[1] 20w Young people's risk of suicide attempts after contact with a psychiatric department -a nested case-control design using Danish register data

[1] 66w Contact with psychiatric department was analyzed as a risk factor for suicide attempt. Among cases, 1,194 (34.46%) had been hospitalized prior to the index suicide attempt. By comparison, only 3,015 (4.35%) of the control had been hospitalized. Individuals who had at some point had contact with a psychiatric department had a risk of attempting suicide that is 12.26 times (p < .0001) higher than other individuals.

[2] 133w After controlling the risk factors relating to contact with psychiatric department for parental confounder effects, the adjusted analysis showed results similar to those of the crude analysis, but some factors became insignificant. The suicide attempt risk was still very high in the first weeks after last contact. The peak in risk after last contact was found to be immediately after discharge. Still, individuals with long-duration contact with a psychiatric department were less at risk than individuals with short duration. Individuals with more than 10 contacts were significantly more at risk than individuals with one or two previous contacts. After controlling diagnoses for all other factors, depression, neurotic, emotionally unstable personality and other personality disorders were still significant independent risk factors. OCD became a significant protective factor in the adjusted analysis (see Table 2).

[3] 45w The effect of drugs on suicide attempt risk was analyzed in an adjusted model. Antiepileptic drugs became an insignificant risk factor and anxiolytics became a significant independent protective factor. The results indicate a high correlation between the different types of psychopharmacological drugs (see Table 3).

[4] 100w An analysis of parental income as moderator of risk after a child/adolescent's contact with a psychiatric department showed that the risk was highest for individuals whose biological father's income belonged in the lowest third. The risk was 8.57 times higher compared to individuals with no contact and father's income in the lowest third (see Table 4). The risk was significantly lower (non-overlapping confidence intervals) for individuals with contact and father's income in the medium or highest third (rate ratios = 1.93 and = 2.56 respectively). The analysis of biological mother's level of income as moderator of risk showed similar results.

[5] 123w The frequency distributions for the two variables are shown in Figure 1. We can see that compared to controls, suicide attempters are more likely to have been given several different diagnoses and to have bought prescription psychopharmacological drugs. Table 5 shows the risk ratio for the two variable codes as dummies. The suicide attempt risk is significantly higher for individuals who have been given several different diagnoses and bought several different psychopharmacological drugs from a pharmacy prior to the attempted suicide. The Pearson correlation coefficient for the number of different diagnoses given and the number of different drugs used was calculated at .58. The coefficient was significantly different from zero (p < .0001). Only data on cases were used to calculate this correlation.

[1] 123w This study shows that a child/adolescent's risk of suicide attempt peaks immediately after discharge from last contact with a psychiatric department. It also shows that the risk of suicide attempt is highest for children and adolescents suffering from personality disorders, emotionally unstable personality disorder, affective disorder, depressive disorder and disorder due to substance use. Children and adolescents who have previously had contact with a psychiatric department and whose parents' income is in the lowest third have a significantly higher risk of suicide attempt compared to other groups. The risk of suicide attempt was correlated with use of psychopharmacological drugs, and suicide attempters were more likely to have been given several different diagnoses and used several different drugs. These two variables were highly correlated.

[2] 94w Pilot studies have shown that not all suicide attempts in Denmark are correctly registered as contact code E4. Therefore, it must be assumed that not all suicide attempts made by the cohort during the follow-up period were registered. It was not possible to check if these suicide attempts corresponded with WHO's definition of suicide attempt. A possible consequence is that the case population is a selected subgroup of all suicide attempters, and that some suicide attempters are used as controls by mistake. This could in theory influence our results, making them somewhat less transparent.

[3] 59w In this study 'parent' refers to biological parent. Only data for biological parents and not stepparents or adoptive parents were analyzed. Our results are not adjusted for living with a single parent or in a foster home. Many children live with one biological parent only and may have a more or less sporadic relationship with the other biological parent.

[4] 78w The starting point of the study was a national population cohort of every child born in the period 1983-1989. All individuals fitting our definition of suicide attempt were extracted from the cohort, and 20 randomly selected controls were matched to every case. We expect the controls to be representative of Danish children/adolescents who have never attempted suicide. All the risk factors used in this study were created by using data from registers whose estimates are without recall bias.

[5] 109w All registers used are longitudinal national population registers. The registers contain all contacts with medical and psychiatric hospital units, including dates and diagnoses, all drugs bought on prescription, including dates and ATC codes, all deaths in Denmark, including time and cause of death. Other registers used contain information about changes in socio-economic and demographic conditions. All citizens in Denmark have equal access to health care services free of charge, but it is not possible to obtain data on mental illnesses that are not treated in a medical or psychiatric hospital unit. Some psychopharmacological drugs bought on prescription can be a good proxy for milder mental illnesses not requiring hospitalization.

[6] 47w By virtue of the size of the population used, this study has high power and valid estimates of risk factors for suicide attempts. We adjusted the risk for confounder effects and analyzed risk from mental illness at an adjusted level, in order to avoid type 1 errors.

[7] 51w This study confirms findings in the existing literature that psychiatric illness is a significant risk factor for suicide attempts among children and adolescents (Beautrais, 2000); and that among them, it is young people suffering from psychoses, depression, substance abuse or personality disorders who have the highest risk (Gould et al., 2003).

[8] 74w Past studies provides us with evidence that the same is the case for completed suicides (Fleischmann, Bertolote, Belfer, & Beautrais, 2005). The adjusted analyses show that, after adjusting for other factors, depression and personality disorders remain independent risk factors. This means that the effects of these illnesses on the risk of attempting suicide cannot alone be explained by number of previous contacts, duration of contact, increased risk after discharge and effects of parental confounders.

[9] 391w Young people suffering from emotionally unstable personality disorder are often self-harming and attempt or commit suicide due to a lack of ability to control impulses (Lynch, Trost, Salsman, & Linehan, 2007;Oumaya et al., 2008). In this study, mild depression is a high-risk factor for suicide attempts, even higher than moderate and severe depression. It is possible that mild depression contains a higher degree of agitated depression and more existential cognition about hopelessness, which may provide the energy to act on suicidal impulses. A high degree of agitation may lead to more breakups of relationships and more stressors than among other people of the same age. Moderate and severe depression may lead to a higher degree of melancholy, a lower degree of energy and increased fatigue; on the other hand, psychiatrists might be more likely to diagnose children and adolescents' depression as mild. Other studies have shown some association between OCD and suicidality, but in this study OCD turned out to be a significant independent protective factor (Hawgood & De Leo, 2008). Our analysis may have included important confounders that are capable of making an unbiased estimate of OCD's effect on suicide attempt risk in a young population. Anxiolytics also became an independent significant protective factor for suicide attempts. Anxiolytics can have the same anti-suicidal effect as lithium by reducing aggression (Mu ¨ller-Oerlinghausen & Lewitzka, 2010). Young people with many previous contacts with psychiatric departments are particularly at risk of attempting suicide. Many contacts may indicate more severe mental illness, which impedes recovery, but admission itself can be a significant stressor. The suicide attempters were more likely to be given multiple diagnoses and many different types of medical treatment. It could be that the symptoms of their mental illness are less distinct, which may result in less structured therapy: offering patients a variety of therapeutic and pharmacological treatments without knowing which is the most effective, thus impeding recovery. Past literature contributes evidence that individuals suffering from co-morbid substance use disorder are more likely to have suicidal behavior than other mental illness patients (Christiansen & Jensen, 2009). Perhaps these individuals suffer from more severe mental illness, have more serious personality problems or interpersonal relationship problems, more difficulties maintaining treatment and/or less access to treatment. Only a few of the Danish treatment centers are able to handle young people with co-morbid substance use disorder.

[10] 79w Our analysis shows that prolonged hospitalization or out-patient treatment is a protective factor, compared to all other contacts. Prolonged hospitalization or out-patient treatment might give the child/adolescent the necessary time to recover from their mental illness and time to develop skills that enable them to cope with new or old stressors. Maintaining adolescents suffering from personality disorder and high impulsivity in prolonged treatment can be difficult, and this group may require specially targeted out-patient treatment (Lynch et al., 2007).

[11] 184w The analysis of parents' level of income as a protective factor for children/adolescents' risk of suicide attempts after contact with a psychiatric department shows that young people from families with parental income in the lowest third have the highest risk of suicide attempts. The level of risk in this group is significantly different from the level of the other groups. There is probably an overrepresentation in this group of parents suffering from mental illness or substance abuse, who exhibit suicidal behavior, are unemployed or receive invalidity or early retirement pension, have a low level of education and a low level of socio-economic status (SES). These parents may lack the ability to show compassion and support their children after discharge from psychiatric units (Geckova, Van Dijk, Stewart, Groothoff, & Post, 2003;Huurre, Eerola, Rahkonen, & Aro, 2007;Mickelson & Kubzansky, 2003). Lack of social support may increase the suicide attempt risk but may also have delayed treatment of the child/adolescent's illness. The social authorities and the secondary healthcare system need to focus on especially these families in order to prevent suicide attempts after discharge from psychiatric units.

[1] 300w Denmark has a large number of longitudinal administrative and research registers. Originally, the registers were used for administration of payment between authority and service provider, but they also provide a unique opportunity for longitudinal register-based research. Every individual in Denmark has a civil personal registration number (CPR number) (Pedersen, Gotzsche, Moller, & Mortensen, 2006). CPR numbers were used to merge data from various registers. All personal data were extracted from longitudinal national registers, which included The Danish Fertility Database (Knudsen, 1998), Cause-of-Death register (Juel & Helweg-Larsen, 1999), The National Patient Registry (Andersen, Madsen, Jorgensen, Mellemkjoer, & Olsen, 1999), The Danish Psychiatric Central Register (Munk-Jorgensen & Mortensen, 1997), Register of Families and Household, Register of Unemployment, The National Register of Persons and The Register of Medicinal Product Statistics. Population A complete extraction of data on every individual born in the period 1983-1989 and ever living in Denmark was made. A total of 403,431 individuals were extracted. Of these 403,431 individuals, 3,465 had attempted suicide, according to our definition (see below), during the period from his or her 10th birthday to 31 December 2005. For all individuals registered as having attempted suicide, the date of the index suicide attempt was identified. In order to control for the risk in time for different individuals and for confounder effects from gender, age and calendar-time, a nested casecontrol study was designed. Each index case was matched with 20 non-suicidal controls for age and gender and being alive at the time of the index suicide attempt. A total population of 72,765 individuals was used to analyze the risk of suicide attempts after contact with a psychiatric department. A link to the child/adolescent's biological parents was established, and a population of parents created. In some cases it was not possible to identify the biological father or mother.

[2] 116w A suicide attempt was defined as closely as possible to WHO's definition of suicide attempts (Bille-Brahe et al., 1995). That is, it was defined as contact with medical services given the contact code E4 (the reason for contact with medical services was suicide attempt) and an ICD-10 diagnosis in the range of (S617-S619, X60-X84, T36-T60, T65). Where ICD-8 was used, suicide attempt was defined as a diagnosis in the range of E9500-E9599. Contact with psychiatric hospital given a suicide attempt diagnosis (X60-X84) was also defined as a suicide attempt. Only contacts after a child's 10th birthday were analyzed, to avoid inclusion of children admitted to hospital due to a parent's and not their own suicide attempt.

[3] 77w Contact with a psychiatric department was used as the independent variable in our models. The term refers to use of psychiatric services provided by psychiatric hospital units in Denmark, including hospitalization, outpatient treatment, emergency psychiatric services and all other day and night services. Services provided by general practitioners and non-hospital psychiatric practitioners were not included. Time since last contact with psychiatric department was defined as the time since end of treatment or contact with psychiatric care unit.

[4] 69w Admissions to psychiatric wards on the day of an attempted suicide were excluded from the analysis, to ensure that contact was not an outcome of a suicide attempt. As a consequence, suicide attempts made on the first day of admission to a psychiatric department were excluded from the analysis, which means that the suicide attempt risk during contact with psychiatric department is likely to be underestimated in the analysis.

[5] 82w For individuals having previous contact with a psychiatric department, the time-span from the most recent contact to the suicide attempt was calculated. The individuals were then divided into 13 time-span groups according to time since last contact (no contact, unknown date of discharge, suicide attempt during contact, one week, two weeks, four weeks, three months, six months, one year, two years, three years, five years and more than five years). Individuals with no contact with a psychiatric department were used as baseline.

[6] 34w The number of previous contacts with a psychiatric department was counted and divided into five groups (no contact, one or two times, three to five times, six to ten times, more than ten times).

[7] 34w The length of the most recent contact with a psychiatric department was calculated, and individuals were divided into six groups (no contact, 0-7 days, 8-30 days, 1-3 months, 3-12 months, more than a year).

[8] 64w Fourteen groups of diagnosis in accordance with ICD-10 were created, in order to make it possible to decompose the ICD-10 main groups into smaller subgroups (see Table 1). Diagnoses given in the period when ICD-8 was used were converted into ICD-10 diagnosis groups on the basis of WHO's conversion Table 3 ( WHO, 2004). Diagnoses given during previous contacts were included in the groups.

[9] 99w It was also possible to extract data about prescription drugs prescribed for individuals in the period 1995 to 2006. The follow-up period was therefore not the same for every case in this study, but was equal for every individual forming the case's stratum. The drugs prescribed were divided into six groups by using the drugs' anatomical therapeutic chemical code (ATC). The groups were 'any kind of drugs', 'antipsychotics' (ATC: NO5A), 'anxiolytics' (ATC: N05B), 'psychoanaleptics' (ATC: N06, which includes, among others, all kinds of antidepressants and drugs used for ADHD), 'antiepileptics' (ATC: N03), and 'drugs for substance dependence' (ATC: N07B).

[10] 89w For each individual in the study we counted the number of different diagnoses given and the number of different drugs used. For cases and controls the frequency distribution of the two variables was calculated and results plotted on a chart. The two variables were recorded as dummy and analyzed in a conditional logistic regression, as a predictor of suicide attempt. For cases, we calculated the Pearson correlation coefficient for the number of different diagnoses given and number of different drugs used in the period prior to the suicide attempt.

[1] 293w Background: There seems to be an increased risk of children and adolescents committing or attempting suicide after contact with a psychiatric department. Children and adolescents living in families with low socio-economic status (SES) might have an especially increased suicide attempt risk. Methods: A complete extraction of Danish register data for every individual born in the period 1983-1989 was made. Of these 403,431 individuals, 3,465 had attempted suicide. In order to control for confounder effects from gender, age and calendar-time, a nested case-control study was designed. A total population of 72,765 individuals was used to analyze the risk of suicide attempts after contact with a psychiatric department. The case-control data were analyzed using conditional logistic regression. Results: This study shows that a child/adolescent's risk of suicide attempt peaks immediately after discharge from last contact with a psychiatric department. The risk of suicide attempt is highest for children and adolescents suffering from personality disorders, depression and substance use disorders. Children and adolescents with previous contact with a psychiatric department and parental income in the lowest third have a significantly higher risk of suicide attempt. Suicide attempters were more likely to have been given several different diagnoses and several different psychopharmacological drugs prior to their attempted suicide. Conclusions: The findings in this study highlight the need for psychopathology assessment in every case of attempted suicide. This study also shows that well-known risk factors such as contact with a psychiatric department do not affect all individuals in the same way. Individuals from families with low SES had the highest risk. This suggests that the presence of factors influencing both vulnerability and resiliency, e.g., family level of SES, needs to be included in the assessment. Keywords: Suicide attempts, contact with psychiatric department, resiliency, low socio-economic status, psychopharmacological drugs.

[2] 99w Although it has been established that the risk of suicide and suicide attempt is a multifactorial problem, it seems that mental illness could be the strongest risk factor (Agerbo, Nordentoft, & Mortensen, 2002). It is also well established that for adults the risk of suicide and suicide attempts peaks in the first weeks after admission and discharge from psychiatric care (Christiansen & Jensen, 2009;Gould, Greenberg, Velting, & Shaffer, 2003;Qin & Nordentoft, 2005). Similarly, there might be an increased risk of children and adolescents committing or attempting suicide after admission and discharge from a psychiatric ward (Stenager & Qin, 2008).

[3] 172w Not all mental disorders carry the same risk of suicide attempts. For young people, substance use, affective and personality disorders are among the strongest risk factors, and the risk increases among those with co-morbid disorders (Beautrais, 2000). So when trying to predict the risk of suicide attempt in children and adolescents, it is important to screen for mental illnesses; however, the clinical value is low because many young people suffer from mental illness and only some of them will have attempted suicide (Steele & Doey, 2007). Knowledge about which mental illnesses increase the risk of suicidal behavior among children and adolescents is therefore important. A number of studies based on populations of children and adolescents have analyzed psychiatric diagnoses as risk factors for suicide attempts, but only a few of them have been population based and focused on high-risk periods (Spirito & Esposito-Smythers, 2006). The first weeks after admission and discharge from psychiatric care might be the period when a child or adolescent is at a very high risk of suicidal behavior.

[4] 99w Even supposing the first weeks after discharge are a high-risk period for attempted suicide, not all young people may be affected to the same extent. This raises the interesting question about which individuals have, and which individuals do not have, increased vulnerability after discharge from psychiatric care. It has been suggested that social support and family connectedness may decrease the risk of suicidal behavior (Borowsky, Ireland, & Resnick, 2001;Kandel, Raveis, & Davies, 1991;Lewinsohn, Rohde, & Seeley, 1993). Resiliency explanations suggest a search for protective factors that may mitigate the risk of suicide attempts after discharge from a psychiatric ward.

[5] 53w Low income is normally associated with low socioeconomic status, including unemployment, low educational achievement, invalidity and early-retirement pension. Parental level of income therefore seems to be a good proxy for parents' resources and ability to support a child or adolescent diagnosed with mental illness (Fergusson, Beautrais, & Horwood, 2003;Fergusson, Woodward, & Horwood, 2000).

[6] 75w The vulnerability and resiliency explanations are to some extent interchangeable (Fergusson et al., 2003). For example, hospitalization in a psychiatric department might increase vulnerability to suicide attempts, but for some individuals contact with a psychiatric department might also be a protective factor. These facts need to be included in our study of the process from mental illness to suicidal behavior. One way to do this is by examining the interaction between risk and protective factors.

[7] 8w The aim of this study was to analyze:

[8] 66w 1. children and adolescents' risk of suicide attempts after contact with a psychiatric department; 2. mental illness, measured by use of psychopharmacological drugs and contact with a psychiatric department in children and adolescents as risk factors for suicide attempts; 3. the interaction between parental socio-economic status and the child/adolescent's risk of suicide attempt after discharge from psychiatric care; 4. the interaction between diagnoses and psychopharmacological drugs.

[9] 42w The parental level of income was measured as the income in the year before the child/adolescent's suicide attempt. The income was adjusted for inflation and divided into thirds. Missing information about income was coded as a dummy and included in the analysis.

[10] 76w All factors were analyzed at a crude level (simple regression) and at an adjusted level where all factors were included in the model (multiple regression). In order to adjust the rate ratios for potential confounder interaction from parents' contact with a psychiatric department, level of income and education, use of psychopharmacological drugs, death and parents' marital status, these factors were included in the adjusted model, in an attempt to avoid Type 1 errors (false positive conclusions).

[11] 87w A conditional logistic regression model with all firstdegree interactions included was used to analyze parents' socio-economic status, measured by level of income, as moderators of risk of suicide attempt. The dichotomous variable 'suicide attempt' (0 = non-attempter, 1 = attempter) was the outcome. The dichotomous variable P 'contact with psychiatric department' (0 = No, 1 = Yes) was the predictor and a categorical variable M (father's or mother's level of income) was the moderating variable (0 = level one, 1 = level two, 2 = level three).

[12] 31w Let p be the probability of the outcome (suicide attempt = 1). Then logit p 1Àp À Á is the logit function of the odds. The conditional logistic regression then is:

[13] 6w This will return the following coefficients:

[14] 5w Contact with psychiatric department (P)

[15] 57w A meaningful interpretation of the coefficients is then: b0/0 is baseline (reference group), (no contact and moderator at lowest level) b1/0 (contact and moderator at lowest level) b0/1 (no contact and moderator at medium level) b1/1 (contact and moderator at medium level) b0/2 (no contact and moderator at highest level) b1/2 (contact and moderator at highest level)

[16] 17w The effect of the moderator is estimated by examining for sign and significance of b1/0, b1/1, b1/2.

[17] 78w The analysis of time since last contact with a psychiatric department showed a considerable increase in risk in the first weeks after last discharge date. The risk decreased as time since last contact passed. The risk was highest immediately after discharge from last contact. The rate ratios were in the range of 2.21 to 362.97 with some overlapping confidence intervals. The risk of suicide attempt during hospitalization or out-patient treatment was high (rate ratios 18.27, p < .001).

[18] 30w The risk of attempted suicide went up with increasing numbers of previous contacts. Individuals with more than 10 contacts had a 70.9 times higher risk than individuals with no contact.

[19] 31w The risk of attempted suicide fell with increasing length of contact. Individuals with contact length of more than 1 year had a 4.45 times higher risk than individuals with no contact.

[20] 67w The suicide attempt risk was not equal across all diagnoses, but all diagnoses were significant risk factors at a crude level. Personality disorder, emotionally unstable personality disorder, affective disorder, depression and disorder due to substance use showed the highest level (RR > 20) at the crude level. Organic mental disorders, OCD and behavioral disorder were the groups of disorders with the lowest rate ratio (RR < 10).

[21] 42w Table 3 shows that cases were more likely to have bought prescription drugs. The suicide attempt risk was not equally distributed across all kinds of psychopharmacological drugs. Antipsychotics, psychoanaleptics and drugs for substance dependence showed the strongest correlation with suicide attempt risk.

[22] 134w The findings of this study highlight the need for psychopathology assessment in every case of attempted suicide. A valid present-state examination system could be a good support tool for identification of psychopathology symptoms. The assessment should include a judgment of the risk of suicidal behavior. As the mental illnesses of suicide attempters perhaps present with less distinct symptoms, there is a risk that any subsequent treatment becomes less effective, thus impeding recovery. This study shows that well-known risk factors, such as admission to a psychiatric department, do not affect all individuals in the same way. Only individuals from families with low SES were at a high risk in this analysis. This suggests that the presence of factors influencing both vulnerability and resiliency, e.g. family level of SES, needs to be included in the assessment.

[23] 166w The findings also highlight the need for coherent treatment involving the secondary healthcare system and the social authorities. In Denmark, it is parents who have the basic responsibility for their child's development and well-being, but the social authorities have a duty to supervise the child's wellbeing. This includes an obligation to offer advice and guidance, and to take action if the parents are unable to take care of their child. Very few places in Denmark have formal co-operation between psychiatric care units and social services on the treatment of suicidal children/adolescents. As a consequence, recently diagnosed children risk being discharged from psychiatric care without proper support. Our study shows that support measures must start immediately after discharge, in order to prevent suicide attempts. It is vital that the secondary healthcare system and the social authorities uphold the principle of following a recently discharged young person closely, offering support and making sure that vulnerable young persons are never left to cope with their problems on their own.