Abstract
It has now been widely acknowledged that cognitive impairment is a core feature in schizophrenia [1]. Cognitive impairment accounts for a considerable part of the rising costs caused by schizophrenia [2]. These impairments have not only been found in chronic, but also in first episode patients [3], and there is also evidence that they, at least partially, exist in the premorbid and prodromal phases of patients with schizophrenia as well as in their first degree relatives [4]. However, the actual impact of these impairments on their daily lifestyle is unclear. It has thus become a major goal of research to determine the factors that predict performance in social and occupational functioning, or, as Green [5] put it, to identify the ‘rate limiting factors’. Moreover, since trends to shorten duration of hospitalization as well as the global processes of de-institutionalization have led to increasing development of outpatient structures over recent years, these issues have increasingly gained importance, as they may decide how well patients will access outpatient care.
In two comprehensive reviews of a series of replicated findings, Green [5] and Green et al. [6] concluded that there are significant associations between neurocognitive impairments and functional outcome. Dividing the literature into three areas of social functioning or outcome, these two studies found some consistent results with respect to the association between neurocognitive and social functioning: verbal memory predicted community functioning, problem-solving skills, and skill acquisition; vigilance predicted social problem-solving and skill acquisition; and executive functioning, as measured by card sorting, was consistently associated with performance on measures of community outcome. He concluded that data concerning the ability of negative symptoms to predict social and community functioning were inconsistent.
Some more recent studies found similar relationships between social problem-solving and neurocognitive impairment as well as with negative symptoms [7, 8], and a number of studies were able to demonstrate significant associations between better performance on the Wisconsin Card Sorting Test (WCST) [9] and higher global levels of occupational functioning [10, 11]. However, other studies have demonstrated a stronger relationship between social functioning and symptom level than with cognitive impairment. In one study applying the method proposed by Liddle and Morris [12] of a tripartite distinction between psychomotor poverty, disorganization and reality distortion, disorganization showed the most reliable relationship to community functioning [13]. Further studies were able to demonstrate an association between cognitive impairment and poor symptomatic outcome [14–16]. These findings suggest that both cognitive impairment and symptom level may play a role as ‘rate-limiting factors’.
Among the most commonly found cognitive deficits in schizophrenia, executive functions are acknowledged to play a pivotal role. ‘Executive function’ is a term that is generally used to describe the processes whereby cognitive systems are coordinated for the successful performance of complex tasks. These processes are important in the planning and execution of complex behaviours, in the generation of a strategic approach to complex problems, in the monitoring of performance and in the revision of strategies and behaviours that cease to be appropriate. It can be hypothesized that they are crucial in the unstructured environment where one has to draw on experience to make the critical judgements necessary for autonomous, independent functioning. Baddeley [17] suggested that a superordinate ‘central executive’ system of working memory, a key component of executive functions, was responsible for intact functioning, and therefore introduced the term ‘dysexecutive syndrome’ as a functional characterization of patients with deficits in those dimensions.
Over many years, the WCST has become a kind of sine qua non among tests of executive functions. However, even if the majority of studies support findings of deficits on the WCST in patients suffering from schizophrenia [18], results are heterogeneous, with some authors reporting normal performance [19, 20]. Patients suffering from schizophrenia are reported to show deficits on other widely used measures of executive functioning such as the Stroop [21, 22], the Trail Making Test (TMT) [23], and verbal fluency tests [24, 25]. Another measure of executive functioning is the Behavioural Assessment of Dysexecutive Functioning (BADS) [26], which is a relatively new test that has only been studied in schizophrenic patients on a rare number of occasions [27, 28]. It was developed to aim at predicting everyday problems that arise from executive impairments. Its forerunners were two tasks developed by Shallice and Burgess [29], the Six Elements Test and the Multiple Errands Test. Wilson et al. [26] modified the test and developed the BADS, which consists of six subtests measuring different aspects of executive functions such as flexibility in thinking and skills for planning, initiation and execution of tasks. In a study of 24 schizophrenic patients, Krabbendam et al. [28] found a moderate correlation between test performance and social adjustment.
The literature cited above raises several issues. First, reports such as normal performances of schizophrenic patients on the WCST [19, 20] suggest that executive functions are not merely simple, but complex constructs with an array of different dimensions that may show individual variation [27, 30]. Second, although associations between impairments in executive functions and social adjustment are found, these associations are of quite varying strength. and last, degree of psychopathology may play an influential role on social adjustment.
In this study, we therefore challenged our hypothesis that not all patients with schizophrenia are equally impaired on measures of executive functions, and investigated whether the latter depended on sociodemographic variables. We equally examined whether executive functions were associated with social adjustment, and whether symptom levels could be shown to account for any such associations.
Method
Subjects
All participants in this cross-sectional study were recruited from a specialized hospital unit of the Psychiatric University Hospital Geneva. The unit treated patients between 16 and 65 years of age and provided two sub-programmes, one for multiple episode patients with chronic evolution of illness, and another for patients with recent onset schizophrenia (< 3 years). Patients fulfilled criteria for DSM-IV schizophrenia according to Structured Clinical Interview for DSM-IV (SCID-IV) [31], and presented no history of traumatic brain injury, epilepsy, or other known neurological disorders, nor other significant medical conditions considered to affect cognitive performance. They neither had a history of drug or alcohol dependence, nor current drug or alcohol abuse as assessed with urine drug screening. An IQ of no less than 70, normal eyesight and familiarity with the French language were required. Patients were only included when acute psychotic symptoms had resolved and when plans for discharge had already been resumed.
After completely describing the study to subjects, we obtained written consent. All assessments of symptom level and of executive and social functioning were done within the same 2 days for each patient.
Thirty-eight patients (23 males, 15 females) were included in this study. Mean age was 24 years (SD ± 7; range 16–38). Mean National Adult Reading Test (NART) [32] score was 100 (SD ± 12). None of the patients had ever been married. Level of education was judged low (obligatory school level) in 22 (58%), middle (high school degree or apprenticeship) in 14 (37%), and high (university degree) in 2 patients (5%). Among the ‘active patients’ (n = 13; 34%), 4 (11%) were employed, 7 (18%) were students, and 2 (5%) had sheltered work. Among the ‘inactive patients’ (n = 25; 66%), 4 (11%) were looking for a job, 12 (32%) were on a disability pension, and 9 (24%) were without a situation. Eight (21%) patients were living alone, 18 (47%) with their parents, 2 (6%) with a partner or a friend, and 10 (26%) were living in sheltered homes. Mean length of illness was 62 months (SD ± 63; range 1–240) with a mean admission rate of 5 (SD ± 7; range 1–30) and a chronic course of illness (< 3 years) being identified in 18 (47%) patients. Of the 36 (95%) patients receiving antipsychotic medication, 5 (13%) were receiving typicals, 31 (82%) atypicals, with an overall of 24 (67%) patients receiving antipsychotic judged to have relevant sedative effects. Mean dose in chlorpromazine equivalents was 310 (SD ± 207; range 62,5–800). Anticholinergic medication was administered in 20 (53%) patients.
Assessment
For the assessment of symptom level, we used the Positive and Negative Syndrome Scale (PANSS) [33]. Based on the method applied by Lindenmayer et al. [34], we calculated the scores for the positive and negative syndromes by subtracting the score of the negative symptom from the score of the positive symptom scale. Furthermore, we applied the Calgary Depression Scale (CDS) [35], which was specifically designed for assessment of level of depression in people with schizophrenia. Both PANSS and CDS were administered by an experienced principal investigator (AS).
For the assessment of psychosocial adjustment, we applied the social dimension of the Global Assessment of Functioning (GAF) [36], the Life Skills Profile (LSP) [37], the Social Behaviour Schedule (SBS) [38], as well as the Dysexecutive Questionnaire (DEX) [26]. The LSP was designed with a clinical interest in reintegration of individuals with schizophrenia in that it assesses constructs of relevance to survival and adaptation in the community. We used the French version [39] which was completed by a key worker upon observable behaviours and social functioning. The ratings range from 1 to 4, the latter expressing the highest degree of social adjustment. The SBS was initially developed for the assessment of behavioural dimensions in chronic schizophrenia. Detailed and anchored ratings of 21 items of behaviour are scored on a scale of 0–4. This schedule was rated from information given by the key worker under supervision of a principal investigator (AS) after the assessment of symptom level and of the LSP had taken place. For each patient, the predominant behaviour of the last month was assessed. The DEX is part of the BADS [26] manual, however, it is not used in the calculation of the profile score for the battery. It is a 20-item questionnaire constructed in order to sample the range of problems commonly associated with the dysexecutive syndrome. Two forms of the DEX exist, one to be completed by the patient, another by a relative or caregiver who has close, preferably daily contact with the patient. The answers range from a score of 1–5, the latter meaning the highest degree of changes.
The assessment of executive measures included an array of tests that tap different dimensions of executive function. The WCST requires patients to sort cards according to shape, colour, and number. There is an arbitrary correct sorting principle that patients must learn without feedback from the tester. The principle changes once the subject has responded correctly for 10 consecutive trials, indicating that the patient has learned the principle and to shift set according to the new rules. In our study, we used a computerized version which was finished after 128 presented cards or as soon as a patient had completed six categories, respectively. We analysed Perseverative Errors (i.e. when the patient continues to sort according to a principle despite negative feedback), and Total Categories Correct (i.e. the sum of trials completed correctly), both of which have been shown to account for most of the variance of WCST performance [9]. For the TMT, we calculated the centiles of the part B, and for the Stroop, we calculated the time ratio between an interference and a non-interference task. In the category fluency task, patients had to name as many animals as possible during one minute, whereas in the letter fluency task, they had to name as many words as possible beginning with the letter ‘M’ during one minute. Detailed descriptions of the administration and scoring methods of the BADS can be obtained from the test manual. For each of the six subtests, a ‘summary profile score’ is obtained. The ratings range from 0 to 4, the latter expressing the highest level of performance. These profiles scores are then summed to produce an overall battery score (maximum = 24), which we entered for data analysis.
Statistical analysis
Using the SPSS for Windows, we applied a stepwise multiple regression analysis to explore putative predictors of executive function as the dependent variable. Independent variables included age, gender, level of education, occupational status, living circumstances, length of illness, medication status and premorbid IQ. For data reduction, Pearson correlations were next calculated between the total scores of the measures of psychosocial adjustment, and a factorial analysis performed, which yielded a one-factor model. Pearson correlations were then calculated between this factor and measures of executive function and symptom level, respectively. p-values of less than 0.05 were regarded as significant. To test whether the BADS may be used as a measure of psychosocial adjustment given its development as instrument to aim at predicting everyday problems that arise from executive impairments, Pearson correlations were calculated between BADS total score and the other measures of executive function tasks. In addition, we applied a regression analysis to investigate putative predictors of the psychosocial adjustment factor as the dependent variable, first entering symptom level and then symptom level and measures of executive function as independent variables.
Results
Table 1 presents clinical characteristics of the study group. Scores on measures of psychosocial adjustment as well as executive function are shown in Table 2. On the WCST, three quarters of the patients performed below the mean norm of 74% (SD ± 10) for Total Categories Correct and one standard deviation below the mean norm of 13% (SD ± 6) for Perseverative Errors, corresponding to about 15% of the general population. Similarly, only one-quarter of the patients showed normal scores on the TMT part B as well as in category and letter fluency. With regard to the Stroop, an interference effect is considered above an index of 2, which accounted for half of the study population. Whereas a quarter of the patients showed severe impairment on total BADS scores, another quarter had normal scores as compared to normed total scores of 100 (SD ± 15). Only two of the BADS subtests, the Rule Shift Cards and the Action Program Tests, were overall performed within normal score ranges.
Clinical details of 38 patients with schizophrenia in an inpatient programme
Measures of psychosocial adjustment and executive functions of 38 patients with schizophrenia in an inpatient programme
Stepwise multiple regression analysis entering demographic and clinical characteristics as independent variables revealed that BADS (total score) was related to duration of illness and level of education (F = 7.45, df = 35, p = 0.002, adjusted R 2 = 27%; mean score (SD) for recent onset patients was 96 (16) versus 76 (24) for chronic patients; mean score for subjects with low level of education was 82 (25) vs 93 (18) for subjects with middle level of education). Whereas category fluency was related to duration of illness (F = 10.96, df = 35, p = 0.002, adjusted R 2 = 22%; mean words (SD) for recent onset patients was 18 (4) vs 13 (5) for chronic patients), neither letter fluency, WCST, TMT, nor Stroop were associated with demographic and clinical variables. The NART was not correlated to any measures of executive function.
Impairments were mild to moderate on the LSP and on the SBS. On the DEX caregiver version, the total mean figured around the 90th percentile, which demonstrates that key workers estimated patients’ executive functioning to be severely impaired. No correlation was found between the DEX caregiver and the patient version (r = 0.05, p = 0.77), a finding that we attributed to the lack of insight among patients. In agreement with the BADS manual, only the DEX caregiver version was considered.
The total scores of the three scales measuring psychosocial functioning (GAF, SBS, LSP) were highly correlated (Pearson correlations between GAF-SBS: −0.68; GAF-LSP: −0.67; SBS-LSP: −0.73), whereas correlations between the DEX and these other three scales were weak to moderate (r = 0.360–0.486). We entered the three highly correlated measures of psychosocial adjustment into a factorial analysis, which yielded a one-factor model (hereafter named psychosocial adjustment index) that explained 80% of the variance. The loadings of the GAF, SBS and LSP scores on this index ranged from 0.88 to 0.90. Table 3 shows Pearson correlations between the psychosocial adjustment index and executive functions and symptom level. Behavioural Assessment of Dysexecutive Functioning total score partially showed moderate to high correlations with other measures of executive function (Pearson correlations between BADS-WCST Total Categories Correct: 0.65, p < 0.01; BADS-WCST Perservative Errors: −0.66, p < 0.01; BADS-category fluency: 0.48, p < 0.01).
Pearson correlation coefficients between psychosocial adjustment, measures of executive functions and symptom level (n = 38)
A regression analysis entering the psychosocial adjustment index as the dependent and symptom level as independent variables revealed that two-thirds of the variance of psychosocial adjustment was explained by symptoms (adjusted R 2 = 63%, F = 16.75, df = 4, p = 0.0001). A further regression analysis entering the symptom level and measures of executive functions as independent variables revealed that 91% of the variance was thus explained (F = 18.66, df = 22, p = 0.0001).
Discussion
The results of this study support our hypothesis that patients with schizophrenia show different degrees of dysexecutive syndrome, with some patients performing within normal ranges. This is in line with findings of other studies [19, 20] that have emphasized heterogeneous test performances among schizophrenic patients. Interestingly, none of the executive tests were influenced by age, gender, or by medication status, and only category fluency and the BADS were predicted by prolonged illness duration, the latter being equally predicted by level of education. These findings suggest that at least some part of the dysexecutive syndrome is an independent feature in patients with schizophrenia and may not be influenced sufficiently by current medication regimes [13, 40]. Since a substantial part of our study cohort (53%) had an illness duration of less than 3 years, these results further support earlier findings of partial dysexecutive syndrome deficits at the onset of illness [4, 41].
With regard to our study aim to investigate correlations between measures of executive function and psychosocial adjustment, our results indicate that these are weak to moderate and, thus, partially consistent with previous reports [5–8]. However, it is noteworthy that unlike reports from other studies [5, 6], ‘standard’ tests such as the WCST and the TMT hardly showed any correlations with psychosocial adjustment.
Several explanations may underlie these findings, although our study cohort was too small to conclusively investigate any of them. First, although the applied tests are all considered to measure executive functions, they may assess different dimensions of executive functions. Our findings may suggest that the degree of complexity of executive tasks leads to different profiles of the dysexecutive syndrome [27, 30]. For example, the BADS consists of six subtests, each tapping other subsets of executive functions.
Second, almost all individual neuropsychological tests of higher mental processes involve a number of different functions. Therefore, poor performance can be attributed to any variety of functional deficits, and poor performance in social functioning can be accounted for by an array of impaired cognitive dimensions such as sustained attention, memory and learning capacities. For example, the WCST may be a multidimensional task that requires the integrity of numerous complex neural substrates and cognitive processes for optimal performance [42–45]. Additionally, as Pantelis et al. [43] outlined, the fact that many patients with chronic schizophrenia already fail at the intradimensional part of the WCST and are not able to complete even one category, make the WCST a merely binary discrimination task between those who can grasp the requirements of the test and those who cannot. Similarly, Goldstein et al. [20] suggested that when WCST performance was impaired, that impairment was accompanied by a broad array of cognitive deficits. This is in line with findings by Brenner et al. [46] who could show that an integrated psychological therapy focusing on early information-processing deficits had pervasive effects on other cognitive as well as behavioural dimensions.
Third, it could be argued that the applied measures of psychosocial adjustment may be too insensitive to investigate daily functioning. However, three of the applied measures, the LSP, SBS and GAF, were strongly intercorrelated, suggesting that common aspects of daily functioning were captured. An interesting finding were the correlations between the BADS total score and a number of executive functions, suggesting that the BADS may be used as a sensitive measure for psychosocial adjustment.
Fourth, our most robust finding that symptom level accounted for two-thirds of the variance of psychosocial adjustment suggests that symptom level plays a pivotal role as ‘rate-limiting factor’. The association of symptom level with psychosocial adjustment is consistent with other reports [13]. In our study, correlations were particularly noteworthy for general and negative symptoms. Several studies were able to show associations between negative symptoms and psychosocial adjustment [5–8]. Whereas one study found that anxiety was a main reason for social withdrawal in hallucinating patients with schizophrenia [47], another study reported disorganized symptoms to best predict psychosocial adjustment [13]. The finding that symptom level together with executive functions accounted for 91% of the variance of psychosocial adjustment suggests a non-negligible impact of executive functions on daily living skills as well as an overlap of symptoms and executive functions [14–16].
Finally, our findings of heterogeneous performance may give some support to the hypothesis that schizophrenia is a heterogeneous illness with different clinical and outcome patterns [48]. The varying presentations of the illness have directed more recent research to identify endophenotypes in schizophrenia [48, 49]. Our results may suggest that different endophenotypes may cause different levels of executive functioning, but that their impact on psychosocial functioning is strongly influenced by symptom level.
More knowledge rising from this field of research may help the clinician to obtain exact knowledge of the patient's deficits and, subsequently, plan more appropriate intervention models. Recent studies have brought up hopes that cognitive remediation in patients suffering from schizophrenia may open new perspectives for psychosocial treatment [46, 50]. To the extent that these findings lead to improvements in psychosocial and pharmacological interventions, they could also reduce the need for future service use and, thus, a considerable part of the costs caused by schizophrenia.
Footnotes
Acknowledgements
We thank Christos Pantelis for his valuable comments.