Suicide and Life-Threatening Behavior 44 (1) February 2014 © 2013 The American Association of Suicidology DOI: 10.1111/sltb.12064

101

The Temporal Association Between Self-Injurious Thoughts and Psychotic Symptoms: A Mobile Phone Assessment Study JASPER E. PALMIER-CLAUS, PHD, PETER J. TAYLOR, PHD, DCLINPSYCH, JOHN AINSWORTH, MSC, MATTHEW MACHIN, MIET, GRAHAM DUNN, PHD, AND SHON W. LEWIS, MD, FMEDSCI

The relationship between psychotic symptoms and self-injurious thoughts (SITs) remains unclear. The short-term temporal associations between psychotic symptoms and SITs were explored. A sample of 36 people with a diagnosis of a psychotic disorder or at-risk mental state completed mobile phone-based measures at multiple times each day for 1 week. Clustered regression with timelagged variables supported a relationship between paranoia and subsequent SITs. Hallucinations did not predict these thoughts when controlling for paranoia. The role of specific psychotic symptoms in triggering SITs is highlighted and the importance of considering these factors in risk management is discussed. The high prevalence of suicide and selfinjury in schizophrenia makes them of significant clinical concern (Haw, Hawton, Sutton, Sinclair, & Deeks, 2005; Hawton, Sutton, Haw, Sinclair, & Deeks, 2005; Pierzchniak et al., 1999), particularly those in the early stages of the disorder (Harvey et al., 2008). Ideation about self-injury and suicide (i.e. thoughts, images, or intentions) is also common in psychotic populations (Fialko et al., 2006; Kim et al., 2010; Upthegrove et al., JASPER E. PALMIER-CLAUS, Division of Clinical Psychology, School of Psychological Science, the University of Manchester, Manchester, UK; PETER J. TAYLOR , Division of Clinical Psychology, Institute of Psychology, University of Liverpool, Liverpool, UK; JOHN AINSWORTH, MATTHEW MACHIN, and GRAHAM DUNN, Institute of Population Health, the University of Manchester, Manchester, UK; SHON W. LEWIS, Institute of Brain, Behaviour and Mental Health, the University of Manchester, Manchester, UK. Address correspondence to Jasper E. Palmier-Claus, 2nd Floor, Zochonis Building, Division of Clinical Psychology, University of Manchester, Manchester, M139PL; E-mail: jasper.palmier-claus@ manchester.ac.uk

2010). Moreover, there is evidence of an underlying continuum of self-injurious thought and behavior (Hargus, Hawton, & Rodham, 2009; McAuliffe, 2002) whereby self-injurious thoughts (SITs) may precede (e.g., O’Connor, Rasmussen, & Hawon, 2009) and predict (Hawton et al., 2005; Nock, Prinstein, & Sterba, 2010) subsequent self-injurious behaviors.). There is a need for a greater understanding of the factors precipitating and triggering SITs in nonaffective psychosis to design better prevention and risk management strategies. Several risk factors for suicide and self-harm in those with psychosis have been identified, including substance misuse, emotional instability, and feelings of hopelessness (Haw et al., 2005; Palmier-Claus, Taylor, Gooding, Dunn, & Lewis, 2012; Pompili, Amador, & Girardi, 2007). The case for whether positive psychotic symptoms themselves can trigger SITs is less clear. Past research has tended to explore risk factors via cross-sectional and long-term longitudinal studies, which may miss more fine-grain temporal relationships between triggers and ideation. In the current

102

PSYCHOTIC SYMPTOMS

study we aimed to extend this literature by exploring the short-term temporal relationship between psychotic symptoms and SITs. We use the term SITs to refer to thoughts of self-harm irrespective of suicidal motive (Figure 1). There are theoretical reasons to predict an association between SITs and psychotic symptoms. The presence of psychotic symptoms may be a stressful experience for many individuals diagnosed with schizophrenia, leading to aversive feelings such as worthlessness, hopelessness, shame, or entrapment, which may in turn provoke SITs (Taylor, Gooding, Wood, & Tarrier, 2011; Williams & Williams, 1997). Such aversive experiences and affective states may initially trigger SITs as an escape, coping, or experiential avoidance mechanism (Baumeister, 1990; Chapman, Gratz, & Brown, 2006). It is also plausible that SITs may themselves act as a trigger for psychotic symptoms. Psychotic symptoms may be in part driven by sensitivity to everyday irritations and hassles (Palmier-Claus, Dunn, & Lewis, 2012; Palmier-Claus, Taylor, et al.,

AND

SELF-INJURIOUS THOUGHTS

2012). SITs may represent a substantive source of life stress for some individuals, thus leading to worsening symptoms. Moreover, many individuals will find that the experience of SITs clashes with their own cultural and personal values (e.g., prohibitions against suicide; Leenaars, 2003). Such perceived transgressions and associated guilt may fuel derogatory hallucinations, because some hallucinations may take on the characteristics of critical authority figures (Gilbert, McEwan, Bellew, Mills, & Gale, 2009). It has also been suggested that paranoia serves a protective function in psychosis by allowing the externalization of thoughts discrepant with an “ideal self ” concept (Bentall, Kinderman, & Kaney, 1994). SITs may therefore come to be externalized to form fears regarding threats from others (or persecutory beliefs). Research examining the association between suicidality and psychosis has had inconsistent results. Hawton and colleagues’ (2005) seminal systematic review identified two studies showing positive and two studies showing negative associations between psychotic symptoms and suicide risk. When

Figure 1. A diagram summarizing the temporal relationship between variables. Paranoia significantly predicted increased suicidal ideation on the following day. Suicidal ideation did not predict subsequent levels of psychotic symptoms. *p < .05.

PALMIER-CLAUS

ET AL.

specific symptoms were considered, hallucinations and delusions were associated with lower, but paranoia with greater, risk of suicide. Contrary to this, a more recent review has suggested that psychotic symptoms are associated with greater suicide risk (Hor & Taylor, 2010). Similarly, while some studies have found an association between overall psychotic symptoms and suicidal ideation in those with psychosis (Fialko et al., 2006), other studies have failed to support this relationship (Kim et al., 2010). Research has suggested that improvement in positive psychotic symptoms is related to a reduction in self-harming thoughts and behaviors (Van Os et al., 1999). Despite mixed evidence, the clinical importance of a link between specific psychotic symptoms and self-harm has been emphasized by some commentators (e.g., Sandor, 2002). Disparities in findings may be related to how symptoms are assessed (e.g., symptom clusters vs. specific symptoms), the design of the research, and whether depression was controlled for in the analyses. While questionnaire studies are useful in determining trends over time, they tell you very little about the factors immediately precipitating or triggering the onset of suicidal ideation. This might best be examined through momentary assessment techniques, where phenomena are recorded several times per day when prompted by an electronic device (Csikszentmihalyi & Larson, 1987). Momentary assessment holds several advantages over traditionally used measures. For example, it (1) provides data with high ecological validity; (2) reduces the confounding effects of retrospective recall bias; and (3) allows short-term temporal associations to be assessed (Palmier-Claus et al., 2011). In this study we examined secondary analysis of data from a validation study for a mobile phone-based assessment for psychosis (Palmier-Claus, Ainsworth, et al., 2012). The aim was to explore the temporal relationship between SITs and psychotic symptoms in individuals at different stages of psychotic disorder. It was hypothesized that hallucinations and paranoia would pre-

103 dict increased subsequent levels of SITs. It was also predicted that suicidal ideation would not predict subsequent levels of hallucinations and paranoia, and would not constitute protective factors.

METHODS

Participants Three different patient groups were recruited. Group one consisted of 12 remitted patients meeting the criteria for a Diagnostic and Statistical Manual of Psychiatric Disorders (DSM-IV American Psychiatric Association, 1994) diagnosis of schizophrenia, schizoaffective disorder, delusional disorder, or schizophreniform disorder, who exhibited mild or absent psychotic symptoms, and had been stable on antipsychotic medication for at least 3 months. Diagnosis was initially made by the referring clinical service and verified against DSM-IV criteria on the client’s admission to the study by the research team. Group two consisted of 12 acute patients with these DSM-IV diagnoses, who were within 4 weeks of starting, restarting, or changing medication, or within 4 weeks of an acute inpatient admission. Group three consisted of 12 individuals who had met the criteria for being ultra-high risk (UHR) of psychosis (i.e., potentially prodromal) according to the Comprehensive Assessment of At-Risk Mental State (CAARMS; Yung et al., 2005) at some point in the past year. These clients, who did not meet the criteria for a DSM-IV diagnosis of a full psychotic disorder, were not on antipsychotic medication. This was established by the participant’s clinical service as part of their routine assessment. All participants were reassessed with the CAARMS upon entry to the study by the lead author who has considerable experience in using this interview in research and clinical settings. Fifty percent of individuals still met UHR criteria, whereas 50% were under threshold. Thus, data from 36 patients were entered into this analyses. The mean age of the sam-

104

PSYCHOTIC SYMPTOMS

ple was 31.4 (SD = 10.2), and the majority of participants were male (n = 28) and White British (n = 29). Assessment Momentary assessment scales were validated in a previous study (PalmierClaus, et al., Under review). Participants were required to indicate the degree to which they agreed or disagreed with statements relating to their symptoms since the last entry on touch-screen analogue scales. The first entry of the day related to the period of time since wakening. The presented questions depended on an individual’s previous response (the items were branched). There were two sets of items, which alternated across the day. Therefore, participants completed each set of items a maximum of three times per day. Hallucinations, depression, and suicidal ideation were measured in set one, whereas paranoia was measured in set two. Eight items measured hallucinations. Participants were asked whether they had heard voices (i.e., “I have heard voices”), which, if endorsed, they then rated for related preoccupation (i.e., “I have found it difficult to concentrate on other things”), disruption to activities (i.e., “This stopped me from doing things”), and resultant distress (i.e., “Hearing the voice upset me”). Visual hallucinations were assessed in the same manner using the items “I have seen things that other people can’t see,” “I have found it difficult to concentrate on other things” (branched), “This stopped me from doing things” (branched), and “Seeing these things upset me” (branched). Paranoia was assessed through the items “I have worried about saying too much,” “I have been suspicious,” “I have felt like someone or something meant me harm,” “This has stopped me from spending time with others” (branched), “This has stopped me from doing things” (branched), and “I have found it difficult to concentrate on other things” (branched).

AND

SELF-INJURIOUS THOUGHTS

In the original validation study, the item “I have had thoughts about harming myself” was included in the depression momentary assessment scale. For the purpose of this analysis and other papers (Palmier-Claus, Ainsworth, et al., 2013), this item was included as a separate scale to measure SITs. Logistic regression suggested that this momentary assessment item was a strong predictor of suicidality measures on the Calgary Depression Scale (Addington, Addington, & Schissel, 1990; OR = 2.40, SE = .790, p = .008), attesting to convergent validity of this item. The remaining depression momentary assessment items were retained. These were as follows: “I have felt sad,” “I have felt cheerful” (reversed), “I have felt motivated to do things” (reversed; branched), and “my mood has affected my appetite or sleep” (branched). Procedure The researcher met with the participant to obtain informed consent and demographic information and to provide training in how to operate the device. Sampling started on the morning following the briefing session. Participants were required to complete self-report questions on a native smartphone software application (i.e., ClinTouch, Manchester, UK) when prompted by an alarm sounding at six pseudo-random times of the day for 1 week, which occurred between 0900 and 2100 hours. Entries were made on a touch-screen analogue scale. Forced entry times meant that participants had to complete all entries within 15 minutes of an alarm going off, although they did have the option to activate a 5minutes reminder (a “snooze function”). Data were stored on the mobile phone handsets and uploaded onto the researcher’s computer at the end of the week. Statistics All analyses were conducted in Stata 10.0 (StataCorporation, 2007). The ClinTouch software automatically converted

PALMIER-CLAUS

ET AL.

the analogue responses into 7-point scales for the purpose of analysis. First, we calculated a mean score from across the momentary assessment item measuring each construct at each time point (e.g., item 1, item 2, etc.). The mean score of this mean was then calculated for each day of sampling. Thus, each participant has seven scores for each construct, one for every day of sampling. Time-lagged variables were created representing individuals’ symptom score on the previous day of sampling (t-1). Momentary assessment data are almost certainly nested and therefore violate the assumption of independent observations, which is necessary for parametric analyses. We therefore used regression with clustering (Rogers, 1993), where participant number was entered as the random effect. This had the effect of imposing robust standard errors on the analyses. Six models were tested. In Model 1, the covariates (depression t0, depression t-1, age, gender, and group) were entered as predictors of SITs. In Model 2, hallucinations at both t0 and t-1 were added to the covariates as predictors of SITs, therefore examining whether hallucinations predicted SITs beyond that of the covariates. Model 3 was the same as Model 2 but with the paranoia variables replacing hallucinations. To examine whether hallucinations or paranoia was a better predictor of SITs, both sets of variables were entered into the regression analysis in Model 4. The rest of the analyses explored whether SITs led to greater subsequent levels of psychotic symptoms. In Model 5, SITs at t0 and t-1 were entered as predictors of hallucinations at t0. Paranoia (t0 and t-1) and the covariates were controlled for in this analysis. Lastly, in Model 6, SITs, the covariates, and hallucinations (t0 and t1) were entered as predictors of paranoia at t0. Bootstrapping (reps 1,000) was used in all analyses to account for the inclusion of nonnormally distributed variables. This has been suggested as a suitable alternative to nonparametric analyses, when the necessary assumptions are not met (Mooney, Duval, & Duval, 1993).

105 RESULTS

Summary statistics are provided in Table 1. There were 157 occasions where a participant reported SITs (scoring ≥ 3), occurring on different 87 days; 47 assessments where an individual endorsed the maximum score (i.e., 7); and 25 of the 36 participants reported SITs (scoring ≥ 3) at least once. There were also a high number of occasions where participants experienced auditory (n = 195) or visual (n = 130) hallucinations. Hallucinations occurred on 116 different days in 24 different participants. The results of the regression analysis can be seen in Table 2. Of the covariates entered into Model 1, only concurrent levels of depression significantly predicted SITs, although there was a nonsignificant trend of group. Model 2 shows that hallucinations predicted concurrent and subsequent levels of SITs, although the time-lagged variable just failed to reach statistical significance. When paranoia was entered in the place of hallucinations (Model 3), this variable only predicted subsequent and not concurrent levels of SITs. When both paranoia and hallucination variables were entered into the same analysis (Model 4), the only significant predictor of SITs was paranoia measured on the previous day (t-1). Thus, much of the effect of hallucinations on SITs may be explained by shared variance with paranoia. SITs were then entered as a predictor of concurrent and subsequent levels of psychotic symptoms, which is shown in Table 3. SITs did not significantly predict concurrent or subsequent levels of hallucinations, nor TABLE 1

Summary Statistics Variable Depression Paranoia Hallucinations Self-injurious thoughts Age

Mean

SD

Min

Max

3.6 2.8 2.5 2.4

1.5 1.8 1.7 1.8

1.0 1.0 1.0 1.0

6.6 7.0 7.0 7.0

31.4

10.0

18.0

51.0

106

PSYCHOTIC SYMPTOMS

AND

SELF-INJURIOUS THOUGHTS

TABLE 2

Predictors of Self-injurious Thoughts (at t0) Across Days 95% Conf. interval Independent variables

Std. b

SE

p

Model 1

Depression at t0 Depression at t-1 Age Gender Remitted group (yes/no) Ultra-high-risk group (yes/no)

.25 .10 .01 .05 .29 .34

0.083 0.091 0.219 0.187 0.177 0.265

.002 .251 .976 .784 .101 .201

0.091 0.074 0.436 0.315 0.638 0.859

0.417 0.283 0.422 0.418 0.057 0.181

Model 2

Hallucinations at t0 Hallucinations at t-1

.34 .20

0.136 0.108

.013 .067

0.073 0.014

0.605 0.408

Model 3

Paranoia at t0 Paranoia at t-1

.13 .51

0.182 0.186

.462 .006

0.223 0.146

0.491 0.874

Model 4

Hallucinations at t0 Hallucinations at t-1 Paranoia at t0 Paranoia at t-1

.16 .09 .13 .45

0.150 0.144 0.189 0.209

.276 .553 .498 .030

0.130 0.368 0.243 0.043

0.457 0.197 0.499 0.864

Model

Lower

Upper

Note. t0, concurrent predictor; t-1, predictor on previous day; SE, standard error.

paranoia. Interestingly, paranoia (t-1) predicted subsequent levels of hallucinations (t0), and hallucinations (t-1) predicted subsequent levels of paranoia (t0). This analysis is summarized in Figure 1.

DISCUSSION

The data in this study partly support the authors’ hypotheses that paranoia and hallucinations would significantly predict greater levels of SITs on the following day, but that SITs would not predict subsequently higher levels of psychotic symptoms. Paranoia and hallucinations predicted subsequent levels of SITs even when controlling for gender, age, stage of disorder, and depression. However, when controlling for the effect of paranoia, the trend between hallucinations and SITs disappeared. Thus, the relationship between hallucinations and SITs may be driven by associated levels of paranoia. SITs may serve a number of functions for those experiencing paranoia. For

example, self-harm may be considered as a means of self-attack, linked to a perception of the self as inferior and vulnerable from powerful others (e.g., Gilbert et al., 2009); self-harm may be considered as a means of coping with aversive and distressing paranoid states through processes of avoidance, dissociation, or escape (e.g., Chapman et al., 2006); or self-harm may be perceived as a means of protecting oneself (e.g., by reducing anger, signaling a need for help or pushing others away; Gratz, 2003). Alternatively, the distress associated with psychotic symptoms may also fuel suicidal thinking, which may manifest in the form of SITs. Other factors such as sleep disturbance and cannabis use may be causing both the paranoia and the later SITs. Reinherz and colleagues (2006) have suggested that suicidal ideation may worsen mental health outcomes. It is therefore important to note that although SITs did not immediately exacerbate psychotic symptoms, they may lead to a worse course of disorder in the longer-term.

PALMIER-CLAUS

ET AL.

107

TABLE 3

Predictors of Symptoms Across Days 95% Conf. interval Model Model 5. Predictors of hallucinations

Model 6. Predictors of paranoia

Independent variables

Std. b

SE

p

Self-injurious thoughts at t0 Self-injurious thoughts at t1 Paranoia at t0 Paranoia at t-1 Depression at t0 Depression at t-1 Age Gender Remitted group (yes/no) Ultra-high-risk group (yes/no) Self-injurious thoughts at t0 Self-injurious thoughts at t-1 Hallucinations at t0 Hallucinations at t-1 Depression at t0 Depression at t-1 Age Gender Remitted group (yes/no) Ultra-high-risk group (yes/no)

.02 .05 .23 .52 .17 .11 .04 .02 .04 .10 .14 .04 .21 .37 .00 .16 .17 .13 .02 .01

0.087 0.080 0.165 0.143 0.089 0.095 0.229 0.110 0.146 0.207 0.079 0.065 0.093 0.085 0.081 0.079 0.178 0.104 0.119 0.197

.792 .559 .165

The temporal association between self-injurious thoughts and psychotic symptoms: a mobile phone assessment study.

The relationship between psychotic symptoms and self-injurious thoughts (SITs) remains unclear. The short-term temporal associations between psychotic...
125KB Sizes 0 Downloads 0 Views