Skip to main content
NCBI home page
Elsevier Sponsored Documents logoLink to Elsevier Sponsored Documents
. 2023 Jan 26;8(2):e99–e108. doi: 10.1016/S2468-2667(22)00292-4

Multilevel interactions between family and neighbourhood socioeconomic indices in childhood and later risks of self-harm and violent criminality in Denmark: a national cohort study

Linda Ejlskov a,d,k,*, Sussie Antonsen a,d, Jesper N Wulff b, Esben Agerbo a,c,f, Oleguer Plana-Ripoll a,c,k, Clive E Sabel d,e, Chun Chieh Fan g, Wesley K Thompson h, Pearl L H Mok i, Carsten Bøcker Pedersen a,c,d,f, Roger T Webb j,l
PMCID: PMC9896147  PMID: 36709062

Summary

Background

A socioeconomically disadvantaged childhood has been associated with elevated self-harm and violent criminality risks during adolescence and young adulthood. However, whether these risks are modified by a neighbourhood's socioeconomic profile is unclear. The aim of our study was to compare risks among disadvantaged young people residing in deprived areas versus risks among similarly disadvantaged individuals residing in affluent areas.

Methods

We did a national cohort study, using Danish interlinked national registers, from which we delineated a longitudinal cohort of people born in Denmark between Jan 1, 1981, and Dec 31, 2001, with two Danish-born parents, who were alive and residing in the country when they were aged 15 years, who were followed up for a hospital-treated self-harm episode or violent crime conviction. A neighbourhood affluence indicator was derived based on nationwide income quartiles, with parental income and educational attainment indicating the socioeconomic position of each cohort member's family. Bayesian multilevel survival analyses were done to examine the moderating influences of neighbourhood affluence on associations between family socioeconomic position and sex-specific risks for the two adverse outcomes.

Findings

1 084 047 cohort members were followed up for 12·8 million person-years in aggregate. Individuals of a low socioeconomic position residing in deprived neighbourhoods had a higher incidence of both self-harm and violent criminality compared with equivalently disadvantaged peers residing in affluent areas. Women from a low-income background residing in affluent areas had, on average, 95 (highest density interval 76–118) fewer self-harm episodes and 25 (15–41) fewer violent crime convictions per 10 000 person-years compared with women of an equally low income residing in deprived areas, whereas men of a low income residing in affluent areas had 61 (39–81) fewer self-harm episodes and 88 (56–191) fewer violent crime convictions per 10 000 person-years than men of a low income residing in deprived areas.

Interpretation

Even in a high-income European country with comprehensive social welfare and low levels of poverty and inequality, individuals residing in affluent neighbourhoods have lower risks of self-harm and violent criminality compared with individuals residing in deprived neighbourhoods. More research is needed to explore the potential of neighbourhood policies and interventions to reduce the harmful effects of growing up in socioeconomically deprived circumstances on later risk of self-harm and violent crime convictions.

Funding

European Research Council, Lundbeck Foundation Initiative for Integrative Psychiatric Research, and BERTHA, the Danish Big Data Centre for Environment and Health funded by the Novo Nordisk Foundation Challenge Programme.

Introduction

Young people who grow up in financially disadvantaged households are at an elevated risk of engaging in self-harm and interpersonal violence.1 Experiencing familial deprivation can negatively affect a child's socio-emotional, behavioural, and cognitive development,2, 3 and increase the likelihood of aggressive behaviour.4 For example, some parents with a low income might be more likely to have psychiatric disorders, abuse substances, and have impaired parenting behaviours because of poverty-related stress.5 The higher frequency and severity of such problems in these families increases the risk of both inward-directed (self-harm)6 and outward-directed (violence)7 aggression. The overlapping causes between self-harm and interpersonal violence prompt a more integrated approach to research and prevention.8 Why do some young people engage in self-harm, some in interpersonal violence, and some in neither behaviour?9

Research in context.

Evidence before this study

We searched Pubmed, PsycINFO, Embase, and Web of Science for peer-reviewed studies published in English from database inception up to Jan 5, 2022, with a follow-up search including Jan 6 to Oct 14, 2022, with article titles that included the following terms: (“SES” OR “SEP” OR “educat*” OR “employ” OR “unemploy*” OR “income*” OR “occupation*” OR “poverty” OR “poor” OR “social class*” OR “disadvantage*” OR “social factor*” OR “economic” OR “socio-economic*” OR “socioeconomic*” OR “inequalit*” OR “labour” OR “labor” OR “work” OR “depriv*”) AND (“multilevel” OR “neighbo*” OR “social disorgani*” OR “structural characteristic*” OR “environmen*” OR “contextual” OR “area*” OR “geograph*” OR “place*” OR “region*” OR “county” OR “counties” OR “ward*” OR “city” OR “cities” OR “district*” OR “countr*”) AND (“self-harm*” OR “self harm*” OR “suicid*” OR “self-poison*” OR “self poison*” OR “self-inj*” OR “self inj*” OR “poison*” OR “parasuicid*” OR “intent*” “inj*” OR “overdos*” OR “violen*” OR “offend*” OR “crime*” OR “crimin*” OR “forensic*” OR “offense*” OR “offence*” OR “prison*” OR “imprison*” OR “incarcer*” OR “homicid*” OR “kill*” OR “murder*”) AND (“Interact*” OR “moderat*” OR “cross-level” OR “links” OR “modif*” OR “buffer”). Our literature search showed that both individual-level and area-level deprivation are associated with elevated risks for both internalised and externalised violence. The evidence base is larger and stronger for violent behaviour compared with self-harm, but there is consistent evidence in relation to both adverse outcomes. How risks by childhood deprivation vary by neighbourhood socioeconomic profile is unclear. Despite the heavy societal burden of self-harming and violent behaviour, we found no studies investigating cross-level interactions between socioeconomic position levels at the neighbourhood level and individual level for self-harm and violent behaviour. However, a US-based residential relocation intervention study, titled Moving to Opportunity, investigated the changes in risk of committing violence, revealing that the risk of violence decreased in the first few years after moving to a less deprived neighbourhood. A further evaluation of this intervention, this time examining suicide as the outcome, found no evidence for a moderating effect of the neighbourhood. Another study found some evidence that area-level deprivation moderated the association between individual-level socioeconomic position and attitudes towards violence. Lastly, one study found evidence that adolescent boys in a low socioeconomic position were more likely to display antisocial behaviour if they had grown up alongside more affluent neighbours.

Added value of this study

Analysing the national cohort data from more than a million Danish people who were followed up for 12·8 million person-years in aggregate we compared the risks of hospital-treated self-harm and violent criminality in relation to parental income and educational levels during childhood. Among cohort members of similar individual-level socioeconomic position, we also investigated whether risks for these two adverse outcomes varied according to the socioeconomic profile of their residential neighbourhood when they were aged 15 years. This novel study has shown that men and women who are deprived residing in affluent neighbourhoods have lower self-harm and violent criminality risks than their peers who are similarly deprived residing in deprived neighbourhoods.

Implications of all the available evidence

Consistent with existing literature, we identified a social gradient in the incidence of self-harm and violent criminality. Additionally, the findings suggest that residing in a more affluent neighbourhood might have a moderating influence on the detrimental effect of growing up in socioeconomically disadvantaged families. The reason for this might be that protective factors are more salient and risk factors less pronounced in affluent neighbourhoods compared with deprived areas. This evidence supports the notion that policies and interventions aimed at improving neighbourhood characteristics might have the additional benefit that they will reduce the detrimental effect of growing up in socioeconomically deprived circumstances.

In addition to personal and familial factors, differential social environmental exposures might explain the heterogeneity in outcomes among otherwise similar individuals.9 Earlier research has found that the characteristics of the neighbourhood in which an individual grows up (eg, services and norms to support daily living such as educational, cultural, and health facilities and neighbourhood deprivation) might influence later risks of engaging in self-harm10 and violent criminality.11 Several studies2, 12 have shown how neighbourhoods both directly and indirectly can influence children's propensity to be healthy and prosperous. Less attention has been given to the potential moderating role that neighbourhoods might have.13 Despite the sparse evidence base, policy makers have enabled the development of economically mixed neighbourhoods,14 thereby creating opportunities for low-income families to move to higher income neighbourhoods15 to potentially ameliorate the detrimental effect of growing up in a family of a low income.

Childhood adversities might have a more harmful effect in low-income households, indicating a moderating effect of family socioeconomic position on the association between childhood adversities and self-harm risk.16 One question is whether a neighbourhood's socioeconomic profile also exerts a moderating influence. Published evidence on this topic is inconsistent. Some studies have indicated that young people with a low socioeconomic position growing up in affluent neighbourhoods might have lower risks than those in less affluent neighbourhoods of adverse outcomes because of a higher quality and availability of favourable communal sociocultural resources such as public services, extracurricular activities, job opportunities, and social support.13, 17 There might also be different norms regarding the use of aggressive behaviour among local peers.18, 19 Access to better resources in the neighbourhood might exert a beneficial influence directly or through influencing parental behaviour and parenting practices.17 A contrasting perspective suggests that young people of a low socioeconomic position raised in affluent neighbourhoods might have a lower relative social standing, higher stigma, and social stress compared with those from less affluent neighbourhoods. For these young people, a lower sense of self-worth and little autonomy could lead to higher risks for engaging in self-harm and interpersonal violence compared with their disadvantaged peers living in deprived neighbourhoods.20 A third perspective suggests it might be that the association between family composition and low household socioeconomic position and later risks of aggressive behaviour is not influenced by the context of the neighbourhood, as one study showed in relation to the survival of patients with common cancers.21 Here, we aimed to investigate whether the associations between socioeconomic disadvantage and self-harm and violent criminality risks are buffered by the level of neighbourhood affluence during childhood.

Methods

Study design and participants

We performed a national cohort study, using Danish national registers that contain information on all residents, with unique personal identification numbering enabling accurate linkage across a broad array of health, social, and geographical indices.22 In this study, the cohort members were all people born in Denmark between Jan 1, 1981, and Dec 31, 2001, with two Danish-born parents, who were alive and residing in the country when they were aged 15 years. We chose to restrict the study population to individuals with two Danish-born parents because parental variables have high amounts of missing data for many first-generation immigrants.22 The cohort was additionally restricted to individuals with accurate geographical coordinates for their residential address on that date (1 084 047 [99·6%] of 1 088 531 participants had complete information on geographical coordinates). Cohort members were followed up to Dec 31, 2018, with the longest possible follow-up time being exactly 22 years (with the oldest participant being aged 37 years). At the time of performing this analysis, this was the most updated data available for research. Similar to Mok and colleagues,1 we identified the first hospital-treated self-harm episodes after the age of 15 years using the Psychiatric Central Research Register and the National Registry of Patients. First violent crime convictions were identified using the National Crime Register, with 15 years being the age at which criminal responsibility commences in Denmark. Follow-up ended when a self-harm episode or a violent crime conviction first occurred, or at emigration, death, or Dec 31, 2018, whichever came first (appendix p 2). This study was approved by the Danish Data Protection Agency. No further ethical approval is required for registry-based research in Denmark. Data were accessed and analysed in a pseudo-anonymous form on secure servers located at Statistics Denmark.

Family socioeconomic position

We generated two markers of family socioeconomic position pertaining to the calendar year before the cohort members turned 15 years: parental income level and parental educational level. The parental income level consisted of maternal and paternal annual gross income quartiles, obtained from the Income Statistics Register; the parental educational level was based on the highest maternal or paternal educational attainment, obtained from the Population Education Register, and categorised into these four groups using the International Classification of Education coding: lowest (mandatory primary schooling), short (secondary school and vocational education), medium (short-cycle higher education below Bachelor degree level), and higher (Bachelor degree level or higher; appendix p2). In addition to measuring income in the year before the start of follow-up (at age 14 years), we further investigated parental income measured at multiple ages during childhood: ages 1, 4, 7, and 10 years; and median parental income (the sum of mother's and father's annual inflation-adjusted income) throughout childhood.

Neighbourhood affluence indicator

Using nationwide longitudinal residential histories and accurate geographical coordinates for each residence, Denmark has been divided into 1885 data zones (neighbourhoods) with similar population sizes (with a mean of 2867 residents).23 Within any given neighbourhood, the affluence indicator denotes the proportion of all its inhabitants aged 25 years and older with income in the highest quartile based on the national population older than 25 years (appendix p 3). Neighbourhood affluence pertains to the year before cohort members turned 15 years. We further performed a sensitivity analysis changing the age of measurement to an overall childhood mean for ages 1–14 years.

Statistical analysis

All analyses were conducted separately for men and women because of differences in sex-specific self-harm and violent criminality incidence rates.8 Because of the sex-specific incidence rates being markedly different, particularly in relation to violent offending, conducting separate analyses ensured that potential notable differences in incidence rates between socioeconomic position groups were discernible. Incidence rate ratios (IRRs) were estimated from log-linear Poisson multilevel regression models adjusted for 1-year increment age groups and calendar year as time-dependent variables. These adjusted models are equivalent to the Cox proportional hazards model, assuming piecewise constant incidence rates.23 We nested the cohort members in their residential neighbourhoods at age 15 years, specifying fully Bayesian multilevel models. The uncertainty surrounding the IRR is illustrated by the highest density interval, denoting the 95% of IRR values with the highest posterior probability (95% highest density interval). The Bayesian prior for the incidence rate was set based on estimates reported from previous studies on self-harm and violent criminality.8 Because missing observations for all covariates were less than 5% of the total study cohort, we used listwise deletion.24 A detailed description of the modelling approach including the model equation, previous values applied, and R code can be found in the appendix (p 6). We modelled parental income and parental education as ordinal predictors using the appropriate monotonic effect parameterisation. Monotonic modelling assumes the effect is consistently negative or positive across the full range of an ordinal variable but allows the size of changes to vary across ordinal categories by a substantial amount.25 Neighbourhood affluence was modelled as a continuous predictor. The statistical modelling was conducted using the brms26 package in R version 1.4.1106 with a burn-in of 750 iterations and a further 1750 iterations. Model convergence and mixing was assessed using graphical representations. The reporting of this study conforms to the STROBE statement (appendix p 20).

To examine variation in risks between disadvantaged cohort members who grew up in deprived versus affluent neighbourhoods, we fitted an interaction term between each family socioeconomic position marker and neighbourhood affluence in separate models, adjusted for age, to investigate cross-level interactions. We then calculated counterfactual posterior prediction incidence rates to investigate the cross-level interactions further. The posterior predicted incidence rates were calculated for the four different quartiles of the family socioeconomic position marker (parental education and parental income) separately, setting neighbourhood affluence to either a low value (7% of all neighbourhood residents belonging to the highest income group), which we denoted as deprived, and a high value (60% of all neighbourhood residents belonging to the highest income group), which we denoted as affluent. We opted to compare these two values at each end of the neighbourhood affluence spectrum in line with the growing policy interest in mixing communities of different socioeconomic position origin.27 These policies often focus on the most deprived areas and whether relocation to more affluent neighbourhoods can ameliorate some of the harmful effects of living in the most deprived areas. However, to assess the robustness of the choice, we calculated similar rates but for ten different points along the entire continuum of the continuous neighbourhood affluence indicator (appendix pp 18–19). We calculated all predicted incidence rates for the modal value age and person-years at risk of 4·5. To ensure that the estimated models did not entail over-adjustment, the main results of the paper were presented with only age adjustment. However, we conducted a sensitivity analysis investigating the potential effect of adjusting for known risk factors for the two outcomes (appendix pp 14–17). We chose not to assess psychiatric diagnosis in young people as a potential confounder because developing these conditions lies on the causal pathway between family socioeconomic position and later risks of self-harm28 and violent criminality.29 Because we restricted the study cohort to Danish-born individuals with Danish-born parents, we did not include ethnicity. Since family composition might account for some of the associations, we also conducted a sensitivity analysis in which only first-born singletons were included in the analyses to account for potential within-family clustering effects (appendix pp 14–17). R version 1.4.1106 was used for all statistical analyses.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

The 1 084 047 cohort members, who were born in Denmark between Jan 1, 1981, and Dec 31, 2001, were followed up for 12·8 million person-years in aggregate. 556 279 (51·3%) were men and 527 768 (48·7%) were women. Ethnicity was not included. 30 684 cohort members received a violent crime conviction (26 766 [87·2%] men and 3918 [12·8%] women), and 27 731 (16 556 [59·7%] women and 11 175 [40·3%] men) had a hospital-treated self-harm episode. There was a higher proportion of both men and women of a low socioeconomic position who engaged in either type of harmful behaviour at the age of 15 years or older than men and women of a higher socioeconomic position, and on average they resided in more deprived neighbourhoods (table 1). Further descriptive information on the study cohort is presented in the appendix (pp 7–12).

Table 1.

Socioeconomic characteristics of the cohort members and neighbourhood areas

Self-harm episodes
 Violent crime convictions
No episodes At least one episode Incidence rate No convictions At least one conviction Incidence rate
Women
Adverse events 507 824 16 556 .. 523 850 3918 ..
Parental income quartile
Q1 (lowest) 40 040/495 384 (8·1%) 2351/15 793 (14·9%) 48·6 42 101/510 743 (8·2%) 756/3674 (20·6%) 15·0
Q2 97 970/495 384 (19·8%) 3906/15 793 (24·7%) 32·8 101 625/510 743 (19·9%) 1038/3674 (28·3%) 8·5
Q3 151 365/495 384 (30·6%) 4713/15 793 (29·8%) 26·6 156 002/510 743 (30·5%) 1027/3674 (28·0%) 5·4
Q4 (highest) 206 009/495 384 (41·6%) 4823/15 793 (30·5%) 19·3 211 015/510 743 (41·3%) 853/3674 (23·2%) 3·3
Parental educational
Primary school (lowest) 31 613/485 269 (6·5%) 2282/15 129 (15·1%) 55·1 33 493/500 088 (6·7%) 840/3436 (24·4%) 19·4
Secondary school or vocational education 243 253/485 269 (50·1%) 8616/15 129 (57·0%) 27·9 251 578/500 088 (50·3%) 2024/3436 (58·9%) 6·4
Short-cycle higher education 28 280/485 269 (5·8%) 700/15 129 (4·6%) 22·2 28 996/500 088 (5·8%) 131/3436 (3·8%) 4·1
Bachelor degree or higher education 182 123/485 269 (37·5%) 3531/15 129 (23·3%) 16·9 186 021/500 088 (37·2%) 441/3436 (12·8%) 2·1
Mean percentage of people within the highest income quartile in a given neighbourhood 25·3 23·1 .. 25·2 22·2 ..
Median percentage of people within the highest income quartile in a given neighbourhood 22 20·3 .. 22 19·6 ..
Men
Adverse events 544 440 11 175 .. 529 513 26 766 ..
Parental income quartile
Q1 (lowest) 43 911/530 979 (8·3%) 1671/10 640 (15·7%) 31·8 41 547/516 803 (8·0%) 4111/25 460 (16·1%) 81·8
Q2 105 564/530 979 (19·9%) 2721/10 640 (25·6%) 21·3 101 947/516 803 (19·7%) 6498/25 460 (25·5%) 52·2
Q3 161 739/530 979 (30·5%) 3129/10 640 (29·4%) 15·9 157 736/516 803 (30·5%) 7359/25 460 (28·9%) 38·3
Q4 (highest) 219 765/530 979 (41·4%) 3119/10 640 (29·3%) 11·7 215 573/516 803 (41·7%) 7492/25 460 (29·4%) 28·6
Parental educational
Primary school (lowest) 34 558/520 011 (6·6%) 1763/10 167 (17·3%) 39·4 31 961/503 574 (6·3%) 4455/24 219 (18·4%) 105·7
Secondary school or vocational education 260 762/520 011 (50·1%) 5889/10 167 (57·9%) 17·9 252 257/503 574 (50·1%) 14 740/24 219 (60·9%) 46·1
Short-cycle higher education 30 201/520 011 (5·8%) 387/10 167 (3·8%) 11·6 29 707/503 574 (5·9%) 910/24 219 (3·8%) 27·6
Bachelor degree or higher education 194 490/520 011 (37·4%) 2128/10 167 (20·9%) 9·6 192 649/503 574 (38·3%) 4114/24 219 (17·0%) 18·8
Mean percentage of people within the highest income quartile in a given neighbourhood 25·3 22·6 .. 25·4 22·9 ..
Median percentage of people within the highest income quartile in a given neighbourhood 22 20·1 .. 22·1 20·2 ..
Open in a new tab

Data shown as n (%) for categorical variables and mean or median for numerical variables. The incidence rate denotes the number of events in each category per 10 000 person-years. Missing data not shown.

Table 2 shows a decrease in the incidence of both first self-harm episodes and violent crime convictions among men and women for each quartile increase in parental income level. For men, each incremental increase in parental income was, on average, associated with a 25% (IRR 0·75; 95% credibility interval 0·73–0·76) lower incidence of self-harm and a 27% (IRR 0·73; 0·72–0·74) lower incidence of violent criminality. For women, each increase in parental income was associated with a 24% (IRR 0·76; 0·75–0·77) lower incidence of self-harm and 37% (IRR 0·63; 0·61–0·65) lower incidence of violent criminality (the specific monotonic effects are shown in the appendix p 13). For both sexes, increasing neighbourhood affluence both in the parental income and parental education models was associated with a lower incidence of both self-harm and violent criminality, ranging from approximately 10–15% for each 10% increase in neighbourhood affluence. The association between family socioeconomic position in childhood and risks for both adverse outcomes were similar, albeit somewhat stronger, when measured by parental education rather than parental income (table 2; model A and C). Further adjustment for known risk factors and restricting the study population to first-born singletons showed marginally attenuated results (appendix pp 14–17).

Table 2.

Models of family socioeconomic position and neighbourhood affluence effect on self-harm and violent criminality IRR estimates

Men
 Women
Self-harm, IRR (95% CI) Violent criminality, IRR (95% CI) Self-harm, IRR (95% CI) Violent criminality, IRR (95% CI)
Parental income: model A
Individual level: one quartile increase in parental income level 0·75 (0·73–0·76) 0·73 (0·72–0·74) 0·76 (0·75–0·77) 0·63 (0·61–0·65)
Neighbourhood level: 10% rise in neighbourhood affluence 0·85 (0·83–0·88) 0·87 (0·85–0·88) 0·88 (0·87–0·90) 0·85 (0·81–0·89)
Parental income: model B
Cross-level interaction: increasing parental income level and 10% rise in neighbourhood affluence 0·96 (0·94–0·98) 0·95 (0·93–0·96) 0·97 (0·95–0·98) 0·93 (0·90–0·96)
Parental education: model C
Individual level: one quartile increase in parental education level 0·64 (0·63–0·66) 0·58 (0·57–0·58) 0·69 (0·67–0·70) 0·48 (0·46–0·50)
Neighbourhood level: 10% rise in neighbourhood affluence 0·87 (0·85–0·89) 0·91 (0·89–0·92) 0·89 (0·88–0·91) 0·89 (0·85–0·93)
Parental education: model D
Cross-level interaction: increasing parental education level and 10% rise in neighbourhood affluence 1·00 (0·98–1·02) 0·98 (0·96–0·99) 0·99 (0·97–1·00) 0·94 (0·90–0·98)
Open in a new tab

IRRs were adjusted for age at self-harm episode or violent crime conviction with Bayesian CIs containing the 95% most credible values given the model, data, and Bayesian prior values. Models A and C include the direct effects only, whereas models B and D have an additional interaction term between individual-level and neighbourhood-level socioeconomic position. See the appendix (pp 14–17) for estimates of the fully adjusted models and a range of sensitivity analyses. Because parental socioeconomic position was modelled as a monotonic variable, the estimates shown here should be interpreted at the average decrease in risk for each quartile increase in parental socioeconomic position levels. The appendix (p 13) shows the differences between each parental socioeconomic position level. 95% CI=95% credibility interval. IRR=incidence rate ratio.

Figure 1 presents the mean predicted incidence rates for self-harm and figure 2 presents the mean predicted incidence rates for violent offending, for deprived versus affluent neighbourhoods grouped by parental education and income level. Consistent with the general results shown in table 2, in both sexes, the lower the parental socioeconomic position, the higher the estimated predicted incidence of both adverse outcomes. This pattern held true for all cohort members with equivalent family socioeconomic position levels having lower incidences of self-harm and violent criminality with higher levels of affluence in their residential neighbourhood (appendix pp 7–11). Focusing specifically on cohort members with a low parental socioeconomic position, those individuals residing in deprived neighbourhoods had a higher incidence of both self-harm (figure 1) and violent criminality (figure 2) compared with equivalently disadvantaged peers residing in affluent areas.

Figure 2.

Figure 2

Open in a new tab

Predicted violent criminality incidence rates for young men and women aged 15–19 years in deprived versus affluent neighbourhoods grouped by family socioeconomic position in childhood characteristics

Age-adjusted predicted incidence rate for those youngest at diagnosis (aged 15–19 years) and for a mean person-years of 4·5. Deprived neighbourhoods were those where 7% of all neighbourhood residents were in the highest national income quartile. Affluent neighbourhoods were those where 60% of all neighbourhood residents were in the highest national income quartile. For parental education level, 1 indicated lowest (mandatory primary schooling), 2 was short (secondary school and vocational education), 3 was medium (short-cycle higher education below Bachelor degree level), and 4 was higher (Bachelor degree level or higher).

Figure 1.

Figure 1

Open in a new tab

Predicted self-harm incidence rates for young men and women aged 15–19 years in deprived versus affluent neighbourhoods grouped by family socioeconomic position in childhood characteristics

Age-adjusted predicted incidence rate for those youngest at diagnosis (aged 15–19 years) and for mean person-years of 4·5. Deprived neighbourhoods were those where 7% of all neighbourhood residents were in the highest national income quartile. Affluent neighbourhoods were those where 60% of all neighbourhood residents were in the highest national income quartile. For parental education level, 1 indicated lowest (mandatory primary schooling), 2 was short (secondary school and vocational education), 3 was medium (short-cycle higher education below Bachelor degree level), and 4 was higher (Bachelor degree level or higher). Parental income was shown in quartiles.

Figure 3 shows the number of predicted fewer incident cases per 10 000 person-years comparing disadvantaged men and women with residence in affluent versus deprived neighbourhoods. The difference in incidence rates between growing up in an affluent versus deprived neighbourhood was greater when family socioeconomic position was measured by parental educational level compared with parental income level at age 14 years: on average, men with parents on a low-income had 88 (highest density interval 56–131) fewer violent crime convictions per 10 000 person-years and 61 (39–84) fewer self-harm episodes per 10 000 person-years. For women with low-income parents, there was a small difference in the incidence of violent criminality among young women residing in affluent versus deprived neighbourhoods, with on average 25 (15–41) fewer convictions per 10 000 person-years, but a marked difference in relation to self-harm incidence, with 95 (76–118) fewer incidences. Comparing those living in affluent versus deprived neighbourhoods, we found lower predicted numbers of incident cases of self-harm and violent criminality were found at all ages at which parental income was measured, with higher predicted differences for parental income measured at younger ages during childhood than older ages, as can be observed in figure 3.

Figure 3.

Figure 3

Open in a new tab

Number of predicted fewer events per 10 000 person-years of self-harm episodes and violent criminality for young men and women aged 15–19 years of a low socioeconomic position residing in affluent compared with deprived neighbourhoods

The mean number of fewer incident events per 10 000 person-years is based on the age-adjusted predicted incidence rate comparing young men and women residing in an affluent neighbourhood compared with a deprived neighbourhood. Deprived neighbourhoods were those where 7% of all neighbourhood residents were in the highest national income quartile. Affluent neighbourhoods were those where 60% of all neighbourhood residents were in the highest national income quartile. HDI=highest density interval.

Overall, residents of a low socioeconomic position in affluent neighbourhoods subsequently had lower incidence rates of both self-harm and violent criminality than their counterparts residing in deprived neighbourhoods. For self-harm, this was especially true when family socioeconomic position in childhood was measured by parental education; there were on average 176 (148–200) fewer episodes per 10 000 person-years for women and 124 (88–159) fewer episodes per 10 000 person-years for men. For violent criminality, men had 166 (118–215) fewer conviction events per 10 000 person-years, whereas women had 13 (8–17) fewer convictions per 10 000 person-years (figure 3).

Discussion

This study has shown that people of a low socioeconomic position residing in affluent neighbourhoods had lower incidence rates of both self-harm and violent criminality compared with those residing in deprived neighbourhoods. Cross-level interaction analyses revealed that this pattern was consistent for both men and women, regardless of whether family socioeconomic position was assessed by parental education or parental income levels and was consistent across the socioeconomic position spectrum. We also found that, although disadvantaged individuals residing in affluent neighbourhoods were subsequently at an elevated risk compared with their counterparts whose families were not of low socioeconomic position, they had a lower incidence than their disadvantaged peers who were raised in deprived neighbourhoods. Although we found similar results for a range of different ages for parental income during childhood, there are many reasons why individuals of a low income might move to more affluent neighbourhoods. Thus, potential selection bias might be present and causal inferences cannot be made. However, one Danish study that was conducted using a quasi-experimental design30 revealed an apparent positive effect of living in a more affluent neighbourhood, supporting the observational results from this study.

To our knowledge, this is the first longitudinal study to use cross-level interactions to assess whether people whose families were of low socioeconomic position when they were children have lower or higher incidences of self-harm and violent criminality during adolescence and young adulthood according to the affluence of the neighbourhood in which they grew up. The observed socioeconomic gradients of rising risks in relation to both adverse outcomes concur with what has been reported previously.1 Additionally, our investigation showed that young people of a low socioeconomic position have lower incidence rates of hospital-presented self-harm or violent criminality if they reside in affluent neighbourhoods compared with deprived neighbourhoods. Our findings add to a growing body of evidence suggesting a notable moderating effect of neighbourhood on the association between family socioeconomic position and adverse outcomes.13, 30 It is notable that such marked differences in the incidence of self-harm and violent criminality are observable within one of the world's more egalitarian nations. As such, stronger associations might be found in countries with greater social inequality and less comprehensive welfare policies.

We observed similar patterns of risk between self-harm and violent criminality. This finding concurs with what we have reported previously from our Danish registry studies1, 17 of these two adverse outcomes, in relation to parental death during childhood,31 for instance. The similarity in patterns of risk is plausibly explained by shared psychological mechanisms32 and commonly experienced environmental risk factors during childhood.33 In Switzerland, analysis of the longitudinal Zurich Project on Social Development from Childhood to Adulthood revealed several factors linked with risk elevations of similar magnitude for self-harm, violence, and dual harm (ie, the same person has physically harmed themselves and other people). These shared risk factors included childhood sensation seeking, parental divorce, and peer violence victimisation.33

Earlier we posited that people of a low socioeconomic position in their childhood who were otherwise at a higher risk of engaging in internalised or externalised violent behaviour might benefit from favourable communal sociocultural resources. This hypothesis is supported by what we observed. For violent criminality, a particularly relevant mechanism that might partly explain our findings pertains to criminal opportunities and offending likelihood; in other words, violent offending can only arise if there is an opportunity for it to happen. Barnes and Jacobs11 showed that among US adolescents, genetic risk had a stronger association with violent behaviour when exposed to neighbourhood disadvantage. That is, a risk factor might have a small effect when a low level of environmental risk is present, but as the environmental risk is increased so is the effect of the risk factor at the whole population level. Our investigation focused on the potentially modifying effect of residing in an affluent neighbourhood on the association between socioeconomic disadvantage during childhood and subsequent risks of internalised and externalised violence. The lower incidence of violent criminality among disadvantaged men and women residing in affluent neighbourhoods might have arisen because violent crime incidence is lower in more affluent neighbourhoods per se.19 Similarly, in affluent neighbourhoods there might be different prevailing attitudes and social norms regarding violent behaviour. For instance, gang culture and the normalisation of violence are both more common in deprived neighbourhoods.18 As such, a lower tolerance for and exposure to interpersonal violence might benefit all young people who grow up in more affluent areas, in educational settings and when engaging in leisure activities with peers. Peer influences are an important factor in the development of aggressive behaviour34 and thus might be a potential mediating mechanism between neighbourhood affluence and violent criminality risk. Self-harm might also be a socially contagious behaviour in some contexts, with elevated risk clustering among people who are closely connected socially.35 Because fewer young people harm themselves in affluent neighbourhoods, it follows that it is less likely that someone predisposed to harming themselves is exposed to the behaviour and is thereby influenced to act on potential impulses.

We found a positive association between growing up in an affluent neighbourhood for both sexes, with a larger absolute reduction in self-harm risk for women of a low socioeconomic position and a larger absolute reduction of violent criminality risk for men of a low socioeconomic position compared with those in a disadvantaged neighbourhood. Although not much evidence exists on whether women and men are differentially affected by neighbourhood social environments, a US-based social experiment, the Moving To Opportunity study, in which families were randomly assigned to receive a housing voucher, showed that moving to a more affluent neighbourhood generally resulted in better long-term mental health and risk behaviour for girls but worse mental health for boys.36 Marked differences between Denmark and the USA in terms of absolute and relative levels of deprivation, inequalities, social structures, and cultures could explain these conflicting findings. However, it is also worth noting that violent behaviour is more often caused by factors other than poor mental health.33 Although the results from this study support the favourable resources hypothesis, which suggests that the higher quality and availability of favourable communal sociocultural resources in more affluent neighbourhoods might lead to better outcomes for people who are socioeconomically disadvanted, we investigated only violent criminality and self-harm episodes. It is possible that if mental disorders had been investigated, we would find similar results to the Moving To Opportunity study, supporting the relative social standing hypothesis, which suggests that being raised in a more affluent neighbourhood results in a lower relative social standing, higher experienced stigma, and social stress. When looking specifically at violent convictions in the Moving to Opportunity study, there was a large reduction for both men and women in the first few years after the move,36 which concurs with what we have observed. A UK study37 found that among 1600 disadvantaged children growing up with more affluent neighbours, more antisocial behaviour was observed for boys but not girls, compared with otherwise similar peers growing up in deprived areas, which contrasts with what we found. However, the focus of that study was mental health rather than violent crime convictions, the investigators did not have equivalently comprehensive data sources, and they did not conduct longitudinal multilevel analyses.

We chose to only include Danish-born people with two Danish-born parents because of the inclusion of multiple parental-level variables that have missing data for many first-generation immigrants.22 Foreign migrants and their descendants have elevated risks for suicidal behaviour and violent offending.38 Therefore, by restricting the study population, the results presented are not influenced by unmeasured confounding by variable socioeconomic position between people born in Denmark to Danish-born parents and descendants of immigrants. A key strength of our investigation was the use of interlinked registry data from an entire national birth cohort, although an important limitation was its observational nature; the residual confounding of individual-level or family-level characteristics that influence both neighbourhood selection and life outcomes is plausible.39 Thus, we cannot rule out that the residents of a low socioeconomic position residing in affluent neighbourhoods simply had more unmeasured resources (eg, supportive social networks) to begin with compared with residents of a low socioeconomic position residing in deprived neighbourhoods. Although we included a range of different ages at which we measured parental income during childhood, this inclusion does not completely counteract potential selection bias, meaning that causal inferences cannot be made. Furthermore, the neighbourhood affluence indicator that we derived could be somewhat misclassified. The neighbourhoods were derived using a semi-automated process that generated smaller data zones that were only based on residential densities.23 However, such misclassification would attenuate results toward the null, and thus our reported estimates are conservative. Additionally, we could only examine hospital-treated self-harm episodes and violent crime convictions because events of a less serious nature were not recorded in the registers.

In conclusion, across the spectrum of family socioeconomic position during childhood, Danish-born men and women growing up in affluent neighbourhoods had lower incidences of later self-harm and violent criminality compared with their peers with equivalent family socioeconomic positions residing in deprived areas. A possible explanation for this risk pattern is that protective factors are more pronounced and risk factors are less pronounced in affluent neighbourhoods compared with deprived areas. Policies and interventions aimed at improving neighbourhood characteristics might therefore have the additional benefit of ameliorating the detrimental effect of growing up in deprived circumstances on increasing subsequent risks of self-harm and violent criminality. Future research should focus on establishing whether the observed association is causal, on pinpointing which specific protective and risk factors are instrumental, and whether these findings extend to other adverse health outcomes.

Data sharing

Although the tailored datasets that were generated for conducting this study are not publicly available, the source registry data can be accessed by making an application to the Danish Health Data Authority (www.sundhedsdatastyrelsen.dk) and Statistics Denmark (https://www.dst.dk/en).

Declaration of interests

We declare no competing interests.

Acknowledgments

Acknowledgments

This study was funded by the European Research Council, Lundbeck Foundation Initiative for Integrative Psychiatric Research, and BERTHA, the Danish Big Data Centre for Environment and Health funded by the Novo Nordisk Foundation Challenge Programme (grant number NNF17OC0027864).

Acknowledgments

Contributors

LE, RTW, and CBP conceived and designed the study. SA curated data for epidemiological analyses. CBP, CES, WKT, and CCF developed the data zones used in this study. LE analysed the epidemiological data with input from JNW, EA, and CBP. LE, OP-R, RTW, PLHM, CBP, and JNW wrote the first draft of the paper. All authors contributed to the interpretation of data and writing of the paper. All authors revised and approved the final manuscript. LE, SA, and CBP had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Supplementary Material

Supplementary appendix
mmc1.pdf (780.7KB, pdf)

References

  • 1.Mok PLH, Antonsen S, Pedersen CB, et al. Family income inequalities and trajectories through childhood and self-harm and violence in young adults: a population-based, nested case-control study. Lancet Public Health. 2018;3:e498–e507. doi: 10.1016/S2468-2667(18)30164-6. [DOI] [PubMed] [Google Scholar]
  • 2.Cooper K, Stewart K. Does household income affect children's outcomes? A systematic review of the evidence. Child Indic Res. 2021;14:981–1005. [Google Scholar]
  • 3.Allen J, Balfour R, Bell R, Marmot M. Social determinants of mental health. Int Rev Psychiatry. 2014;26:392–407. doi: 10.3109/09540261.2014.928270. [DOI] [PubMed] [Google Scholar]
  • 4.Pitkänen J, Bijlsma MJ, Remes H, Aaltonen M, Martikainen P. The effect of low childhood income on self-harm in young adulthood: mediation by adolescent mental health, behavioural factors and school performance. SSM Popul Health. 2021;13 doi: 10.1016/j.ssmph.2021.100756. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Berger LM, Paxson C, Waldfogel J. Income and child development. Child Youth Serv Rev. 2009;31:978–989. doi: 10.1016/j.childyouth.2009.04.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Björkenstam E, Kosidou K, Björkenstam C. Childhood household dysfunction and risk of self-harm: a cohort study of 107 518 young adults in Stockholm County. Int J Epidemiol. 2016;45:501–511. doi: 10.1093/ije/dyw012. [DOI] [PubMed] [Google Scholar]
  • 7.Sariaslan A, Larsson H, D'Onofrio B, Långström N, Lichtenstein P. Childhood family income, adolescent violent criminality and substance misuse: quasi-experimental total population study. Br J Psychiatry. 2014;205:286–290. doi: 10.1192/bjp.bp.113.136200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Mok PLH, Pedersen CB, Springate D, et al. Parental psychiatric disease and risks of attempted suicide and violent criminal offending in offspring: a population-based cohort study. JAMA Psychiatry. 2016;73:1015–1022. doi: 10.1001/jamapsychiatry.2016.1728. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Brent DA, Melhem NM, Wilcox HC. Violent offending and suicidal behavior have common familial risk factors: a rejoinder to Tolstoy. JAMA Psychiatry. 2016;73:1005–1007. doi: 10.1001/jamapsychiatry.2016.1785. [DOI] [PubMed] [Google Scholar]
  • 10.Cairns J-MM, Graham E, Bambra C. Area-level socioeconomic disadvantage and suicidal behaviour in Europe: a systematic review. Soc Sci Med. 2017;192:102–111. doi: 10.1016/j.socscimed.2017.09.034. [DOI] [PubMed] [Google Scholar]
  • 11.Barnes JC, Jacobs BA. Genetic risk for violent behavior and environmental exposure to disadvantage and violent crime: the case for gene-environment interaction. J Interpers Violence. 2013;28:92–120. doi: 10.1177/0886260512448847. [DOI] [PubMed] [Google Scholar]
  • 12.Ribeiro AI. Public health: why study neighborhoods? Porto Biomed J. 2018;3:e16. doi: 10.1016/j.pbj.0000000000000016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ribeiro AI, Fraga S, Severo M, et al. Association of neighbourhood disadvantage and individual socioeconomic position with all-cause mortality: a longitudinal multicohort analysis. Lancet Public Health. 2022;7:e447–e457. doi: 10.1016/S2468-2667(22)00036-6. [DOI] [PubMed] [Google Scholar]
  • 14.Lees L. Gentrification and social mixing: towards an inclusive urban renaissance? Urban Stud. 2008;45:2449–2470. [Google Scholar]
  • 15.Lindberg RA, Shenassa ED, Acevedo-Garcia D, Popkin SJ, Villaveces A, Morley RL. Housing interventions at the neighborhood level and health: a review of the evidence. J Public Health Manag Pract. 2010;16(suppl):S44–S52. doi: 10.1097/PHH.0b013e3181dfbb72. [DOI] [PubMed] [Google Scholar]
  • 16.Pitkänen J, Remes H, Aaltonen M, Martikainen P. Experience of maternal and paternal adversities in childhood as determinants of self-harm in adolescence and young adulthood. J Epidemiol Community Health. 2019;73:1040–1046. doi: 10.1136/jech-2019-212689. [DOI] [PubMed] [Google Scholar]
  • 17.Galster GC. In: Neighbourhood effects research: new perspectives. van Ham M, Manley D, Bailey N, Simpson L, Maclennan D, editors. Springer; Dordrecht: 2013. The mechanism(s) of neighbourhood effects: theory, evidence, and policy implications; pp. 23–56. [Google Scholar]
  • 18.Scott D, Bennett S. Endorsing the street code: the impact of neighborhood gang activity on incarcerated youth. J Youth Stud. 2022;25:595–615. [Google Scholar]
  • 19.Kearns A, Livingston M, Galster G, Bannister J. The effects of neighbourhood offender concentrations on the number, type and location of crimes committed by resident offenders. Br J Criminol. 2019;59:653–673. [Google Scholar]
  • 20.Stafford M, Marmot M. Neighbourhood deprivation and health: does it affect us all equally? Int J Epidemiol. 2003;32:357–366. doi: 10.1093/ije/dyg084. [DOI] [PubMed] [Google Scholar]
  • 21.Cheng E, Soulos PR, Irwin ML, et al. Neighborhood and individual socioeconomic disadvantage and survival among patients with nonmetastatic common cancers. JAMA Netw Open. 2021;4 doi: 10.1001/jamanetworkopen.2021.39593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Pedersen CB, Gøtzsche H, Møller JO, Mortensen PB. The Danish Civil Registration System. A cohort of eight million persons. Dan Med Bull. 2006;53:441–449. [PubMed] [Google Scholar]
  • 23.Pedersen CB, Antonsen S, Timmermann A, et al. Urban-rural differences in schizophrenia risk: multilevel survival analyses of individual- and neighborhood-level indicators, urbanicity and population density in a Danish national cohort study. Schizophr Bull Open. 2022;3 doi: 10.1093/schizbullopen/sgab056. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Wulff J, Ejlskov L. Multiple imputation by chained equations in praxis: guidelines and review. Electron J Bus Res Methods. 2017;15:41–56. [Google Scholar]
  • 25.Bürkner P-C, Charpentier E. Modelling monotonic effects of ordinal predictors in Bayesian regression models. Br J Math Stat Psychol. 2020;73:420–451. doi: 10.1111/bmsp.12195. [DOI] [PubMed] [Google Scholar]
  • 26.Bürkner PC. brms: an R package for Bayesian multilevel models using Stan. J Stat Softw. 2017;80:1–28. [Google Scholar]
  • 27.Galster G, Andersson R, Musterd S. Neighborhood social mix and adults' income trajectories: longitudinal evidence from Stockholm. Geogr Ann, Ser B. 2016;98:145–170. [Google Scholar]
  • 28.Cybulski L, Ashcroft DM, Carr MJ, et al. Risk factors for nonfatal self-harm and suicide among adolescents: two nested case–control studies conducted in the UK Clinical Practice Research Datalink. J Child Psychol Psychiatry. 2022;63:1078–1088. doi: 10.1111/jcpp.13552. [DOI] [PubMed] [Google Scholar]
  • 29.Copeland WE, Miller-Johnson S, Keeler G, Angold A, Costello EJ. Childhood psychiatric disorders and young adult crime: a prospective, population-based study. Am J Psychiatry. 2007;164:1668–1675. doi: 10.1176/appi.ajp.2007.06122026. [DOI] [PubMed] [Google Scholar]
  • 30.Foverskov E, White JS, Frøslev T, Sørensen HT, Hamad R. Risk of psychiatric disorders among refugee children and adolescents living in disadvantaged neighborhoods. JAMA Pediatr. 2022;176:1107–1114. doi: 10.1001/jamapediatrics.2022.3235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Carr MJ, Mok PLH, Antonsen S, Pedersen CB, Webb RT. Self-harm and violent criminality linked with parental death during childhood. Psychol Med. 2020;50:1224–1232. doi: 10.1017/S0033291719001193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Shafti M, Taylor PJ, Forrester A, Pratt D. The co-occurrence of self-harm and aggression: a cognitive-emotional model of dual-harm. Front Psychol. 2021;12 doi: 10.3389/fpsyg.2021.586135. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Steinhoff A, Bechtiger L, Ribeaud D, Eisner M, Shanahan L. Self-, other-, and dual-harm during adolescence: a prospective-longitudinal study of childhood risk factors and early adult correlates. Psychol Med. 2022 doi: 10.1017/S0033291722000666. published online March 17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Powers CJ, Bierman KL. The multifaceted impact of peer relations on aggressive-disruptive behavior in early elementary school. Dev Psychol. 2013;49:1174–1186. doi: 10.1037/a0028400. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Hawton K, Saunders KE, O'Connor RC. Self-harm and suicide in adolescents. Lancet. 2012;379:2373–2382. doi: 10.1016/S0140-6736(12)60322-5. [DOI] [PubMed] [Google Scholar]
  • 36.Sciandra M, Sanbonmatsu L, Duncan GJ, et al. Long-term effects of the Moving to Opportunity residential mobility experiment on crime and delinquency. J Exp Criminol. 2013;9:451–489. doi: 10.1007/s11292-013-9189-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Odgers CL, Donley S, Caspi A, Bates CJ, Moffitt TE. Living alongside more affluent neighbors predicts greater involvement in antisocial behavior among low-income boys. J Child Psychol Psychiatry. 2015;56:1055–1064. doi: 10.1111/jcpp.12380. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Webb RT, Antonsen S, Mok PLH, Agerbo E, Pedersen CB. National Cohort Study of suicidality and violent criminality among Danish immigrants. PLoS One. 2015;10 doi: 10.1371/journal.pone.0131915. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Ludwig J, Duncan GJ, Gennetian LA, et al. Neighborhood effects on the long-term well-being of low-income adults. Science. 2012;337:1505–1510. doi: 10.1126/science.1224648. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary appendix
mmc1.pdf (780.7KB, pdf)

Data Availability Statement

Although the tailored datasets that were generated for conducting this study are not publicly available, the source registry data can be accessed by making an application to the Danish Health Data Authority (www.sundhedsdatastyrelsen.dk) and Statistics Denmark (https://www.dst.dk/en).

RESOURCES