Clinical characteristics in schizophrenia patients with or without suicide attempts and non-suicidal self-harm - a cross-sectional study

Clinical psychologists or medical doctors, with formal training in use of the assessment protocol, interviewed the patients when in a stable phase. The Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I) [21] was used for diagnostic purposes. Mean overall kappa for SCID diagnoses as assessed in the training course was 0.77. Age of onset was defined as age at the first SCID-verified psychotic episode. Clinical symptoms were assessed using the Positive and Negative Syndrome Scale, PANSS [22]. All interviewers participated in inter-rater reliability testing that entailed rating of patient videos. Inter-rater reliability was acceptable with intra-class correlation coefficients [23] for PANSS subscales ranging from 0.71 to 0.73. The PANSS item G14 (Disordered regulation and control of action on inner urges/emotions) was used as a proxy for impulsive aggression in the past week. Suicidal ideation last week was measured using five items from the InterSePT scale for suicidal thinking [24]. Duration of untreated psychosis was measured as time from psychosis onset until start of adequate treatment for psychotic disorder [25].

The concept self-harm (SH) [26] was subdivided into SA, defined as ‘self-harm with the intent to die’, and NSSH, defined as ‘self-harm without suicide intent’. NSSH thus encompasses both non-suicidal self-injury and self-poisoning for purposes of self-injury without suicidal intent. Information on lifetime episodes of self-harm was based on a semistructured interview including the following question adopted from a previous European study of self-harm (CASE-study): “Have you ever deliberately taken an overdose (e.g., of pills or other medication) or tried to harm yourself in some other way (such as cut yourself)?” [26, 27]. The response options were “no”, “yes” and if yes, “the specific number of times” with the follow-up question: “how many times did you try to kill yourself?” (SA) and “How many times did you harm yourself without a wish to die” (NSSH). The presence of any suicide intent led to classification of the behavior as a suicide attempt, also in the presence of seemingly contradictory additional reasons given for the behavior (such as “I wished to get help from someone”). In addition to the self-harm question, the classification was based on an open description of the participants’ most recent episode of self-harm, whether they considered the behavior a suicide attempt (yes, no, uncertain), and on answers to questions about what they wanted to achieve with the act (“I wished to die” (yes/no), “I wished to get help from someone” (yes/no), “I wished to escape an unbearable emotion” (yes/no), ”Other reasons (describe)”). The principal author reviewed all classifications of self-harm variables and in the few cases (n = 6) with a discrepancy, a consensus was reached with the senior author. This defined three groups relative to the presence of SA: 1) NoSA (patients with no SH and NSSH only), 2) SA only (participants with history of SA, but no NSSH) and 3) SA + NSSH (participants with both SA and NSSH) (see Figure 1).

Group differences were analyzed using Chi-square tests for categorical variables, one-way ANOVAs (post-hoc Scheffé’s tests) for normally distributed continuous variables and Kruskal-Wallis test (post hoc Mann–Whitney U test) for non-normally distributed continuous variables. Pairwise comparisons were performed comparing the SA + NSSH group with the SA only and with the NoSA groups, based on the research questions. All tests were two-tailed with α < 0.05. Two multinomial logistic regression models were generated to answer the two research questions. Backward stepwise procedures (p in = 0.05: p out = 0.10) were used to search for the models that best differentiated between participants with SA + NSSH, SA only, and NoSA history. DUP and impulsive aggression (G14) were skewed and dichotomized at the median (DUP > 52 weeks/G14 > 1) in the regression analysis. We also included variables that were differentially distributed over groups and thus could confound relationships. Current age and age of first psychotic symptoms were highly correlated and since age of first psychotic symptoms was of primary interest, this measure was retained in the models. We used the PASW statistics v.18 software (SPSS Inc., Chicago, Illinois).

A total of seventy-five patients (30%) reported one or more episodes of NSSH. The median number of NSSH episodes was six (min = 1, max ≥ 50), 22 patients (9% of total sample) reported 50 episodes or more. Women were significantly more likely to report at least one NSSH episode (43% (n = 46) vs. 20% (n = 29), χ2 = 16.00, DF = 1, p < 0.001). When the most recent self-harm episode was NSSH (n = 54), it had little or no risk of death (n = 52/96%), the behavior had been planned for less than an hour in 35 cases (65%), and the most commonly used method was cutting (n = 42/78%). Among those with SA + NSSH, almost all patients reported cutting as method, when the most recent episode was NSSH (n = 17/94%), indicating that NSSH in this group primarily consists of self-injury. One or more episodes of SA were present in 88 patients (35%), the median number of SA among these was two (min = 1, max = 7). Significantly more women reported history of SA (47% (n = 50) vs. 26% (n = 38), χ2 = 11.82, DF = 1, p = 0.001). When the most recent self-harm was SA (n = 63), the lethality was rated as moderate to high in 45 cases (77%), and the most frequent method was self-poisoning (n = 34/54%). Patients with a lifetime history of NSSH were significantly more likely to report SA (n = 36/48%) than patients without NSSH (n = 52/30%, χ2 = 7.9, DF = 1, p = 0.005). Fifty-two patients (21%) had a history of SA only while 36 (14% of total) had a history of SA + NSSH. The SA + NSSH group was significantly more likely to have repeated SA (n = 28/78%) compared to the SA only group (n = 17/33%, χ2 = 17.3, DF = 1, p < 0.001). Patients with SA + NSSH were significantly more often female and younger at the time of participation than patients with SA only or NoSA (Table 1).

Patients with SA + NSSH were significantly younger at psychosis onset than participants with SA only or NoSA (mean difference 6.7 and 4.9 years respectively), and had longer DUP (Table 1). SA + NSSH patients were also significantly younger at first contact with mental health services and more likely to have had their first mental health contact for other reasons than psychotic disorder than the other patients. Furthermore, they were significantly more likely to report to have had depressive episode(s) before the age of 18.

The SA + NSSH group had significantly higher levels of depressive symptoms and were more likely to report current impulsive aggression or suicidal ideation compared to the SA and NoSA groups (Table 2). There were no group differences in current levels of negative symptoms or total PANSS score. There was, however, a trend towards group differences on positive symptoms (p = .08), and significant group differences were observed on the delusions and hallucinatory behavior items from the positive symptoms score, with those with SA + NSSH scoring significantly higher than both other groups on current hallucinatory behavior. One third of the SA + NSSH patients did not use any antipsychotic medication; a significantly higher fraction than among NoSA patients. Regarding recent SA, there were no significant differences between the SA + NSSH and SA only groups in lethality (moderate to high: 75% vs 77%, χ2 = 0.0, DF = 1, p = 0.854), degree of hospital treatment (41% vs 49%, χ2 = 0.3, DF = 1, p = 0.583), degree of premeditation (less than an hour: 41% vs 43%, χ2 = 0.0, DF = 1, p = 0.887), or in method used (self-poisoning: 59% vs 51%, χ2 = 0.3, DF = 1, p = 0.583).

In the first model we investigated the multivariate relationships between group memberships and illness history (Aim 1, i.e. ages at onset of psychosis and depression) together with potential confounding variables of this relationship (here DUP and gender). All four variables had a significant contribution to the model (Table 3). The SA + NSSH group could be differentiated from the NoSA group by a) being younger at the time of onset of their first psychotic symptoms, b) more often having had episodes of depression before the age of 18, c) being more often female and d) more often having a DUP of more than a year. The SA only group could only be differentiated from the NoSA group by more often having had episodes of depression before the age of 18. To test whether the SA + NSSH and SA only groups were significantly different on these measures, we performed identical multinomial analyses with the SA + NSSH group as reference category (see Additional file 1). These analyses confirmed that the SA + NSSH group differed significantly from the SA only group by a) being younger at first psychotic symptoms, b) being more often female and c) more often having had a DUP of more than a year. There were no significant differences between the SA + NSSH and SA only groups in having had episodes of depression before the age of 18.

In Model 2 we added current symptoms and behavior to the illness history variables entered in model 1 to address the second aim of this study. Levels of current impulsive aggression and depressive symptoms contributed to differentiate between the groups. The SA + NSSH group had higher scores on current impulsive aggression and depressive symptoms compared to the NoSA group. Current depressive symptoms also differentiated the SA only group from those with NoSA, while current impulsive aggression did not. Once current depressive symptoms were added, length of DUP no longer contributed significantly to the model (Model 2, Table 3). Current medication did not significantly differentiate between groups in the multivariate analyses. The supplementary analysis with SA + NSSH as reference category showed that the SA + NSSH group scored significantly higher on both current impulsive aggression and depressive symptoms compared with the SA only group (Additional file 1).

Strengths: The strengths of the study include the relatively large and representative sample recruited from publicly funded catchment area based services serving all socio-economic classes as well as the use of comprehensive clinical assessments performed by qualified and trained clinical interviewers. Limitations include the cross-sectional design preventing inferences about causal relationships, the general problems with recall bias for retrospective data, the lack of data on the age of onset of SA + NSSH and the use of PANSS item G14 as a proxy measure for impulsive aggression. The relatively small size of the groups of interest increases the risk of Type II error. Findings must thus be interpreted with some caution.