Schizophrenia patients with and without Post-traumatic Stress Disorder (PTSD) have different mood symptom levels but same cognitive functioning


  • Please also see editorial comment to this paper by K. Mueser in this issue, Acta Psychiatr Scand 2013;127:440–441.

Dawn E. Peleikis, Department of Psychiatry, Akershus University Hospital, Alna Outpatient Clinic, 1478 Lørenskog, Norway.




To investigate differences in cognitive function and level of psychopathology in patients with schizophrenia (SZ) with or without psychological traumatization/post-traumatic stress disorder (PTSD). We hypothesized that traumatized patients with or without PTSD would have more severe cognitive impairments because of the neuropathological changes associated with PTSD, and more severe psychopathology compared with non-traumatized SZ patients.


Seventy-five SZ patients with traumatization and 217 SZ patients without traumatization were evaluated regarding the symptoms and cognitive functioning, using standard symptom scales (PANSS; CDSS) and a neuropsychological test battery (IQ, verbal memory, attention, working memory, psychomotor speed, and executive functioning).


No significant differences were observed between the groups in cognitive test performance. The patients in the traumatized group with PTSD showed significantly more current depression than the non-traumatized group (P = 0.012).


The findings did not support the hypothesis that the presence of comorbid PTSD/traumatization in SZ is associated with increased cognitive impairment. The increase in current depression in SZ with comorbid traumatization suggests that more severe psychopathology is associated with traumatization.

Significant outcomes

  • No significant differences were observed between the groups in cognitive test performance.
  • Schizophrenia patients with trauma and post-traumatic stress disorder (PTSD) had significantly more current depression than the non-traumatized group.


  • Given the nature of the cross-sectional design, causal relationships between traumatization/PTSD, schizophrenia, and functional measures cannot be drawn.
  • Some questionnaires have a low response rate Calgary Depression Scale for Schizophrenia (CDSS).
  • Heterogeneity within the trauma group with regard to the type and age when the traumatic event happened complicates the interpretation of our results.


Numerous studies have documented high rates of exposure to traumatic events in patients with schizophrenia (SZ), which often leads to high rates of comorbid post-traumatic stress disorder (PTSD) [1-3]. In contrast, not much is known about the associations between traumatization and the clinical symptoms and cognitive dysfunctions in SZ. The prevalence of psychological traumatization (PT), here, defined as in DSM-IV: ‘The person experienced, witnessed, or was confronted with an event or events that involved actual or threatened death or serious injury, or a threat to the physical integrity of self or others. The person's response involved intense fear, helplessness, or horror’. The prevalence of PT is high in the general population. American epidemiological studies show a lifetime prevalence of 40–50%. The National Comorbidity Survey (NCS) found that violence and sexual abuse in childhood and adolescence were strongly associated with mental disorders in adulthood [4, 5].

The prevalence of PT in SZ is still somewhat uncertain, although the first study to assess prevalence of PTSD in this population was written in the late 1980s [6, 7]. Another important study [8] focusing on people with the first episode of psychosis showed lifetime prevalence of PT and PTSD in a cohort of 426 patients initially hospitalized for psychosis. The prevalence of PT was 68.5%, and the prevalence of PTSD was 14.3% in the full sample and 26.5% in those with PT. The prevalence of PTSD in the sample of patients with SZ was 10%. A weighted average of data from several studies indicates a 29% prevalence of PTSD in patients with SZ [9]; however, most of these studies assessed the rates of current PTSD, not lifetime PTSD.

A substantial body of evidence has documented that patients with SZ display cognitive deficits, especially in the domains of attention, working and verbal memory, and executive function [10-13]. Similarly, research in patients with PTSD has identified the basic deficits in attention, memory, and executive function [14, 15], with some overlap with SZ patients in affected brain regions. In particular, structural and functional abnormalities in frontal and temporal regions are found in patients with SZ and in patients with PTSD [16-18].

With regard to PTSD, the hippocampus is a region of particular interest because noradrenergic activity during the stress response directly affects the hypothalamic–pituitary–adrenal (HPA) axis, disrupting the normal activities of the limbic structures [16]. Magnetic resonance imaging (MRI) studies have documented bilateral, hippocampal volume reductions in trauma survivors with PTSD [19, 20], although it is not clear whether this is the result of PTSD or a risk factor existing prior to the development of PTSD [18]. Also frontal lobe abnormalities are reported in PTSD, including hypoactivity of the mediolateral and dorsolateral prefrontal cortex [21, 22], consistent with documented deficits in attention and executive functioning in individuals with PTSD [19, 23, 24]. Building on advances in social cognitive and effective neuroscience, it has been suggested that a core set of brain regions, including cortical midline structures, amygdala, insula, posterior parietal cortex, and temporal poles, is vulnerable to psychological trauma. Patients with PTSD have exhibited behavioural and neurobiological impairments impacting emotional self-awareness, emotional regulation, social–emotional processing, and self-referential processing as compared with healthy controls [25]. Studies have also found explicit memory deficits in PTSD [26-29].The findings have been reported in various populations such as college students [30], combat veterans [23, 31, 32], Holocaust survivors [33], victims of childhood sexual abuse [34], and victims of natural disasters [35].

A trauma history contributes to increased psychiatric symptoms and poorer functional outcomes in patients with SZ compared with those without trauma [36, 37]. Similarly, a recent study [38] hypothesize that the cognitive impairment associated with traumatization/PTSD could further affect the cognitive impairment in SZ, because of a limited cognitive reserve in individuals with SZ to compensate for the HPA stress response associated with PTSD. The documented cognitive impairments may be the result of HPA abnormalities arising from an extreme stressor or reflect premorbid neurological abnormalities that predispose an individual to the development of PTSD.

Aims of the study

The aim of the study was to assess psychological traumatization/post-traumatic stress disorder (PTSD), cognitive function, and level of psychopathology in a sample of patients with schizophrenia (SZ) using a self-reported history of trauma exposure. We hypothesized that traumatized patients with or without PTSD would have more severe cognitive impairments because of the neuropathological changes associated with PTSD, and more severe psychopathology compared with non-traumatized SZ patients.

Material and methods

The study included patients from the ongoing Thematically Organized Psychosis (TOP) Study in Norway. The TOP Study is a large collaborative effort including the University of Oslo and hospitals in the Oslo region, including Oslo University Hospital, Lovisenberg and Diakonhjemmet Hospital, and two hospitals in neighboring counties of Oslo. The project is catchment-area-based, and aims at recruiting all patients within the relevant diagnostic groups (schizophrenia, bipolar disorders, and other psychosis). Recruitment started in 2002 and is still ongoing. In addition to clinical and cognitive studies, specific subprojects investigate genetic factors and brain imaging.

The patient sampling was conducted between January 2002 and September 2010 and consecutively recruited from in- and out-patient psychiatric units in the five catchment areas of Oslo City, covering inner city as well as suburban areas. The sample consisted of 292 patients (161 men and 131 women) with a schizophrenia spectrum diagnosis, in the following named schizophrenia (SZ). This includes schizophrenia (n = 225), schizophreniform disorder (n = 8), and schizoaffective disorder (n = 59).

Inclusion criteria were age 23–67 years at intake; a main diagnosis of schizophrenia, including schizophreniform disorder and schizoaffective disorder according to the DSM-IV [39]. Exclusion criteria were a history of traumatic brain injury, organic brain disorders, diagnosis of mental retardation, primary substance dependence disorder, or poor neuropsychological test effort. Inadequate test effort among participants was determined based on the Forced Choice Recognition subtest from the California Verbal Learning Test – Second Edition CVLT-II [40]. A cutoff of <15 on this subtest was used. Seven participants were excluded. To assure valid cognitive test performance and self-rating of depressive symptoms, all participants had to have Norwegian as their first language or had received their compulsory schooling in Norway. Patients with first episode psychosis are not included in the current sample.

The patients were assessed by the Structured Clinical Interview for DSM-IV (SCID-I) to confirm the diagnosis of SZ or schizoaffective disorder [41]. Lifetime trauma and PTSD were assessed with a Structured Clinical Interview M.I.N.I., version 5.0 to diagnose DSM-IV Axis I disorder for PTSD [42] This particular method does not require participants to respond with a yes or no to each traumatic event from a list, but rather asks participants to indicate whether they have experienced any events after reading an extended list of traumatic events. Also information from the patients' hospital medical records was collected by a psychiatrist or clinical psychologist.

All interviewers were trained in SCID assessment based on the programme at the University of California, Los Angeles, and had regularly meetings where diagnostics in severe mental disorders were discussed with an experienced clinical researcher in the field.

Ethics and approvals

All participants gave written informed consent. Responders participated in an altogether 3-h interview divided into three sessions. Participants were paid 89 USD for participating and were provided with test results, post-test counseling, and referrals for follow-up and treatment as needed. The study was approved by the Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate.

Symptom assessment

The Positive and Negative Syndrome Scale (PANSS) [43] assessed the severity of general, positive and negative psychotic symptoms. The PANSS is a semistructured interview using a 7-point rating scale over 30 items and is a commonly used tool in schizophrenia research. Functional level was assessed with the Global Assessment of Functioning Scale, split version-function score (GAF-F) [44], with an ICC inter-rater reliability of 0.81. Calgary Depression Scale for Schizophrenia (CDSS) [45] was used to measure the severity of depression and suicidality; CDSS is a 9-item semistructured interview designed for use for individuals with a diagnosis of schizophrenia. The scale assesses the levels of depressive symptoms independent of positive and negative symptoms of schizophrenia and any effects of medication [46]. Current depression was assessed by CDSS total score, with an internal consistency Cronbach's coefficient alpha of 0.82.

Neuropsychological assessment

All participants underwent testing with a comprehensive neuropsychological test battery, carried out by trained psychologists. The tests were selected a priori based on the domain specific cognitive impairments commonly reported to be central to SZ in general [47]. As measures of verbal memory, California Verbal Learning Test [48] and a Norwegian research version of Logical Memory from the Wechsler Memory Scale [49] were included. Attention was assessed with the Digit Span Forwards, and working memory with the Digit Span Backwards from the Wechsler Adult Intelligence Scale – Third Edition, WAIS-III [50]. Psychomotor speed was measured with the Digit Symbol from the WAIS-III. As measures of executive functioning, tests from the Norwegian version of the Delis–Kaplan Executive Function System [51] were administered. From the Letter-Colour Inhibition test, the ‘Stroop’ condition was used, that is, the third subtest where the subject is instructed to name the color of the ink of color words that are written in ink of a different color. From the Verbal Fluency test scores, Letter Fluency, Category Fluency (animals and boys' names), and Category Switching (where the subject is instructed to shift between naming as many items of fruits and furniture as possible within 60 s) were used. Different aspects of executive functioning are covered when these tests are included in the analyses: inhibition of behaviour (the ‘Stroop’ task), initiation of behaviour (Category Fluency), and set shifting (Category Switching).

Current IQ was assessed with a Norwegian research version [52] of the Wechsler Abbreviated Scale of Intelligence (WASI) [53]. Mean current IQ was 100.3 (range: 59–136; SD = 15.9) in the SZ group without PT and 97.6 (range: 63–129; SD = 17.1) in the SZ group with trauma. Premorbid IQ was estimated from the NART score [54]: mean premorbid IQ in the SZ group without PT was 109.3 (SD = 5.90) and 109.0 (SD = 7.40) in the SZ group with trauma.

Statistical analysis

The data were analyzed with PASW for PC version 18.0. The chosen outcome instruments have continuous and categorical scales, and the analyses focused on comparing average scores between the groups. When the data were normally distributed with equal variance, analyses were based on parametric tests. When comparing groups, independent sample t-tests and anova statistics were used for continuous variables, and chi-squared test was applied for categorical variables. However, when there was no normal distribution, nonparametric, alternative Mann–Whitney U-test was used. An alpha level of P < 0.05 was used throughout the study, and all tests were two-sided.


Description of sample, comparisons between groups and cognitive correlates of trauma exposure

In the sample of 292 patients with a schizophrenia spectrum diagnosis, 217 (74%) patients (124 men and 93 women) had no trauma, and 75 (25%) patients (37 men and 37 women) had experienced trauma (Table 1). Of the patients with trauma exposure, 21 (7%) patients fulfilled the criteria for the diagnosis PTSD, and the type of trauma in relation to age is further specified in Table 2. There were no significant difference in the living situation between the no trauma, trauma, and PTSD groups. No significant difference was seen in substance use 2 weeks prior to study entry between the three SZ groups. The majority of the patients with SZ were living on social welfareillness benefit. There were few patients from the sample without any psychiatric hospitalization (Table 1).

Table 1. Demographical and clinical characteristics of patients with schizophrenia: No trauma, trauma and PTSD groups
 All patients N = 292No trauma n = 218Trauma n = 54PTSD n = 21 P
  1. a

    anova, P < 0.05, PTSD > no trauma, trauma statistically sign. by post hoc test (Tukey's test). Mean and Standard Deviation (M, SD). Positive and Negative Syndrome Scale (PANSS). Global Assessment of Functioning (GAF), Symptom-(GAF-S) and Function-(GAF-F) subscale.

  2. b

    Chi-squared test, two-sided for categorical variables, anova one-way analysis for scaled variables.

  3. c

    Fisher's exact test.

  4. d

    Use of tablets, alcohol, cannabis, amphetamine, ecstasy, or cocaine during the last 2 weeks before inclusion.

  5. e

    The Calgary Depression Scale (CDSS) (item 8): Any current suicidal thoughts, plans, or attempts (score ≥1).

  6. f

    The psychopathology mean score from the three-factor model of Kay et al. [43] (Item G1–G16).

  7. g

    Dysphoric mood factor from five-factor PANSS model of White et al. [56] (Item G1, G2, G3, G4, and G6).

  8. h

    Intelligence Quotient (IQ) assessed by the Wechsler Abbreviated Scale of Intelligence (WASI).

Measuren (%)n (%)n (%)n (%) 
Gender: Male161 (55.1)124 (57.1)24 (45.3)13 (61.9)0.244b
Employed216 (74.0)22 (10.2)4 (7.4)1 (5.0)0.823c
Living on their own217 (74.3)98 (45.2)23 (42.6)6 (31.6)0.238c
Substance use, currentd202 (69.2)13 (6.4)3 (5.8)1 (5.3)0.810c
CDSS (item 8)e120 (41.1)16 (26.2)8 (17.4)3 (23.1)0.568c
Not hospitalized217 (74.3)25 (11.5)5 (9.3)2 (10.5)0.547
 n (%)M (SD)M (SD)M (SD) P b
Education in years284 (97.3)12.8 (2.6)12.4 (3.1)12.0 (4.9)0.373
PANSS total252 (86.3)62.7 (17.8)63.6 (17.9)64.8 (13.1)0.850
Positive scale256 (87.7)15.1 (6.1)16.0 (6.3)14.6 (4.6)0.559
Negative scale258 (88.4)15.8 (6.5)14.8 (6.9)17. 9 (6.3)0.206
Psychopathologyf257 (88.0)31.6 (8.8)31.4 (10.7)32.3 (6.9)0.925
G.2 (anxiety)287 (98.2)3.0 (1.4)3.0 (1.5)3.8 (1.8)0.083
G.6 (depression)287 (98.2)2.5 (1.5)2.6 (1.3)3.2 (1.6)0.172
Dysphoric moodg285 (97.6)2.3 (0.9)2.3 (0.9)2.6 (0.9)0.263
CDSS (current depression, total)128 (43.8)4.0 (4.0)4.1 (4.2)7.4 (4.3)0.012a
GAF-S292 (100)43.2 (11.6)43.2 (10.8)43.6 (6.4)0.987
GAF-F292 (100)43.5 (11.1)41.6 (9.5)43.1 (6.8)0.281
Current IQh211 (72.3)99.5 (16.5)97.3 (17.5)94.4 (19.6)0.377
Premorbid IQh211 (72.3)101.5 (15.8)99.2 (17.0)94.4 (16.8)0.160
Table 2. Type of trauma in relation to age
Type of traumaTrauma (not PTSD) (n = 54)PTSD (n = 21)
ChildaAdultAge unknownTotalChildaAdultTotal
  1. a

    Child of age <16 years.

Child sexual abuse (CSA)6006000
CSA+ violence5005404
Violence + neglect4004303
CSA+ neglect8008404
CSA+ neglect+ violence140014808
Trauma, not specified6039011

In Global Assessment of Functioning (GAF) score, split version, symptoms and function in all three groups were comparable. Respectively, there were no significant differences between the three groups' no trauma, trauma, and PTSD groups in years of education (Table 1). Neither the total PANSS score nor factor analyses of the PANSS [55, 56] using subscales of positive, negative, psychopathology, and dysphoric mood did show any significant differences between the groups (Table 1). There were no significant differences when comparing all three groups (no trauma, trauma, and PTSD) on all measures (Table 1), except for CDSS (total) current depression, with CDSS scores of mean (SD) of 4.0 (4.0) no trauma, 4.1 (4.2) trauma, and 7.4 (4.3) PTSD, P = 0.012. CDSS was completed by 44% of our sample. We have also investigated the difference in sociodemographic or clinical variables between the groups and found no significant difference between with (n = 128) and without (n = 164) CDSS (data not shown). Current suicidal thoughts, plans, and attempts (CDSS-item 8) were rated and showed no significant differences between groups (Table 1).Yet, SZ patients with trauma and PTSD had significantly more current depression (CDSS total) than the other groups (P = 0.012).

Cognitive test performance

Regarding the cognitive test performance, including the premorbid IQ measure, no significant differences between the three groups were found. Most cognitive measures were within the normal range of performance compared with established norms of healthy controls, except for psychomotor speed or processing speed (Digit Symbol from the WAIS) with performances 1 SD below the mean in all groups. Because no significant differences were found regarding age, education, and positive and negative symptoms, there was no need for adjustments in the further analyses (Table 3). As can be seen in this table, there is no significant difference between the groups on the current IQ measure, consisting of the subtests Vocabulary, Similarities, Block Design, and Matrix Reasoning of the WASI, where the patients with SZ in no trauma, trauma, and PTSD performed equally (P = 0.377).

Table 3. Cognitive test performance for Schizophrenia between patients with no trauma, trauma, and PTSD
Neuropsychological testsNo traumaTrauma (without PTSD)PTSD P
Mean (SD)Mean (SD)Mean (SD)
  1. WMS-III LM I, Wechsler's Memory Scale-III Logical Memory I; CVLT-II, California Verbal Learning Task-II; LDR, long delay–free recall.

  2. All neuropsychological values were converted to z (or t) score (mean 10; SD 3), except the CVLT-II Total (number of correct words recollected) and Digit Span Forward and Backward (raw scores).

Verbal memory
WMS-III LM I8.26 (2.89)8. 30 (3.05)8.24 (3.33)0.996
CVLT-II Total A1-546.72 (11.84)46.89 (13.34)46.76 (13.57)0.997
CVLT-II LDR−0.44 (1.27)−0.52 (1.37)−0.36 (1.35)0.883
Digit span forward5.85 (1.06)5.71 (1.08)5.33 (1.46)0.121
Working memory
Digit span backward4.15 (1.07)3.98 (1.10)4.19 (1.21)0.611
Psychomotor speed
Digit symbol6.36 (2.14)6.78 (2.29)6.76 (2.81)0.456
Executive functioning
Inhibition7.38 (3.68)6.82 (3.93)7.19 (4.25)0.688
Letter fluency8.92 (3.94)8. 23 (4.02)8.40 (3.44)0.544
Category fluency8. 62 (4.15)8.18 (4.22)8.05 (3.59)0.729
Inhibition/Switching7.81 (3.61)7.43 (3.46)7.52 (4.08)0.809
Vocabulary47.06 (12.35)45.55 (12.9143.81 (16.86)0.493
Similarities48.92 (11. 32)47.89 (11.30)46.65 (12.10)0.653
Block design50.33 (10.41)49.23 (11.73)46.25 (11.93)0.275
Matrix reasoning51.08 (11.63)48.76 (12.54)45.57 (12.73)0.102
WASI VIQ97.39 (16.59)95.52 (17.13)94.80 (20.29)0.707
WASI PIQ101.52 (15.82)98.77 (16.96)94.40 (16.81)0.160
WASI FIQ 4 subtests99.54 (16.48)97.30 (17.51)94.35 (19.64)0.377


The main finding of the current study was no significant differences in cognitive functioning between schizophrenia (SZ) spectrum disorder patients with trauma, post-traumatic stress disorder (PTSD), or no trauma, but clear psychopathological differences. The study of Duke et al., 2010 [38], did not find any association between PTSD and cognitive abnormalities already present in patients with SZ. This result was in line with our main findings where no significant differences in cognitive functioning between SZ patients with trauma, PTSD, or no trauma were seen. However, some studies have found an association of comorbid PTSD with cognitive dysfunction in schizophrenia [14, 57]. Fan and colleagues [14] suggest that patients with comorbid PTSD (n = 15) suffer more cognitive impairments especially in the domains of attention, working memory, and executive functioning, compared with those who have not developed PTSD.

Previous studies of cognitive function in SZ patients with traumatization/PTSD have some limitations including a small number of participants with limited evaluation of cognitive abilities and reliance of clinical diagnoses of SZ and PTSD. It may be speculated whether the cognitive deficits associated with traumatization are sufficiently severe to be differentiated from those associated with SZ development. There is also a large variation in cognitive functioning in SZ, which suggests that some of the discrepant findings in this field could be due to differences in statistical power. Due to the correlation design, the current findings cannot be used for causative inferences. However, it may be speculated that trauma has less effect on neurocognitive functioning, which is often regarded as a result of abnormal neurodevelopment. Trauma seems more related to mood symptoms, which are generally regarded as a state measure. To our knowledge, our study provides the most extensive evaluation of cognitive function in the largest traumatized SZ sample to date.

As assumed, we found a significantly higher level of depression in the SZ group with PTSD compared with both the no trauma and the trauma group (Table 1). Our findings of current depression (CDSS total) were in line with a recent study [38], which shows that trauma and PTSD have been related to a variety of adverse outcomes in patients with SZ. Patients with PTSD tend to have more severe psychotic symptoms [58], increased suicidality [59], and to use more psychiatric services [60]. Our study, however, did not find any significant differences between the groups concerning psychotic symptoms, suicidality, or the use of psychiatric services.

A series of studies have been conducted on the prevalence of PTSD in patients with SZ [7, 37, 61-64], with some discrepancies in the prevalence estimates. Apparently, 20–29% of psychological traumatization results in the development of PTSD [2, 3, 64, 65]. In our study, we found a prevalence of trauma in the total SZ group of 25%, of which only 7% had current PTSD. The prevalence of PTSD in our study was low, but comparable to Neria and coworkers [8] who found a prevalence of PTSD in their sample of patients with schizophrenia of approximately 10%.

There are several limitations in the present study. Given the nature of the cross-sectional design, causal relationships between traumatization/PTSD, SZ, and functional measures cannot be drawn. Our sample size with the groups containing 292 participants with SZ, 75 traumatized patients, and 21 with a comorbid PTSD may lack adequate power to detect between-group differences. Also some questionnaires have a low response rate (CDSS). The reason for CDSS not being used for the total sample is that CDSS was first included in the protocol after the first phase of the TOP recruitment; thus, it is missing in 164 patients. However, there is no selection bias in these two subsamples, because there is no reason to believe that patient characteristics changed over time in Oslo. There were no significant differences in clinical or sociodemographic characteristics between the CDSS and no CDSS groups (data not shown).

A limitation of the method for assessing trauma was that the approach probably led to underreporting of traumatic events that are common in the general population [66], such as the sudden, unexpected death of a loved one. These are even more common in people with severe mental illness [67]. This is a significant limitation for the current study. However, we did obtain self-reports about witnessing traumatic events, but these were rare. Another limitation of the current study was the use of a semistructured clinical interview to assess lifetime trauma exposure. This may have the disadvantage of ‘false low’ rate of lifetime trauma exposure.

Heterogeneity within the trauma group with regard to the type and also age when the traumatic event happened complicates the interpretation of our results. There is some evidence that suggests that certain types of trauma differentially contribute to cognitive impairment and that the age at which the traumatic event took place may influence the impact of trauma on cognitive function [29].

The theoretical approach of the vulnerability–stress model can be used as a starting point for studying the role of PTSD and post-traumatic symptoms in SZ.

In a clinical perspective, an individual with schizophrenia can be considered to have a ‘dual vulnerability’ with respect to traumas, with both types of vulnerability leading to the same functional deficits in the patients. These deficits include cognitive deficits and interpersonal and intrapersonal deficits that lead to difficulties in comprehension of situations and increased tendency to misinterpret the intentions of others.

Further longitudinal investigations are needed to delineate the relationship between PTSD comorbidity in SZ and the cognitive deficits that potentially could impact the level of distress, subjective quality of life, and poorer treatment outcomes in these patients.


A special thanks to Mari H. Vårdal for helpful discussions and advice regarding the statistical analysis concerning this article. We would also like to thank Alna Outpatient Rehabilitation Unit for recruiting patients and clinical social worker Irma Helgesen for her dedication and assistance throughout the study. The study was supported by South–East Norway Health Authority (2010-074) and Research Council of Norway (#167153/V50).

Declaration of interest