A total of 88 youths aged 13 to 18 years, 53 girls and 35 boys, were consecutively referred from 1 March 1997 to 31 December 1999 for psychiatric treatment to the adolescent psychiatric inpatient unit of Kuopio University Hospital, Finland. Inclusion criteria in the present study were that the adolescent was: 1) between 13 and 18 years of age, 2) not referred only for brief intervention in response to a temporary crisis (less than four weeks), 3) able to co-operate, and 4) suffered from a clinically significant psychiatric disorder. Adolescents exhibiting primarily substance abuse or dependence and those who needed only a brief intervention were referred to another treatment setting. Sixty-five adolescents met the inclusion criteria, 1 female and 2 males declined to participate 24. All patients and their parents provided written informed consent before entering the study. The study was approved by the Ethics Committee of Kuopio University Hospital and the University of Kuopio.

At study entry, during a semi-structured interview, a psychiatrist screened the adolescents for suicide attempts before the inpatient treatment (no/yes/when). We excluded inpatients with psychotic disorders (n = 7). The sample (n = 55) was then divided into those who had attempted suicide during the previous 12 months (suicide attempters; n = 16) and those with no previous suicide attempts (non-suicidal patients; n = 39).

The treatment was part of an existing clinical service and the assessment of relationship and outcome variables were included in the service. Treatment schedules were individualized from a range of psychosocial interventions, and included individual psychotherapy, psychotropic medication when appropriate and sport, occupational, and art group therapy. Each adolescent received individual psychotherapy, two 45-minute sessions weekly, conducted by the case manager nurse, who was a member of a multidisciplinary team. The nurses had been trained in a two-year course on psychodynamic psychotherapy as a part of multidisciplinary treatment and psychotherapy was conducted under supervision. Case manager nurses also interacted with other patients in the ward but did not have regular sessions with them. Family sessions were conducted once every three weeks, mostly led by a social worker. The treatment included specialized hospital school. The adolescents spent every second weekend at home. Adolescents' treatment schedules were focused and evaluated individually at entry, after two weeks and then once every six weeks by the multidisciplinary team. To determine the appropriate time for discharge from hospital the medical staff team assessed psychosocial functioning, suicidal, and the safety of the foster environment. They also discussed with the adolescent and her/his parents whether it was safe to enter outpatient treatment. The decision to discharge the patient was made by a psychiatrist after these assessments.

The two-group pre-post test prospective study design included assessments by a psychiatrist, a psychologist and medical staff members as well as self-rated measures collected at entry and on discharge. The medical staff of the hospital ward was trained to perform all procedures and assessments in a standardized manner for research purposes. DSM-III-R diagnoses were assigned using the SCID 25, and thereafter transformed to DSM-IV diagnoses by checking each DSM-IV diagnostic criterion from the SCID-interview, patient files and other data collected for the study. Multiple psychiatric diagnoses were allowed. General IQ was 70 or more in all participants, and the participants had not been diagnosed as having any neurological diseases.

Data on parental basic socio-demographic and socio-economic backgrounds included: 1) marital status of the biological parents (married, divorced, separated (each: no/yes)), 2) family on living allowances, 3) mental health problems among family members and 4) death of the mother or father (no/yes). Data on family characteristics were collected during semi-structured interviews with the parents conducted by staff members. The patients' academic achievement, including behavioral problems (many/some or none), were assessed at entry together with hopelessness (one item yes/no) and alcohol and drug misuse in a semi-structured interview by a psychiatrist. Data on indirect self-destructive behavior defined as repeated exposure to life-threatening danger without suicidal intent and having been bullied at school during the previous six months were collected at entry. Prior psychiatric treatment and the time interval between suicide attempts and referral to inpatient treatment were assessed. Anhedonia, current depressive mood, and suicidal ideation were assessed using the questions of the SCID I interview. Medical staff members recorded the number of treatment interventions daily (therapy sessions with the individual case manager nurse, assessments conducted by a psychiatrist, a psychologist, a social worker and an occupational therapist, family sessions, treatment meetings of the staff-team, group therapy sessions and number of school days).

A team of 6–8 staff members assessed the psychosocial functioning of the patients at entry and on discharge using the Global Assessment Scale (GAS) 26. The adolescents' cognitive performance (FIQ) was assessed by a clinical psychologist using five sub-tests of the Wechsler Adult Intelligence Scale – Revised, adapted for use in the Finnish population 27. The verbal intellectual level (VIQ) was assessed using three scales (Similarities, Vocabulary and Digit Span) and nonverbal performance intelligence (PIQ) was assessed using two scales (Digit Symbol and Block Design). Standard points were adjusted to correspond to the IQ 27. The overall sum of standard points was the combined sum. The sum of standard points (Similarities, Vocabulary, and Digit Span) was multiplied by 6/3, while the sum of standard points (Digit Symbol and Block Design) was multiplied by 5/2. To be consistent, we converted the raw scores from sub-tests to scaled score equivalents using the corresponding norms for the ages of 16–17 years. The subscales of Similarities, Digit Span, Digit Symbols and Block Design that were used in this study are suitable for emotionally disturbed patients, and their scaled scores do not differ from those of the Wechsler Intelligence Scale for Children 2829. A Logical Prose Subtest, Story recall A, from the Wechsler Memory Scale WMS; 30 was used as a measure of immediate and delayed logical verbal memory. The first score was based on the units immediately recalled. The delayed score was based on the units recalled from the story 60 minutes later. The List Learning Test (LLT) was used to assess learning and recall 31. The participants learned and recalled aloud 15 semantically unrelated words in four consecutive trials, using version A at entry and B after six months or on discharge. The word lists were used alternatively, eliminating the effect of learning.

The Offer Self-Image Questionnaire (OSIQ) for adolescents was used to assess the self-images of the patients 323334. The OSIQ has been validated among Finnish adolescents 3536. In this study, Psychological Self (impulse control, emotional tone, body image), Social Self (social relations, vocational and educational goals) and Familial Self (family relationships) were assessed. A higher score in the OSIQ indicates better functioning.

The student's t-test, chi-squared test, Fisher's exact test and exact Mann-Whitney U-test were used to compare suicide attempters and non-suicidal subjects. Paired samples t-test or exact Wilcoxon signed ranks test were used to analyze differences between entry and discharge and repeated measures analyses of variance with covariates (suicide attempt or not, current depressive mood, anhedonia at entry, number of days attending school, number of individual therapy sessions, psychotropic medication on discharge) was used to analyze changes in cognitive performance and self image. All tests were applied on a two-tailed basis. Normality and homoscedasticity were visually checked.

Univariate logistic regression analyses were applied to define variables to be included in the final multivariate model associated with the dependent variable (improvement 10 points or more in GAS score). All the explanatory variables were first entered into a univariate model as single predictors (age, gender, patient in the suicidal group, number of therapy sessions, duration of inpatient treatment, GAS score at entry, an improved body-image, number of team sessions, MDD diagnosis (no/yes), number of group sessions, number of Family sessions, Verbal cognitive performance score on discharge, medication on discharge, Psychological Self, Social Self and Familial Self). Being classified as a suicide attempter or not and the predictors showing statistical significance of less than 0.2 in the univariate analysis were included in the multivariate stepwise logistic regression model. Backward selection procedure was used in order to find the final reduced model. Logistic regressions were used also when considering whether suicide attempt history related to categorical clinical variables (suicidal ideation, indirect self-destructive behaviour, depressive mood, anhedonia) adjusted for baseline suicidal ideation. The results are expressed as ORs (odds ratios) with their 95% confidence intervals. In interpreting the results of the final model, the level of significance was set at 0.05. Statistical analyses were performed with the statistical package SPSS for Windows 14.0, and post hoc sample size estimation was carried out with the nQuery Advisor^® 4.0 program.

The proportion of girls was 88% (14/16) in the suicide attempter group, and 59% (23/39) among the non-attempters. Suicide attempts had taken place during the previous 12 months (median 27 days). Eight (50%) of these patients had attempted suicide during the previous month. In the suicide attempter group, there were seven adolescents with one, two with two, and one with three or more suicide attempts in the prior year. No one of the non-attempters had attempted suicide prior to the 12-month timeframe. The mean age of the suicide attempters was 15.4 years (SD 0.8, range 14–17 years) and that of the non-attempters 15.1 years (SD 1.0, range 13–17 years; p = ns). The principal psychiatric diagnoses (DSM-IV) in the groups (suicide attempters/non-attempters) were major depressive disorder (n = 13/15, z = 2.88, p = 0.004), dysthymia (n = 0/3, z = 1.14, p = ns.), bipolar disorder (n = 2/6, z = 0.29, p = ns.), conduct disorder (n = 1/10, z = 1,62, p = ns.), eating disorders (n = 0/3, z = 1.14, p = ns.) and anxiety disorders (n = 0/2, z = 0.92, p = ns.). Eleven (69%) of the suicide attempters and 23 (59%) of the non-attempters had comorbid psychiatric disorders (z = 0.69, p = ns.).

Suicidal ideation, hopelessness, current depressive mood, indirect self-destructive behavior, and alcohol and drug misuse were more common among the suicide attempters than non-attempters (Table 1). There was no statistically significant between-group difference in psychosocial functioning at entry (mean GAS score ± SD: suicide attempters 41.8 ± 8.3 versus non-attempters 39.3 ± 8.7, p = ns.). Fourteen (87.5%) suicide attempters and 32 (82.1%) non-attempters had previously utilized psychiatric services. A family history of being on living allowances was more common among suicide attempters than among non-attempters (Table 1).

Except for use of psychotropic medication during treatment, no statistically significant differences were found in treatment variables between adolescents with and without suicide attempts (Table 2). There was not statistically significant difference in the mean duration of inpatient treatment between suicide attempter and non-attempter adolescents (M = 175 days, SD = 99 vs. M = 145 days, SD = 81, respectively). Non-verbal cognitive performance scores at entry were better among the suicide attempters than among the non-attempters (Table 3; PIQ; p = 0.040). The Psychological Self score (p = 0.003), including the emotional tone (p = 0.003) and body-image sub-scale scores (p = 0.001), were significantly poorer among the suicide attempters compared with the non-attempters at entry (Table 4).

There was a statistically significant improvement in mean (± SD) GAS scores from admission to discharge in both groups: among suicide attempters from 41.8 (8.3) to 52.1 (9.1) (p = 0.003) and among those with no suicide attempts from 39.1 (9.7) to 51.2 (8.1) (p = 0.001). GAS scores did not differ between the groups on discharge. Two patients in the suicide attempter group (12.5%) and one (2.6%) in the non-attempter group attempted suicide during the treatment. All the attempters were females. The timing of the suicide attempts was close to the end of treatment among two subjects. The third patient made several suicide attempts during the treatment that lasted for five months. All the three patients received individual psychotherapy and they had family sessions. One of the attempters had bipolar I disorder, her medication consisted of mirtazapine and a mood stabilizer. The other attempter received citalopram and thioridatzine, while the third attempter received only anxiolytic medication. Seven (43.8%) suicide attempters and 7 (17.9%) non-attempters received SSRI medication (p = 0.046). Five (31.3%) attempters and 2 (5.1%) non-attempters received antidepressants remeron from the NASSA-group (p = 0.008). One (6.3%) attempter and 3 (7.7%) non-attempters received mood stabilizer (p = ns.). Six (37.5%) attempters and 8 (20.5%) received antipsychotic medication (p = ns.).

When adjusted for baseline suicidal ideation the rate of positive change in suicidal ideation was 55.6% among patients with suicide attempts and 21.1% among non-attempters with no statistically significant difference between the two groups (OR = 0.62, 95% CI 0.40–9.65). The rate of positive change in indirect self-destructive behavior was 46.7% among suicide attempters and 21.6% among non-attempters (OR = 2.6, 95%CI 0.50–13.7) and in hopelessness 80.0% among suicide attempters and 51.1% among non-attempters (OR = 6.28, 95% CI 0.61–64.9). Depressive mood declined in 35.7% among suicide attempters and in 33.3% of non-attempters (OR = 0.87, 95% CI 0.18–4.08), and anhedonia in 28.6% among the attempters and in 31.3% of the non-attempters, respectively (OR = 0.43, 95% CI 0.01–2.30).

In unadjusted comparisons general cognitive performance (FIQ), as well as verbal (VIQ) and nonverbal cognitive performance (PIQ), improved during treatment in both patient groups (Table 3). On discharge, suicide attempters performed tentatively better in the similarities subscale of verbal cognition than the non-suicidal patients (Table 3; p = 0.060). IQ was at the common average levels 27 both at entry and on discharge in both patient groups. Non-attempters' improved learning memory functioning during treatment.

In adjusted comparisons, the improvement in verbal cognitive performance, and in the vocabulary subscale, was more pronounced among the suicide attempters (Table 3, column 'Subjects Effects'). More severe depressive mood associated with less improvement in general cognition [F (1, 36) = 4.14, p = 0.049] and non-verbal cognition [F (1,36) = 7.26, p = 0.011] including the block design subscale [F (1, 36) = 6.95, p = 0.012] as well as the digit symbol subscale [F (1, 36) = 7.92, p = 0.008]. Use of medication associated with improvement in the non-verbal cognitive performance [F (1, 36) = 4.11, p = 0.050]. General cognition [F (1, 36) = 6.64, p = 0.014], and in non-verbal cognitive performance [F (1, 36) = 6.24, p = 0.017] and the digit symbol subscale [F (1, 36) = 4.15, p = 0.049] changed in both groups during treatment and the change were similar in the patients groups. The number of therapy sessions associated with improvement in immediate logical memory (Table 3). The higher the number of therapy sessions (Table 3 column "Main Effect") and the number of days attending school the better was improvement in the vocabulary subscale during treatment [F (1, 36) = 5.62, p = 0.023].

In unadjusted comparisons the total OSIQ scores in the Psychological Self, including impulse control, emotional tone and body image, improved during treatment in both patient groups. (Table 4). Among subjects with no suicide attempts, also Social self and Familial Self scores improved. In adjusted comparisons there were no between-group differences (Table 4, column "Subjects Effects"). The high number of therapy sessions associated with improvement in Psychological Self, the Body image subscale and Familial Self (Table 4, column "Main Effect"). More severe depressive mood tended to associate with less improvement in the Psychological Self [F (1, 36) = 4.02, p = 0.053] including improvement in the impulse control subscale [F (1, 36) = 5.06, p = 0.031]. The more severe anhedonia the lower was the change in the social relationships subscale [F (1, 36) = 4.85, p = 0.034].

Ten patients (63%) in the suicide attempter group improved in their GAS score by 10 points or more compared with 22 (56%) in the non-attempter group. The predictors showing statistical significance of less than 0.2 in the univariate analysis (number of therapy sessions, duration of inpatient treatment, GAS score at entry, an improved body-image and number of team sessions) and being classified as a suicide attempter or not were entered into the multivariate stepwise logistic regression model. Backward selection was used in order to find the final reduced model.

The final multivariate logistic regression model explained 26,2% of the variance in the change in psychosocial functioning (GAS score) during treatment (Table 5). Having an improved body-image associated significantly with a higher probability of improvement in psychosocial functioning while higher GAS score at entry was associated with lower probability of functional improvement.