Neuropsychological and Socio-Occupational Functioning in Young Psychiatric Outpatients: A Longitudinal Investigation

Background Clinical symptoms and neuropsychological deficits are longitudinally associated with functional outcome in chronic psychiatric cohorts. The current study extended these findings to young and early-course psychiatric outpatients, with the aim of identifying cognitive markers that predict later socio-occupational functioning. Methods At baseline, 183 young psychiatric outpatients were assessed. Ninety-three returned for follow-up (M = 21.6 years old; SD = 4.5) with an average re-assessment interval of 21.6 months (SD = 7.0), and primary diagnoses of major depressive disorder (n = 34), bipolar disorder (n = 29), or psychosis (n = 30). The primary outcome measure was cross-validated with various other functional measures and structural equation modelling was used to map out the interrelationships between predictors and later functional outcome. Results Good socio-occupational functioning at follow-up was associated with better quality of life, less disability, current employment and being in a romantic relationship. The final structural equation model explained 47.5% of the variability in functional outcome at follow-up, with baseline neuropsychological functioning (a composite of memory, working memory and attentional switching) the best independent predictor of later functional outcome. Notably, depressive and negative symptoms were only associated with functioning cross-sectionally. Diagnosis at follow-up was not associated with functional outcome. Conclusions Neuropsychological functioning was the single best predictor of later socio-occupational outcome among young psychiatric outpatients. Therefore, framing psychiatric disorders along a neuropsychological continuum is likely to be more useful in predicting functional trajectory than traditional symptom-based classification systems. The current findings also have implications for early intervention utilising cognitive remediation approaches.


Introduction
Traditionally in mental health, there has been a strong emphasis on defining recovery in terms of symptom resolution. However, there is increasing recognition that defining remission solely on the basis of clinical symptoms is inadequate. Frequently, these 'remitted' patients with diagnoses of major depression [1], bipolar disorder [2] or psychosis [3] remain functionally impaired. Zimmerman et al. [1] demonstrated that half of patients with major depression who were classified as remitted based on rating scale criteria did not consider themselves to be in remission.
Critically, these individuals were significantly more impaired in their work performance and social relationships. It is likely that subsyndromal symptoms, particularly depression [2] and negative symptoms [4], contribute to these 'residual' functional deficits across the major psychiatric disorders. Of significance, cognitive vulnerabilities may additionally contribute to persisting disability given evidence that neuropsychological dysfunction predicts later socio-occupational functioning, over and above symptom levels, as observed in depression [5], bipolar disorder [6], and schizophrenia [7].
While in chronic cohorts psychotic disorders are associated with more pronounced neuropsychological and psychosocial disability than affective disorders [8], neuropsychological performance appears to predict functional outcome irrespective of diagnosis [8]. Accordingly, neuropsychological deficits have been proposed as endophenotypes that cut across traditional diagnostic boundaries [9]. This idea of intermediate traits spanning the major psychiatric disorders is not new, and is buttressed by findings of neurometabolic [10], neuroanatomical [11], and neurophysiological [12] factors common to major depression, bipolar disorder and psychosis. The overarching premise is that affective and psychotic disorders may exist along a continuum [13], and exploring shared dimensional features is likely to be more informative prognostically, and for the understanding of pathophysiology (e.g. for identifying suseptibility genes) [14], than categorising individuals according to their idiosyncratic symptomatology [15]. Indeed, within-disorder heterogeneity is not uncommon in the current diagnostic classification systems. Furthermore, separating patients based on syndromes is often confounded by varying illness trajectories, as well as the finding that diagnoses can be relatively unstable, particularly in the early stages of psychiatric illness [16].
To date, there is strong evidence that cognitive dysfunction is shared across early-course major depression [17], bipolar disorder [18], and schizophrenia [19]. Consistent with studies in wellestablished cohorts, neuropsychological deficits appear longitudinally predictive of functional outcome over and above symptomatology in early psychosis [20] and possibly, in first-episode mania [21]. To our knowledge, such relationships have not been explored in first-episode major depression. Identification of cognitive markers with predictive value may ultimately help to direct early intervention strategies. For example, some deficits may be indicative of underlying structural and functional brain changes, which in turn may be amenable to novel pharmacotherapies, or cognitive remediation, with the latter already showing some promise as an early intervention strategy [22].
The aim of the current study was to determine whether neuropsychological functioning longitudinally predicts socio-occupational functioning over and above symptomatology in a young mixed psychiatric sample with major depression, bipolar disorder or psychosis. We hypothesised that neuropsychological functioning, as well as depressive and negative symptoms, would contribute to later functional outcome. Structural equation modelling techniques were used to estimate the interrelationships among clinical variables more accurately than past studies utilising multivariate regression approaches [23].

Ethics Statement
The current study was approved by the University of Sydney Human Research Ethics Committee. After complete description of the study to the subjects at each assessment time-point, those aged 16 years and above gave written informed consent to participate, whereas for those under the age of 16 years both the participant and their legal guardian gave written informed consent in accordance with Human Research Ethics Committee guidelines and Australian law.

Subjects
Participants were recruited from specialised tertiary referral services ('Treatment Centre') for the assessment and early intervention of mental health problems in young people (Youth Mental Health Clinic, YMHC, at the Brain & Mind Research Institute, BMRI; and headspace, Campbelltown, Sydney, Australia; Inner West Area Health Service First Episode Psychosis Intervention Services) [24]. Inclusion criteria were restricted to persons aged 12 to 35 years seeking professional help primarily for a depressive (unipolar or bipolar) and/or psychotic syndrome. Subjects were excluded if they were diagnosed with a neurological condition, had insufficient English-language skills, or intellectual disability (premorbid IQ,70). Patients were assessed by a psychiatrist according to DSM-IV TR [American Psychiatric Association, 25] criteria at two time points (and diagnoses were subsequently confirmed through case-note review), and underwent a comprehensive clinical and neuropsychological assessment (including self-report questionnaires) by clinical neuropsychologists or trained research psychologists at baseline. A repeat clinical assessment (with self-report questionnaires) was also conducted at follow-up.

Measures
The BMRI Structured Interview for Neurobiological Studies was conducted to gather key demographic and clinical information [26]. Symptoms were rated using the expanded version of the Brief Psychiatric Rating Scale (BPRS-E) [27]. Subscores were calculated based on a previously derived component structure (depressive, positive, negative and manic symptoms) [28].
Various socio-occupational measures were administered. The Social and Occupational Functioning Scale (SOFAS) rated by the treating psychiatrist and research psychologist [39] was the primary functional outcome measure used. Additional patientrated information relating to work and study, financial and living situation, and relationship status, and structured questionnaires, including total scores from the World Health Organization Quality of Life BREF version (WHOQOL-BREF) [40] and the Disability Assessment Scale II (WHODAS-II) [41], were included to cross-validate the clinician-rated SOFAS.

Data Analysis
Statistical analyses were conducted using Statistical Package for the Social Sciences Version 20.0 (SPSS 20.0) and Analysis of Moment Structures Version 20.0 (AMOS 20.0). Participants were initially dichotomised into two groups, as adapted from similar methodology [42], based on their follow-up SOFAS scores. Scores greater than or equal to 61 were classed as 'good outcome', whereas a lesser score was categorised as 'poor outcome'. This dichotomy was chosen since it was more meaningful to examine outcome groups anchored to qualitative descriptors (i.e. 'generally functioning well or better' compared with 'moderate impairment or worse') than using correlation coefficients based on a continuous SOFAS score. In other words, dichotomising the SOFAS score means that the sample characteristics are immediately interpretable in reference to whether patients are functioning well or poorly. Analyses of Variance (ANOVAs; Welch's correction where homogeneity of variance was violated according to Levene's test) and chi-squared tests were conducted to crossvalidate the clinician-rated SOFAS score with other objective and subjective functional indices. ANOVAs were also conducted on baseline variables to explore potential factors associated with follow-up SOFAS outcome. Neuropsychological indices that differed according to outcome group were reduced to latent variables by confirmatory factor analysis.
Since structural equation modelling is particularly sensitive to the full range of statistical variability, continuous SOFAS scores were used in this procedure. A two-step approach to structural equation modelling was adopted [43]. First, the measurement model (latent neuropsychological functioning) was tested for model fit. Second, we tested the structural model (the relationships between measured variables) using a model-trimming approach through an iterative process in which non-significant paths with the smallest contribution were sequentially eliminated from a saturated model (where all variables were allowed to correlate) until a best fitting model was ascertained to explain the relationships between baseline symptoms and follow-up functioning. All baseline cross-sectional relationships were treated as correlational due to uncertainty of directionality. The model was improved through the inclusion of follow-up symptoms and compared to the original model using nested-models. Latent baseline neuropsychological functioning was then entered into the model to determine whether it explained functional outcome over and above other predictors, again using nested-models. Overall model fit was determined using the x 2 statistic, Bentler comparative fit index (CFI) [44], Bentler-Bonnett nonnormed fit index (NFI) [45], and Root Mean Square Error of Approximation (RMSEA) [46] with 90% confidence interval. A good-fitting model was indicated by a non-significant x 2 test (indicating nonsignificant difference between the covariance matrix of the data and the model), a CFI and NFI of greater than 0.90 (indicating that the current model was superior to a null model where all paths are constrained to zero), and a RMSEA of less than 0.05 with an upper confidence interval bound of less than 0.08 (indicating that the error of approximation of the model compared with the data was acceptable). Nested-models were compared using Dx 2 with significant changes representative of better model fit. In the case of non-significant Dx 2 , the model with greater degrees of freedom (i.e. more parsimonious) was considered the superior model [47].
Patients did not return for follow-up for various reasons, including those who were disinterested in participation (n = 23), not contactable or had transportation difficulties (n = 54), too busy with full-time work or study (n = 8), too ill (n = 4), or had deceased (n = 1). Compared with those who returned for follow-up, those who did not were not significantly different in terms of diagnoses, age, age of onset, years of education, depressive, positive, negative or manic symptoms, or on any neuropsychological measure (all p.0.14). However, returning patients had significantly higher premorbid IQ [104 vs. 100; F(1, 177) = 6.73, p = .01], although this no longer remained significant once outliers (n = 6) with premorbid IQ less than 80 or greater than 120 were excluded (p..10). Nonetheless, even by including patients with outlying premorbid IQs the difference is unlikely to be clinically significant given the discrepancy of 4 IQ points (equates to a standard deviation difference of 0.27). Thus, these patients were not excluded from subsequent structural equation modelling analyses.

Cross-validating SOFAS at Follow-Up
At follow-up, a high inter-rater agreement was obtained between treating psychiatrists and research psychologists on the SOFAS measure (intraclass correlation coefficient = 0.778, p,.000) [48,49]. Using our qualitatively-defined dichotomy, 53 patients met criteria for 'good outcome', whereas 40 patients were classified as 'poor outcome' (see Table 1

Characteristics of Outcome Groups
Age, age of onset, gender, years of education, and premorbid IQ did not differ between outcome groups (p..05; see Table 2
Entering only baseline SOFAS and symptoms into the structural model did not yield any models with uniformly adequate fit indices. The best possible model included depression and negative symptoms (mania did not predict SOFAS outcome despite being different between good and poor outcome groups). This model was somewhat inadequate and paths from symptoms to follow-up functioning, as measured by the SOFAS, were both nonsignificant (x 2 = 37.33, df = 31, p = .20, CFI = .97, NFI = .86, RMSEA = 0.05, 90% CI = 0.00-0.10). Modifying the model by including follow-up symptoms as predictors (which were in turn predicted by baseline symptoms) yielded a good fitting model with adequate fit indices (x 2 = 23.68, df = 29, p = .74, CFI = 1.00, NFI = 0.91, RMSEA = 0.00, 90% CI = 0.00-0.06). This was a significant improvement over the previous model and was thus adopted (Dx 2 = 213.65, Ddf = 22, p,.001). As the paths from baseline symptoms to follow-up functioning were still not statistically significant, we tested whether these paths were necessary. The subsequent model, removing direct paths from baseline symptoms to follow-up functioning, was an adequate fit and all paths were significant (x 2 = 27.32, df = 31, p = .66, CFI = 1.00, NFI = 0.90, RMSEA = 0.00, 90% CI = 0.00-0.07). In addition, it was not significantly different from the previous model (Dx 2 = 3.64, Ddf = 2, p = .162) and, thus, it was adopted because it was more parsimonious. Finally, the neuropsychological latent variable was entered into the model as a predictor of followup functioning. This model was a good fit with all paths statistically significant (x 2 = 19.00, df = 30, p = .94, CFI = 1.00, NFI = 0.93, RMSEA = 0.00, 90% CI = 0.00-0.02), fitting better than the previous model (Dx 2 = 28.32, Ddf = 21, p,.005) and, thus, was adopted as the final model since the significant increase in model fit justified the reduction in parsimony [47]. The final model (Figure 1) explained 47.5% of follow-up functioning, with the latent baseline neuropsychological variable being the strongest predictor, accounting for 10.9% of the variance. Since proportions of diagnoses differed between outcome groups when initially dichotomising the SOFAS scores, we decided to test whether being diagnosed with bipolar disorder or psychosis (the two diagnoses that were differentially represented) might have better accounted for the outcomes seen at follow-up. The results revealed that diagnoses did not improve model fit and, to the contrary, made the model less parsimonious (Dx 2 = 21.85, Ddf = 22, p = .40). In addition, neither the paths from bipolar disorder nor psychosis to follow-up functioning were significant (r = .07 for both predictors, p..40), while baseline neuropsychological functioning remained the best predictor. Therefore, the final model in Figure 1 was retained.

Discussion
In young psychiatric outpatients relatively early in their illness, moderate or greater functional impairment was associated with poorer quality of life, greater disability, unemployment and single relationship-status. Functional outcome was most strongly predicted by baseline neuropsychological functioning, controlling for initial level of functioning and concurrent symptoms. Baseline verbal and visuospatial memory, spatial working memory, and attentional switching were instrumental to socio-occupational functioning 6 to 48 months later, consistent with findings from more chronic cohorts, which have identified baseline memory [6,7,50] and executive functioning [5][6][7] to be predictive of later functioning across our diagnostic groups. Neurobiologically, this may reflect underlying frontotemporal changes, which underpin later functional outcome early in the course of affective and psychotic illness. In contrast, baseline neuropsychological functioning was not associated with symptomatology at baseline or follow-up, in keeping with the literature [21]. Thus, memory and executive functioning are likely independent of clinical state in early-course psychiatric illness, underscoring their potential as endophenotypes.
In contrast, baseline depressive and negative symptoms did not directly predict later functional outcome. These symptoms were, however, associated with functional outcome cross-sectionally, in line with previous findings [51]. Therefore, symptoms and functioning do not appear to be causally linked, but instead may be outcomes of a common pathophysiology. Similarly, the utility of diagnosis in predicting functional outcome was questioned by the current findings. Diagnosis was unrelated to level of functioning over and above the effects of cognition and clinical symptoms, implying that categorising patients based on traditional classification systems is unlikely to be informative in regards to functional trajectory. Taken together, these results strongly suggest that a traditional, or sole, focus on symptom factors is inadequate in characterising prognosis and recovery. On the other hand, the longer-term outcome of young patients presenting to mental health services with a range of disorders is best predicted by their initial level of neuropsychological functioning.
The strong cross-sectional relationship currently found between neuropsychological and socio-occupational functioning contradicts previous reports in more chronic patient populations. For instance, Simonsen et al. [8] found that most neuropsychological measures were not singly predictive of functional outcome in a group of chronic bipolar disorder and schizophrenia patients in their midthirties. Moreover, overall neuropsychological and socio-occupational functioning were only modestly correlated. This discrepancy with the current data suggests that the influence of neuropsychological functioning on later functional outcome may be greater in the early stages of illness. That is, over time the impact of neuropsychological deficits on functioning may be diminished as the chronicity of disability is spontaneously maintained. Although this is speculative, it highlights the potential importance of early intervention, since interventions aimed at reducing disability may be particularly effective within this initial, critical period from psychiatric onset [52].
A major strength of the current study was its longitudinal design, providing stronger, directional evidence of the predictive utility of neuropsychological functioning over traditional crosssectional approaches. Structural equation modelling also permitted a more sophisticated representation of how clinical symptoms, cognition and socio-occupational functioning are inter-related over time. In addition, the use of a naturalistic sample of patients suffering from a range of severe mental illnesses, and frequently with other psychiatric comorbidities, strengthens the ecological validity of the current findings.

Limitations and Future Directions
A number of limitations must be acknowledged. The current study was restricted by its modest sample size. As such, we were unable to explore whether the relationships between cognition and functional outcome differed according to traditional diagnostic categories using multiple-group modelling procedures [53]. Future studies need to directly test whether diagnosis moderates the currently observed effects, as previously demonstrated crosssectionally in more chronically ill samples [4]. Moreover, studies should include mediators of the relationship between neuropsychological and socio-occupational functioning, which have been identified in cross-sectional studies of chronic schizophrenia and bipolar disorder patients. Indeed, there is evidence suggesting that social competence [54] and social cognition [55] mediate this relationship. Separately, the suggestion that memory and executive functioning may be potential endophenotypes of major psychiatric morbidity needs to be verified in future studies including cohorts of unaffected family members and healthy comparison individuals.

Summary
The current study was the first longitudinal study utilising structural equation modelling to demonstrate a strong link between neuropsychological and later socio-occupational functioning in young and early-course psychiatric outpatients. Memory and executive dysfunction were linked to later disability, highlighting the potential of reducing longer-term disability through early neuroprotective strategies targeting frontal and temporal brain regions. Thus, psychiatric illness may be better viewed along a neuropsychological continuum, with those on the impaired end more likely to follow a less favourable functional trajectory. A traditional and sole focus on symptoms or syndromes is unlikely to be useful in predicting functional recovery. Compared with more chronic stages, neuropsychological functioning may be a more robust predictor of later functioning in the early stages of psychiatric disease, suggesting that cognitive remediation strategies may be most efficacious in effecting functional change in this early 'critical period'.