دانلود مقاله ISI انگلیسی شماره 36842
عنوان فارسی مقاله

پیش بینی اعتبار HCR-20 برای بیماران بستری خودآسیبی

کد مقاله سال انتشار مقاله انگلیسی ترجمه فارسی تعداد کلمات
36842 2014 13 صفحه PDF سفارش دهید محاسبه نشده
خرید مقاله
پس از پرداخت، فوراً می توانید مقاله را دانلود فرمایید.
عنوان انگلیسی
Predictive validity of the HCR-20 for inpatient self-harm
منبع

Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)

Journal : Comprehensive Psychiatry, Volume 55, Issue 8, November 2014, Pages 1937–1949

کلمات کلیدی
- پیش بینی اعتبار - بیماران بستری - خودآسیبی
پیش نمایش مقاله
پیش نمایش مقاله پیش بینی اعتبار HCR-20 برای بیماران بستری خودآسیبی

چکیده انگلیسی

Abstract Background Few instruments have been developed to assess the risk of self-harm by psychiatric patients and the evidence for their predictive validity is limited. Given that individuals who self-harm may also engage in other-directed aggression, and that the behaviour can be a precursor to violence, we tested whether, and for which groups, the commonly used violence risk assessment HCR-20 demonstrated predictive validity for self-harm. Procedures A pseudo-prospective cohort study (N = 504) was conducted in a UK secure/forensic mental health setting using routinely collected data. HCR-20 assessments were completed by the clinical team and incidents of self-harm during the 3 months following assessment were coded from patient records.

مقدمه انگلیسی

. Introduction 1The structured professional judgment approach to risk assessment has primarily focused on the prediction of aggression and violence due to its potentially serious consequences [1]. However, adequate care of mental health inpatients requires the consideration of a range of risk behaviours including self-harm [2]. A recent review found that between 1 and 69% (mean 17.4%) of psychiatric inpatients had self-harmed during their stay; rates were especially high (mean 42.9%) among patients resident on forensic wards [3]. Patients at increased risk for self-harm are younger, female, and those diagnosed with EUPD [4]. It is important to accurately predict patients' engagement in self-injurious behaviour because it has obvious deleterious sequelae for themselves; 12–20% of incidents are classified as severe, resulting in deep cuts, fractures, or internal injuries [3]. In addition, it affects the emotional wellbeing of carers, and contributes to therapeutic nihilism [3] and staff absence [5]; Nijman et al. [5] reported that, in 1 year, 84% of psychiatric nurses witnessed self-harm, 68% a suicide attempt, and 28% a completed suicide. Witnessing such incidents was significantly correlated with number of days absent from work due to illness. The HCR-20 [6] is not intended to assess risk of self-harm. However, version 3 of the HCR-20 manual [7] states that self-injurious or suicidal behaviour should be categorised as violence if, as a result, others may also suffer physical harm. Whilst this may not encompass all acts of self-harm occurring within the inpatient setting, research has found that self-harm can be a precursor to violence, particularly in women [8], and that the majority of inpatients who self-harm also engage in outwardly directed aggressive behaviour [4]. Although more recently developed risk assessment tools have been designed to assess risk for self-harm and suicide, such as the START [2] and the SRAMM [9], the ability of the START to predict self-harm is still a matter of some conjecture [10] and the SRAMM has not been widely researched. Whilst the reasons behind the association between self-harm and externally directed aggression in inpatient settings are unexplored, the empirical link between the two [11] suggests that a violence risk assessment might usefully contribute to a self-harm risk assessment irrespective of the theoretical basis for such a link. Given that the HCR-20 is the most commonly used risk assessment tool in medium secure units in England [12], and that time and resources are often limited in clinical practice, it would be pragmatic to determine if the tool is also predictive of self-directed aggression in order to minimise the need for additional assessments. Very little research has been conducted on the predictive validity of the HCR-20 for inpatient self-harm. Whilst a small number of studies have incorporated self-harm incidents within outcomes categories for ‘any aggression’, a recent meta-analysis of the tool for inpatient aggression [13] could not examine predictive validity for self-harm due to an insufficient number of studies examining the behaviour as a distinct outcome. Those studies that have examined the predictive validity of the HCR-20 for self-harm in isolation have produced mixed results. Gray et al. [14] found that the tool was not a useful predictor of self-harm over a 3-month period; however, Fagan et al. [15] found that it was predictive of self-harm and suicidal ideation over a 6-month period, and Daffern and Howells [16] found that the clinical scale of the HCR-20 was predictive of self-harm occurring in the day after assessment. The aim of the current study was to determine the predictive efficacy for inpatient self-harm of the various HCR-20 scales and summary judgement (SJ) for violence; the ability of the SJ for violence to predict inpatient self-harm was examined as the empirical link between the two behaviours suggests that, at least in some cases, those deemed at risk of violence may also be at risk of self-harm. Given that the efficacy of the HCR-20 for violence prediction (e.g. [13] and [17]), and the prevalence of self-harm [4], has been shown to vary across different clinical and demographic groups, its performance was also evaluated as a function of gender, diagnosis, age, and ethnicity. We also aimed to examine the predictive validity of the individual HCR-20 items, as their relative importance has the potential to inform summary judgements and risk-management plans. Finally, we investigated the degree of overlap between self-harm and aggression against others as this may have implications for the feasibility of using the HCR-20 to examine risk of both outcomes

نتیجه گیری انگلیسی

3. Results 3.1. Participants The mean age at time of assessment was 39.79 (SD = 15.82) and the mean number of days between admission and first HCR-20 assessment was 850.86 (SD = 1334.45). Just over two thirds of the sample were male (n = 348, 69%) and 156 (31%) were female. Approximately half of the sample identified as Caucasian (n = 235, 46.6%), 92 (18.3%) were non-Caucasian and the remainder (n = 177, 35.1%) declined to state their ethnicity. The most common diagnoses were schizophrenia (n = 220, 43.7%), personality disorder (n = 72, 14.3%), co-morbid schizophrenia and personality disorder (n = 42, 8.3%), developmental disorder (n = 33, 6.5%), and organic disorder (n = 32, 6.3%); 105 (20.8%) held other or multiple diagnoses. Seventy four (14.7%) of the sample held a co-morbid substance use disorder diagnosis. The majority (n = 379, 75.2%) resided in conditions of low security and 125 (24.8%) were in medium secure wards. Most patients (n = 273, 54.2%) were admitted under forensic sections of the Mental Health Act (1983, Amended 2007), 191 (37.9%) were admitted under civil sections, and a small number (n = 40, 7.9%) were informal admissions. 3.2. HCR-20 scores Mean HCR-20 total and R5 scores were significantly higher in those that had engaged in any self-harm, severe self-harm, and repeated self-harm compared with those who had not. H10 scores differed significantly for engagement in any and repeated self-harm but not for severe self-harm whilst C5 scores did not differ as a function of any outcome. The SJ was available for 338 patients; for all outcomes, a greater proportion of those who engaged were rated as high risk compared to those who had not engaged (see Table 2). Table 2. Differences in HCR-20 scores and risk levels as a function of engagement in self-harm. Overall Any self-harm Severe self-harm Repeated self-harm Yes No Test Yes No Test Yes No Test Mean HCR-20 scores (SD) Total 27.04 (5.61) 29.20 (5.30) 26.71 (5.59) t(502) = 3.44, p < .001 29.38 (4.93) 26.88 (5.62) t(502) = 2.48, p = .013 29.80 (4.74) 26.78 (5.62) t(502) = 3.45, p < .001 H10 13.69 (3.08) 14.43 (2.62) 13.58 (3.13) t(502) = 2.12, p = .034 14.57 (2.63) 13.63 (3.10) t(502) = 1.71, p = .088 14.69 (2.19) 13.60 (3.14) t(61) = 3.03, p = .004 C5 6.77 (2.36) 7.10 (2.12) 6.71 (2.39) t(502) = 1.27, p = .205 6.89 (2.05) 6.76 (2.38) t(502) = 0.32, p = .747 7.14 (2.06) 6.73 (2.39) t(502) = 1.11, p = .268 R5 6.59 (2.52) 7.68 (2.05) 6.43 (2.54) t(102) = 4.51, p < .001 7.91 (1.79) 6.50 (2.54) t(42) = 4.22, p < .001 7.98 (1.89) 6.46 (2.53) t(59) = 4.92, p < .001 SJ ratingsa Low 94 (27.8%) 5 (12.8%) 70 (32.3%) χ2(2, N = 256) = 14.08, p < .001 2 (9.5%) 73 (31.1%) χ2(2, N = 256) = 12.46, p = .002 1 (4%) 74 (32%) χ2(2, N = 256) = 19.13, p < .001 Medium 158 (46.7%) 15 (38.5%) 100 (46.1%) 7 (33.3%) 108 (46%) 9 (36%) 106 (45.9%) High 86 (25.4%) 19 (48.7%) 47 (21.7%) 12 (57.1%) 54 (23%) 15 (60%) 51 (22.1%) H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment; Any PS, any incidents of self-harm; Severe PS, incidents of level 3 and level 4 self-harm; Repeated PS, 2 or more incidents of self-harm. a n = 338. Table options Approximately half of those with an available SJ (n = 158, 46.7%) were classified as moderate risk, 94 (27.8%) were classified as low risk and 86 (25.4%) as high risk. Mean HCR-20 total (F[2, 335] = 23.34, p < .001) scores differed significantly across the risk levels assigned by the SJ. Post-hoc Tukey tests revealed that HCR-20 scores were significantly lower in the low risk group (24.66, SD = 6.30) compared to the moderate (28.90, SD = 4.78; p < .001) and high (29.52, SD = 5.45; p < .001) risk groups. C5 and R5 scores were also significantly smaller in the low risk group compared to the moderate and high risk groups; H10 scores were only significantly different between the low and moderate risk groups. There were no significant differences between scores in the moderate and high risk groups (see Table 3). Table 3. Differences in HCR-20 scores across assigned risk levels. Mean HCR-20 scores (SD) Low Moderate High Test Total 24.66 (6.30) 28.90 (4.78) 29.52 (5.45) F(2, 335) = 23.34, p < .001 H10 13.19 (3.24) 14.30 (2.59) 14.11 (3.23) F(2, 335) = 4.38, p = .013 C5 5.66 (2.57) 7.14 (2.26) 7.33 (2.08) F(2, 335) = 15.50, p < .001 R5 5.82 (2.54) 7.44 (2.15) 8.06 (2.08) F(2, 335) = 24.73, p < .001 H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment. Table options 3.3. Self-harm base rates There were 407 incidents of self-harm in the 3 month follow up period involving 68 (13.5%) patients; approximately half of these (n = 33, 6.5%) were involved in 67 incidents of severe self-harm. Of those patients who self-harmed, 24 (35.3%) engaged in just one incident whilst 44 (64.7%) engaged in repeated self-harm; the number of incidents among those who engaged ranged from 1 to 52, mean 5.99 (SD = 8.40; median = 2). 3.4. Overlap between self-harm and aggression against others A significantly higher proportion of those who had engaged in each type of self-harm had engaged in aggression against others, compared to those who had not self-harmed. This difference was most pronounced for repeated self-harm, where 77.3% of those who engaged in repeated self-harm had engaged in aggression against others, compared to 32.0% of those who had not engaged in repeated self-harm (χ2[1, N = 504] = 35.83, p < .001). There was greater overlap in the people who had engaged in aggression against others and those who had engaged in self-harm for women compared with men, younger adults compared with older adults, Caucasians compared with non-Caucasians and in people with a personality disorder compared with other diagnoses. 3.5. Predictive validity Results of the overall predictive validity analyses are presented in Table 4. The HCR-20 total, H10 and R5 scales, and SJ significantly predicted all three outcomes whilst the C5 scale was not a significant predictor of any outcome. With the exception of repeated self-harm, where it was exceeded by the SJ, the R5 scale was the strongest predictor. In most cases, the largest AUC values were obtained for the prediction of repeated self-harm. Item–outcome analysis revealed that nine, seven and five individual items significantly predicted any self-harm, repeated self-harm and severe self-harm, respectively. For all outcomes, the only C5 item to significantly predict self-harm was item 4 (impulsivity). The strongest predictor from the R5 scale was item 2 (exposure to destabilizers) and the strongest from the H10 scale was item 9 (personality disorder). Item 6 on the H10 scale (major mental illness) produced AUC values significantly less than .5 for all three outcomes. None of the AUC values reached the threshold for a large effect size (>.75); the strongest predictor for all outcomes was personality disorder (H10:9). Table 4. Overall predictive validity of the HCR-20 for self-harm. Any self-harm Severe self-harm Repeated self-harm AUC 99% CI AUC 99% CI AUC 99% CI HCR-20 totala .638⁎⁎⁎ [.545, .732] .641⁎⁎ [.515, .768] .663⁎⁎⁎ [.557, .768] H10 scale Totala .584⁎ [.492, .676] .602⁎ [.479, .726] .599⁎ [.494, .703] 1b .501 [.403, .598] .492 [.355, .628] .516 [.401, .631] 2c .582⁎ [.489, .675] .595 [.467, .724] .607⁎ [.500, .713] 3d .550 [.453, .647] .556 [.427, .686] .551 [.436, .667] 4e .517 [.413, .622] .562 [.426, .697] .519 [.395, .643] 5b .491 [.396, .586] .504 [.371, .636] .484 [.368, .600] 6f .345⁎⁎⁎ [.247, .442] .349⁎⁎ [.212, .487] .345⁎⁎⁎ [.230, .460] 8g .596⁎ [.508, .685] .581 [.459, .703] .587 [.483, .692] 9h .707⁎⁎⁎ [.624, .790] .740⁎⁎⁎ [.642, .839] .749⁎⁎⁎ [.666, .832] 10i .523 [.427, .618] .497 [.362, .632] .513 [.396, .631] C5 scale Totala .540 [.447, .633] .505 [.382, .628] .538 [.427, .649] 1f .462 [.364, .561] .409 [.270, .547] .429 [.307, .550] 2j .518 [.421, .616] .474 [.334, .614] .503 [.385, .622] 3i .460 [.366, .554] .485 [.353, .618] .484 [.374, .595] 4j .682⁎⁎⁎ [.603, .760] .678⁎⁎⁎ [.571, .784] .704⁎⁎⁎ [.617, .790] 5c .505 [.410, .601] .498 [.375, .621] .518 [.401, .634] R5 scale Totala .641⁎⁎⁎ [.556, .725] .659⁎⁎ [.556, .761] .673⁎⁎⁎ [.579, .767] 1i .615⁎⁎ [.531, .700] .595 [.480, .711] .624⁎⁎ [.526, .722] 2a .615⁎⁎ [.526, .704] .661⁎⁎ [.553, .769] .682⁎⁎⁎ [.588, .777] 3c .582⁎ [.490, .674] .572 [.440, .704] .588 [.474, .703] 4b .556 [.462, .651] .572 [.447, .697] .570 [.451, .690] 5i .586⁎ [.497, .675] .626⁎ [.516, .736] .613⁎ [.511, .716] SJk .630⁎⁎ [.525, .734] .652⁎⁎ [.509, .794] .687⁎⁎⁎ [.573, .802] AUC, area under receiver operating characteristic curve; CI, confidence interval; H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment. a N = 504. b n = 502. c n = 498. d n = 476. e n = 470. f n = 503. g n = 494. h n = 444. i n = 501. j n = 499. k n = 338. ⁎ p < .05. ⁎⁎ p < .01. ⁎⁎⁎ p < .001. Table options 3.6. Gender A significantly higher proportion of women (n = 47, 30.1%) engaged in self-harm than men (n = 21, 6%) (χ2[1, N = 504] = 53.57, p < .001). HCR-20 total scores did not differ as a function of gender; however, women had significantly higher H10 (14.10 vs. 13.51, t[366] = −.219, p = .029) and R5 (6.96 vs. 6.43, t[502] = −2.21, p = .028) scores than men and were more likely to be classified as high risk by the SJ (χ2[2, N = 338] = 21.31, p < .001). Women were less likely than men to be non-Caucasian and more likely to have an unknown ethnicity (χ2[2, N = 504] = 11.38, p = .003); they were also more likely to be aged <40 than men (χ2[1, N = 504] = 5.56, p = .018). Gender differences in diagnosis (χ2[5, N = 504] = 129.65, p < .001) were due to a greater proportion of women having a personality disorder than men and an over-representation of men among the schizophrenia, developmental, and organic groups. Men had a significantly greater mean number of days between admission and HCR-20 assessment (922 vs. 693, t[485] = 2.20, p = .028). Women were more likely than men to be admitted under a civil section of the Mental Health Act and less likely to be under a forensic section (χ2[2, N = 504] = 7.23, p = .027). There were no differences in security level or rates of substance abuse diagnoses as a function of gender; therefore, ethnicity, diagnosis, legal status, age, and time between admission and assessment were controlled for in the subsequent rocreg analysis. Results of the rocreg analyses are presented in Table 5. Only one item differed significantly in its predictive validity as a function of gender, namely item 9 on the H10 scale (personality disorder), which performed significantly better in women. However, The HCR-20 total score, R5 scale score and SJ significantly predicted self-harm in women, whilst none of the scales predicted self-harm for men; item 2 on the H10 scale (young age at first violent incident) was the only significant predictor for this group. Three of the R5 items, one C5 item (impulsivity) and two H10 items significantly predicted self-harm in women, the strongest predictor being H10:9 (personality disorder), which was the only item to exceed the threshold for a large effect size. Item 6 on the H10 scale (major mental illness) produced an AUC value significantly smaller than .5 for this group. The AUC value obtained for the SJ was larger than that for the HCR-20 total in women, but this was not the case for men. Table 5. Predictive validity of the HCR-20 for any self-harm as a function of gender. rocreg Men Women Coefficient 99% CI AUC 99% CI AUC 99% CI HCR-20 total −.027 [−.805, .647] .633 [.466, .768] .625⁎ [.512, .726] H10 scale Total −.061 [−.937, .847] .557 [.373, .733] .540 [.432, .657] 1 −.017 [−.826, .817] .527 [.267, .737] .522 [.343, .701] 2 −.522 [−1.306, .134] .658⁎ [.531, .813] .525 [416, .657] 3 .204 [−.544, 1.028] .542 [.329, .745] .597 [.441, .759] 4 −.300 [−1.169, .490] .573 [.358, .783] .487 [.352, .649] 5 .125 [−.677, .818] .465 [.309, .658] .500 [.383, .631] 6 −.498 [−1.315, .245] .500 [.291, .687] .335⁎ [.192, .483] 8 .126 [−.836, .876] .520 [.361, .698] .550 [.443, .659] 9 .922⁎ [.066, 1.904] .551 [.345, .742] .761⁎ [.671, .856] 10 .208 [−.785, .858] .426 [.278, .624] .475 [.362, .602] C5 scale Total −.226 [−.988, .543] .587 [.396, .754] .524 [.406, .649] 1 −.072 [−.745, 555] .505 [.352, .708] .484 [.360, .614] 2 .099 [−.567, .866] .468 [.313, .630] .496 [.359, .641] 3 −.192 [−.894, .473] .546 [.378, .745] .488 [.377, .622] 4 .137 [−.623, 1.080] .609 [.412, .747] .638⁎ [.534, .720] 5 −.169 [−.920, .646] .545 [.353, .728] .497 [.374, .626] R5 scale Total .068 [−.703, .722] .640 [.482, .797] .658⁎ [.511, .770] 1 .247 [−.465, 1.076] .598 [.419, .775] .662⁎ [.557, .787] 2 .268 [−.353, .927] .567 [.426, .716] .638⁎ [.510, .749] 3 .059 [−.623, .832] .612 [.444, .775] .628⁎ [.525, .765] 4 −.137 [−.857, .712] .592 [.425, .771] .553 [.427, .674] 5 −.009 [−.695, .776] .615 [.439, .781] .613 [.496, .746] SJ .634 [−.416, 1.539] .524 [.314, .738] .700⁎ [.548, .817] AUC, area under receiver operating characteristic curve; CI, confidence interval; H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment. ⁎ Significant at .01 level based on inspection of confidence intervals. Table options 3.7. Diagnosis Those with a personality disorder (n = 26, 36.1%) were significantly more likely, and those with schizophrenia (n = 9, 4.1%) significantly less likely, to have engaged in self-harm (χ2[4, N = 399] = 56.65, p < .001). HCR-20 total and all subscale scores differed significantly across the diagnostic groups; the schizophrenia and personality disorder group had significantly higher HCR-20 total scores than all groups except those with an organic disorder; this group also had the largest H10 scores whilst the organic group had the largest C5 and R5 scale scores. Risk levels assigned by the SJ also differed significantly across groups (χ2[8, N = 399] = 16.77, p = .033); the schizophrenia group was less likely to be categorised as high risk and more likely to be classed as low risk whilst the organic group contained a greater proportion of moderate risk patients and fewer low risk patients. There was an over-representation of women in the personality disorder group and of men in the schizophrenia, developmental and organic groups (χ2[4, N = 399] = 131.25, p < .001). The schizophrenia group was more likely to be non-Caucasian, whilst the personality disorder group was less likely to be non-Caucasian and more likely to have an unknown ethnicity (χ2[8, N = 399] = 31.04, p < .001). Those with a personality disorder or developmental diagnosis were more likely to be aged <40, whilst those with schizophrenia or an organic diagnosis were more likely to be aged ≥40 (χ2[4, N = 399] = 43.85, p < .001). Time between admission and assessment differed across diagnoses (F[4, 394] = 2.70, p = .031); post-hoc Tukey tests revealed that the developmental group had a significantly shorter mean time between admission and assessment compared with the schizophrenia group (286.18 vs. 1033.79, p = .040). Security level, legal status and substance use did not differ as a function of diagnosis; therefore, gender, ethnicity, age and time between admission and assessment were controlled for in the rocreg analyses. The HCR-20 total and C5 scale significantly predicted self-harm in the schizophrenia group and the personality disorder group but not for the remaining diagnoses; the H10 scale and SJ did not predict self-harm for any group and the R5 scale was a significant predictor for all groups except those with an organic diagnosis. Item–outcome analyses revealed that the schizophrenia and organic groups had the fewest number of significant predictors; for the schizophrenia group, only C5:2 (negative attitudes) and R5:4 (noncompliance with remediation attempts) significantly predicted self-harm, whilst R5:1 (plans lack feasibility) and R5:2 (exposure to destabilizers) were the only significant items for the organic group. The personality disorder and the schizophrenia and personality disorder groups had the largest number of significant predictors, which were largely the same between the two groups. The largest AUC value for both groups was obtained from R5:2. The strongest predictors in the developmental group were the same as those for the organic group; H10:5 (substance use problems) produced an AUC value significantly smaller than .5 in this group and the schizophrenia and personality disorder group. AUC values obtained from the SJ were larger than those from the HCR-20 total in the schizophrenia and personality disorder, developmental and organic groups, the opposite being true for the schizophrenia and personality disorder groups (Table 6). Table 6. Predictive validity of the HCR-20 for any self-harm as a function of diagnosis. rocreg Schizophrenia Personality disorder Schizophrenia and personality disorder Developmental Organic Coefficient 99% CI AUC 99% CI AUC 99% CI AUC 99% CI AUC 99% CI AUC 99% CI HCR-20 total −.094 [−.538, .312] .696⁎ [.531, .842] .673⁎ [.535, .790] .650 [.477, .787] .626 [.375, .834] .602 [.250, .879] H10 scale Total −.177 [−.729, .385] .550 [.357, .746] .520 [.387, .639] .490 [.314, .619] .460 [.209, .698] .430 [.128, .763] 1 −.089 [−.771, .547] .637 [.359, .895] .610 [.330, .848] .582 [.287, .844] .555 [.227, .935] .527 [.116, .977] 2 −.041 [−.494, .495] .666 [.479, .854] .655⁎ [.532, .771] .644⁎ [.505, .819] .633 [.418, .878] .622 [.318, .937] 3 .130 [−.497, 1.132] .430 [.137, .661] .461 [.312, .650] .493 [.268, .736] .524 [.177, .834] .555 [.104, .920] 4 −.199 [−.722, .349] .672 [.443, .926] .612 [.437, .792] .548 [.342, .818] .484 [.223, .886] .420 [.101, .899] 5 −.253 [−.929, .251] .457 [.275, .633] .399 [.268, .524] .344⁎ [.163, .482] .292⁎ [.073, .497] .243 [.028, .553] 6 −.193 [−.736, .276] .471 [.246, .691] .405 [.264, .561] .342 [.123, .512] .283 [.037, .569] .229 [.002, .640] 8 .260 [−.198, .837] .437 [.271, .597] .490 [.384, .604] .543 [.418, .666] .596 [.414, .795] .647 [.389, .898] 9 .014 [−.826, 1.346] .632 [.418, .859] .635⁎ [.521, .761] .639⁎ [.504, .921] .642 [.398, .985] .645 [.291, .999] 10 −.066 [−.638, .410] .397 [.236, .574] .383⁎ [.279, .491] .370⁎ [.257, .498] .356 [.166, .551] .343 [.086, .603] C5 scale Total −.395⁎ [−.885, −.045] .715⁎ [.558, .865] .623⁎ [.509, .752] .523 [.341, .662] .421 [.182, .608] .325 [.074, .575] 1 −.047 [−.523, .460] .580 [.343, .775] .564 [.426, .699] .549 [.368, .715] .533 [.228, .789] .518 [.125, .881] 2 −.247 [−.706, .088] .678⁎ [.503, .874] .611 [.459, .742] .540 [.376, .692] .467 [.254, .707] .396 [.138, .725] 3 −.339 [−.795, .044] .652 [.486, .830] .553 [.431, .684] .452 [.292, .646] .353 [.132, .643] .264 [.046, .644] 4 −.040 [−.547, .625] .578 [.336, .715] .570 [.457, .685] .562 [.344, .714] .553 [.256, .767] .545 [.170, .842] 5 −.290 [−.812, .260] .686 [.465, .852] .612 [.456, .736] .535 [.364, .753] .456 [.218, .757] .378 [.097, .804] R5 scale Total .104 [−.357, .705] .699⁎ [.524, .843] .723⁎ [.606, .820] .746⁎ [.613, .869] .767⁎ [.550, .930] .788 [.452, .969] 1 .510⁎ [.081, 1.090] .562 [.356, .747] .682⁎ [.538, .796] .785⁎ [.635, .909] .866⁎ [.695, .974] .922⁎ [.730, .995] 2 .193 [−.329, 1.028] .701 [.494, .937] .752⁎ [.594, .881] .797⁎ [.615, .966] .838⁎ [.592, .997] .873⁎ [.544, 1.000] 3 .077 [−.374, .607] .592 [.406, .743] .610 [.485, .741] .628 [.476, .780] .646 [.401, .853] .663 [.326, .920] 4 −.166 [−.711, .292] .731⁎ [.556, .892] .688⁎ [.568, .817] .643 [.477, .785] .596 [.318, .837] .547 [.176, .878] 5 .096 [−.404, .763] .705 [.460, .891] .729⁎ [.451, .851] .751⁎ [.570, .895] .773⁎ [.525, 953] .793 [.452, .989] SJ .121 [−.580, 1.069] .597 [.310, .864] .631 [.455, .778] .664 [.445, .874] .695 [.333, .969] .725 [.215, .994] AUC, area under receiver operating characteristic curve; CI, confidence interval; H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment. ⁎ Significant at .01 level based on inspection of confidence intervals. Table options 3.8. Ethnicity There were no significant differences in rates of engagement in self-harm between the Caucasian and non-Caucasian groups (n = 37, 15.7% vs. n = 10, 10.9%, respectively). HCR-20 total scores did not differ significantly between the two groups, neither did assignment to risk levels; however, R5 scale scores were significantly larger in the Caucasian group (6.49 vs. 5.88, t[325] = 1.99, p = .047). The non-Caucasian group contained a significantly higher proportion of patients residing in medium security (n = 26, 28.3% vs. n = 43, 18.3%, χ2[1, N = 327] = 3.94, p = .047) and a significantly smaller proportion of women (n = 17, 18.5% vs. n = 71, 30.2%, χ2[1, N = 327] = 4.63, p = .031) than the Caucasian group. The diagnostic breakdown of the two groups also differed significantly (χ2[5, N = 327] = 31.42, p < .001) such that there was a greater proportion of people with a personality disorder or other/multiple disorders in the Caucasian group and a greater proportion of those with schizophrenia in the non-Caucasian group. The Caucasian group were also significantly more likely to be detained under a civil section of the Mental Health Act, whilst the non-Caucasian group were more likely to be detained under forensic sections (χ2[2, N = 327] = 8.87, p = .012). Substance use and time between admission and assessment did not differ between the two groups; therefore, gender, security level, legal status and diagnoses were controlled for in the subsequent analyses. Rocreg analyses revealed that the performance of the HCR-20 did not differ as a function of ethnicity (see Table 7) although the HCR-20 total score, H10 scale and C5 scale produced larger AUC values in the non-Caucasian group compared with Caucasians. R5 was the only scale to significantly predict self-harm in Caucasians whilst none of the scales were significant predictors among the non-Caucasian group. For both groups, AUC values based on the SJ were smaller than those based on the HCR-20 total score. Item–outcome analyses showed that only one item significantly predicted self-harm among the non-Caucasians, namely item 2 on the H10 scale (young age at first violent incident); this item was also a significant predictor among the Caucasian group, along with H10:9 (personality disorder). None of the C5 or R5 items were significant predictors of self-harm in either group. Table 7. Predictive validity of the HCR-20 for any self-harm as a function of ethnicity. rocreg Caucasian Non-Caucasian Coefficient 99% CI AUC 99% CI AUC 99% CI HCR-20 total .139 [−1.122, 1.410] .580 [.422, .726] .619 [.290, .882] H10 scale Total .328 [−.836, 1.742] .494 [.358, .621] .578 [.302, .845] 1 −.295 [−1.795, .880] .465 [.205, .720] .382 [.074, .737] 2 .341 [−.943, 1.265] .656⁎ [.521, .801] .744⁎ [.518, .975] 3 .194 [−1.043, 2.209] .387 [.268, .544] .427 [.101, .719] 4 .266 [−1.101, 1.662] .506 [.331, .696] .581 [.210, .801] 5 .332 [−.856, 2.080] .459 [.302, .645] .562 [.247, .928] 6 .048 [−1.229, 1.150] .357 [.206, .535] .374 [.115, .746] 8 .299 [−.758, 1.534] .470 [.355, .638] .534 [.286, .783] 9 −.540 [−1.820, 1.062] .673⁎ [.541, .787] .552 [.289, .876] 10 −.030 [−2.091, .921] .403 [.261, .535] .397 [.125, .589] C5 scale Total .425 [−.615, 1.617] .476 [.335, .641] .589 [.343, .845] 1 .508 [−.456, 2.007] .459 [.308, .669] .605 [.370, .919] 2 .233 [−.674, 1.631] .423 [.264, .570] .491 [.243, .773] 3 .330 [−1.293, 1.373] .430 [.292, .570] .523 [.179, .799] 4 −.230 [−1.444, .799] .604 [.493, .729] .554 [.333, .744] 5 .138 [−1.446, 1.528] .466 [.296, .637] .501 [.185, .771] R5 scale Total −.253 [−1.531, .803] .640⁎ [.512, .780] .572 [.232, .805] 1 −.405 [−1.923, 1.337] .622 [.458, .755] .515 [.196, .790] 2 −.838 [−2.533, .336] .623 [.487, .777] .423 [.174, .751] 3 .518 [−.451, 1.636] .458 [.329, .578] .592 [.328, .803] 4 −.082 [−1.504, .949] .609 [.450, .780] .588 [.306, .860] 5 .141 [−.953, 1.176] .496 [.365, .608] .526 [.310, .759] SJ −.104 [−1.711, 1.312] .566 [.385, .755] .537 [.180, .895] AUC, area under receiver operating characteristic curve; CI, confidence interval; H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment. ⁎ Significant at .01 level based on inspection of confidence intervals. Table options 3.9. Age A significantly higher proportion of those aged <40 years engaged in self-harm compared to those aged ≥40 years (n = 53, 19.3% vs. n = 15, 6.5%, χ2[1, N = 504] = 17.61, p < .001). HCR-20 total scores did not differ between the two groups, nor did risk levels assigned by the SJ; however, the younger group had significantly higher H10 scores [14.51 vs. 12.72, t(421) = -6.61, p < .001] and smaller C5 scores [6.55 vs. 7.02, t(502) = 2.21, p = .027] than those aged ≥40 years. The group aged <40 years contained a higher proportion of women (χ2[1, N = 504] = 5.56, p = .018) and those with a personality or developmental disorder (χ2[5, N = 504] = 44.48, p < .001) compared to the older group, which were more likely to have schizophrenia or organic diagnoses. Those aged ≥40 also had a significantly longer mean time between admission and assessment (1163 vs. 704 days, t[289] = 4.61, p < .001). Ethnicity, substance use, legal status and security level did not differ significantly between the two groups; therefore, gender, diagnosis and time between admission and assessment were controlled for in the rocreg analyses. Rocreg analyses are presented in Table 8. None of the HCR-20 total, subscales or SJ were significant predictors of self-harm in those aged ≥40 years; the R5 scale was the only scale to significantly predict self-harm in those aged <40. AUC values obtained from the SJ were smaller than that of the HCR-20 total in both groups. Overall, the performance of the HCR-20 did not differ as a function of age, although AUC values were larger in the older group for the HCR-20 total, C5 and R5 scales compared with those aged <40 years. Item–outcome analyses revealed that the performance of item 2 on the H10 scale (young age at first violent incident) was a significantly better predictor in those aged <40 years compared to those aged ≥40; this item was the best predictor in the younger group and achieved a large effect size. The only item to significantly predict self-harm in the older group was item 1 on the C5 scale (lack of insight). Whilst the R5 scale score was a significant predictor of self-harm in those aged <40 years, none of the individual items were significant predictors; the only individual items that achieved a significant AUC value in this group came from the H10 scale. Table 8. Predictive validity of the HCR-20 for any self-harm as a function of age. rocreg <40 years ≥40 years Coefficient 99% CI AUC 99% CI AUC 99% CI HCR-20 total .010 [−.927, 1.046] .621 [.499, .713] .624 [.373, .864] H10 scale Total −.600 [−1.500, .229] .591 [.485, .687] .440 [.236, .643] 1 −.276 [−1.336, .851] .641 [.338, .837] .567 [.313, .820] 2 −1.315⁎ [−2.490, −.382] .790⁎ [.693, .883] .421 [.223, .698] 3 −.259 [−1.241, .797] .461 [.313, .613] .396 [.211, .577] 4 −.445 [−1.368, .467] .667⁎ [.535, .804] .528 [.290, .752] 5 −.139 [−1.037, .669] .452 [.323, .606] .417 [.189, .611] 6 .646 [−.240, 1.636] .383 [.245, .517] .611 [.365, .852] 8 −.140 [−1.145, .836] .454 [.358, .556] .427 [.236, .603] 9 −.439 [−1.700, .513] .676⁎ [.583, .778] .579 [.369, .771] 10 .084 [−.999, 1.832] .480 [.299, .511] .426 [.231, .721] C5 scale Total .686 [−.241, 1.783] .518 [.409, .626] .695 [.465, .882] 1 .632 [−.131, 2.011] .524 [.388, .670] .718⁎ [.519, .909] 2 .203 [−.810, 1.186] .514 [.384, .667] .577 [.292, .831] 3 .447 [−.395, 1.763] .498 [.386, .619] .629 [.329, .883] 4 .548 [−.251, 1.621] .549 [.441, .638] .653 [.498, .810] 5 .769 [−.078, 1.920] .507 [.311, .665] .712 [.496, .899] R5 scale Total .046 [−.948, 1.094] .616⁎ [.502, .710] .628 [.370, .829] 1 −.039 [−.879, .904] .581 [.468, .733] .573 [.399, .805] 2 −.061 [−.916, .952] .584 [.459, .721] .569 [.367, .793] 3 .243 [−.541, 1.004] .485 [.374, .589] .548 [.365, .706] 4 .254 [−.664, 1.145] .578 [.406, .717] .651 [.387, .859] 5 .204 [−.817, 1.595] .517 [.361, .713] .566 [.321, .794] SJ −.394 [−1.319, 1.426] .573 [.401, .679] .465 [.193, .702] AUC, area under receiver operating characteristic curve; CI, confidence interval; H10, Historical subscale of HCR-20; C5, Clinical subscale of HCR-20; R5, Risk-management subscale of HCR-20; SJ, summary judgment. ⁎ Significant at .01 level based on inspection of confidence intervals.

خرید مقاله
پس از پرداخت، فوراً می توانید مقاله را دانلود فرمایید.