Review
Self-reported cognitive biases in depression: A meta-analysis

https://doi.org/10.1016/j.cpr.2020.101934Get rights and content

Highlights

  • We examined the magnitude of self-reported cognitive biases in depression.

  • Sixty-three studies published between 1979 and 2017 were included (N = 9543 participants).

  • Meta-analyses included studies of individuals grouped by either a diagnostic category or a dimensional measure depression.

  • The most consistent evidence of biases was found for catastrophizing and interpretation bias.

  • There were few evidences of the specificity of cognitive biases in depression.

Abstract

Despite the influence of Beck's cognitive models of depression, the presence and magnitude of the specific proposed cognitive biases have not been systematically investigated. After a systematic search in PsycInfo and PubMED, studies reporting self-reported outcomes on cognitive biases and depressive symptoms in depressed and/or healthy groups were included. From a total of 4840 records, two different meta-analyses were conducted. 23 studies on 4865 participants provided data about catastrophising and depression (g = 0.95, 95% CI [0.64; 1.26]) and 40 studies on 4678 participants provided data about interpretation bias in depression (g = 0.78, 95% CI [0.43; 1.13]). Moderation analyses showed that the relationship between catastrophising and depression was higher in studies with more women, when the corresponding author was from a Western country, and when the instrument to measure depression was the DSM criteria, the SCL-90, the BDI, or the DASS. The relationship between interpretation bias and depressive symptoms was significant only in studies comparing depressed and healthy groups, and when using specific instruments to measure symptoms (DSM/RDC criteria plus a scale cut-off score) and cognitive bias (CDQ/CBQ, SCT, AST-D, other). Some limitations are acknowledged, but risk of publication bias was found to be low, and these results support the utility of some self-reported measures of cognitive biases in depression.

Introduction

Depression is considered one of the leading causes of disability in the world and is associated with great social and economic costs (Whiteford et al., 2013). More than 200 million people suffer from major depression (World Health Organization, 2017) and the relapse rate is around 85% (McIntyre & O'Donovan, 2004). Given the impact of this disorder, it is important to understand the variables that influence its development, maintenance and recurrence. Several theories have been proposed to explain the causes of this disorder (Gotlib & Hammen, 2014). Among them, cognitive theories have been very influential both in research and applied fields. These theories point out to dysfunctional thinking as a key causal factor related to the onset and maintenance of these emotional disorders (Barlow et al., 2011; Beck, 1967, Beck, 1976).

Beck's seminal cognitive model of depression (1967, 1976) provided some of the concepts and explanatory heuristics that have been incorporated in most of the current cognitive models of depression. According to this model (see Fig. 1), individuals may develop, early in life, latent cognitive schemas that get activated when facing environmental stressors that are pertinent to the contents of those schemas (e.g., themes of loss). Cognitive schemas are rather abstract representations of the world and determine the way in which information is processed and how events and stimuli are interpreted in a given context (Dalgleish & Power, 2000). When the negative contents of these cognitive schemas are triggered by internal or external events, psychological processes (i.e., memory, interpretation, and attention) operate following negative cognitive biases, like dichotomous thinking or arbitrary inferences (see Table 1). These processes then lead to biased mental products or thoughts about the self, the world and the future (i.e. the cognitive triad). It is important to note the difference between cognitive schemas and cognitive biases. Following Ingram and Kendall's (1986) cognitive taxonomy, cognitive schemas would be structural variables of the system (i.e., broad cognitive frames through which information is filtered, represented and organized), whereas cognitive biases would function as operational variables, or mechanisms, by which cognitive structures work. Although schemas would be the most distal causes of depression (Panzarella, Alloy, & Whitehouse, 2006), biased cognitive operations would also have a critical causal role as they are the action mechanisms used to support and validate the schemas. Overall, this cognitive machinery might be considered as an antecedent of the presence of negative automatic thoughts, negative beliefs (about oneself, the world and/or the future), and ultimately depressive symptoms, which would all be the products (i.e., the tangible outputs) with which clinicians typically work with their clients.

Beck's theoretical account has had a profound impact on both clinical and research fields. For instance, cognitive-behavior therapy (CBT) for depression follows the rationale of the cognitive model (Beck, Rush, Shaw, & Emery, 1979; Greenberg & Padesky, 2015). One of the main aims in CBT is to modify biased thinking (Beck, 1976) as a path to improve symptoms such as negative affect and anhedonia (Joorman & Gotlib, 2010). Also, extensive research on cognition and depression has analysed the relationships between these constructs and depressive symptoms in a variety of samples regarding culture, gender, age and diagnostic severity (e.g. Hu et al., 2015; Cristea et al., 2015; Vîslă, Flückiger, Grosse Holtforth, & David, 2016). Moreover, Beck's proposal has inspired the creation of widely used instruments to measure both symptoms (e.g., the Beck Depression Inventory-II, BDI-II, Beck, Steer, & Brown, 1996) and cognitive constructs like the Cognitive Errors Questionnaire (CEQ, Lefebvre, 1981) or the Cognitive Bias Questionnaire (CBQ, Krantz & Hammen, 1979). In sum, Beck's model has been extraordinarily influential both in research and applied fields in clinical psychology.

Beck's original theory has been further developed during recent decades. Empirical research has supported and refined the model by clarifying the nature of stressors contributing to form early negative schemas (Hammen, 2005), distinguishing different subtypes of schemas (Clark & Beck, 1991), or lately adding new mechanistic processes, like rumination (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008) or overgeneral autobiographical memory (Williams et al., 2007), that have been incorporated in the depression literature for some decades after Beck's initial formulation. Likewise, there have been efforts to inquire into the neurological basis of Beck's theoretical account (Beck, 2008). For instance, researchers have found that cognitive biases seem to be characterized by a maladaptive bottom-up system at the subcortical level, which is reinforced by an attenuated cortical cognitive control unable to regulate them (Disner, Beevers, Haigh, & Beck, 2011).

Despite the relevance of Beck's cognitive model of depression, the specific cognitive biases proposed in it (see Table 1) have not been subjected to a systematic review, except for causal attribution bias (Hu, Zhang, & Yang, 2015). Although the model is still inspiring research on cognitive processes (Gotlib & Joormann, 2010) and has become one of the best validated and most frequently used therapeutic interventions (Cuijpers, Noma, Karyotaki, Cipriani, & Furukawa, 2019), the validity of the cognitive biases explicitly formulated in the model is relatively unknown.

Whereas some meta-analyses have approached the study of cognitive constructs such as irrational beliefs, dysfunctional thinking, or automatic thoughts (Cristea et al., 2015; Vîslă et al., 2016), the magnitude of most cognitive biases in depressed individuals has not been synthesised in the literature. Also, it remains unclear whether these processes are exclusive for depressed individuals, which makes it important to study their relationship with depressive symptoms itself (dimensional studies) but also to compare different types of populations (categorical studies). Thus, the aim of this study was to conduct a meta-analysis of those cognitive biases of depression, as specified in Beck's cognitive model, that have not been meta-analysed in previous studies (i.e. those presented in Table 1 except for Personalization/Internal causal attributions). Random-effect models were used (since sampling variability was expected, Riley, Higgins, & Deeks, 2011) to meta-analyse each cognitive bias, using the standardized mean difference with Hedge's correction (g) as the effect size (ES). The meta-analysis was pre-registered in PROSPERO (CRD42018115365). The first hypothesis was that there would be significantly larger levels of cognitive biases in depression groups than in other comparison groups, such as healthy participants, or individuals with other psychological symptoms, or subclinical levels of depression. Also, it was expected that several variables could moderate this effect. The year of the study and geographic location of the corresponding author were coded to study potential spatial and temporal effects. Based on previous findings, sample characteristics such as age (Reed, Chan, & Mikels, 2014), gender (Kessler & Bromet, 2013), or type of sample (Clark, Beck, & Alford, 1999) were investigated although the direction of these potential moderation effects was not anticipated. Finally, methodological variables such as sample size, type of measure, and the psychometric characteristics of the instruments were also included as potential moderators given previous meta-analytic results (Everaert, Podina, & Koster, 2017).

Section snippets

Eligibility and search criteria

A systematic search was conducted on the databases of PsycINFO and PubMed, until February 2020, combining terms related to the spectrum of depression (depress* OR dysphor* OR mood OR “affective disorder” OR “sad mood” OR sadness), comparison groups based on DSM-5 (American Psychology Association, 2013) categories and type of population (delirium OR dementia OR “neurodevelopmental disorder” OR schizophrenia OR “psychotic disorder” OR “delusional disorder” OR “bipolar disorder” OR “anxiety

Summary measures

Random-effects meta-analyses for all the included cognitive biases were conducted, in SPSS 20 and R 3.5.0 (metafor package Viechtbauer, 2010), using the standardized mean difference (d= X̅1-X̅2/Spooled) with Hedge's correction (g = c(m)*d)) as the effect size (ES). Positive values reflect a higher level of cognitive bias in the experimental group compared to the control group. Hedge's g values can be categorized as small (0.2–0.5), medium (0.5–0.8), or large (>0.8) (Cohen, 1988). All studies

Study selection

The process of selection and inclusion of studies is shown in Fig. 2. From a total of 4840 records (1320 duplicates), 3131 were excluded based on the screening of the title or abstract, while 461 were excluded after a full-text reading. The main reason for exclusion was the lack of a measure of bias (i.e., many studies were focused on constructs different from cognitive biases, such as cognitive schemas or automatic thoughts -see Fig. 1). Many studies were also excluded because they did not use

Overall effect sizes

Effect sizes for catastrophizing bias and interpretation bias (see Fig. 3 and Fig. 4) were large and moderate (g = 0.95, p < 0.001 and g = 0.78, p < 0.001, respectively). Heterogeneity was significant and high in all cases, with the I2 value around 90% (see Table 2, Table 3). Significant effect sizes were maintained after the removal of 6 outliers for the catastrophising bias analysis, and 13 for interpretation bias. These sensitivity analyses also showed a reduction in heterogeneity. Details

Discussion

The aim of this study was to quantify the evidence of self-reported cognitive biases in depressed individuals as compared to other groups of participants. Based on the relatively scarce amount of empirical studies found in our search (k = 63), analyses only included two categories of bias (catastrophizing and interpretation bias) and comparisons between groups of depressive participants and non-symptomatic groups.

Given that the search covered a large period of time (no limit-2020), it seems

Author Biography

Inés Nieto is a Ph.D. student at Complutense University. She is currently finishing her dissertation on cognitive modification bias in emotional disorders.

Elena Robles is a Ph.D. student at Complutense University. She is currently doing her dissertation on selective attention biases in chronic pain.

Carmelo Vazquez, pH.D. is a Professor of Psychopathology at Complutense University. He has done extensive experimental psychopathology research and has published review papers and meta-analyses on

Declaration of Competing Interest

None.

Acknowledgements

This work was partially supported by MINECO PSI2015-69253-R and Ministry of Science and Innovation PID2019-108711GB-I00 grants to CV and CT17/17-CT18/17 Complutense University predoctoral fellowhip to IN. We also thank Jamie O'Grady for his help in editing the paper and Jonas Everaert for his technical advice when conducting this meta-analysis.

References (67)

  • G.E. Matt et al.

    Mood-congruent recall of affectively toned stimuli: a meta-analytic review

    Clinical Psychology Review

    (1992)
  • S. Samtani et al.

    Assessing maladaptive repetitive thought in clinical disorders: A critical review of existing measures

    Clinical Psychology Review

    (2017)
  • H.A. Whiteford et al.

    Global burden of disease attributable to mental and substance use disorders: Findings from the global burden of disease study 2010

    The Lancet

    (2013)
  • World Health Organization

    Depression and other common mental disorders: global health estimates

  • M. Appelbaum et al.

    Journal article reporting standards for quantitative research in psychology: The APA publications and communications board task force report

    American Psychologist

    (2018)
  • D.H. Barlow et al.

    The unified protocol for transdiagnostic treatment of emotional disorders: Therapist guide

    (2011)
  • A.T. Beck

    Thinking and depression: I. idiosyncratic content and cognitive distortions

    Archives of General Psychiatry

    (1963)
  • A.T. Beck

    Depression: Clinical, experimental, and theoretical aspects

    (1967)
  • A.T. Beck

    Cognitive therapy and the emotional disorders

    (1976)
  • A.T. Beck

    The evolution of the cognitive model of depression and its neurobiological correlates

    American Journal of Psychiatry

    (2008)
  • R. Beck et al.

    Cognitive content-specificity for anxiety and depression: A meta-analysis

    Cognitive Therapy and Research

    (2001)
  • A.T. Beck et al.

    Cognitive therapy of depression

    (1979)
  • A.T. Beck et al.

    Manual for the Beck depression inventory-II

    (1996)
  • M. Borenstein et al.

    Introduction to meta-analysis

    (2009)
  • R.B. Burns

    Feeling good: The new mood therapy

    (1980)
  • S.W. Choi et al.

    Establishing a common metric for depressive symptoms: linking the BDI-II, CES-D, and PHQ-9 to PROMIS depression

    Psychological Assessment

    (2014)
  • D.A. Clark et al.

    Personality factors in dysphoria: a psychometric refinement of Beck’s Sociotropy-autonomy scale

    Journal of Psychopathology and Behavioral Assessment

    (1991)
  • D.A. Clark et al.

    Scientific foundations of cognitive theory and therapy of depression

    (1999)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1988)
  • P. Cuijpers et al.

    Effectiveness and acceptability of cognitive behavior therapy delivery formats in adults with depression: A network meta-analysis

    JAMA Psychiatry

    (2019)
  • S.G. Disner et al.

    Neural mechanisms of the cognitive model of depression

    Nature Reviews Neuroscience

    (2011)
  • Cited by (24)

    • Early indicators of vulnerability to depression: The role of rumination and heart rate variability

      2022, Journal of Affective Disorders
      Citation Excerpt :

      Given these premises, identifying factors and mechanisms that elevate the risk for depression, especially in young adults, is of fundamental importance for effectively preventing and early identifying depression. Several features characterizing the risk of developing major depressive disorder (MDD) have been recognized, such as some personality traits (Malinowski et al., 2017; Sedlinská et al., 2021; Takahashi et al., 2021), blunted neural response to rewards (Messerotti Benvenuti et al., 2019; Moretta et al., 2021; Weinberg et al., 2015), and dysfunctional cognitive biases (Keller et al., 2019; Murrough et al., 2011; Nieto et al., 2020; Platt et al., 2017). When studying vulnerability factors for depression as possible targets for early identification and prevention programs, an important question is whether these factors can help answer why depression is heritable.

    • Disentangling the mediating role of modifying interpretation bias on emotional distress using a novel cognitive bias modification program

      2021, Journal of Anxiety Disorders
      Citation Excerpt :

      As Fried (2017) recently pointed out, the standard instruments used to assess problems like depression poorly reflect the wide variety of symptoms that characterize the disorders, which limits the validity of research results. Similarly, regarding cognitive biases, a recent meta-analysis has also found that there is a high variety of self-report instruments and unfortunately, many of them have limited evidence to support their psychometric robustness (Nieto et al., 2020). Dysfunctional attitudes significantly changed after the CBM-IClin, which was explained by both a direct effect of the intervention and an indirect effect through a change in interpretation bias.

    • The relationship between conventional clinical assessments and momentary assessments of symptoms and functioning in schizophrenia spectrum disorders: A systematic review

      2021, Schizophrenia Research
      Citation Excerpt :

      If an individual is experiencing depressive symptoms during the clinic visit interview, they may overreport depressive symptoms over the past week, overlooking days in which their mood was more positive (Urban et al., 2018). This cognitive bias of overgeneralization (i.e., to see a single negative event as representative of all events) (Nieto et al., 2020) is a commonly described phenomena in depression (Weeks et al., 2017) and an example of how a single, cross-sectional measure may impact assessment validity. Momentary assessment mitigates this by asking individuals to report their symptoms in the moment, rather than recalling earlier symptoms.

    View all citing articles on Scopus
    View full text