خوداظهاری بیماران از خطر آینده آنها برای رفتار خشونت آمیز و خودآسیبی: بیماران بستری آینده نگر و مطالعه پیگیری پس از ترخیص در یک واحد روانی حاد

Abstract Recently patients' responsibility for and ownership of their own treatment have been emphasised. A literature search on patients'' structured self-reported assessment of future risk of violent, suicidal or self mutilating behaviour failed to disclose any published empirical research. The present prospective naturalistic study comprised all involuntary and voluntary acutely admitted patients (n = 489) to a psychiatric hospital during one year. Patients' self-reported risks of violence and self-harm at admission and at discharge were compared with episodes recorded during hospital stay and 3 months post-discharge. Patients' predictions were significant concerning violent, suicidal and self-injurious behaviour, with AUC values of 0.73 (95%CI = 0.61–0.85), 0.92 (95%CI = 0.88–0.96) and 0.82 (95%CI = 0.67–0.98) for hospital stay, and 0.67 (95%CI = 0.58–0.76), 0.63 (95%CI = 0.55–0.72) and 0.66 (95%CI = 0.57–0.76) after 3 months, respectively. Moderate or higher risk predictions remained significant in multivariate analysis, and risk of violence even after gender stratification. Self-harm predictions were significant for women. Moderate or higher risk scores remained significant predictors of violence one year post-discharge. Controlling for readmissions the results remained the same. Low sensitivity limits the clinical value, but relatively high positive predictive values might be clinically important. Still future research is recommended to explore if self prediction is a valid adjuvant method to established risk assessment procedures.

1. Introduction Self-harm and violent behaviour by psychiatric patients are important. Several studies and reviews have revealed increased violence among persons with major mental illnesses (Brennan et al., 2000, Colasanti et al., 2008 and Fazel and Grann, 2006). Most persons who commit suicide have a present mental illness, and a majority have symptoms of depression (Hawton and van Heeringen, 2009). Recently the focus on patients' responsibility for and ownership of their own treatment has increased. Traditionally, the patients' own opinion has been taken into account in the clinical risk management of suicidal behaviour. Self-report questionnaires for patients have been developed for both suicidal and violence risk assessment (Helfritz et al., 2006, Huth-Bocks et al., 2007, Kroner and Loza, 2001, Loza et al., 2007 and Nimeus et al., 2006). However, instead of measuring patients' perception of their own risk, these tools have been developed to obtain risk estimates by clinicians, or by computerized soft-ware programs. Moreover, there has been controversy about the reliability of self-report questionnaires (Doyle and Dolan, 2006, Gaynes et al., 2004, Hart, 1995 and Loza, 2007). Our literature search failed to show any empirical research on patients' self-reported “direct” opinion of subsequent violent and self-harm behaviour. Hence we set up a prospective study in the acute psychiatric unit at Ålesund Hospital. Other parts of the study were biological markers (serum lipids and platelet serotonin) and two risk assessment screens (Hartvig et al., 2007 and Sheehan et al., 1998).

3. Results 3.1. Characteristics of the samples The demographic and clinical characteristics of included and missing patients at hospital stay and after discharge are shown in Table 1. One hundred and forty nine (31%) patients had a history of previous suicidal attempts, 127 (26%) had a history of previous violence, and 277 (57%) had experienced suicidal ideation the last months before admission. The sub-sample of patients (n = 28) reporting won't answer the risk of violent behaviour was characterized by more involuntary admissions (P < 0.001), mandatory aftercare (P = 0.001), substance abuse (P = 0.050), and less affective disorders (P = 0.002) compared to the rest of the sample. Gender, age, hospital stay, anxiety disorders and personality disorders did not differ. More psychotic disorders in the “won't answer” sample were close to significance (P = 0.056). The subsample of patients reporting don't know the risk of suicide (n = 29) had a significant higher frequency of readmissions (P = 0.042) than the rest of the sample. Including all admissions (n = 63) in the analyses, the don't know group had also more patients with personality disorders (P = 0.015) and anxiety disorders (P = 0.008). 3.2. SRS ratings and prevalence of episodes A great majority (60–82%) of the patients rated “no risk” on all SRS for the hospital stay and the first 3 months post-discharge, and 1.9–7.0% rated the risk as high or very high. “Don't know” and “won't answer” made a total of 5.3–10%. The rates of occurred episodes, and the distribution of episodes related to the SRSs, are shown in Table 2. Two patients, one discharged into the community and one to a district psychiatric centre, committed suicide during the first three months after discharge. Table 2. Distribution of the recorded episodes (threats and acts) compared with patients' risk estimates. Risk prediction SRS No risk Low risk Moderate High Very high Don't know Won't answer Threats + acts all episodes (%) At hospital stay (episodes/predictions) Suicide 0/322 2/47 2/6 0/3 0/1 3/15 2/16 0 + 9 (2.2%) Self-injury 1/306 0/43 3/21 3/5 0/2 0/17 0/16 0 + 7 (1.7%) Violent threats 9/337 2/27 4/13 1/4 1/2 0/7 7/19 12 + 12 (5.9%) Violent acts 13/353 1/24 3/5 0/1 1/2 0/5 6/18 12 + 12 (5.9%) 3 months post-discharge (episodes/predictions) Suicidea 28/200 6/34 4/8 2/4 0/1 4/12 1/7 31 + 14 (17%) Self-injury a 10/187 3/33 3/16 3/4 1/5 3/13 0/7 9 + 14 (8.6%) Violent threats 22/211 6/26 4/9 1/1 2/2 3/8 4/9 26 + 16 (16%) Violent acts 24/221 6/22 2/4 0/0 2/2 3/8 5/9 26 + 16 (16%) a 14 patients (5.3%) had been recorded with both suicidal and self-injurious behaviour after discharge. Table options 3.3. Predictive validity and characteristics The SRS for violent threats and SRS for violent acts predicted violent behavior with equal accuracy (Table 3). In the following we give only the results of SRS of violent threats. For hospital stay and at 3 months after discharge, SRSs of suicidal, self-injurious (SIB) and violent behavior correlated significantly with occurred episodes (Table 3). The correlations remained unchanged when data from all readmissions were included in the analyses. SRS predicted violent behaviour for both genders. Sucidal and self-injurious behaviours were predicted significantly only for women. Table 3. Predictive validity of self-prediction (SRS) at hospital stay and 3 months post-discharge. Prediction of At hospital stay 3 months post-discharge Patients, n = 422 All admissions, n = 582 Patients, n = 266 Readmissions only, n = 160 n a AUC (95%CI) P n a AUC (95%CI) P n a AUC (95%CI) P n a AUC (95%CI) P Violent threats 24 0.73 (0.61–0.85) < 0.001 31 0.69 (0.58–0.80) < 0.001 42 0.67 (0.58–0.76) < 0.001 20 0.70 (0.60–0.83) 0.005 men 15 0.68 (0.52–0.84) 0.016 24 0.64 (0.51–0.77) 0.024 woman 9 0.82(0.64–0.99) 0.002 18 0.72 (0.57–0.87) 0.006 Violent acts 24 0.68 (0.55–0.81) 0.003 31 0.65 (0.53–0.76) 0.006 42 0.64 (0.55–0.73) 0.001 20 0.67 (0.53–0.81) 0.015 Suicidality 9b 0.92 (0.88–0.96) < 0.001 9 0.90 (0.85–0.94) < 0.001 45 0.63 (0.55–0.72) 0.002 54 0.61 (0.53–0.71) 0.012 woman 7 0.91 (0.85–0.97) < 0.001 18 0.65 (0.53–0.77) 0.023 Self–injurious 7b 0.82 (0.67–0.98) 0.003 10 0.82 (0.71–0.93) < 0.001 23 0.66 (0.57–0.76) 0.001 38 0.65 (0.58–0.79) 0.003 Woman 7 0.80 (0.65–0.95) 0.007 18 0.67 (0.60–0.77) 0.002 a Number of recorded episodes of violence, suicidal behavior and self-injurious behaviour, respectively. b Suicidal and self-injurious acts. Table options SRS of violence was significant even after one year (AUC = 0.61, 95%CI = 0.53–0.69, P = 0.005), and with severe violent acts as the only significant category (AUC = 0.69, 95%CI = 0.51–0.88, P = 0.024). For the subgroup patients recorded with both suicidal and self-injurious behavior, AUC values at 3 months were 0.71 (0.54–0.90, P = 0.019) when predicted by the SRS of suicide, and 0.83 (0.72–0.95, P < 0.001) when predicted by the SRS of SIB. This subgroup (n = 14) was more frequently rehospitalised than the subgroups with suicidal behaviour only (n = 31) or SIB only (n = 9); the mean number of rehospitalisations for these patients at 3 months were 2.0, 1.0 and 0.3, respectively. 3.4. Controlling for other factors For each of the outcome measures, all univariate significant factors were entered in a multivariate analysis. The factors that remained significant were again entered in multivariate analysis. The results from the last analyses are shown in Table 4. SRS for all three outcome measures remained significant at hospital stay in multivariate analyses. SRS of violent behaviour remained significant after 3 months, and moderate or higher risk of violence was even significant after one year (OR = 4.3, 95%CI = 1.2–16, P = 0.027). Table 4. Significant factors in multivariate analyses. Inpatient 3 months outpatient OR (95%CI) p OR (95%CI) p A. Violent behavioura SRS scale (no risk = baseline) 0.026 0.014 Moderate or higher risk 6.3 (1.5–23) 0.011 8.1 (1.6–40) 0.010 Won't answer about risk 5.7 (1.5–22) 0.011 5.2 (1.2–23) 0.029 Involuntary admission 15 (1.5–45) < 0.001 4.8 (2.2–10) < 0.001 History of violence 3.6 (1.0–13) 0.043 4.9 (2.0–13) < 0.001 B. Self-injurious behaviourb SRS scale (no risk = baseline) 0.085 0.218 Moderate or higher risk 24 (2.2–256) 0.009 2.5 (0.88–7.0) 0.085 Don't know risk 35 (2.4–510) 0.009 3.4 (1.1–11) 0.033 Female all female 2.0 (0.88–4.8) 0.097 Bipolar disorders 63 (6.6–601) <.001 1.3 (0.48–3.3) 0.643 C. Suicidal behaviourc SRS scale (low risk¹ = baseline) 0.021 0.129 Moderate or higher risk 9.6 (1.1–78) 0.040 3.3 (0.88–12) 0.078 Don't know risk 12 (1.7–84) 0.012 3.3 (1.1–9.6) 0.032 Won't answer risk 19 (2.3–149) 0.006 0.78 (0.07–6.7) 0.749 Personality disorder 0.27 (0.03–2.5) 0.255 6.6 (2.1–21) 0.001 Inpatient suicidality 8.9 (0.85–94) 0.069 History of suicide attempts 8.3 (0.90–78) 0.062 1.6 (0.83–3.3) 0.158 a b c Univariate significant factors, but not significant in multivariate analysis: a = length of stay (inverse), involuntary hospitalisation, involuntary aftercare, F1 × substance abuse, F2 × psychotic disorders, F3 × affective disorders (inverse), F4 × anxiety disorders (inverse), gender, inpatient violence, b = history of suicidal attempts, suicidal ideation last month, personality disorders, age, bipolar disorders (inpatient significant), inpatient violence, –suicidality, –self-injury, c = bipolar disorders, alcohol abuse, age, gender, inpatient violence, – suicidality, –self-injury, suicidal ideation the last month, length of stay (inverse), ¹ = no recorded episodes among no risk ratings. Table options SRS counted for 7–20% of the “model fit” of violent behavior during hospital stay, 5–27% of suicidal behaviour and 7–40% of SIB. After three months the corresponding values were 10–16%, 5–8% and 7–13%, respectively. 3.5. Other predictive measures At hospital stay, false positives and false negatives for violent behaviour were 26 patients (68%) and 12 patients (50%); for suicidal behaviour 37 patients (86%) and 2 patients (25%), and for SIB 24 patients (80%) and 1 patient (14%). Corresponding sensitivity/specificity was 0.50/0.93, 0.75/0.91 and 0.88/0.94 (Table 5). Table 5. Predictive characteristics of self predictions: inpatient and 3 months post-discharge. Inpatient 3 months post-discharge PPV NPV NND PPV NPV NND Violent behaviour 0.32 0.97 3.2 0.43 0.88 2.3 Suicidal behaviour 0.14 0.99 7.1 0.36 0.86 2.8 Self-injurious behaviour 0.20 0.99 5.0 0.25 0.93 4.0 Table options Three months post-discharge false positives and false negatives were 16 (57%) and 28 (79%) for violent behaviour, 21 (64%) and 33 (73%) for suicidal behaviour, and 18 (75%) and 16 (73%) for SIB. Corresponding sensitivity/specificity was 0.30/0.93, 0.27/0.91 and 0.27/0.93. Excluding “don't know” and “won't answer” increased the PPV's of suicidal behaviour to 0.20 and 0.46, respectively, and to 0.50 for 3 months violent behaviour, but it had no effect on the results for SIB and inpatient violence.