Suicide schemas in non-affective psychosis: An empirical investigation

Behaviour Research and Therapy 48 (2010) 1211e1220

Contents lists available at ScienceDirect

Behaviour Research and Therapy journal homepage: www.elsevier.com/locate/brat

Suicide schemas in non-affective psychosis: An empirical investigation Daniel Pratt*, Patricia Gooding, Judith Johnson, Peter Taylor, Nicholas Tarrier School of Psychological Sciences, University of Manchester, Manchester M13 9PL, UK

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 October 2009 Received in revised form 16 August 2010 Accepted 20 August 2010

Suicide is the leading cause of premature death among individuals experiencing psychosis. The risk of suicide is proposed to increase with a greater potential for activation of suicide related schemas. Empirical representations of suicide schemas were compared between individuals experiencing nonaffective psychosis, with and without a history of suicidal behaviour. Employing a cross-sectional between-groups comparison design, 84 participants, previously diagnosed with a non-affective psychotic disorder, were recruited from community mental health services. Participants completed a demographic questionnaire and clinical measures of psychopathology. To assess participants’ suicide schemas, a series of direct and indirect cognitive tasks were designed and administered. Pathfinder analysis enabled the construction of empirically derived representations of the groups’ suicide schemas based on responses to the cognitive tasks. The suicide group achieved significantly greater scores on measures of anxiety, depression, hopelessness and suicidality than the non-suicide group, but not on measures indicative of the severity of psychosis. The suicide schema for the suicide group was more elaborate and extensive than for the non-suicide group, even when clinical measures were taken into account. Clinical and theoretical implications are discussed. Ó 2010 Elsevier Ltd. All rights reserved.

Keywords: Psychosis Schizophrenia Suicide schema Semantic network

Introduction Suicide is the leading cause of premature death among individuals experiencing psychosis with the rate of suicide ranging from 147 to 750 per 100,000 persons per year (Heila et al., 1997; Palmer, Pankratz, & Bostwick, 2005). Approximately 40% of patients diagnosed with schizophrenia report suicidal ideation (Tarrier, Barrowclough, Andrews, & Gregg, 2004), 20e40% make at least one suicide attempt during the illness phase and 5e13% end their lives by suicide (Barraclough, Bunch, Nelson, & Sainsbury, 1974; Harris & Barraclough, 1997). The lifetime risk of completing suicide is estimated to be 20e50 times higher than in the general population (Caldwell & Gottesman, 1992). An established literature now exists describing factors reliably shown to be associated with an increased risk of suicide among people experiencing psychosis, including previous suicide attempts and comorbid mental health problems (Hawton, Sutton, Haw, Sinclair, & Deeks, 2005; Hu et al., 1991; Roy & Draper, 1995). The identification of such risk factors is a major strategy for predicting and preventing suicide (Tatarelli, Pompili, & Girardi, 2006). * Corresponding author. Division of Clinical Psychology, Zochonis Building, Brunswick Street, University of Manchester, UK. Tel.: þ44 161 306 0400; fax: þ44 161 306 0406. E-mail address: [email protected] (D. Pratt). 0005-7967/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.brat.2010.08.005

However, it remains a much more difficult task to prospectively evaluate which individual will eventually complete suicide (Bolton, Gooding, Kapur, Barrowclough, & Tarrier, 2007). In order to develop viable psychological interventions for suicide in psychosis, a better understanding is required of the underlying mechanisms. However, there are few well-articulated, theoretically driven and empirically tested models to explain suicidal behaviour in general (O’Connor & Sheehy, 2000) and in psychosis, in particular (Bolton et al., 2007). One theoretical model that attempts to explain suicidal behaviour is the Cry of Pain model (Williams, 1997). In brief, events, either directly or indirectly related to psychosis, can present as the necessary stressors for suicide risk. Specifically, situations of social rejections, failure to achieve valued roles or negative self-evaluation may be appraised in terms of defeat. Information processing biases, a negative schema and problem-solving deficits may influence appraisals such that inflexible negative perceptions of the self or negative responses to others become more likely. Positive, constructive exits or escape routes subsequently become limited. This process then elevates a felt sense of pessimism, worthlessness, and helplessness resulting in intractable feelings of entrapment. A real or perceived absence of rescue factors, in the form of social support resources that are available and important, accentuate the effects of this process. Finally, the ‘Cry of Pain’ can only be acted upon in the presence of imitation models and access to available means.

1212

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

Williams’ (1997) Cry of Pain model supports a common multifactor mechanism approach and, as such, should be equally applicable to a range of mental disorders. In recognition of the strengths and limitations of this model, we have developed the Schematic Appraisal Model of Suicide (SAMS; Johnson, Gooding, & Tarrier, 2008). The SAMS model extends the focus upon concepts of defeat, entrapment and ‘no rescue’ by specifying the key underlying cognitive and behavioural processes associated with suicidal behaviour. Negative information processing biases are thought to feed into a semantic memory system or ‘suicide schema’ and a multi-stage appraisal system (current, historical, future, self, agency). These latter two systems interact and determine goal directed escape behaviour towards suicide. Our empirical research so far has supported the SAMS model (Johnson, Tarrier, & Gooding, 2008; Tarrier, Gooding, Gregg, Johnson, & Drake, 2007; Taylor, Wood, Gooding, Johnson, & Tarrier, 2009). The ‘suicide schema’ can be seen as an example of a semantic network of interconnecting stimulus, response and emotional stored information pertaining to suicide. When activated, this schema will trigger thoughts of suicidal behaviour as an escape strategy from an intolerable emotional or situational state (Bower, 1981). According to spreading activation theories, each time the suicide schema is activated, it becomes strengthened and embellished as it incorporates further cognitive, emotional or stimulus elements, such as experiential psychotic symptoms and associated emotional, cognitive states or consequences (Teasdale, 1988). The more extensive and elaborate the suicide schema becomes, the greater its potential to be re-activated and subsequently even more refined, persistently adding to the individual’s risk of eventual suicide. Repeated activation of the schema will lead to associations with a wider range of mood states and contexts; thus increasing the risk of suicidal behaviour in the future (Williams, Crane, Barnhofer, & Duggan, 2005). It is expected that suicide schemas will vary from individual to individual, since the differential activation model suggests that people differ in the ease with which small changes in mood can reactivate particular networks of self-referent, negative thoughts (Williams et al., 2005). However, Rudd, Joiner, and Rajab (2001) hypothesised consistency across individuals in terms of categories or themes comprising the suicide schemas. Whilst the conceptual notion of schemas has been intuitively appealing since it was introduced into clinical applications by Beck (1967), empirical descriptions of individuals’ actual schemas are rare in the literature. The main aim of the current study was to construct an empirical representation of a suicide schema typical of individuals with psychosis and a history of suicidal behaviour (suicide group) and to compare that suicide schema with a suicide schema typical of individuals with psychosis but with no previous suicidal behaviour (non-suicide group). People with a history of suicidal behaviour are predicted to have activated their suicide schema more often than people without such a history (Lau, Segal, & Williams, 2004). Therefore, the suicide schema generated by the suicide group was hypothesised to be more extensive and elaborate, compared to the non-suicide group. Since the current study was exploratory in nature, two additional investigations were conducted to examine potential alternative explanations to any differences found between the suicide and nonsuicide groups. To investigate the potential influence of psychopathology on suicide schemas (Hawton et al., 2005), the current study examined whether differences in the groups’ suicide schemas could be explained by measures of psychopathology. A second exploratory hypothesis suggested the suicide schema generated for individuals with histories of multiple suicide attempts would be more extensive and elaborate than the suicide schema for individuals with one or no previous suicide attempts (Hu et al., 1991; Roy & Draper, 1995).

Method Design The study employed a cross-sectional between-groups design with a suicide group, comprising of individuals with a self-reported history of suicidal behaviour, compared with a non-suicide group, with no such history. All participants had previously received a diagnosis of a non-affective psychotic disorder. Participants and recruitment Participants were recruited via contact with the adult community mental health, early intervention and assertive outreach services within an NHS trust in the North West of England. Local voluntary sector mental health organisations also supported recruitment. Recruitment was conducted between October 2008 and April 2009. Eligible participants referred to the study by their care team were interviewed by a research psychologist (DP, JJ, PT) at a mutually convenient time and place. Following written consent, clinical measures and cognitive tasks were completed. The inclusion criteria for participants were (i) aged over 18 years; (ii) a chart diagnosis (ICD-10 criteria) of a non-affective psychosis (schizophrenia, schizophreniform disorder, schizoaffective psychosis, delusional disorder or psychosis not otherwise specified); (iii) under the care of an appropriate clinical team; and (iv) a sufficient grasp of the English language or English as first language to enable the completion of the measures. Participants were excluded from the study if (i) substance misuse or organic disorder was a primary diagnosis or judged to be the major cause of their psychotic experiences, (ii) they were currently acutely suicidal or considered a danger to themselves or others by the clinical team, or (iii) unable to give informed consent (e.g. displaying severe thought disorder). A self-reported history of at least one previous suicide attempt informed group allocation. Assessments and measures Anxiety The Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, 1988) was used to assess the severity of participants’ anxiety. The BAI is a 21-question multiple-choice self-report inventory that asks participants how much they have been “troubled by each symptom during the past week including today” (such as numbness, hot and cold sweats, or feelings of dread). Items are scored on a 0e3 point scale. The BAI total score ranges from 0 to 63, with higher scores indicating greater severity of anxiety. The BAI has previously been used in psychotic populations (Kuipers et al., 1997; Startup, Freeman, & Garety, 2007). Depression The revised version of the Beck Depression Inventory (BDI-II; Beck, Steer, & Brown, 1996) is a 21-item multiple-choice self-report inventory. Participants rate how they were feeling for the past fortnight on a four point scale (0e3). The items relate to depressive symptoms, cognitions, and physical symptoms. Responses are summed to provide an overall score ranging from 0 to 63, with higher scores indicating greater severity. The BDI-II is one of the most widely used instruments for measuring the severity of depression and has been used in samples of participants with psychosis (Birchwood, Iqbal, Chadwick, & Trower, 2000; Smith et al., 2006). Hopelessness The Beck Hopelessness Scale (BHS) is a 20-item, self-report inventory for measuring three major aspects of hopelessness;

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

negative attitudes about the future, loss of motivation and expectations (Beck, 1988). Higher scores indicating a greater degree of hopelessness and a total score of 9 or above is often used to identify those participants at risk of suicidal behaviour (Beck, Brown, Berchick, Stewart, & Steer, 1990). The BHS converged with clinician ratings of hopelessness at between r ¼ 0.62 and 0.74, has a reported alpha coefficient of 0.93 and a test-retest reliability of 0.85 over three weeks (Holden & Fekken, 1988). Among adult clinical samples, the BHS has repeatedly been found to be associated with repeated self-harm behaviours (Scott, House, Yates, & Harrington, 1997) and eventual suicide (Beck, Steer, Kovacs, & Garrison, 1985). The BHS is recommended for measuring extent of negative attitudes in clinical and research settings, and has been used in psychotic populations (Kuipers et al., 1997; Power et al., 2003). Suicidality The Beck Scale for Suicide Ideation (BSI; Beck & Steer, 1991) is a 21-item self-report instrument for detecting and measuring the current intensity of the individuals’ specific attitudes, behaviours, and plans to complete suicide during the past week. Individual items assess characteristics such as wish to die, desire to make an active or passive suicide attempt and amount of actual preparation for a contemplated attempt. The final two items assess the number of previous suicide attempts and the seriousness of the intent to die associated with the last attempt. The BSI has been validated for use with participants diagnosed with schizophrenia, schizoaffective, or bipolar disorders demonstrating an alpha coefficient of 0.96 and test-retest reliability over 1 week of 0.88 (Pinninti, Steer, Rissmiller, Nelson, & Beck, 2002) Psychotic symptoms The current study used the expanded version of the Brief Psychiatric Rating Scale (BPRS-E; Lukoff, Nuechterlein, & Ventura, 1986), which was specifically developed for evaluation of individuals diagnosed with schizophrenia. The BPRS-E comprises of 24 items to assess the broad range of symptoms indicative of schizophrenia, such as suspiciousness, hallucination and grandiosity. The BPRS-E was administered in an interview format with the participant. The first 14 items were rated on the basis of the individual’s self-report and the last 10 items rated according to observed behaviour and speech during the interview. Intra-class correlation coefficients between ratings by the three research psychologists for a subset of interviews (n ¼ 19) ranged between 0.87 and 0.91, suggesting good inter-rater reliability. Data regarding participants’ clinical histories were also obtained from participants and, if necessary, from their care coordinator (with prior consent). This information helped to determine the participant’s diagnosis and thus confirm eligibility for the study. Other details included age, gender, age at onset of psychotic experiences, length of illness and number of previous suicide attempts. Participants who self-reported previous use of illicit drugs were considered to have a ‘history of substance misuse’, and participants self-reporting a typical daily alcohol consumption of more than the recommended guidelines (2e3 units for women and 3e4 units for men; DoH, 2008) were considered to have a ‘history of alcohol misuse’. Measurement of suicide schemas Adopting the structural approach (Goldsmith, Johnson, & Acton, 1991), a series of cognitive tasks were designed to quantify the content and structure of participants’ suicide schema.1

1

Further details of these tests are available from the first author.

1213

Animal Category Fluency Task An Animal Category Fluency Task (ACF) was presented first to orientate participants to the cognitive tasks. The task was also used to assess possible differences between the groups on their ability to generate information (Lezak, Howiesson, & Loring, 2004), such that any participants experiencing difficulty in completing the ACF task were offered the opportunity not to continue with the remaining cognitive tasks. Participants’ responses to this task were recorded in the order they were produced, and the total number of words generated was calculated minus errors (i.e. category inappropriate words). Suicide Category Fluency Task To encourage participants to access their suicide schema, the study used the abstract concept of ‘Suicide’ as the category in a second fluency task. Whilst the fluency task paradigm traditionally uses concrete concepts as the category, e.g. animal, jobs, furniture, etc, this task has previously been adapted using abstract categories (see MacLeod, Pankhania, Lee, & Mitchell, 1997; MacLeod, et al., 1998). Participants’ responses in the Suicide Category Fluency Task (SCF) were recorded verbatim by the researcher in the order they were produced and a total was calculated. Suicide Category Sort Task Whilst the ‘direct’ technique of the SCF was used to assess participants’ schemas, this approach relied on the availability of the components of the schema to both introspection and articulation (Olsen & Rueter, 1987). Therefore, a complementary ‘indirect’ technique was also administered; the Suicide Concept Sort Task (SCS). The SCS required participants to sort a set of suicidal concepts into an order which they viewed as most related through to least related to the overarching concept of suicide. To develop this task, 100 concepts with varying degrees of relevance to suicide were identified from the academic and clinical literature. These concepts were collated into a questionnaire so that each concept’s relatedness to suicide could then be rated. A sample of 12 volunteers agreed to rate the ‘suicidality’ of the words listed, prior to the current study. Each word was rated on a 0e3 scale, with 0 indicating the word is not related to suicide at all, through to 3 indicating the word is closely related to suicide. Ratings of the 100 words by the volunteers were then collated and an overall rank of suicide relevance was established for the set of words. The SCS task was designed to be sensitive to capture concepts located centrally within the suicide schema, at the periphery of the schema, and between these two extremes. Concepts for this task were selected from a range of relatedness scores given by the volunteers. The 10 most related and 10 least related words were excluded from the list as they were expected to be, either, too centrally located or beyond the extreme periphery of all participants’ suicide schemas to be of use when discriminating between the two groups. Therefore, to allow sufficient variance amongst participants, the 10 words for the SCS task comprised of 3 words randomly selected from those rated 11th to 20th in relatedness, four words from those rated 46th to 55th, and 3 words from those rated 81st to 90th in relatedness. Examples of words include ‘beliefs’, ‘self-hate’, ‘hopeless’ and ‘relieving pain’. Participants were allowed up to five minutes to arrange the 10 words.

Procedure Each assessment interview started with a proforma to collect demographic information, followed by the BAI, BDI, BHS and BSS.

1214

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

Participants were then presented with the three cognitive tasks; ACF, SCF and then SCS. Finally, the BPRS was completed.

Results Sample characteristics

Sample size The current study proposed the novel use of a methodology in an area of research previously untried; therefore the expected number of participants was informed by previous studies that used similar methodologies, albeit in different areas of research. Also, for the solutions generated from the analyses to be stable and reliable, a widely-held heuristic “rule-of-thumb” suggested there should be no less than 3 participants for every item contained within the constructed solution (Davison, 1983; Kruskal & Wish, 1978). Since 10 concepts were presented to participants in the SCS task, it was proposed that the suicide and non-suicide groups should comprise of 30 participants each. This sample size was in keeping with other studies (Paulsen et al., 1996; Prescott, Newton, Mir, Woodruff, & Parks, 2006). Methodological innovations in the current study2 Data produced from the ACF, SCF and SCS tasks consisted of lists of words generated, in production order, by each participant. It is widely thought that the ordering of words in these lists is indicative of the semantic distance between items within the schema (Crowe & Prescott, 2003). Semantically similar words appear closer to each other in participants’ lists than dissimilar words (Neely, 1991). Participants’ responses to the three tasks were transformed into a matrix of proximity data, where the number in each cell represented the distance (i.e. in order of response) between two items. Thus, the higher the number, the greater the distance between the items reported. Individual proximity matrices were then combined into a group matrix for each group, using Prescott et al.’s (2006) Mean Cumulative Frequency (MCF) metric.3 The Pathfinder technique (Schvaneveldt, 1990) was then used to generate graphical descriptions or representations of the underlying structures of each group’s proximity data. For each Pathfinder network, a measure of Coherence was computed to indicate the “internal consistency” of the data. A very low coherence measure (<0.20) indicates a lack of consistency and renders interpretation of the networks meaningless. Comparisons of the Pathfinder networks were then conducted. Firstly, the MCF metric for each pair of items within the group’s networks were correlated (Goldsmith & Davenport, 1991), with higher correlations indicative of greater network similarity. Secondly, Goldsmith and Davenport’s (1991) Closeness statistic was calculated to measure the proportion of shared links for matching items across two different networks, where greater values indicate increasing similarity. Finally, as recommended by Prescott et al. (2006), a multivariate analysis of variance (MANOVA) was conducted of the MCF metric and standard deviation of the MCF metric as dependent variables and group as the independent variable. Data were analysed using SPSS version 14.0 for Windows and the software package Pathfinder for Windows, Version 6.2 (Interlink ltd, Gilbert, Arizona, USA) was used to generate graphical representations of the networks.

2 Further details of the Pathfinder technique and associated analytic procedures are available from the first author. 3 The transformation of the fluency and sort data into proximity matrices was performed by custom software written by Prof. Prescott, University of Sheffield, who provided access to this software for the current study.

A total of 90 participants were recruited to the study, although data for 5 participants were excluded from the analyses as the participants failed to meet the inclusion criteria, specifically with respect to psychiatric diagnosis (2  bipolar affective disorder, 2  recurrent depressive disorders, 1  no recorded diagnosis). A sixth participant withdrew from the study before completing any measures. The final sample of 84 participants comprised of 61 (73%) males with an average age of 42.3 years (SD ¼ 11.84), ranging from 19 to 73 years (median ¼ 42). Twenty-six (31%) participants reported no previous suicide attempts and were allocated to the non-suicide group. Fifty-eight (69%) participants reported at least one previous suicide attempt and were, therefore, allocated to the suicide group. Forty (48%) participants reported multiple previous suicide attempts. The mean number of previous suicide attempts in the suicide group was 2.6 (SD ¼ 1.98) with a minimum of 1 and a maximum number of attempts of 11 (median ¼ 2). The demographic and clinical characteristics of the sample are given in Table 1. The mean age of the suicide group (M ¼ 42.22, SD ¼ 10.60) did not differ significantly from the mean age of the non-suicide group (M ¼ 42.35, SD ¼ 14.46; t[83] ¼ .043, n.s.) and no significant differences were found between the two groups, in terms of gender (Fisher’s exact test, p ¼ .119). In terms of ethnicity, the suicide group had a greater percentage of white participants and a lesser percentage of participants of an Asian ethnicity (c2 ¼ 9.871, df ¼ 4, p ¼ .043). The most common diagnosis for both groups was Schizophrenia (suicide: n ¼ 52, 90%; non-suicide: n ¼ 23, 89%), and the range of psychiatric diagnoses were similar across groups (c2 ¼ 0.963, df ¼ 2, n.s.). No significant differences were found between the groups in terms of previous alcohol misuse (suicide: n ¼ 12, 22%; nonsuicide: n ¼ 3, 12%; Fisher’s exact test, n.s.) or substance misuse (suicide: n ¼ 34, 59%; non-suicide: n ¼ 13, 50%; Fisher’s exact test, n.s.). Furthermore, there were no group differences for age at onset of psychosis (suicide: M ¼ 24.11, SD ¼ 9.24; non-suicide: M ¼ 24.63, SD ¼ 11.22; t[79] ¼ .217, n.s.), duration of untreated psychosis (suicide: median ¼ 50, range ¼ 1e1100; non-suicide: median ¼ 25, range ¼ 1e1200; ManneWhitney U ¼ 522.0, n ¼ 76, n.s.), duration of Illness (suicide: M ¼ 17.72, SD ¼ 9.68; non-suicide: M ¼ 16.67, SD ¼ 13.52; t[79] ¼ .396, n.s.), and time since last relapse (suicide: median ¼ 150, range ¼ 1e1250; non-suicide: median ¼ 200, range ¼ 1e2150; ManneWhitney U ¼ 652.5, n ¼ 79, n.s.). Clinical measures As would be expected, the suicide group scored significantly higher on the BAI (M ¼ 16.97, SD ¼ 12.39) than the non-suicide group (M ¼ 8.41, SD ¼ 9.57; t[83] ¼ 3.172, p ¼ .002), and on the BDI (suicide: M ¼ 17.70, SD ¼ 11.88; non-suicide: M ¼ 9.89, SD ¼ 9.89; t[82] ¼ 2.963, p ¼ .004) (see Table 2). Similarly, the suicide group also scored significantly higher on the BHS (M ¼ 8.26, SD ¼ 5.57) than the non-suicide group (M ¼ 5.37, SD ¼ 5.23; t[82] ¼ 2.266, p ¼ .026) indicating greater levels of hopelessness amongst the suicide group. Also, scores on the BSI were significantly higher for the suicide group (median ¼ 4, range ¼ 0e25) than the non-suicide group (median ¼ 0, range ¼ 0e16; ManneWhitney U ¼ 118.5, n ¼ 83, p < .001). Finally, scores on the BPRS for the suicide group (M ¼ 43.60, SD ¼ 10.99) were not significantly different from scores for the non-suicide group (M ¼ 39.04, SD ¼ 8.18; t[82] ¼ 1.655, n.s.). Therefore, the two groups did not differ on measures indicative of

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

1215

Table 1 Demographic and clinical characteristics of the sample.

c2

df

p

85 15

2.725

1

0.119

19 2 3 1 1

73 8 12 4 4

9.871*

4

0.043

90 7 3

23 1 2

89 4 8

0.963

2

0.618

12 3

22

3

12

1.092

1

0.370

History of Substance Misusea

34

59

13

50

0.541

1

0.485

Age at Onset of Psychosis (yrs) Missing

M 24.11 1

(SD) (9.24)

M 24.63 2

SD (11.22)

t 0.217

df 79

p 0.829

Duration of Untreated Psychosis (wks)b Missing

138.35 3

(227.76)

128.09 5

(269.67)

522.0

Duration of Illness (yrs) Missing

17.72 1

(9.68)

16.67 2

(13.52)

0.396

Time since last relapse (wks)b Missing

263.25 3

(294.06)

343.83 2

(524.20)

652.5

Suicide

Non-suicide

n ¼ 58

%

n ¼ 26

%

Gendera Male Female

39 19

67 33

22 4

Ethnicity White Black Asian Other Unknown

51 6 0 1 0

88 10 0 2 0

Psychiatric Diagnosis Schizophrenia Schizoaffective Disorder Psychosis NOS

52 4 2

History of Alcohol Misusea Missing

0.517 79

0.693 0.936

*p < 0.05 **p < 0.01 ***p < 0.001. a Fisher’s exact test. b ManneWhitney U test, as variables not normally distributed.

the severity of psychosis, i.e. BPRS scores and duration of untreated psychosis.

could not simply be accounted for by the suicide group being more (or less) cognitively fluent.

Animal Category Fluency Task

Suicide Category Fluency Task

The Animal Category Fluency Task (ACF) was the first of the cognitive tasks. Eighty-one participants completed the Animal Category Fluency Task (ACF), comprising of 25 (96%) of the 26 participants from the non-suicide group, and 56 (97%) of the 58 participants from the suicide group. Across both groups, a total of 329 distinct animal concepts were generated and the most frequently named animals were dog and cat, reported by 99% and 98% of participants, respectively. The mean number of animal concepts named by the suicide group (M ¼ 25.29, SD ¼ 10.92) was not significantly different from the mean for the non-suicide group (M ¼ 21.72, SD ¼ 9.50; t[79] ¼ 1.411, n.s.). This result suggests that subsequent findings

The Suicide Category Fluency (SCF) task was completed by 80 participants; 25 (96%) from the non-suicide group and 55 (95%) from the suicide group. A total of 484 distinct suicide words or phrases were listed by all participants combined. Tablets (43%), cut self (36%), death (33%), hanging (31%) and depressed (28%) were the most commonly named concepts. The number of words generated on the SCF task, i.e. SCF scores, for two participants from the suicide group and two from the non-suicide group were recognised as being more than 3 standard deviations away from the group mean. Considered to be outliers (Osborne & Overbay, 2004), these data were subsequently excluded from the analysis (Barnett & Lewis, 1994). Following the removal of the outliers, the adjusted mean

Table 2 Comparison of clinical measures between-groups. Measure

BAI BDI BHS BSIa BPRS

Suicide (n ¼ 58)

Non-suicide (n ¼ 26)

Mean

(SD)

Range

Mean

SD

Range

16.97 17.70 8.26 7.34 43.60

(12.39) (11.88) (5.57) (6.66) (10.99)

0e59 0e47 0e18 0e25 26e76

8.73 10.19 5.58 1.11 39.62

(9.61) (9.96) (5.22) (3.31) (8.18)

0e35 0e39 0e20 0e16 24e60

*p < 0.05 **p < 0.01 ***p < 0.001. a ManneWhitney U test, as variables not normally distributed.

t

df

p

3.004** 2.803** 2.078* 116.0*** 1.655

82 81 81

0.004 0.006 0.041 <0.001 0.102

82

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

SCF score for the suicide group (M ¼ 11.89, SD ¼ 7.37) was significantly greater than for the non-suicide group (M ¼ 8.57, SD ¼ 5.12; t[59.074] ¼ 2.257, p ¼ .028).4 To control for group differences in psychopathology found in earlier analyses, an analysis of covariance was conducted on the SCF scores, entering BAI, BDI, BHS, BSI and BPRS scores as covariates. A significant main effect for group was obtained (F(1,69) ¼ 3.952, p ¼ .05). Therefore, the suicide group’s mean SCF score (adjusted M ¼ 12.05, SE ¼ 0.98) was significantly greater than the mean SCF score for the non-suicide group (adjusted M ¼ 8.20, SE ¼ 1.56) when group differences in anxiety, depression, hopelessness, current suicidal ideation and psychopathology scores were taken into account. Participants with more extensive histories of suicidal behaviour, as evidenced by multiple suicide attempts, were expected to have more developed suicide schemas than individuals with a single previous suicide attempt. The group with no previous suicidal behaviour were expected to have the least developed suicide schemas. To test this hypothesis, the suicide group was split into two subgroups of participants reporting a single previous suicide attempt and those with multiple previous attempts. The 4 outliers were also excluded from this analysis. Of the remaining 76 participants, 23 (30%) participants remained in the non-suicide group, 15 (20%) were allocated to the single attempt group, and 38 (50%) to the multiple attempts group. As expected, the non-suicide group achieved the lowest mean SCF score (M ¼ 8.57, SD ¼ 5.12), then the single attempt group (M ¼ 10.73, SD ¼ 6.81) and the multiple attempt group achieved the highest mean score (M ¼ 12.35, SD ¼ 7.62) although the group differences failed to achieve statistical significance (F(2,73) ¼ 2.22, p ¼ .116). Since the study was exploratory in nature, post hoc analyses were conducted to examine the potential of a trend across the three groups. Indeed, the multiple attempts group named significantly more concepts than the non-suicide group (Dunnett’s test (multiple attempts versus non-suicide): p ¼ .036), but no significant differences were found between the multiple attempts and single attempt groups or the single attempts group and the non-suicide group. A tentative exploratory analysis was conducted to investigate the trend indicated in the group mean SCF scores. The error bar chart in Fig. 1 shows there was considerable overlap in the 95% confidence intervals for the mean SCF scores of the three groups. There was also a significant positive correlation between the SCF scores and the number of previous suicide attempts (Spearman’s rho ¼ .287, p ¼ .012). These results may be indicative of the ongoing development of participants’ suicide schemas. Goldsmith et al. (1991) recommended a subset of high frequency items is selected when generating representations of semantic networks. Therefore, only items named by at least 10% of participants were included in the Pathfinder analysis. Participants’ responses were then transformed into proximity data using Prescott et al.’s (2006) MCF metric and input into a Pathfinder analysis. The resulting networks can be seen in Figs. 2 and 3. There was a strong degree of coherence for the non-suicide data (0.60) and the coherence measure for the suicide aggregated data (0.48) was acceptable. Therefore, meaningful interpretations can be made from the groups’ network structures. Comprising of 16 nodes, the suicide group’s network appears more expansive than the network for the non-suicide group, which comprises of 9 nodes. Also, the suicide group’s network appears more dense and complex in structure with 5 nodes having more than two links with

4 Equality of variances was not assumed as Levene’s test was significant (F ¼ 4. 079, p ¼ .047).

15

Mean SCF scores

1216

12

9

6

non-suicide

single attempt

multiple attempts

Fig. 1. Error bar chart of mean SCF Scores for the three groups.

neighbouring nodes, whereas the non-suicide group’s network is fairly simple with the minimum number of links between items. Both networks show a clear organisational focus on methods of suicidal behaviour. Notably, two jumping methods were adjacently located and linked with the pairing of rope and tablets for both the suicide and non-suicide groups. The centrality of these three methods (jumping, hanging and overdose) in both networks may reflect the frequency of such methods used by the general population, accounting for over 80% of self-inflicted deaths in 2008 (ONS, 2009). Whilst the network for the non-suicide group is limited to methods, the suicide group’s network extends to associated feelings (depressed), consequences (funeral, death) and the impact on others (relationships), perhaps indicative of a deeper level of understanding of suicide. For a comparison to be made between the two networks, a subset of common items stated by both the suicide and nonsuicide group was required. For fluency tasks with concrete categories, such as Animals, this requirement is usually met. However, the use of the abstract category of ‘Suicide’ in a fluency task brought this requirement under threat, with only a small number of items named by both groups. Of the original 28 items named by 10% of either group, only 7 items were named by both groups. The subset of common items comprised of cut self, hanging, knife, jump from height, jump in front of vehicle, rope and tablets, which will be denoted SCF7. The groups’ responses on the SCF7 items were then compared. The correlation between the inter-item distances, measured by the MCF metric, for the two networks provided an index of network (dis)similarity. No significant correlation was found (Pearson’s r ¼ .237, n.s.), which was interpreted as the two schemas being dissimilar (Goldsmith & Davenport, 1991). The Closeness statistic, C, also failed to indicate similarity between the two networks (C ¼ 0.333, n.s.). A comparison of the underlying structure of the proximity data for the SCF7 items was achieved by examining the variability in the inter-item distance measure (MCF) for each item pair. For a network to consist of stronger inter-item relationships, the suicide group were predicted to have a significantly lower mean standard deviation of the MCF metric than for the non-suicide group. Multivariate Analysis of Variance (MANOVA) of these data was employed to compare the variability in the inter-item distance measures for pairs of the SCF7 items. Although Box’s M test was significant (Box’s M ¼ 12.925, p ¼ .007), roughly equal group sizes

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

1217

Fig. 4. Pathfinder network of SCS items for the suicide group.

However, the lack of group differences in terms of the variability of the proximity data suggests there was no evidence that the suicide schema generated by participants with a history of suicidal behaviour was more organised or structured. Suicide category sort task Fig. 2. Pathfinder network of SCF items named by the suicide group.

indicated the robust Pillai’s trace criterion would be an accurate multivariate test statistic (Field, 2000). The MCF metric and its standard deviation were treated as dependent variables. A significant multivariate difference between the two groups was found (F(2,36) ¼ 3.78, p ¼ .032, Pillai’s trace ¼ .174). Univariate tests revealed a significant group difference on the MCF metric (suicide: M ¼ 0.382, SD ¼ 0.119; non-suicide: M ¼ 0.514, SD ¼ 0.199; F(1,37) ¼ 6.591, p ¼ .014), but no significant group difference on the standard deviation of the MCF metric (suicide: M ¼ 0.206, SD ¼ 0.107; non-suicide: M ¼ 0.164, SD ¼ 0.101; F(1,37) ¼ 1.536, n.s.). Therefore, the inter-item distances between concepts within the suicide group’s suicide schema were significantly smaller than for the non-suicide group, indicating a more dense network.

Fig. 3. Pathfinder network of SCF items named by the non-suicide group.

A total of 79 participants completed the SCS task, comprising of 24 (92%) of the 26 participants from the non-suicide group, and 55 (95%) of the 58 participants from the suicide group. The Pathfinder technique was used to create diagrammatic representations of the proximity data generated from the SCS task. The resulting networks can be seen in Figs. 4 and 5. There was a very strong degree of coherence for both the suicide (0.76) and the non-suicide (0.77) aggregated data, hence interpretations can be made from the groups’ network structures. Whereas the two networks previously derived from the fluency data presented clear differences between the suicide and non-suicide groups, the networks derived from the sort data appear more similar. The trio of personality, beliefs and sinful form a clique in both networks and there also appears a common organisational focus upon the 5 items; self-esteem, hopeless, suffering, relieving pain and self-hate. Whilst these five items do not share the same inter-item pairings, they can be seen to form a grouping in the centre of both networks. Notably, the psychosis item occupies a central location within the non-suicide group’s network but more peripheral in the suicide group’s network, whilst death sits towards the outside of both networks. A significantly high correlation was found between the interitem distances in the two networks (Pearson’s r ¼ .681, p < .001), which is interpreted as the two schemas being similar. The

Fig. 5. Pathfinder network of SCS items for the non-suicide group.

1218

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

Closeness statistic, C, also indicated similarity between the two networks (C ¼ .294) although this measure failed to achieve statistical significance. As for the SCF data, a MANOVA was employed to compare the variability in the inter-item distance measures for pairs of items from the SCS task. The sensitive Box’s M test was again significant (Box’s M ¼ 13.80, p ¼ .004), hence Pillai’s trace was used as the test statistic (Field, 2000). The test revealed there was no significant multivariate difference between the two groups (F(2,87) ¼ 1.352, n.s.). Univariate analyses of variance of the MCF metric (suicide: M ¼ 0.467, SD ¼ 0.079; non-suicide: M ¼ 0.467, SD ¼ 0.099; F(1,88) ¼ .01, n.s.) and the standard deviation of the MCF metric (suicide: M ¼ 0.261, SD ¼ 0.019; non-suicide: M ¼ 0.254, SD ¼ 0.026; F(1,88) ¼ 2.353, n.s.) found no significant differences between the suicide and non-suicide group. These tests failed to reveal a significant difference in the variability in the normalised distance measures generated from the non-suicide group compared to the suicide group. Discussion The main aim of the study was to construct an empirical representation of a suicide schema typical of individuals with psychosis and a history of suicidal behaviour, which could be compared with that of individuals with psychosis but with no previous suicidal behaviour. The main hypothesis predicted the schema for the suicide group would be more extensive and elaborate than for the non-suicide group. Additional hypotheses predicted group differences in the suicide schemas would not be explained by measures of psychopathology, and individuals with histories of multiple suicide attempts would have more extensive and elaborate suicide schemas. In the current study, more than two-thirds (69%) of individuals with non-affective psychosis reported a history of suicidal behaviour, with almost half (48%) reporting multiple previous suicide attempts. These figures are noticeably higher than reported in the literature, with reported estimates of individuals with psychosis reporting a previous suicide attempt ranging from 18% to 55% (Bolton et al., 2007; Siris, 2001). The current sample were on average into their forties and had been living with psychosis for more than 16 years, which may partially explain the high frequency of previous suicidal behaviour. The high proportion may have also reflect a methodological bias with referrers tending to approach potential participants with such a history. The two groups were found to differ on anxiety, depression, hopelessness and suicidality, as is consistent with studies of the risk factors for suicide (Barraclough et al., 1974; Harkavy-Friedman et al., 1999; Tarrier et al., 2004). Similarly, the lack of a group difference on a measure of psychotic symptomatology was consistent with a systematic review of risk factors for suicide in schizophrenia (Hawton et al., 2005). Previous studies have found substance misuse to be significantly more likely amongst suicide attempters, both with and without psychosis (Beautrais et al., 1996; Hawton et al., 2005; Tarrier, Haddock, Lewis, Drake, & Gregg, 2006), although this difference was not found in the current study. On the Animal Category Fluency (ACF) task, the current study was consistent with previous findings in terms of the number of animals named by participants with schizophrenia (Fossatia, Amarb, Raouxc, Ergisc, & Allilairea, 1999) and the lack of a difference in the ability to generate items from a common semantic category between individuals with and without previous suicidal behaviour (MacLeod, Rose, & Williams, 1993). Therefore, participants can be seen as representative of the population of people with non-affective psychoses. On the Suicide Category Fluency Task (SCF), the suicide group generated a significantly greater number of concepts than the non-

suicide group. When statistically controlling for measures of anxiety, depression, hopelessness, suicidality and an overall measure of psychotic symptomatology, group differences remained. Therefore, differences in the size of each group’s suicide schema could not be better explained by the presence of specific mental health problems or disorders, previously shown to be key risk factors for suicide in psychosis (Heila et al., 1997). Hence, the suicide schema may be a central component of a trans-diagnostic causal mechanism for suicidal behaviour (Bolton et al., 2007). When the suicide group was split into multiple and single attempters, which were then compared with the non-suicide group, there was some evidence to suggest the single attempt group ‘bridged the gap’ between the other two groups, with the suicide schema in a transitional stage of development. This interpretation was supported by a significant positive correlation between the SCF scores and the number of previous suicide attempts. Such findings are in accord with Teasdale’s (1988) differential activation theory, which predicts that as a person’s repertoire of suicidal behaviour expands, their underlying suicide schema, which structures their understanding of these experiences, develops and expands accordingly. Pathfinder networks, generated from the proximity data generated from the cognitive tasks, were used to represent the suicide schemas. The suicide group’s network was shown to be more elaborate and extensive than the network for the non-suicide group, which suggests people with a history of suicidal behaviour possess more extensive suicide schemas than individuals with no such history (Tarrier et al., 2007). The SAMS model (Johnson, Gooding, et al., 2008) introduced the suicide schema as comprising of a wide range of cognitive and behavioural events from ideation, planning, intention to act, and actual self-harm, that become entangled and intertwined in a network of associations, which are further strengthened each time the schema is re-activated. According to this model, a thorough assessment of suicide risk would, necessarily, pay attention to the individual’s suicide schema, as this component may provide the link between the acute episodes of suicidal behaviour observable to the clinician. For some individuals, the resolution of an acute suicidal crisis does not mean that their enduring suicide risk is low. The clinician needs to remain aware of chronic risk, in addition to acute suicide risk, by constructing a formulation of the underlying relationship between the recently subsided acute risk and the maintaining factor of the individual’s suicide schema. The study has a number of limitations. Group allocation was dependent upon participant’s response to the question “have you ever attempted suicide?” Self-report assessments of previous suicidal behaviour have been shown to be reliable and valid (Beck & Steer, 1991), however, such assessments assume participants share a common definition of ‘suicide’. Therefore, the current study is limited by the definition of suicide assumed by participants, which may reflect the practical, clinical and theoretical confusion between suicide and self-injurious behaviour that exists in the literature (Ellis, 2006). There has been considerable debate in the literature as to how best to interpret group differences in performance on fluency tasks by participants with schizophrenia. Some studies have suggested that differences in performance are attributable to general cognitive impairments, which lead to difficulties in accessing or searching a schema, rather than its organisation, per se (Bokat & Goldberg, 2003). The current study consisted of two groups with non-affective psychosis, therefore any cognitive difficulties affecting performance were expected to be randomly distributed across both groups, and would therefore have no effect on group comparisons. Also, the use of fluency tasks to assess suicide schema could potentially introduce a confounder if individuals with

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

a history of suicidal behaviour perform differently on such tasks compared to those without such a history. The literature investigating executive functioning deficits amongst suicide attempters appears mixed, with some evidence to suggest the performance of suicide attempters on tests of executive functioning is diminished compared with non-attempters (Bartfai, Winborg, Nordstrom, & Asberg, 1990; Keilp et al., 2001;), but other studies have failed to identify any neuropsychological differences between such groups, especially on tests of verbal fluency (Ellis, Berg, & Franzen, 1992; King et al., 2000; Marzuk, Hartwell, Leon, & Portera, 2005). The current study failed to identify any group differences in performance on the Animal Category Fluency task, suggesting equivalent executive functioning capabilities across the two groups, although future studies should assess the impact of such confounders in more explicit detail. The current study investigated the cognitive architecture underlying suicidal behaviour in individuals with non-affective psychosis, focussing on the role of the suicide schema. Adopting a novel and innovative approach to the empirical description of suicide schemas, this study has made an important contribution to the understanding of these cognitive mechanisms. To be able to develop viable psychological treatments for suicidality, a better understanding is required of the key psychological processes underlying suicidal behaviour (Tarrier, Taylor, & Gooding, 2008). Evidence from the current study suggests the Schematic Appraisal Model of Suicide (Johnson, Gooding, et al., 2008) shows considerable promise, both theoretically and clinically. Ethics committee approval Ethical approval was granted by Stockport NHS Research Ethics Committee. Role of the funding source The study was funded by the ClinPsyD training contract awarded to the Division of Clinical Psychology, University of Manchester by the North West Strategic Health Authority and was undertaken as part of a ClinPsyD programme by DP. Acknowledgements We are grateful to all participants who gave their time to this study, and to the Mental Health Research Network, Making Space, MIND and Creative Support for their assistance in the recruitment of participants. We would like to acknowledge Prof Tony Prescott, University of Sheffield, for access to his software and advice with analysis of the fluency data. References Barnett, V., & Lewis, T. (1994). Outliers in statistical data (3rd ed.). New York: Wiley. Barraclough, B., Bunch, J., Nelson, B., & Sainsbury, P. (1974). A hundred cases of suicide: clinical aspects. British Journal of Psychiatry, 125, 355e373. Bartfai, A., Winborg, I. M., Nordstrom, P., & Asberg, M. (1990). Suicidal behavior and cognitive flexibility: design and verbal fluency after attempted suicide. Suicide and Life Threatening Behavior, 20, 254e266. Beautrais, A. L., Joyce, P. R., Mulder, R. T., Fergusson, D. M., Deavoll, B. J., & Nightingale, S. K. (1996). Prevalence and comorbidity of mental disorders in persons making serious suicide attempts: a case-control study. American Journal of Psychiatry, 153, 1009e1014. Beck, A. T. (1967). Depression. New York: Harper and Row. Beck, A. T. (1988). Manual for the Beck hopelessness scale. San Antonio, Texas: Psychological Corporation. Beck, A. T., Brown, G., Berchick, R., Stewart, B., & Steer, R. (1990). Relationship between hopelessness and ultimate suicide: a replication with psychiatric outpatients. American Journal of Psychiatry, 147, 190e195.

1219

Beck, A. T., Epstein, N., Brown, G., & Steer, R. (1988). An inventory for measuring clinical anxiety: psychometric properties. Journal of Consulting and Clinical Psychology, 56, 893e897. Beck, A. T., & Steer, R. A. (1991). Manual for the Beck scale for suicide ideation. San Antonio, Texas: Psychological Corporation. Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck depression inventoryeII. San Antonio, Texas: Psychological Corporation. Beck, A. T., Steer, R. A., Kovacs, M., & Garrison, B. (1985). Hopelessness and eventual suicide: a 10-year prospective study of patients hospitalized with suicidal ideation. American Journal of Psychiatry, 142, 559e563. Birchwood, M., Iqbal, Z., Chadwick, P., & Trower, P. (2000). Cognitive approach to depression and suicidal thinking in psychosis I. Ontogeny of post-psychotic depression. British Journal of Psychiatry, 177, 516e528. Bokat, C. E., & Goldberg, T. E. (2003). Letter and category fluency in schizophrenic patients: a meta-analysis. Schizophrenia Research, 64, 73e78. Bolton, C., Gooding, P., Kapur, N., Barrowclough, C., & Tarrier, N. (2007). Developing psychological perspectives of suicidal behaviour and risk in people with a diagnosis of schizophrenia: we know they kill themselves but do we understand why? Clinical Psychology Review, 27, 511e536. Bower, G. (1981). Mood and memory. American Psychologist, 36, 129e148. Caldwell, C. B., & Gottesman, I. I. (1992). Schizophrenia e a high risk factor for suicide: clues to risk reduction. Suicide and Life Threatening Behavior, 22, 479e493. Crowe, S. J., & Prescott, T. J. (2003). Continuity and change in the development of category structure: insights from the semantic fluency task. International Journal of Behavioral Development, 27, 467e479. Davison, M. L. (1983). Multidimensional scaling. New York: Wiley. Department of Health. (2008). Drinking, you and your mates: How does your night end? London: HMSO. Ellis, T. E. (Ed.). (2006). Cognition and suicide. Theory, research and therapy. Washington, DC: American Psychological Association. Ellis, T. E., Berg, R. A., & Franzen, M. D. (1992). Neuropsychological performance and suicidal behavior in adult psychiatric patients. Perceptual and Motor Skills, 75, 639e647. Field, A. (2000). Discovering statistics using SPSS for windows: Advanced techniques for the beginner. London: Sage. Fossatia, P., Amarb, G., Raouxc, N., Ergisc, A. M., & Allilairea, J. F. (1999). Executive functioning and verbal memory in young patients with unipolar depression and schizophrenia. Psychiatry Research, 89, 171e187. Goldsmith, T. E., & Davenport, D. M. (1991). Assessing structural similarity of graphs. In R. W. Schvaneveldt (Ed.), Pathfinder associative networks: Studies in knowledge organization (pp. 1e31). Norwood, NJ, USA: Ablex. Goldsmith, T. E., Johnson, P. J., & Acton, W. H. (1991). Assessing structural knowledge. Journal of Educational Psychology, 83, 88e96. Harkavy-Friedman, J. M., Restifo, K., Malaspina, D., Kaufman, C. A., Amador, X. F., Yale, S. A., et al. (1999). Suicidal behaviour in schizophrenia: characteristics of individuals who had and had not attempted suicide. American Journal of Psychiatry, 156, 1276e1278. Harris, E. C., & Barraclough, B. (1997). Suicide as an outcome for mental disorders; a meta-analysis. British Journal of Psychiatry, 170, 205e228. Hawton, K., Sutton, L., Haw, C., Sinclair, J., & Deeks, J. J. (2005). Schizophrenia and suicide: systematic review of risk factors. British Journal of Psychiatry, 187, 9e20. Heila, H., Isometsa, E. T., Henriksson, M. M., Heikkinen, M. E., Martinnen, M. J., & Loonqvist, J. K. (1997). Suicide and schizophrenia: a nationwide psychological autopsy study on age and sex specific clinical characteristics of 92 suicide victims with schizophrenia. American Journal of Psychiatry, 154, 1235e1242. Holden, R. R., & Fekken, G. C. (1988). Test-retest reliability of the hopelessness scale and its items in a university population. Journal of Clinical Psychology, 44, 40e43. Hu, W. H., Sun, C. M., Lee, C. T., Peng, S. L., Lin, S. K., & Shen, W. W. (1991). A clinical study of schizophrenic suicides: 42 cases in Taiwan. Schizophrenia Research, 5, 43e50. Johnson, J., Gooding, P., & Tarrier, N. (2008). Suicide risk in schizophrenia: explanatory models and clinical implications, the Schematic Appraisal Model of Suicide (SAMS). Psychology and Psychotherapy: Theory, Research and Practice, 81, 55e77. Johnson, J., Tarrier, N., & Gooding, P. (2008). An investigation of aspects of the cry of pain model of suicide risk: the role of defeat in impairing memory. Behaviour Research and Therapy, 46, 968e975. Keilp, J. G., Sackheim, H. A., Brodsky, B. S., Oquendo, M. A., Malone, K. M., & Mann, J. J. (2001). Neuropsychological dysfunction in depressed suicide attempters. American Journal of Psychiatry, 158, 735e741. King, D. A., Conwell, Y., Cox, C., Henderson, R. E., Denning, D. G., & Caine, E. D. (2000). A neuropsychological comparison of depressed suicide attempters and non attempters. Journal of Neuropsychiatry and Clinical Neurosciences, 12, 64e70. Kruskal, J. B., & Wish, M. (1978). Multidimensional scaling. Newberry Park, CA: Sage. Kuipers, E., Garety, P., Fowler, D., Dunn, G., Bebbington, P., Freeman, D., et al. (1997). London-East Anglia randomised controlled trial of cognitive-behavioural therapy for psychosis. I: effects of the treatment phase. British Journal of Psychiatry, 171, 319e327. Lau, M. A., Segal, Z. V., & Williams, J. M. G. (2004). Teasdale’s differential activation hypothesis: implications for mechanisms of depressive relapse and suicidal behaviour. Behaviour Research and Therapy, 42, 1001e1017. Lezak, M. D., Howiesson, D. B., & Loring, D. W. (2004). Neuropsychological assessment (4th ed.). New York: Oxford University Press. Lukoff, D., Nuechterlein, K. H., & Ventura, J. (1986). Manual for the expanded BPRS. Schizophrenia Bulletin, 12, 594e602.

1220

D. Pratt et al. / Behaviour Research and Therapy 48 (2010) 1211e1220

MacLeod, A. K., Tata, P., Evans, K., Tyrer, P., Schmidt, U., Davidson, K., et al. (1998). Recovery of positive future thinking within a high-risk parasuicide group: results from a pilot randomized controlled trial. British Journal of Clinical Psychology, 37, 371e379. MacLeod, A. K., Pankhania, B., Lee, M., & Mitchell, D. (1997). Parasuicide, depression and the anticipation of positive and negative future experiences. Psychological Medicine, 27, 973e977. MacLeod, A. K., Rose, G. S., & Williams, J. M. G. (1993). Components of hopelessness about the future in parasuicide. Cognitive Therapy and Research, 17, 441e455. Marzuk, P. M., Hartwell, N., Leon, A. C., & Portera, L. (2005). Executive functioning in depressed patients with suicidal ideation. Acta Psychiatrica Scandinavia, 112, 294e301. Neely, J. H. (1991). Semantic priming effects in visual word recognition: a selective review of current findings and theories. In D. Besner, & G. W. Humphreys (Eds.), Basic Processes in Reading: Visual Word Recognition (pp. 264e336). Hillsdale, NJ: Lawrence Erlbaum Associates. O’Connor, R. C., & Sheehy, N. (2000). Understanding suicidal behaviour. Oxford: British Psychological Society - Blackwell. Office for National Statistics. (2009). Mortality statistics: Deaths registered in 2008. London: Office for National Statistics. Olsen, J. R., & Rueter, H. H. (1987). Extracting expertise from experts: methods for knowledge acquisition. Expert Systems, 4, 152e168. Osborne, J. W., & Overbay, A. (2004). The power of outliers (and why researchers should always check for them). Practical Assessment, Research and Evaluation, 9, 1e10. Palmer, B. A., Pankratz, V. S., & Bostwick, J. M. (2005). The lifetime risk of suicide in schizophrenia e a reexamination. Archives of General Psychiatry, 62, 247e253. Paulsen, J. S., Romero, R., Chan, A., Davis, A. V., Heaton, R. K., & Jeste, D. V. (1996). Impairment of the semantic network in schizophrenia. Psychiatry Research, 63, 109e121. Pinninti, N., Steer, R. A., Rissmiller, D. J., Nelson, S., & Beck, A. T. (2002). Use of the Beck scale for suicide ideation with psychiatric inpatients diagnosed with schizophrenia, schizoaffective, or bipolar disorders. Behaviour Research and Therapy, 40, 1071e1079. Power, P. J. R., Bell, R. J., Mills, R., Herrman-Doig, T., Davern, M., Henry, L., et al. (2003). Suicide prevention in first episode psychosis: the development of a randomised controlled trial of cognitive therapy for acutely suicidal patients with early psychosis. Australian and New Zealand Journal of Psychiatry, 37, 414e420. Prescott, T. J., Newton, L. D., Mir, N. U., Woodruff, P. W., & Parks, R. W. (2006). A new dissimilarity measure for finding semantic structure in category fluency data with implications for understanding memory organization in schizophrenia. Neuropsychology, 20, 685e699.

Roy, A., & Draper, R. (1995). Suicide among psychiatric hospital in-patients. Psychological Medicine, 25, 199e202. Rudd, M. D., Joiner, T. E., & Rajab, M. H. (2001). Treating suicidal behavior: An effective, time-limited approach. New York: Guilford. Schvaneveldt, R. W. (Ed.). (1990). Pathfinder associative networks: Studies in knowledge organization. Norwood, NJ: Ablex. Scott, J., House, R., Yates, M., & Harrington, J. (1997). Individual risk factors for early repetition of deliberate self-harm. British Journal of Medical Psychology, 70, 387e393. Siris, S. G. (2001). Suicide and schizophrenia. Journal of Psychopharmacology, 15, 127e135. Smith, B., Fowler, D. G., Freeman, D., Bebbington, P., Bashforth, H., Garety, P. I., et al. (2006). Emotion and psychosis: links between depression, self-esteem, negative schematic beliefs and delusions and hallucinations. Schizophrenia Research, 86, 181e188. Startup, H., Freeman, D., & Garety, P. A. (2007). Persecutory delusions and catastrophic worry in psychosis: developing the understanding of delusion distress and persistence. Behaviour Research and Therapy, 45, 523e537. Tarrier, N., Barrowclough, C., Andrews, B., & Gregg, L. (2004). Suicide risk in recent onset schizophrenia: the influence of clinical. Social, self-esteem and demographic factors. Social Psychiatry and Psychiatric Epidemiology, 39, 927e937. Tarrier, N., Gooding, P., Gregg, L., Johnson, J., & Drake, R. (2007). Suicide schema in schizophrenia: the effect of emotional reactivity, negative symptoms and schema elaboration. Behaviour Research and Therapy, 45, 2090e2097. Tarrier, N., Haddock, G., Lewis, S., Drake, R., & Gregg, L. (2006). Suicide behaviour over 18 months in recent onset schizophrenic patients: the effects of CBT. Schizophrenia Research, 83, 15e27. Tarrier, N., Taylor, K., & Gooding, P. (2008). Cognitive-behavioural interventions to reduce suicide behaviour: a systematic review and meta-analysis. Behavior Modification, 32, 77e108. Tatarelli, R., Pompili, M., & Girardi, P. (2006). Suicide in schizophrenia. New York: Nova Science. Taylor, P. J., Wood, A. M., Gooding, P., Johnson, J., & Tarrier, N. (2009). Are defeat and entrapment best defined as a single construct? Personality and Individual Differences, 47, 795e797. Teasdale, J. D. (1988). Cognitive vulnerability to persistent depression. Cognition and Emotion, 2, 247e274. Williams, J. M. G. (1997). The cry of pain. London: Penguin. Williams, J. M. G., Crane, C., Barnhofer, T., & Duggan, D. (2005). Psychology and suicidal behaviour: elaborating the entrapment model. In K. Hawton (Ed.), Prevention and treatment of suicidal behaviour: From science to practice (pp. 71e89). Oxford: Oxford University Press.