Developing mixed-latent-state-trait-models to analyse the temporal characteristics of affect and self-esteem in borderline personality disorder

点击翻译按钮获取中文摘要

Affective instability is hypothesized to be the core pathology in patients with borderline personality disorder (BPD) and the ICD-10 even lists BPD under the category of emotionally unstable personality disorders. Surprisingly, recent work using e-diaries to assess affect over time did not reveal affective instability as a disorder-specific mechanism, as other disorders did impress by similar instability. Accordingly, it has been hypothesized that statically modeling the dynamic interplay between affect and self-esteem is necessary to discriminate between transdiagnostic and disorder-specific mechanisms. In more details, to better understand affective instability in daily life in patients with BPD we have to dynamically model affect and self-esteem over time, taking additionally into account affect-changing events, reactivity to these events, as well as specific affect regulation mechanisms, like self-mutilating behavior. Therefore statistical models are necessary which allow: a) decomposing random from true variability, b) modeling multiple subcomponents of affective processes at the same time, c) modeling between-subject and group-specific differences in affective response and regulation processes, as well as d) enabling estimations on psychometric properties and the appropriateness of the sampling strategies used.Accordingly our grant proposal follows two main strategies: The clinical goal is to simultaneously model the dynamical characteristics of affect and self-esteem including their interplay and specific regulation mechanisms. We will use electronic diaries to assess affect, self-esteem, and specific regulation strategies in everyday life in 125 patients with BPD, 125 patients with anxiety disorders, as well as 125 healthy controls. We do hypothesize that the covariation of affect and self-esteem is specific for BPD as well as that self-mutilating behavior is predicted by a synchronous increasing negative affect and decreasing self-esteem. As current statistical models are limited for this kind of analyses, our second main goal is to develop and systematically test advanced statistical models. In details, we will expand ordinary mixed-latent-state-trait-models to multigroup-multitrait-mixed-autoregressive-latent-state-trait-growth-curve-models. Beyond that, we will use Monte Carlo simulations studies to investigate under which preconditions these models are fruitful usable in ambulatory assessment research.