Predictors of psychosocial functioning in euthymic patients with bipolar disorder: A model selection approach

Abstract

Background

Functional impairment is a major target in the treatment of bipolar disorder (BD), but the magnitude and type of functional difficulties differ across patients. Findings on predictors of overall functioning and specific functional areas are inconsistent. We aimed to characterize functional difficulties and identify factors associated with global functioning and individual domains in euthymic patients.

Methods

The Functional Assessment Short Test (FAST) was used to assess overall psychosocial functioning and specific functional domains in 80 euthymic participants with BD. Participants also completed a clinical interview and a cognitive assessment. Model selection with elastic net regression was performed to identify predictors of global functioning. We then examined the association of these predictors with individual functional domains using correlation.

Results

FAST scores indicated moderate or severe impairment for 54% of the sample, with occupational functioning showing the highest impairment rate. Elastic net regression selected a model with three variables (higher residual depressive symptoms, lower executive functioning, more perceived cognitive deficits) as predictors of overall functioning. No significant associations were found between these predictors. Depressive symptoms were associated with interpersonal relationships and leisure time, executive skills with occupational functioning, and perceived deficits with cognitive functioning.

Conclusions

Residual depressive symptoms were the strongest predictor of overall functioning which highlights the importance of assessing and targeting subthreshold symptoms for recovery. Executive difficulties were associated with functioning, particularly occupational skills, independently of depressive symptoms. Interventions targeting these difficulties, such as cognitive and functional remediation, may be key treatment options towards facilitating functional recovery.

Introduction

Bipolar disorder (BD) is a chronic affective condition with significant impact on multiple areas of functioning, including work, family, and social life (Sanchez-Moreno et al., 2009). Functional difficulties often persist beyond the acute phases of the illness, while functional recovery is not always achieved after mood episodes (or achieved with a significant time-lag behind remission of mood symptoms) even for patients receiving adequate mood stabilization treatment (Vieta et al., 2018). An observational study following-up patients two years after treatment for an acute episode showed that although 64% had achieved clinical remission, only 34% had achieved functional recovery (Haro et al., 2011). This indicates a substantial gap between remission and recovery which has recently ignited research interest in assessment and treatment options for functional outcomes (Sanchez-Moreno, Martinez-Aran, & Vieta, 2017).

An important issue related to functional recovery is how to define and measure it in a standardized way (Harvey, 2006). Although several different tools have been used to assess psychosocial functioning, the Functional Assessment Short Test (FAST) is probably the first scale specifically examining difficulties experienced by patients with BD (Rosa et al., 2007) and has been recommended for clinical research in BD (Miskowiak et al., 2017). Although the original validation of the FAST established a single cut-off score to differentiate between impaired and non-impaired patients (Rosa et al., 2007), a recent study provided multiple severity thresholds and classified euthymic patients across different categories of functional impairment severity based on their FAST score: no impairment (44%), mild (15%), moderate (35%) or severe impairment (6%) in global functioning (Bonnín et al., 2018). This classification may be useful for assessing the effects of interventions targeting psychosocial functioning in more homogeneous patient subgroups, but these interventions could be more tailored towards specific therapeutic targets and potentially more effective if we had a better understanding of the factors contributing to functional impairment and delaying recovery. Informing tailored interventions could further benefit from exploring impairment in specific functional domains and the correlates of difficulties in these domains.

Depressive symptoms and cognitive deficits are the only variables consistently showing an association with functional outcomes in euthymic patients (Gitlin and Miklowitz, 2017). Residual depressive symptoms affect multiple functional domains, including occupational outcomes (Gilbert and Marwaha, 2013; O'Donnell et al., 2017) and poor quality of life (Bonnín et al., 2019). Several studies have suggested that cognition may also predict psychosocial functioning (Ehrminger et al., 2019; López-Villarreal et al., 2020), with verbal memory and executive functions found to be significantly associated with employment outcomes (Tse et al., 2014). Although the impact of these factors on functional outcomes is well-supported, more evidence is required on the relationship between these two variables in association with functioning. Studies have suggested a mediating role of cognition on the effect of residual depressive symptoms on functioning (Bonnin et al., 2014) or relative independence of mood symptoms (Roux et al., 2018).

Reviews also suggest that other factors contribute to functioning, including clinical course and illness severity characteristics (Léda-Rêgo et al., 2020), but findings are inconsistent between studies, probably due to the variety of factors considered and the choice of analytic methods. Most studies select a priori certain candidate predictors without considering others or test multivariable linear models which creates multiple comparison problems and increases the risk of model overfitting (i.e., high variance, low predictive value). A potential solution is using advanced statistical procedures, such as regularized regression (Zou and Hastie, 2005), on a larger number of potential predictors in order to select factors to be further investigated, as these approaches allow the inclusion of multiple predictors (e.g., sociodemographic, clinical, and cognitive variables) without the limitations of power and overfitting.

This study will a) characterize functional difficulties in euthymic patients with bipolar disorder with reference to global functioning and specific functional domains of the FAST, and b) identify factors associated with the FAST total and different domain scores using a model selection approach. We anticipated replicating the well-documented association with residual depressive symptoms and cognition, as well as identifying further factors associated with psychosocial functioning.