Depressive disorders are highly disabling, prevailing as one of the three leading causes of non-fatal health loss for nearly three decades (The Institute for Health Metrics and Evaluation, 2017). Despite extensive research, the etiology of depression is not fully understood. Epidemiologically, depression is highly comorbid with anxiety (Brady and Kendall, 1992; Lewinsohn et al., 1994) with 50 % of participants with Major Depressive Disorder having anxiety (Fava et al., 1997). Another important contributing factor is personality traits, categorized as neuroticism (tendency to experience negative emotions), extraversion(−introversion) (quantity and intensity of interpersonal interactions and positive emotions), openness to experience (appreciation of experience), agreeableness (prosocial orientation toward others), and conscientiousness (organization, motivation, and persistence in achieving goals) (McCrae and Costa, 2008, McCrae and Costa, 2013). Previous research showed that high neuroticism, introversion, and low conscientiousness predicted depression and anxiety (Bienvenu et al., 2004; Boudouda and Gana, 2020; Clark et al., 1994; Duberstein et al., 2008; Enns and Cox, 1997; Hakulinen et al., 2015; Hayward et al., 2013; Karsten et al., 2012; Kendler et al., 2006; Koorevaar et al., 2013; Kotov et al., 2010; Lyon et al., 2020).

However, research to date has used either a categorical or dimensional approach when assessing the relationships between personality traits and depression. Because the categorical analytical approach examines a single personality dimension at a time, it cannot assess multiple relationships between personality traits and psychopathology across a severity spectrum. To avoid this limitation, it is important to evaluate how the collective personality profile is linked to these disorders. Furthermore, prior dimensional research has not examined the distinctive relationships between depression and anxiety in relation to personality traits (Bienvenu et al., 2001, Bienvenu et al., 2004, Bienvenu et al., 2007; Griffith et al., 2010; Karsten et al., 2012; Kotov et al., 2007, Kotov et al., 2010; Lyon et al., 2020; Nikčević et al., 2021) and how these relationships change across development. Finally, additional factors such as cognition, behavioral constructs and physical measures should be accounted for. Such an investigation can be performed using machine learning approaches.

This study sought to address these above limitations using a large community sample, the Nathan Kline Institute Rockland Sample (NKI-RS), with deep phenotypic data collected across the lifespan and carrying out novel analytical approaches including machine learning. The wide age range in this cohort (ages 12–85) enabled us to examine depression, anxiety, and personality traits across different developmental stages while studying both categorical (DSM diagnoses) and dimensional approaches (symptom severity measures). Our aims were: 1) to compare the five personality dimensions across the lifespan in healthy controls (HC) and among three categorical diagnostic groups: depression, anxiety, comorbid depression and anxiety; 2) to investigate personality traits in relation to depression and anxiety symptoms in adolescence, adulthood, and senescence, and; 3) to train a machine learning model to predict depression using personality traits and various behavioral assessments, cognitive tasks, and physical measurements shown to be related to depression. We hypothesized that: 1) compared to HC, depression and anxiety would share the same personality traits while comorbid depression-anxiety would be associated with increased personality vulnerability; 2) common and unique associations between personality traits and depressive and anxiety symptoms would be present and would be age-dependent (Eysenck and Fajkowska, 2018); 3) a machine learning model can predict depression with personality constituting a major contributing factor.