Article Text

Download PDFPDF

Risk for attempted suicide in children and youths after contact with somatic hospitals: a Danish register based nested case–control study
  1. E Christiansen1,
  2. E Stenager2
  1. 1Clinic of Suicide Prevention and Treatment for Children and Adolescents, Department of Child and Adolescent Psychiatry, Odense University Hospital, Odense C, Denmark
  2. 2Odense University Hospital, Department of Psychiatry and University of Southern Denmark, Institute of Public Health, Odense C, Denmark
  1. Correspondence to E Christiansen, Clinic of Suicide Prevention and Treatment for Children and Adolescents, Department of Child and Adolescent Psychiatry, Odense University Hospital, Bjergegade 15, 1st floor, DK-5000 Odense C, Denmark; erik.christiansen{at}ouh.fyns-amt.dk

Abstract

Background A range of studies have found an association between some somatic diseases and increased risk of suicide and attempted suicide. These studies are mostly analyses of adult populations and illnesses related to adulthood.

Objectives To study the risk of attempted suicide in children and youths with a somatic diagnosis, and to assess a possible association from a somatic perspective.

Methods From a cohort of 403 431 individuals (born 1983–89), 3465 children and youths who had attempted suicide were identified. Each case was matched with 20 population controls. 72 765 children and youths constituted the case–control population. All data were obtained from national population registers and analysed in a nested case–control design.

Results Contact of children and youths with a somatic hospital is correlated with increased risk of attempted suicide; the risk peaks in the time immediately after contact. Risk factors were treatment for injury caused by violence, epilepsy, asthma and malformation for males; and spontaneous and medical abortions, treatment for injury caused by violence, epilepsy, asthma, insulin dependent diabetes mellitus and malformation for females. Not all the mentioned diagnoses were significant in the adjusted model.

Conclusions Based on the results of the study a strategy to minimise the risk of attempted suicide among children and youths must be implemented. The strategy should mainly focus on children at high risk—that is, children from families with low socioeconomic status, and children with a psychiatric history, a history of previous suicide attempts and with an unstable somatic disease subsequently causing many admissions.

  • Case control Me
  • children
  • suicide SI
View Full Text

Statistics from Altmetric.com

Introduction

Suicide is the third most common cause of death in children and youths. Depression, alcohol and drug abuse and a number of psychiatric diseases are associated with an increased risk of suicide.1 A recent Finish study found an association between psychiatric difficulties such as hyperkinetic and emotional problems in childhood, and later suicidal behaviour in adolescence or early adulthood.2 Other studies have found associations between physical and sexual abuse and suicidal behaviour.3

A Danish register based study has shown that psychiatric admission is a strong risk factor for completed suicide in adolescents and young adults, especially females. For both women and men the risk peaked in the time immediately after admission to or discharge from a psychiatric hospital.4 Other studies have found that somatic diseases such as asthma, cancer, insulin dependent diabetes mellitus (IDDM) and epilepsy likewise increase the risk of suicide.5–8 However, these studies are all characterised by methodological problems associated with control groups, the applied statistics or the size of the sample. A number of studies have found an increased risk of suicidal thoughts in children with asthma.5–9 Children and youths diagnosed with IDDM have an increased risk of suicide; of the many sudden unexpected deaths in IDDM patients, some could be suicides.10–12 Goldston found that a significant proportion of the suicidal children diagnosed with IDDM were from families with a low socioeconomic status.13

Studies examining the risk of suicide in children diagnosed with cancer show contradicting results. The general tendency however, is that there is no association between children diagnosed with cancer and later increased risk of suicide,14–17 whereas such an association can be found among adults.18 Finally, a range of studies has found an association between epilepsy and increased risk of suicide and attempted suicide.19–23

Consequently, it seems highly relevant to conduct a study on the risk of attempted suicide in children and youths with a somatic diagnosis, and to assess a possible association from a somatic perspective.

Based on data from Danish national registers prior to an attempted suicide, the objectives of the study are:

  • To assess the risk of attempted suicide when a child or youth has been in contact with a somatic department.

  • To investigate the association between risk of attempted suicide and time since last contact with the somatic department, number of admissions and duration of contact.

  • To assess the association between risk of attempted suicide and the following diagnoses: epilepsy, asthma, IDDM, cancer, malformation, abortion and exposure to violence as indicators of risk behaviour.

  • To investigate the association between risk of attempted suicide in children and youths after contact with a somatic department and parents' socioeconomic status indicated by annual income.

  • To calculate the attributable risk for contact with somatic department and risk of attempted suicide in children and youths.

Method

Data

Denmark has many longitudinal administrative registers. They are used to administer payment between public authorities and service providers, but also provide a unique opportunity for longitudinal register based research. All citizens in Denmark have a unique personal identification number which can be used to merge data from different registers. All treatments at the emergency ward, psychiatric and somatic departments, and private practitioners are recorded in central registers and contain information such as, for example, date of treatment, reason for the contact, and diagnoses. Moreover, all Danish citizens are registered in the Danish Civil Register which contains information on date of birth, sex and links to biological and legal parents as well as siblings. It is therefore possible to obtain information on the person's parents and family. All data are at an individual level and almost all citizens in Denmark can be identified in the Danish Civil Register. Moreover links to at least the mother are available for almost everybody born in Denmark in 1960 or later, gradually decreasing backwards with birth year.

All data for this study were obtained from the following national population registers: the National Registry of Patients (Landspatient Registeret), the Danish Psychiatric Central Register, the Fertility database, the Integrated Database for Labour Marked Research (IDA) and the Danish Civil Register.24 25

Population

From a cohort of 403 431, children and youths who had attempted suicide were identified. The cohort consisted of all individuals born in the time span 1983–89 and who had a Danish personal identification number. All individuals were followed from date of birth until 31 December 2005; in this period of time all suicide attempts registered at a somatic or psychiatric department were recorded.

Contact with a somatic department was recorded as an attempted suicide if one of the following classification codes were used: E4 (attempted suicide) and an ICD-10 diagnosis of S617–S619 (open wound of wrist and hand), X60–X84 (intentional self-harm), T36–T60 (poisoning by drugs, medicaments and biological substances) or T65 (toxic effect of other and unspecified substances); or an ICD-8 diagnosis of E9500–E9599 (suicide and self-inflicted poisoning or injury). Contact with a psychiatric hospital was recorded as attempted suicide if the classification code ICD-10 diagnosis of X60–X84 (intentional self-harm) had been used. Several individuals had more than one attempted suicide. In such cases, the chosen index attempted suicide from the individuals was 10 years up until 31 December 2005.

Each suicide case was then matched with 20 population controls from the cohort with identical age and sex who were alive at the time of the case attempted suicide. This is referred to as a nested case–control design. The matching eliminates the confounder effect of age and sex. In total 3465 cases were identified and matched with 69 300 controls. Consequently, 72 765 children and youths constituted the case–control population. All links to biological parents were established, except in a few cases where the link to the father was missing.

Variables

All variables in the study are register-based. Subsequently only retrospective variables were included—that is, events prior to the case attempted suicide. Prevalence of attempted suicide among children and youths was coded as a dummy variable with one for cases and two for controls.

Variables containing information on the children

All somatic inpatient and outpatient, and emergency department contacts with somatic hospitals (ie, non-psychiatric hospitals), from the date of the child's first birthday to the time of the attempted suicide were analysed. The following dummy variables were created:

  1. The first variable indicated whether the child had had contact with a somatic department prior to the matching date.

  2. The second set of dummy variables was contact divided into eight diagnosis classifications (see table 1).

  3. The third variable was time since latest contact with a somatic department divided into 10 dummy variables, each with a unique time span: no contact, 1–7 days, 8–14 days, 15–30 days, 31–90 days, 91–180 days, 181–360 days, 1–2 years, 2–5 years, >5 years.

  4. The fourth variable was the duration of the last contact divided into six dummy variables: no contact, 1 day, 2–8 days, 9–15 days, 16–30 days, 31–90 days, >90 days.

  5. The fifth variable was number of contacts to a somatic department divided into five dummy variables: no contact, 1–2 contacts, 3–5 contacts, 6–10 contacts, >10 contacts.

  6. The sixth variable indicated whether the child had contact with a psychiatric department prior to the matching date.

Table 1

Classification of somatic diagnoses

Variables containing information on the parents

  1. The first variable indicated whether at least one of the parents had had contact with a psychiatric department.

  2. The second explanatory variable indicated whether at least one of the parents had cashed a prescription of psychopharmacological drugs in the time prior to the matching date.

  3. The third variable were the parents' level of income.

  4. The fourth variable was the highest achieved level of education.

  5. The fifth variable indicated death of a least one of the parents prior to the matching date.

  6. The sixth variable specified whether the parents lived together the year prior to the matching date.

Statistical analysis

A conditional logistic regression model was undertaken containing the described explanatory variables and attempted suicide as the dependent variable. Both crude and adjusted analysis was carried out for all the explanatory variables related to somatic contact. The adjusted model contains all variables related to somatic contact as well as all confounders—that is, the child's contact with a psychiatric department, parents' contact with a psychiatric department, parents' use of psychopharmacological drugs, parents' level of income, parents' level of education, death among parents and parents' marital status. The explanatory variables were analysed for interactions with sex.

Using a conditional logistic regression model including all first level interactions, parents' level of income was analysed as a moderator on the risk of attempted suicide after contact with a somatic department. Let π denote the probability of attempted suicide, then logit(π) equals the logit function for the probability. Let S denote the dummy variable of contact with a somatic department, and I denote the dummy variable of parents' level of income; then the moderator model is abbreviated as:logit(π)=β1S+β2I+β3SI

If level of income is divided into three fractals, the logit model returns coefficients detailed in table 2.

Table 2

Coefficients returned by the logistic regression model

A meaningful interpretation of the coefficients is then:

  • β00 is baseline (reference group) (no contact and income at lowest level)

  • β10 (contact and income at lowest level)

  • β01 (no contact and income at medium level)

  • β11 (contact and income at medium level)

  • β10 (no contact and income at highest level)

  • β11 (contact and income at highest level)

The effect from income on risk after contact with the somatic department is estimated by examining β10, β11 and β12.

The confounder analysis of the variable ‘contact with somatic department’ was performed by stepwise adding potential confounders to the regression model containing the variable. The analysis included three levels: level 1, where the variable was analysed in a univariate regression model; level 2, where the variable was analysed in a multiple regression model containing all relevant confounders, including somatic diagnoses; and level 3, where the variable was analysed in a multiple regression model containing all relevant confounders, including factors relating to previous contacts with somatic departments (see table 3 for more details).

Table 3

Confounder analysis of contact with a somatic department

The attributable risk of contact with the somatic department was calculated.26 The attributable risk is a measure of the reduction in incidence that would be observed if the population was entirely unexposed to the risk factors, compared with its current exposure pattern.

Data were analysed with the PhReg procedure available in SAS V.9.1.3 for Windows. Because we sampled controls from individuals at risk for attempted suicide at the time—that is, risk set sampling—the estimated ORs in this study can be interpreted as RRs; p values and CIs were also calculated.

Results

Table 4 contains a crude analysis of all explanatory variables related to contact with a somatic department. Table 5 contains an adjusted analysis of all explanatory variables related to contact with a somatic department as well as confounders. Table 6 contains an analysis of parents' level of income as a moderator on the risk of attempted suicide after contact with a somatic department, and table 3 contains a confounder analysis of contact with a somatic department.

Table 4

Factors related to young people's contact with a somatic department as risk factors for attempted suicide; crude level

Table 5

Factors related to young people’s contact with a somatic department as risk factors for attempted suicide; adjusted level

Table 6

Income in the family as moderator on risk for suicide attempts after contact with somatic department; incidence RR and CI

Due to the method of sampling, cases and controls have the same age and sex distribution. Males constituted 21% of the population, females 79%. The population were all in the age span 10–24 years, with males on average older than females at the time of the index suicide (males: median 18 years, mean 17.75, SD 2.39; females: median 17 years, mean 16.79, SD 2.34).

In the crude analysis (table 4), contact with a somatic department was a risk factor for attempted suicide for both sexes. According to the incidence RRs the risk of attempted suicide was three times higher for children and youths who had had contact with a somatic department compared to children and youths without such contact. The risk peaked in the weeks right after the contact with a somatic department, then gradually decreased with time; however it was still significant up to 5 years after the last contact. Especially short contacts increased the risk of attempted suicide. Furthermore the risk increased with number of contacts. Analysis of the diagnosis documented that for males, treatments for injury caused by violence, epilepsy, asthma and malformation were significant risk factors for attempted suicide. For females, spontaneous and medical abortions were significant risk factors for attempted suicide as well as treatment for injury caused by violence, epilepsy, asthma, IDDM and malformation. Among the diagnoses the highest risk appeared for exposure to violence.

In the adjusted analysis (table 5), the risk of attempted suicide peaked in the first couple of weeks after contact with a somatic department. For females the risk of attempted suicide declined over time, whereas for males no clear tendency emerged which can be explained by the relatively low number of males in the population. Moreover, long admissions at a somatic department decreased the risk of attempted suicide, whereas several previous contacts with a somatic department increased the risk of attempted suicide. For males, treatment of injury caused by violence was a significant risk factor for attempted suicide, and for females, medical abortion and treatment injury caused by violence likewise increased the risk of attempted suicide. No significant risk of attempted suicide was associated with the additional diagnoses.

The analysis of parents' level of income as a moderator on the risk of attempted suicide after contact with a somatic department showed that risk was only significant for the group of children and youths where the parents' income belonged to the lowest third or were unknown. Children and youths who had had contact with a somatic department and where the parents' income belonged to the highest third were protected against attempted suicide compared to those at baseline (no contact and lowest third).

In table 3 we can see that the effect from the factor contact with a somatic department was lowered in level 2 and became insignificant in level 3.

The attributable risk was calculated as 0.65, which mean that the incidence of suicide attempts would be reduced by 65% if the entire population was unexposed to the risk factor ‘contact with a somatic department’.

Discussion

The use of register-based data results in a highly reliable study, but at the same time limitations apply to interpretation of the results. Moreover, as the incidence of attempted suicide is rare among children and youths, the study is limited by small subgroups and thus the risk of type II errors. However, based on the design of the study, the data and the size of the population, the results of the study are, in our opinion, highly reliable and valid.

Not every attempted suicide is correctly registered in the National Registry of Patients, and it has been noted by Nordentoft and Soegaard that only 37% of all suicide attempts were correctly registered. They also concluded that a contact reason code E4 reflected ‘that the patient has inflicted self-harm’.27 We therefore expect the case population to reflect real attempted suicides, but also that some attempted suicides are by mistake being used as controls. The author has earlier studied risk for repetition of attempted suicide and found that the risk was high in the weeks after the index attempt (age >14).28 Reconciling this knowledge with the chance for incorrect registration of attempted suicide, we expect to have some hidden suicide attempts in the weeks after the index attempt.

This comprehensive study based on data from Danish longitudinal registers shows a correlation between children and youths' contact with a somatic department and increased risk of attempted suicide, with the risk peaking in the time right after contact. The risk was higher for females than for males. The risk was adjusted for the child's psychiatric history. Another Danish study has shown that adolescents and young adults have an increased risk of suicide following an admission to a psychiatric hospital.3 These two studies thus show that contact with a somatic, as well as a psychiatric department, poses an independent risk of later suicidal behaviour in children and youths.

Furthermore, the risk of attempted suicide was highest in the weeks following the latest contact with a somatic department. The risk peaked with number of contacts to the somatic department, whereas the risk declined with the length of admission to a somatic department. When exploring this association in detail, analysis showed that long admissions among children with malformations (a heterogeneous group constituted by children with minor orthopaedic problems as well as children with severe physical and mental congenital health problems) was a protective factor. A possible interpretation of this result could be that children with severe handicaps are not capable of carrying out an attempted suicide.

The analysis of the possible association between risk of attempted suicide and being diagnosed with epilepsy, asthma, cancer or IDDM, documented that when adjusted for relevant confounders, none of the diagnoses were independent risk factors for attempted suicide in children and youths. Previous register-based studies among adult patients have found an increased risk of suicide in young patients with IDDM, and in cancer patients and patients suffering from epilepsy.8 18 Other studies show conflicting results on the association between cancer and risk of suicide in children and youths, whereas an unambiguous association between epilepsy and increased risk of suicidal behaviour in children and youths has been demonstrated.19–21 To our knowledge there is a lack of studies on asthma and suicide among adults with a coherent methodological design.18 The variations in these results can be explained by different methodology such as the applied statistics, small samples (especially among children and youths) and the heterogeneity of the risk groups of suicide and attempted suicide.

When adjusting for risk factors such as number of contacts and psychiatric history, epilepsy turned out to be an insignificant risk factor for attempted suicide. Thus children with a psychiatric history and admitted several time due to epilepsy, indicating the instability of the disease, had an increased risk of suicide attempts.

After adjustment for number of contacts, treatment for injury caused by violence remained a significant risk factor of attempted suicide. When further adjusting for the person's psychiatric history and cohabitating parents, treatment for injury caused by violence was still a significant risk factor, although at a lower level. Thus children exposed to violence are more likely to attempt suicide, but it is out of the remit of this study to interpret the interlinked association between somatic and psychiatric history and social circumstances. Exposure to violence can indicate undiagnosed psychiatric problems—that is, personality disorders, various forms of abuse and constraining social circumstances. Thus our study confirms previous findings of an association between increased risk of suicidal behaviour and exposure to physical and sexual abuse.3 29 30 After adjustment for confounders, medical abortion remained a risk factor for attempted suicide. This association calls for the same interpretation as the association between exposure to violence and increased risk of attempted suicide in children and youths.

The purpose of the confounder analysis was to investigate whether contact with a somatic department was a significant independent explanatory factor of risk for attempted suicide, among young people. At level 2 the effect from the factor was lowered, but it was still significant. Therefore the factor is not entirely explained by confounders such as somatic and psychiatric diagnoses and parental risk factors. The factor became insignificant at level 3, which indicates that the factor is not an independent significant factor but is highly correlated with other factors, related to contact with a somatic department. Therefore the contact itself is not as important as other factors related to the contact (diagnoses and severity of illness, time since last contact, etc). In particular, time since last contact was a significant factor in this analysis, as it was highly correlated with attempted suicide, even after controlling for other factors. A part of this correlation might be attributed to hidden suicide attempts, as mentioned in the limitations; nevertheless it documents a need for risk assessment and coherent treatment between sectors, when discharging young people from somatic departments.

Offspring of parents in the lowest third of level of income had an increased risk of attempted suicide compared to the rest of the group, and children from low-income families had a higher frequency of somatic contacts. Interpreting income as an indicator for parents' capability, the association could indicate that these parents were able to minimise their children's suffering to a lesser degree, and consequently these children were to a larger extent constrained by the somatic disease.

The results of the study highlight the importance of taking measures to prevent attempted suicide among the high number of children and youths being discharged from somatic departments. The attributable risk of 0.65 indicates that if the risk of attempted suicide associated with somatic contacts was eliminated, 65% of suicide attempts in the population could be prevented. This is a theoretic percentage and is mostly attributed to the high number of youths who have been in contact with a somatic department, rather than a high risk. As we are analysing lifetime chances for contact with a somatic department prior to the attempted suicide, we expect a high proportion of the population to have received treatment from a somatic department.

Based on the results of the study a strategy to minimise the risk of attempted suicide among children and youths must be implemented. The strategy should mainly focus on children at high risk—that is, children from families with low socioeconomic status, children with a psychiatric history including those with a history of previous suicide attempts, and children with an unstable somatic disease resulting in many subsequent admissions. Among children with a somatic diagnosis, special attention must be directed at children diagnosed with epilepsy, especially if they also have a psychiatric history. Moreover the strategy should focus on children and youths exposed to violence, as well as young women in the time following a medical abortion.

In order to prevent suicidal behaviour in children and youths, is it important to coordinate the relevant treatment measures. In a number of places in Denmark, paediatric hospitals, children and youth psychiatric hospitals, the social security system, the school and the general practitioner are working together in order to offer an overall solution to children with suicidal behaviour.31 Another important issue is to make sure that professionals—doctors, nurses, teachers, psychologists and social workers—are well-informed on groups of children and youths with an increased risk of attempted suicide, and are capable of offering relevant treatments.

What is already known on this subject

  • Children and youths, especially females, are at high risk for attempted suicide.

  • The risk for suicide peaks during the time immediately after contact with a psychiatric department.

  • The risk for suicide is significant higher among adults diagnosed with insulin dependent diabetes and epilepsy.

What this study adds

  • The risk of attempted suicide in children and youths peaks during the time immediately after contact with a somatic department.

  • Children and youths who have had contact with a somatic department and whose parents' income is low are at greatest risk for attempted suicide.

Acknowledgments

We wish to show appreciation to psychologist Kim Juul Larsen for help with putting forward the hypothesis and designing the study.

References

View Abstract

Footnotes

  • All authors had full access to all the data.

  • This study is registered at the Danish Data Protection Agency j.nr.: 2007-41-0896 and is subject to the Act on Processing of Personal Data. The data is owned by Statistics Denmark and the authors are allowed to access that data but not to share the data with other than staff at the Clinic of Suicide Prevention and Treatment for Children and Adolescents. It is possible for other researchers to reuse data, but this requires approval from the Danish Data Protection Agency, Statistics Denmark and the authors.

  • Funding This study has been financed be the Danish Ministry of Social Affairs (MSA). The MSA had no role other than financing the study.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.