Long-term economic and social outcomes of youth suicide attempts

Abstract

Background

Youth who attempt suicide are more at risk for later mental disorders and suicide. However, little is known about their long-term socioeconomic outcomes.

Aims

We investigated associations between youth suicide attempts and adult economic and social outcomes.

Method

Participants were drawn from the Quebec Longitudinal Study of Kindergarten Children (n = 2140) and followed up from ages 6 to 37 years. Lifetime suicide attempt was assessed at 15 and 22 years. Economic (employment earnings, retirement savings, welfare support, bankruptcy) and social (romantic partnership, separation/divorce, number of children) outcomes were assessed through data linkage with government tax return records obtained from age 22 to 37 years (2002–2017). Generalised linear models were used to test the association between youth suicide attempt and outcomes adjusting for background characteristics, parental mental disorders and suicide, and youth concurrent mental disorders.

Results

By age 22, 210 youths (9.8%) had attempted suicide. In fully adjusted models, youth who attempted suicide had lower annual earnings (average last 5 years, US$ −4134, 95% CI −7950 to −317), retirement savings (average last 5 years, US$ −1387, 95% CI −2982 to 209), greater risk of receiving welfare support (risk ratio (RR) = 2.05, 95% CI 1.39 to 3.04) and were less likely to be married/cohabiting (RR = 0.82, 95% CI 0.73 to 0.93), compared with those who did not attempt suicide. Over a 40-year working career, the loss of individual earnings attributable to suicide attempts was estimated at US$98 384.

Conclusions

Youth who attempt suicide are at risk of poor adult socioeconomic outcomes. Findings underscore the importance of psychosocial interventions for young people who have attempted suicide to prevent long-term social and economic disadvantage.

Background

Suicide attempts in young people are a leading public health concern worldwide.¹ It is estimated that approximately 4–10% of youth aged 12–25 attempt suicide before reaching adulthood.^1,2 This is likely to be an underestimation since a substantial proportion of young people who attempt suicide do not seek specialist services.¹ Additionally, emergency department visits for youth who have attempted suicide have increased in several countries in recent years.^3,4 Previous studies have shown that youth attempted suicide is associated with increased risk of suicide attempt repetition⁵ and suicide mortality,⁶ as well as metal disorders.⁷ For example, longitudinal studies show that young people who attempted suicide are more likely to experience depression, substance use and antisocial behaviours in adulthood.⁸ However, negative outcomes of youth suicide attempts may go beyond mental health, and also include reduced life chances in terms of economic and social participation.^7,9–11

If youth suicide attempt carries an additional risk of poor long-term social and economic outcomes above and beyond concurrent mental disorders, then it is vital to quantify this burden so that enhanced prevention efforts can be justified and implemented for at-risk youth. To our knowledge, the available evidence documenting the socioeconomic outcomes of youth suicide attempts is restricted to a handful of studies^7,9–11 that have produced contradictory results and therefore limit firm conclusions.

First, prior studies have mostly relied on self-reported economic outcomes that are vulnerable to reporting and recall bias,⁷ such as non-random drop-out (for example individuals from lower socioeconomic backgrounds), social desirability, deliberate or non-deliberate misreporting (such as recall failure or lack of knowledge about income or welfare benefits).¹² The use of administrative data is therefore preferable as they are usually supplied by impartial third parties (for example employers or government agencies) and consequently provide more reliable estimates of individual and family economic circumstances especially over long follow-up periods.¹² Second, definitions of suicide attempt used in previous studies are inconsistent. In particular, some studies include youth who self-harmed irrespective of the intentionality of the act.^7,9 This may create heterogeneity, given that evidence shows that later outcomes of youth self-harming behaviour differ depending on the intentionality of the act.⁹ Third, it is unclear to what extent future negative socioeconomic outcomes of youth who attempt suicide are because of contemporaneous mental disorders and substance use. Results of previous studies are mixed, with some reporting that concurrent mental disorders fully explained the association between suicide attempts and socioeconomic outcomes,^7,9 and others reporting an increased risk over and above concurrent mental disorders.^10,11 Fourth, no study to date has quantified the individual and societal economic burden of youth suicide attempts across an individual's full working career.

Aims

The present study sought to address these limitations in the existing literature by examining the association between youth suicide attempt and adult social and economic outcomes (including earnings, welfare and partnership status). We relied on a large population-based sample followed up from age 6 to 37 years with data on suicide attempts collected at ages 15 and 22 years via structured interviews and social and economic outcomes obtained from government administrative data.

Method

Participants

Participants were members of the Quebec Longitudinal Study of Kindergarten Children (QLSKC), a longitudinal cohort of children attending kindergarten in Quebec's French-speaking public schools (Canada) between 1986 and 1987.¹³ The cohort initially included 3020 children whose parents were born in Canada and whose mother tongue was French. The vast majority (88%) were of non-Hispanic White ethnicity. Of these children, 2000 (1000 boys and 1000 girls) were selected using a random sampling procedure stratified by administrative region, school board size and gender to be representative of the population (representative sample). The remaining 1020 participants (600 boys and 420 girls) were oversampled for children exhibiting disruptive behaviours (disruptive sample). Children were followed up annually from age 6 to age 13 years and again at ages 15 and 22 years.

Federal tax returns records from ages 22 to 37 years (2002–2017) were linked to the cohort. Written informed consent was obtained from the children's parents at each year of follow-up prior to participation. The study was approved by the University of Montreal, McGill University and St-Justine Hospital ethics boards and Statistics Canada.

The present study used data from 2140 participants (1060, 49.5% males) for whom data on suicide attempts were available (Table 1). Participants included in the sample were more likely to have more educated mothers, to be male, to have a father with a mental disorder and to have a family history of suicide. To account for this differential attrition, inverse-probability weighting was used.

Table 1 Sociodemographic characteristics of the participants^a

a. Count variables are rounded to base 10 according to Statistics Canada data protection regulations. P-values are based on chi-square or t-tests.

Assessment of suicide attempts

Suicide attempts were assessed at ages 15 and 22 years with structured interviews administered by a trained researcher. Assessments at age 15 were obtained from parental or youth responses to the Diagnostic Interview Schedule for Children (DISC), version 2.¹⁴ Participants and their parents were asked: ‘have you/your child tried to kill yourself/themselves?’. Assessments at age 22 years was made using the same questions from the Diagnostic Interview Schedule (DIS)¹⁵ answered by the participant. We derived a lifetime suicide attempt variable, coded 1 if the participant or his/her parent reported a suicide attempt at either age 15 or 22 years, and 0 if not.

Economic and social outcomes

Outcome data were obtained annually from federal tax return records from age 22 to 37 years (2002–2017) and linked to the QLSKC cohort data.¹² Economic outcomes were as follows.

1. (a) Personal earnings, measured as all pretax wages, salaries and commissions, not including income from capital gains.

2. (b) Retirement savings, measured as the contribution to Registered Retirement Savings Plans and contributions to employer-sponsored retirement plans. For each of these outcomes, the mean of the five most recent tax return records was used (ages 32–37 years). All financial data were converted to US$ prior to analysis (CA$1 = US$0.75).

3. (c) Welfare support (i.e. social assistance), which provides ‘last resort’ financial help to people without income who are no longer eligible for unemployment insurance,¹⁶ excluding individuals with severely limited capacity for work, was dichotomously coded for each year of follow-up (received in the past year 1, not received in past year 0); scores were then summed to create a count variable representing the total number of years in which the participant received welfare support.

4. (d) Bankruptcy, which was extracted from the tax declaration and defined as having ever declared bankruptcy between 2002 and 2017 (binary variable).

Social outcomes included the following.

1. (a) Partnership status, coded dichotomously based on reported marital status in the past year (married/cohabiting 1, single/separated/divorced 0); the final outcome was a count variable representing the number of years in which the participant declared being married/cohabitating (hereafter referred to as partnered).

2. (b) Separation/divorce, binary variable obtained by coding marital status as 1 if participants ever reported being separated/divorced, and 0 otherwise.

3. (c) Number of children, extracted from the last available data point (2017, age 37 years), included any children living in the household (i.e. not limited to biological offspring).

Covariates

Four sets of covariates were included.

Background, child and family characteristics

This set of covariates included gender of the child; childhood socioeconomic adversity between ages 6 and 13 years (a composite measure including mother-reported information on parental education, occupational prestige and family income); low birth weight (<2500 g); and child verbal IQ at age 13, measured using the Sentence Completion Task. This measure consists of 13 sentences containing one missing word that must be selected by the participant from a list of five options to complete the sentence (mean 9.03, s.d. = 2.13, range 0–13); it correlates highly with other verbal and non-verbal measures of intelligence.^{17, 18}

Family history of mental disorder and suicide

This set of covariates included maternal or paternal suicide attempt or death by suicide, assessed when the participant was 15 years of age, and maternal and paternal lifetime history of diagnosed anxiety and/or mood disorders (i.e. panic, generalised anxiety, depressive, dysthymic disorder and episodes of mania), assessed with the DISC when the participant was 15 years of age.

Participant mental disorders

This set of covariates included meeting diagnostic criteria for any of the following disorders at ages 15 or 22 years, assessed with the DISC and the DIS, respectively: depressive disorder, dysthymic disorder, manic episode/bipolar disorder, generalised anxiety disorder, simple phobia, social phobia, agoraphobia, panic disorder, conduct problems, attention-deficit hyperactivity disorder, oppositional defiant disorder.

Substance use

This set of covariates included: substance use, including high alcohol use, defined as consuming alcohol >3 days per week; cannabis use, defined as ever using cannabis; hard drugs use, defined as any consumption of illicit drugs (i.e. cocaine, psychedelics, sedatives such as barbiturates, opioids and inhalants). Substance use was self-reported with reference to the past 6 months and were assessed at age 15 and 22. For each substance, individuals reporting substance use at any of these time points were coded 1 or 0 otherwise.

Statistical analysis

Analyses were performed with Stata version 16. First, descriptive statistics for outcome variables were provided for all the available data (ages 22 to 37 years) for participants who attempted suicide and those who did not. Second, we investigated the association between suicide attempt and economic and social outcomes using generalised linear models with robust standard errors. To investigate the association between suicide attempt and earnings and retirement savings (continuous outcomes), we used Tobit regression left censured at US$1000. To investigate the association between suicide attempt and welfare and partnership (count outcomes), we used negative binomial regressions to account for overdispersion of the outcome variable. To investigate the association between suicide attempt and welfare receipt, bankruptcy and partnership (binary outcomes), we used Poisson regression.

To examine to what extent the selected covariates explained the association between suicide attempt and outcomes, we fitted five models with different adjustment levels: model 1, unadjusted association; model 2, adjusted for child gender (to account for known gender differences in suicide attempts and outcomes); model 3, adjusted for background, family, child characteristics and parental history of mental disorders and suicide; model 4, additionally adjusted for youth lifetime mental disorders assessed concurrently with suicide attempt; model 5, additionally adjusted for substance use. Additionally, all models were adjusted for study design variables, namely, the sample (representative versus disruptive) and age at entry to the cohort.

In supplementary analyses, we re-estimated model 5 using specific mental disorders (i.e. externalising and internalising), and the number of mental disorders as adjustment variables, instead of the variable combining all mental disorders. Results were consistent with the main analyses (Supplementary Table 1 available at https://doi.org/10.1192/bjp.2021.133).

To account for missing data in these covariates, we used multiple imputations: models were estimated across 50 data-sets, and the results pooled. To test the sensitivity of our analyses to missing data, we re-estimated our models for all participants initially enrolled in the QLSKC study using multiple imputation as described above to additionally impute missing data for suicide attempt. Results were reported for males and females combined, as we did not find evidence for gender × suicide attempt interactions (P > 0.05) for any of our outcomes.

Finally, to examine the economic impact of suicide attempt, we estimated the loss of earnings over a 40-year work career as follows:¹⁹

$${\rm Financial}\;{\rm effect} = \mathop \sum \limits_{t = 1}^{40} \displaystyle{{\beta \;} \over {1.03}} = 23.8\;\times \;\beta $$

where β is the estimate obtained from the model and assuming the commonly used annual discount rate of 3%. This financial effect indicates how much would be gained in terms of individual earnings by preventing one suicide attempt, under the strong assumption that suicide attempt was a causal determinant of our outcomes of interest.

Results

A total of 210 (9.8%) youth attempted suicide by age 22 years. Suicide attempts were more common among female participants (n = 140, 13.0%) compared with male (n = 70, 6.6%) participants (risk ratio (RR) = 2.04, 95% CI 1.54–2.76).

At age 15, when both mother- and self-reported suicide attempts were assessed, 100 adolescents had attempted suicide according to either informant. For these, there was an agreement between mother- and self-report for 20 participants, whereas 70 and 10 participants had attempted suicide according to self-reports only or mother-reports only, respectively (Cohen's kappa (κ) = 0.31, revealing low interrater agreement).

Economic and social outcomes for the analysis sample are described in Table 2, and descriptive statistics across follow-up are reported in Fig. 1. At age 22, youth who attempted suicide had similar earnings compared with those who did not attempt suicide. However, over the course of follow-up, annual earnings among youth who attempted suicide rose more slowly compared with those who did not attempt suicide.

Fig. 1 Descriptive statistics for the economic and social outcomes from 22 to 37 years of age by suicide attempt. The figure shows the observed increase in annual earning (a) and retirement savings (b), and the probability of receiving welfare (c) and of marriage/cohabitation (d) by year for the entire follow-up period.

Table 2 Descriptive statistics for the economic and social outcomes in adulthood in the study participants^a

a. Count variables are rounded to base 10 and dollar amounts to base 100 according to Statistics Canada data protection regulations.

b. Average of the last five available years (ages 32 to 37 years), in US dollars.

c. Average count across the whole follow-up period (ages 22 to 37 years).

d. Experienced divorce/separation at least once across follow-up (ages 22 to 37 years).

e. Average number living in the household during the last available year (age 37 years).

As shown in Fig. 2 (see also Supplementary Table 2), youth who attempted suicide earned on average US$ −10 492 (95% CI −14 332 to −6652) less in their mid-thirties than youth who did not attempt suicide. This association was partly explained by background family, child and parental characteristics, as well as by contemporaneous mental disorders and substance use (which accounted for about 60% of the association). After accounting for all these factors, however, we found that youth who attempted suicide still had lower earnings compared with those who did not attempt suicide (adjusted β = US$ −4134, 95% CI −7950 to −317).

Fig. 2 Association between suicide attempt and adult economic outcomes. Estimates are beta coefficients for Tobit regression (a) and (b), and risk ratios for negative binomial (c) and (f) and Poisson (d), (e) and (g) regressions, with 95% confidence intervals.

Estimates are reported for the analysis sample (n = 2140; dark blue) and the multiple-imputed sample (n = 3020; light blue). M1, unadjusted: model adjusted for the variable ‘sample’ (representative versus disruptive) and child age at enrolment in the cohort; M2 adjusted for child gender; M3 additionally adjusted for family socioeconomic status, low birth weight, child verbal IQ, family history of suicide, maternal and paternal mental disorders; M4 additionally adjusted for adolescent mental disorders; M5 additionally adjusted for adolescent substance use.

Over a 40-year career, this corresponds to a loss of earning of US$249 702 (95% CI 158 314–341 090), of which US$ 98 384 (95% CI 7553–189 216) were attributable to attempted suicide after accounting for family and parental background and concurrent mental disorders and substance use.

Furthermore, youth who attempted suicide reported lower annual retirement savings (β = US$ −2689, 95% CI −4215 to −1164). This estimate was reduced by half after accounting for all covariates (adjusted β = −1387, 95% CI −2982 to 209). Over the follow-up period, youth who attempted suicide were more than twice as likely to rely on welfare support compared with those who did not attempt suicide (adjusted RR = 2.05, 95% CI 1.39–3.04).

Figure 1 shows that the gap in the probability of receiving welfare, for participants in the suicide attempt and no suicide attempt groups, was largest when participants were in their twenties but declined and eventually levelled off in their thirties. Finally, we observed that youth who attempted suicide were more likely to declare bankruptcy (RR = 1.94, 95% CI 1.28–2.93), although the estimate did not reach conventional thresholds for statistical significance in the fully adjusted models (RR = 1.44, 95% CI 0.92–2.27). Figure 2 shows that the analyses conducted in the fully imputed sample (n = 3020) are virtually identical to those in the analysis sample, suggesting that attrition did not bias the results.

Figure 1 shows that youth who attempted suicide were more likely to report being partnered in early adulthood (age 22–23 years). However, this tendency was reversed from 27 years onward, when the partnering gap for the two groups began widening. Overall, youth who attempted suicide were almost two times more likely to be ‘unpartnered’ from age 22 to 37 years, compared with those who did not attempt suicide (adjusted RR = 0.82, 95% CI 0.73–0.93). However, no difference was observed regarding the likelihood of being separated/divorced (adjusted RR = 1.21, 95% CI 0.90–1.63) or the likelihood of having children living in the household (RR = 0.88, 95% CI 0.76–1.02) in fully adjusted models. Once again, estimates for the fully imputed sample were consistent with those reported for the analysis sample (Fig. 2).

Discussion

Main findings

Relying on data from official government reports linked to a population-based cohort, we found that youth who attempted suicide were more likely to experience negative economic and social outcomes in adulthood compared with those who did not attempt suicide. These outcomes include personal wealth, social welfare needs and partnership, which underpin key elements of human health and well-being.²⁰ These associations were partially explained by background sociodemographic characteristics, IQ, parental mental disorders, as well as concurrent youth mental disorders and substance use, accounting for up to 60% of the associations. However, our findings suggest that risk of negative economic and social outcomes for youth who attempted suicide were still higher than for those who did not after accounting for these factors. Although these associations do not indicate causal mechanisms, these findings suggest that a suicide attempt signals profound distress that goes beyond concurrent mental health problems and undermines full social and economic participation in society.

Comparison with findings from other studies

Studies of suicide attempt (or self-harm) and subsequent economic and social outcomes are rare in population-based samples. Our results concur with previous work that used stringent definitions of suicide attempt and administrative outcome data¹⁰ but not with a study of adolescent self-harm and self-reported earnings.⁷ Borschmann et al⁷ reported that adolescent self-harm in an Australian population-based sample was associated with self-reported financial hardship at age 35 years but not after mental disorders and substance use were adjusted for (suicide attempts were not assessed). Likewise, no association between self-harm and partnership or welfare receipt was found. In contrast, a study of the Dunedin cohort in New Zealand found that youth who attempted suicide in adolescence (age 17 at first attempt) were more likely to experience longer periods of unemployment and welfare receipt by age 38 years (both measured by official records), compared with those who did not attempt suicide.¹⁰ Thus, when a narrower definition of suicide attempt is used (as an intentional act towards ending one's own life versus self-harm with unspecified intent) in conjunction with objectively measured economic data (rather than self-reports), youth suicide attempts indeed appear to be associated with adverse economic and social outcomes.

Interpretation of our findings

Understanding the mechanisms explaining these associations would be the first step to inform possible interventions to mitigate the negative economic and social outcomes of youth who attempt suicide. First, youth who attempted suicide may be more likely to experience a continuation or aggravation of mental health problems,¹⁰ which in turn can negatively influence their functioning in society.

Second, youth who attempt suicide are more likely to have academic difficulties (including school drop-out),⁹ or to leave the school system for long periods to receive in-patient and/or out-patient mental healthcare.⁴ Such school-related problems may result in decreased job opportunities and increased economic and social difficulties.

Third, although non-fatal, suicide attempts may result in serious injury or disability²¹ that prevents young people's return to education or work, leading to lost income and compromised long-term socioeconomic well-being. For patients without access to insurance or nationalised healthcare the financial burden may be compounded by having to pay for treatment and follow-up care.

Fourth, supportive reactions of family members and peers after a suicide attempt have been linked to improved recovery.²² Consequently, negative reactions of significant people from these important relational environments after a young person's suicide attempt may lead to increased social problems later in life.

Fifth, a suicide attempt may act as a signal to the family and community about the life prospects of the individual. For example, others may stigmatise them as ‘weak’ or ‘ill’²³ and thus justify the withdrawal of personal and material investment, leaving them even more economically and socially exposed than before. Unfortunately, such stereotypes are often internalised by individuals, leading to decreased help seeking behaviours. These potential mechanisms should be tested in future studies.

Our findings suggest that interventions for youth suicide attempts should also take into account that, beyond the risk of repeated suicide attempts, these young people may face socioeconomic disadvantage across adult life that further undermine psychosocial resilience and long-term health and well-being. As social and economic problems are known proximal risk factors for suicide in adults, addressing the socioeconomic outcomes of youth who attempt suicide may reduce later risk of suicide, thus interrupting a vicious cycle between mental health and socioeconomic problems.

Evidence-based psychological treatments show efficacy in reducing suicide attempts in high-risk youth,²⁴ and data from population-based studies shows that psychotherapy for self-harm can reduce long-term suicide risk (including mortality).²⁵ Whether these interventions affect economic and social outcomes remain unclear however.

Future studies should investigate whether the economic and social disadvantages associated with youth suicide attempts are alleviated through effective psychotherapy, or whether further psychosocial interventions are needed. Furthermore, it is worth noting that because only a small proportion of individuals who attempt suicide receive healthcare attention,¹ it is likely that most of the youth who attempted suicide in our sample were not seen in emergency departments and were never treated in hospital. By assessing attempted suicide using interviews, rather than medical records, our findings suggest that all suicide attempts – regardless of their medical severity – may have important socioeconomic consequences.

Strengths and limitations

The study strengths were the long period of prospective follow-up, use of administrative data (which reduce reporting biases that limit studies relying on self-reported earnings), large sample size and use of structured interview to assess suicide attempts. The following limitations should be noted. First, because of attrition, information on attempted suicide was available for a subsample of the initial cohort participants, which could have biased results. However, a series of sensitivity analyses including the application of inverse-probability weighting and multiple imputations suggest that these effects were minimal. Third, suicide-related outcomes vary substantially during adolescence and young adulthood, and previous studies suggest that some individuals have a high suicide attempt risk only in early adolescence, whereas others have a persistently high suicide attempt risk across adolescence and young adulthood.²⁶ Although economic and social outcomes may be different in these two groups, we were unable to investigate such differences because of sample size limitations. We were unable to stratify our analyses by gender for the same reason. Finally, although we controlled for most mental disorders in youth, we were unable to control for psychotic disorders because they were not measured in our cohort.

Implications

Findings from this study, based on a 32-year follow-up of a population-based cohort with linked administrative data, suggest that youth who attempted suicide experience increasing economic and social marginalisation across early adulthood. These outcomes undermine full participation in society and therefore the potential for full human flourishing. For young people who have attempted suicide, psychosocial interventions alongside standard care could improve long-term social and economic participation with benefits for individuals and society.