**3.2 Correlates of PTSD after stroke**

Eight studies examined associations between the severity of PTSD symptoms and potential risk factors. In addition, one study also compared stroke survivors with and without PTSD (Sembi et al., 1998), reporting that those with PTSD had higher levels of neuroticism, anxiety and depression and lower levels of psychological well-being. However, the PTSD group was very small (*n* = 6). A range of significant correlates of PTSD symptom severity has been reported, including demographic variables such as age (Sampson et al., 2003; Sharkey, 2007) and gender (Bruggimann et al., 2006), stroke details including the number of previous strokes (Merriman et al., 2007), time since stroke (Merriman et al., 2007) and post-stroke disability (Wang et al., 2011), personality variables such as neuroticism (Sembi et al., 1998), negative affect (Merriman et al., 2007), emotionalism (Eccles et al., 1999) and alexthymia

Posttraumatic Stress Disorder after Stroke: A Review of Quantitative Studies 251

Table 1. Summary of Studies Examining the Prevalence and Correlates of PTSD after Stroke

(Wang et al., 2011), psychological distress including psychiatric morbidity (Wang et al., 2011) as well as anxiety and depression (Bruggimann et al., 2006; Field et al., 2008; Merriman et al., 2007; Sembi et al., 1998), and cognitive appraisals about the stroke (Bruggimann et al., 2006; Field et al., 2008; Merriman et al., 2007; Sharkey, 2007). A number of additional risk factors have been found to have non-significant associations with PTSD symptom severity, including neurological impairment (Bruggimann et al., 2006), lesion site/hemisphere (Bruggimann et al., 2006, Merriman et al., 2007), memory of stroke (Bruggimann et al., 2006), dissociation (Merriman et al., 2007), and consciousness (Field et al., 2008). The strongest and most consistent correlates of PTSD symptom severity have been anxiety, depression and negative cognitive appraisals about the stroke. Considering each correlate in turn, the significant correlations between generalized anxiety and PTSD symptom severity are not unexpected given that PTSD is an anxiety disorder and that several symptoms overlap in the diagnostic criteria for the two disorders. The significant correlations with depression suggest that there might be high levels of psychological co-morbidity following stroke, with many stroke survivors experiencing both mood and anxiety disorders. For example, comorbidity between anxiety and depression after stroke has been reported to be in the range of 11-18% (Astrom, 1996; Barker-Collo, 2007; Leppavuori et al., 2003; Sagen et al., 2009). In relation to co-morbidity with PTSD, Wang et al. (2011) reported that 93% of their sample of stroke survivors with PTSD scored above the GHQ-28 cut-off for psychiatric morbidity at one month post-stroke, although this figure fell to 50% at three months post-stroke. More generally, large community surveys have revealed that 80-85% of people diagnosed with PTSD also meet the diagnostic criteria for at least one other psychiatric condition (Brady, Killeen, Brewerton, & Lucerini, 2000; Creamer, Burgess, & McFarlane, 2001). While these figures highlight the breadth of psychopathology that may develop following trauma exposure, they may also reflect the lack of specificity of the current PTSD diagnostic criteria (Spitzer, First, & Wakefield, 2007). The significant correlations between negative cognitive appraisals about the stroke and PTSD symptom severity are consistent with psychological models of PTSD (e.g., Ehlers & Clark, 2000; Foa & Rothbaum, 1998) that emphasize that the way in which the trauma is interpreted and processed is important in the development and persistence of PTSD. However, closer inspection of the items used in some of the studies reveals that they may be confounded with Criterion A2 of the DSM-IV (APA, 1994) which states that an event must evoke feelings of intense fear, helplessness or horror to qualify as a traumatic event. Thus, items assessing feelings of hopelessness and helplessness (Bruggimann et al., 2006), fear (Merrimann et al., 2007) and horror (Sharkey, 2007) have been related to the severity of PTSD symptoms after stroke. In contrast, Field et al. (2008) focused on negative cognitions about the self and the world, that do not exhibit this overlap with Criterion A2.

#### **3.3 Methodological issues**

It is difficult to draw strong conclusions regarding the prevalence and correlates of PTSD after stroke because the majority of studies suffer from a number of important methodological limitations. These include a reliance on self-report measures of PTSD, small sample sizes, a preponderance of cross-sectional designs, a lack of representative samples, the assessment of a limited set of potential risk factors and a failure to fully consider the impact of specific features of stroke, medical events and older adults on PTSD symptomatology.


Table 1. Summary of Studies Examining the Prevalence and Correlates of PTSD after Stroke

(Wang et al., 2011), psychological distress including psychiatric morbidity (Wang et al., 2011) as well as anxiety and depression (Bruggimann et al., 2006; Field et al., 2008; Merriman et al., 2007; Sembi et al., 1998), and cognitive appraisals about the stroke (Bruggimann et al., 2006; Field et al., 2008; Merriman et al., 2007; Sharkey, 2007). A number of additional risk factors have been found to have non-significant associations with PTSD symptom severity, including neurological impairment (Bruggimann et al., 2006), lesion site/hemisphere (Bruggimann et al., 2006, Merriman et al., 2007), memory of stroke (Bruggimann et al., 2006), dissociation (Merriman et al., 2007), and consciousness (Field et al., 2008). The strongest and most consistent correlates of PTSD symptom severity have been anxiety, depression and negative cognitive appraisals about the stroke. Considering each correlate in turn, the significant correlations between generalized anxiety and PTSD symptom severity are not unexpected given that PTSD is an anxiety disorder and that several symptoms overlap in the diagnostic criteria for the two disorders. The significant correlations with depression suggest that there might be high levels of psychological co-morbidity following stroke, with many stroke survivors experiencing both mood and anxiety disorders. For example, comorbidity between anxiety and depression after stroke has been reported to be in the range of 11-18% (Astrom, 1996; Barker-Collo, 2007; Leppavuori et al., 2003; Sagen et al., 2009). In relation to co-morbidity with PTSD, Wang et al. (2011) reported that 93% of their sample of stroke survivors with PTSD scored above the GHQ-28 cut-off for psychiatric morbidity at one month post-stroke, although this figure fell to 50% at three months post-stroke. More generally, large community surveys have revealed that 80-85% of people diagnosed with PTSD also meet the diagnostic criteria for at least one other psychiatric condition (Brady, Killeen, Brewerton, & Lucerini, 2000; Creamer, Burgess, & McFarlane, 2001). While these figures highlight the breadth of psychopathology that may develop following trauma exposure, they may also reflect the lack of specificity of the current PTSD diagnostic criteria (Spitzer, First, & Wakefield, 2007). The significant correlations between negative cognitive appraisals about the stroke and PTSD symptom severity are consistent with psychological models of PTSD (e.g., Ehlers & Clark, 2000; Foa & Rothbaum, 1998) that emphasize that the way in which the trauma is interpreted and processed is important in the development and persistence of PTSD. However, closer inspection of the items used in some of the studies reveals that they may be confounded with Criterion A2 of the DSM-IV (APA, 1994) which states that an event must evoke feelings of intense fear, helplessness or horror to qualify as a traumatic event. Thus, items assessing feelings of hopelessness and helplessness (Bruggimann et al., 2006), fear (Merrimann et al., 2007) and horror (Sharkey, 2007) have been related to the severity of PTSD symptoms after stroke. In contrast, Field et al. (2008) focused on negative cognitions about the self and the world, that do not exhibit this overlap with Criterion A2.

It is difficult to draw strong conclusions regarding the prevalence and correlates of PTSD after stroke because the majority of studies suffer from a number of important methodological limitations. These include a reliance on self-report measures of PTSD, small sample sizes, a preponderance of cross-sectional designs, a lack of representative samples, the assessment of a limited set of potential risk factors and a failure to fully consider the impact of specific features of stroke, medical events and older adults on PTSD

**3.3 Methodological issues** 

symptomatology.


Table 1. (Continued)

Posttraumatic Stress Disorder after Stroke: A Review of Quantitative Studies 253

Table 1. (Continued)


Table 1. (Continued)

252 Post Traumatic Stress Disorders in a Global Context

Table 1. (Continued)


Table 1. (Continued)

Posttraumatic Stress Disorder after Stroke: A Review of Quantitative Studies 255

Most studies have used self-report measures to assess the prevalence of PTSD, including the Impact of Events Scale (IES; Horowitz et al., 1979), the Penn Inventory of PTSD (Penn; Hammarberg, 1992), the Post Traumatic Stress Disorder Checklist (PCL-S; Weathers et al., 1993) and the Posttraumatic Diagnostic Scale (PDS; Foa et al., 1997). A major limitation of most of these measures is that they only assess the severity of PTSD symptoms (Criteria B, C and D) and fail to consider the length of time symptoms have been present (Criterion E), the impact of symptoms on daily functioning (Criterion F), or powerful emotional reactions (Criterion A2). Indeed, one of the most frequently used measures, the IES (Horowitz et al., 1979), only assesses two symptom clusters: intrusive thoughts and avoidance. Such measures are therefore likely to over-estimate the prevalence of PTSD as they do not assess

Clinical diagnostic interviews were used in three studies to provide a diagnosis of PTSD. The Clinician-Administered PTSD Scale (CAPS; Blake et al., 1992) was used in two studies in conjunction with self-report measures. Thus, Sembi et al. (1998) first screened stroke survivors using the IES and Penn self-report scales; those scoring above cut-off points on both measures were then interviewed using the CAPS. Similarly, Sharkey (2007) used a "multi-modal" assessment of PTSD using the CAPS in conjunction with the IES and Penn. The Structured Clinical Interview for DSM-IV (SCID; First et al., 1995) was used in one study (Sagen et al., 2009, 2010) to provide a PTSD diagnosis. There was some evidence that studies employing diagnostic interviews to assess PTSD reported lower prevalence estimates than those using self-report PTSD measures. The prevalence rates reported in studies that included a clinical diagnostic interview ranged from 3% (Sagen et al., 2010) to 10% (Sembi et al., 1998), whereas the frequencies reported in studies only employing selfreport measures ranged from 6% (Sampson et al., 2003) to 31% (Bruggiman et al., 2006).

The samples sizes for studies included in the review were small, ranging from 34 (Sharkley, 2007) to 104 (Sagen et al., 2009, 2010). This has important consequences for research on the prevalence of PTSD after stroke. Small sample sizes are likely to lead to large confidence intervals and unreliable prevalence estimates. In addition, they increase the probability that outliers may have a disproportionate impact on prevalence rates (O'Donnell et al., 2003). For example, with a sample size of 50, each additional PTSD diagnosis increases the estimated prevalence rate by 2%. This issue is likely to be exacerbated when cut-off scores on self-report measures, rather than clinical interviews, are used to provide a PTSD diagnosis. Small sample sizes also impact on research on the correlates of PTSD after stroke. In particular, they are likely to lead to many analyses being under-powered, thereby increasing the probability of Type II errors (Cohen, 1992). In addition, outliers may have a disproportionate influence on

One of the most limiting aspects of research on the prevalence of PTSD after stroke is the lack of longitudinal studies. To date, all bar one study (Wang et al., 2011) have employed cross-sectional designs. Moreover, the time since stroke at which PTSD was assessed varied considerably between studies. For example, the average time since stroke ranged from 43.5 days (Sampson et al., 2003) to 62 weeks (Sharkey, 2007). Moreover, there was also considerable variability within studies. For example, the standard deviation for time since

all the DSM-IV criteria (A-F) for a PTSD diagnosis (APA, 1994).

the strength of correlations, leading to potentially spurious findings.

**3.3.1 Assessment of PTSD** 

**3.3.2 Sample sizes** 

**3.3.3 Study design** 

#### **3.3.1 Assessment of PTSD**

254 Post Traumatic Stress Disorders in a Global Context

Table 1. (Continued)

Most studies have used self-report measures to assess the prevalence of PTSD, including the Impact of Events Scale (IES; Horowitz et al., 1979), the Penn Inventory of PTSD (Penn; Hammarberg, 1992), the Post Traumatic Stress Disorder Checklist (PCL-S; Weathers et al., 1993) and the Posttraumatic Diagnostic Scale (PDS; Foa et al., 1997). A major limitation of most of these measures is that they only assess the severity of PTSD symptoms (Criteria B, C and D) and fail to consider the length of time symptoms have been present (Criterion E), the impact of symptoms on daily functioning (Criterion F), or powerful emotional reactions (Criterion A2). Indeed, one of the most frequently used measures, the IES (Horowitz et al., 1979), only assesses two symptom clusters: intrusive thoughts and avoidance. Such measures are therefore likely to over-estimate the prevalence of PTSD as they do not assess all the DSM-IV criteria (A-F) for a PTSD diagnosis (APA, 1994).

Clinical diagnostic interviews were used in three studies to provide a diagnosis of PTSD. The Clinician-Administered PTSD Scale (CAPS; Blake et al., 1992) was used in two studies in conjunction with self-report measures. Thus, Sembi et al. (1998) first screened stroke survivors using the IES and Penn self-report scales; those scoring above cut-off points on both measures were then interviewed using the CAPS. Similarly, Sharkey (2007) used a "multi-modal" assessment of PTSD using the CAPS in conjunction with the IES and Penn. The Structured Clinical Interview for DSM-IV (SCID; First et al., 1995) was used in one study (Sagen et al., 2009, 2010) to provide a PTSD diagnosis. There was some evidence that studies employing diagnostic interviews to assess PTSD reported lower prevalence estimates than those using self-report PTSD measures. The prevalence rates reported in studies that included a clinical diagnostic interview ranged from 3% (Sagen et al., 2010) to 10% (Sembi et al., 1998), whereas the frequencies reported in studies only employing selfreport measures ranged from 6% (Sampson et al., 2003) to 31% (Bruggiman et al., 2006).

### **3.3.2 Sample sizes**

The samples sizes for studies included in the review were small, ranging from 34 (Sharkley, 2007) to 104 (Sagen et al., 2009, 2010). This has important consequences for research on the prevalence of PTSD after stroke. Small sample sizes are likely to lead to large confidence intervals and unreliable prevalence estimates. In addition, they increase the probability that outliers may have a disproportionate impact on prevalence rates (O'Donnell et al., 2003). For example, with a sample size of 50, each additional PTSD diagnosis increases the estimated prevalence rate by 2%. This issue is likely to be exacerbated when cut-off scores on self-report measures, rather than clinical interviews, are used to provide a PTSD diagnosis. Small sample sizes also impact on research on the correlates of PTSD after stroke. In particular, they are likely to lead to many analyses being under-powered, thereby increasing the probability of Type II errors (Cohen, 1992). In addition, outliers may have a disproportionate influence on the strength of correlations, leading to potentially spurious findings.

#### **3.3.3 Study design**

One of the most limiting aspects of research on the prevalence of PTSD after stroke is the lack of longitudinal studies. To date, all bar one study (Wang et al., 2011) have employed cross-sectional designs. Moreover, the time since stroke at which PTSD was assessed varied considerably between studies. For example, the average time since stroke ranged from 43.5 days (Sampson et al., 2003) to 62 weeks (Sharkey, 2007). Moreover, there was also considerable variability within studies. For example, the standard deviation for time since

Posttraumatic Stress Disorder after Stroke: A Review of Quantitative Studies 257

smaller correlations with PTSD symptomatology than more proximal psychological factors (e.g., perceived life threat, dissociation). Such findings are in line with current psychological models of PTSD that emphasize the importance of appraisal and memory processes in the development of PTSD (Brewin & Holmes, 2003). Few studies on PTSD after stroke have drawn on such models to guide the selection of independent variables. Ehlers and Clark's (2000) cognitive model, which according to Brewin and Holmes (2003) provides the most detailed account of PTSD, proposes that PTSD is likely to develop and persist when the trauma and/or its sequelae is processed in such a way that leads to a sense of serious current threat, as a result of (i) making excessively negative appraisals and (ii) disturbances

Considering negative appraisals, only one study has tested the Ehlers and Clark (2000) model in relation to stroke. Field et al. (2008) reported that negative cognitions about the self (e.g., "I am inadequate") and about the world (e.g., "The world is a dangerous place"), assessed shortly after the stroke (*M* = 20 days), were significantly correlated with the severity of PTSD symptoms both cross-sectionally and prospectively three months later. However, the prospective correlations became non-significant after controlling for the effect of initial PTSD symptoms. Considering disturbances in autobiographical memory, Ehlers and Clark (2000) propose that the overwhelming experience of a traumatic event may disrupt peritraumatic cognitive processing resulting in trauma memories that are disorganised and poorly elaborated. This, in turn, may make trauma memories more vulnerable to triggering by matching cues, thereby increasing the frequency of reexperiencing symptoms. Three aspects of cognitive processing during the trauma have been related to poorly elaborated/organised trauma memories and subsequent PTSD (Halligan et al., 2003); namely, (i) engaging in surface level, or data-driven processing (e.g., "It was just like a dream of unconnected impressions following each other"), (ii) a lack selfreferential processing (e.g., "I felt as if it was happening to someone else"), and (iii) dissociation (e.g., reduced awareness of the self, time and/or environment at the time of the trauma). Halligan et al. (2003) reported that measures of these memory processes, assessed within three months after assault, were predictive of the severity of PTSD symptoms at three and six months follow-up. Only two studies have examined memory variables after stroke. Bruggiman et al. (2006) reported no differences in the symptom severity scores of survivors with fragmented versus complete memories of their stroke, whereas Merriman et al. (2007) reported that peritraumatic dissociation was related to the number, but not the severity, of

A related strand of work has noted that people with PTSD have difficulty recalling specific autobiographical memories (e.g., "When I watched the football on the television last Sunday") in response to cue words (e.g., "happy"). Instead, they recall abstract or more general memories that cover several different events or time points (e.g., "When I watch football on the television"). This tendency has been termed overgeneral memory bias (Williams & Broadbent, 1986). An inability to retrieve specific autobiographical memories may prevent the trauma memory from being integrated with other autobiographical memories and with the person's schemas about the self and the world (Kleim & Ehlers, 2008), thereby contributing to the development of PTSD. A number of studies have reported that PTSD is associated with overgeneral memory bias (Bryant et al., 2008; Schönfeld & Ehlers, 2006; Schönfeld et al., 2007). Dalgleish et al. (2008) have presented evidence to suggest that people with PTSD may avoid retrieving specific personal information as a means of affect regulation, in support of a functional avoidance account of overgeneral

in autobiographical memory.

PTSD symptoms after stroke.

stroke reported by Sharkey (2007) was 26.3 weeks, and the time since stroke reported by Merriman et al. (2007) ranged from 27 to 365 days. Only three studies assessed PTSD at fixed time points post-stroke. Sagen et al. (2009, 2010) assessed PTSD at four months post-stroke, Bruggimann et al. (2006) assessed PTSD at one year post-stroke, and Wang et al. (2011) assessed PTSD at approximately one and three months post-stroke, although the mean time since stroke at time 1 was 47.1 days with a standard deviation of 26.0 days. It is therefore difficult to compare reported prevalence rates or to provide an accurate point prevalence of PTSD. As a result, there are no accurate data on the natural course of post-stroke PTSD over time. The preponderance of cross-sectional designs also limits conclusions regarding the direction of relationships between risk factors and PTSD after stroke. For example, the experience of PTSD may result in excessively negative appraisals about the stroke, rather than negative appraisals determining PTSD as proposed in cognitive models of PTSD (Brewin & Holmes, 2003). As result, prospective designs in which potential risk factors are assessed shortly after stroke and related to the development of PTSD at a later time point are essential. To date, only two studies have employed such a design (Field et al., 2008; Wang et al., 2011). Moreover, only three studies (Field et al., 2008; Merrimann et al., 2007; Wang et al., 2011) have employed multivariate analyses to examine associations between potential risk factors and PTSD symptom severity; all other studies only examined bivariate associations.

#### **3.3.4 Sample representativeness**

The number of stroke survivors excluded from participating in the studies was either not reported (Field et al., 2008; Sembi et al., 1998; Sharkey, 2007) or was considerable, ranging from 26% (Bruggimann et al., 2006) to 63% (Eccles et al., 1999). The main exclusion criteria included cognitive impairment, communication difficulties (e.g., aphasia), and poor physical health. Stroke survivors were usually recruited from inpatient stroke wards (Field et al., 2008; Merriman et al., 2007; Sagen et al., 2009, 2010; Sembi et al., 1998), stroke rehabilitation units (Wang et al., 2011) or general hospital wards (Eccles et al., 1999; Sampson et al., 2003). A couple studies also recruited participants using postal questionnaires (Bruggimann et al., 2006; Merriman et al., 2007). Response rates (after exclusions) among studies recruiting from inpatient wards were generally high, ranging from 76% (Sampson et al., 2003) to 95% (Merriman et al., 2007). Studies employing postal questionnaires obtained lower response rates (52%) (Bruggimann et al., 2006; Merriman et al., 2007).

High exclusion rates raise questions regarding the representativeness of the samples recruited as stroke survivors who experienced more severe strokes are likely to have been excluded. None of the studies reported whether the samples were representative of the populations from which they were drawn. As a result, generalisability is limited. Moreover, it is possible that these exclusion criteria may themselves be risk factors for the development of PTSD. If so, this would imply that most studies have underestimated the prevalence of PTSD after stroke.

#### **3.3.5 Assessment of risk factors**

Studies on the correlates of PTSD after stroke have focused on a limited set of variables and have not assessed variables that have been identified as having strong associations with PTSD symptomatology in response to other traumas (Brewin et al., 2000; Ozer et al., 2003). For example, Ozer et al. (2003) conducted a meta-analysis of seven potential predictors of PTSD diagnosis or symptoms and found that more distal characteristics related to the individual or their life history (e.g., prior trauma, family history of psychopathology) had

stroke reported by Sharkey (2007) was 26.3 weeks, and the time since stroke reported by Merriman et al. (2007) ranged from 27 to 365 days. Only three studies assessed PTSD at fixed time points post-stroke. Sagen et al. (2009, 2010) assessed PTSD at four months post-stroke, Bruggimann et al. (2006) assessed PTSD at one year post-stroke, and Wang et al. (2011) assessed PTSD at approximately one and three months post-stroke, although the mean time since stroke at time 1 was 47.1 days with a standard deviation of 26.0 days. It is therefore difficult to compare reported prevalence rates or to provide an accurate point prevalence of PTSD. As a result, there are no accurate data on the natural course of post-stroke PTSD over time. The preponderance of cross-sectional designs also limits conclusions regarding the direction of relationships between risk factors and PTSD after stroke. For example, the experience of PTSD may result in excessively negative appraisals about the stroke, rather than negative appraisals determining PTSD as proposed in cognitive models of PTSD (Brewin & Holmes, 2003). As result, prospective designs in which potential risk factors are assessed shortly after stroke and related to the development of PTSD at a later time point are essential. To date, only two studies have employed such a design (Field et al., 2008; Wang et al., 2011). Moreover, only three studies (Field et al., 2008; Merrimann et al., 2007; Wang et al., 2011) have employed multivariate analyses to examine associations between potential risk factors and PTSD symptom severity; all other studies only examined bivariate associations.

The number of stroke survivors excluded from participating in the studies was either not reported (Field et al., 2008; Sembi et al., 1998; Sharkey, 2007) or was considerable, ranging from 26% (Bruggimann et al., 2006) to 63% (Eccles et al., 1999). The main exclusion criteria included cognitive impairment, communication difficulties (e.g., aphasia), and poor physical health. Stroke survivors were usually recruited from inpatient stroke wards (Field et al., 2008; Merriman et al., 2007; Sagen et al., 2009, 2010; Sembi et al., 1998), stroke rehabilitation units (Wang et al., 2011) or general hospital wards (Eccles et al., 1999; Sampson et al., 2003). A couple studies also recruited participants using postal questionnaires (Bruggimann et al., 2006; Merriman et al., 2007). Response rates (after exclusions) among studies recruiting from inpatient wards were generally high, ranging from 76% (Sampson et al., 2003) to 95% (Merriman et al., 2007). Studies employing postal questionnaires obtained lower response

High exclusion rates raise questions regarding the representativeness of the samples recruited as stroke survivors who experienced more severe strokes are likely to have been excluded. None of the studies reported whether the samples were representative of the populations from which they were drawn. As a result, generalisability is limited. Moreover, it is possible that these exclusion criteria may themselves be risk factors for the development of PTSD. If so, this would imply that most studies have underestimated the prevalence of

Studies on the correlates of PTSD after stroke have focused on a limited set of variables and have not assessed variables that have been identified as having strong associations with PTSD symptomatology in response to other traumas (Brewin et al., 2000; Ozer et al., 2003). For example, Ozer et al. (2003) conducted a meta-analysis of seven potential predictors of PTSD diagnosis or symptoms and found that more distal characteristics related to the individual or their life history (e.g., prior trauma, family history of psychopathology) had

**3.3.4 Sample representativeness** 

PTSD after stroke.

**3.3.5 Assessment of risk factors** 

rates (52%) (Bruggimann et al., 2006; Merriman et al., 2007).

smaller correlations with PTSD symptomatology than more proximal psychological factors (e.g., perceived life threat, dissociation). Such findings are in line with current psychological models of PTSD that emphasize the importance of appraisal and memory processes in the development of PTSD (Brewin & Holmes, 2003). Few studies on PTSD after stroke have drawn on such models to guide the selection of independent variables. Ehlers and Clark's (2000) cognitive model, which according to Brewin and Holmes (2003) provides the most detailed account of PTSD, proposes that PTSD is likely to develop and persist when the trauma and/or its sequelae is processed in such a way that leads to a sense of serious current threat, as a result of (i) making excessively negative appraisals and (ii) disturbances in autobiographical memory.

Considering negative appraisals, only one study has tested the Ehlers and Clark (2000) model in relation to stroke. Field et al. (2008) reported that negative cognitions about the self (e.g., "I am inadequate") and about the world (e.g., "The world is a dangerous place"), assessed shortly after the stroke (*M* = 20 days), were significantly correlated with the severity of PTSD symptoms both cross-sectionally and prospectively three months later. However, the prospective correlations became non-significant after controlling for the effect of initial PTSD symptoms. Considering disturbances in autobiographical memory, Ehlers and Clark (2000) propose that the overwhelming experience of a traumatic event may disrupt peritraumatic cognitive processing resulting in trauma memories that are disorganised and poorly elaborated. This, in turn, may make trauma memories more vulnerable to triggering by matching cues, thereby increasing the frequency of reexperiencing symptoms. Three aspects of cognitive processing during the trauma have been related to poorly elaborated/organised trauma memories and subsequent PTSD (Halligan et al., 2003); namely, (i) engaging in surface level, or data-driven processing (e.g., "It was just like a dream of unconnected impressions following each other"), (ii) a lack selfreferential processing (e.g., "I felt as if it was happening to someone else"), and (iii) dissociation (e.g., reduced awareness of the self, time and/or environment at the time of the trauma). Halligan et al. (2003) reported that measures of these memory processes, assessed within three months after assault, were predictive of the severity of PTSD symptoms at three and six months follow-up. Only two studies have examined memory variables after stroke. Bruggiman et al. (2006) reported no differences in the symptom severity scores of survivors with fragmented versus complete memories of their stroke, whereas Merriman et al. (2007) reported that peritraumatic dissociation was related to the number, but not the severity, of PTSD symptoms after stroke.

A related strand of work has noted that people with PTSD have difficulty recalling specific autobiographical memories (e.g., "When I watched the football on the television last Sunday") in response to cue words (e.g., "happy"). Instead, they recall abstract or more general memories that cover several different events or time points (e.g., "When I watch football on the television"). This tendency has been termed overgeneral memory bias (Williams & Broadbent, 1986). An inability to retrieve specific autobiographical memories may prevent the trauma memory from being integrated with other autobiographical memories and with the person's schemas about the self and the world (Kleim & Ehlers, 2008), thereby contributing to the development of PTSD. A number of studies have reported that PTSD is associated with overgeneral memory bias (Bryant et al., 2008; Schönfeld & Ehlers, 2006; Schönfeld et al., 2007). Dalgleish et al. (2008) have presented evidence to suggest that people with PTSD may avoid retrieving specific personal information as a means of affect regulation, in support of a functional avoidance account of overgeneral

Posttraumatic Stress Disorder after Stroke: A Review of Quantitative Studies 259

symptoms may develop in relation to those aspects of the trauma experience that individuals are able to encode (Creamer et al., 2005). Second, individuals may retrospectively reconstruct memories of the trauma experience, for example from witnesses' reports, which subsequently develop into intrusive memories or flashbacks (Bryant et al., 2009). Third, processing of the trauma experience may occur at an implicit level during

Stroke is a leading cause of severe disability. Survivors typically experience a range of ongoing problems (e.g., weakness or paralysis, cognitive impairment, communication difficulties, problems with balance and coordination). One important question is the extent to which PTSD symptom severity reflects the impact of these ongoing stressors, rather than reactions to the stroke event itself. For example, in relation to MI, Shemesh et al. (2001) found that patients who experienced ongoing physical symptoms (e.g., angina) reported more intrusion and avoidance PTSD symptoms than those who were asymptomatic. In addition, in relation to stroke, Wang et al. (2011) reported that the level of physical disability was related to the severity of PTSD symptoms at three months post-stroke. There are a number of ways in which chronic stressors may contribute to the severity of PTSD symptoms. First, chronic stressors may erode individuals' resources, or their ability, to deal with their psychological reactions to the acute stressor (Adams & Boscarino, 2006). Second, chronic stressors may evoke reminders of the stroke which may, more directly, act as triggering cues for the reexperiencing symptoms of PTSD. Third, the experience of ongoing disability may be perceived by the individual to signify permanent negative change. Fourth, some disabilities experienced after stroke, including cognitive and language impairments, may impede the person's ability to fully process and integrate trauma memories with other autobiographical material. For example, cognitive impairment has been related to difficulties in recalling specific autobiographical memories in the elderly (Phillips & Williams, 1997) and in stroke survivors (Sampson et al., 2003). Fure et al. (2006) reported that cognitive impairment was related to elevated levels of anxiety in stroke patients; however, to date, no studies have examined the relationship between cognitive impairment and PTSD. In addition, Thomas and Lincoln (2008) reported that stroke survivors with aphasia had higher levels of emotional distress at one and six months post-stroke, with more detailed analyses revealing that this was the result of expressive, but not receptive, communication impairments. It is possible that stroke survivors with aphasia may also

The majority of stroke survivors are older adults, with almost 80% of first-ever strokes occurring in people aged 65 years or older (Stroke Association, 2011). Knowledge regarding the prevalence and determinants of PTSD as well as its phenomenology in older adults, more generally, is limited (Averill & Beck, 2000; Cook & O'Donnell, 2005). Moreover, the majority of previous research on PSTD in older adults has focused on holocaust survivors, combat veterans and survivors of natural disasters, rather than on survivors of lifethreatening illnesses such as stroke (Cook & O'Donnell, 2005). Some studies have highlighted differences between younger and older adults in the experience and/or reporting of PTSD symptoms (Acierno et al., 2002; Davidson et al., 1990; Fontana & Rosenheck, 1994), although other work has suggested that their PTSD reactions are quite

periods of impaired consciousness (Bryant, 2001).

**3.3.7 Chronic stressors** 

experience more PTSD symptoms.

**3.3.8 PTSD and older adults** 

memory bias (Williams et al., 2007). Thus, intentional memory searches may be stopped prematurely at an abstract level in order to avoid retrieving potentially distressing material related to the trauma. To date, no studies have examined the relationship between overgeneral memory bias and PTSD after stroke.
