics Committee approved each phase of the E-Risk study. The study has therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.
Below is the link to the electronic supplementary material.
About this article
Cite this article
Latham, R.M., Arseneault, L., Alexandrescu, B. et al. Violent experiences and neighbourhoods during adolescence: understanding and mitigating the association with mental health at the transition to adulthood in a longitudinal cohort study. Soc Psychiatry Psychiatr Epidemiol 57, 2379–2391 (2022). https://doi.org/10.1007/s00127-022-02343-6