Using a novel stress resilience score to construct and investigate self-efficacy networks in high compared to low resilient adults

The following paper has been published on this research project:

Schueler, K., Fritz, J., Dorfschmidt, L., Van Harmelen, A. L., Strömer, E., & Wessa, M. (2021). Psychological Network Analysis of General Self-Efficacy in High vs. Low Resilient Functioning Healthy Adults. Frontiers in Psychiatry. https://doi.org/10.3389/fpsyt.2021.736147

Psychological network analysis has been widely used to model symptom networks of psychiatric disorders with promising results for a better understanding of psychopathology (Borsboom et al., 2017). Only few studies exist which transfer this method to the concept of resilience to stress-related psychopathology. Resilience is the adaptive process that leads to maintenance or quick recovery after stress exposure (Kalisch et al., 2017). Resilience factors, such as general self-efficacy, are psychological (or other) constructs contributing to resilience (Kalisch et al., 2017). So far, research approaches mainly include resilience as a dynamic property of psychopathology symptom networks (PSN) (Borsboom et al., 2017) or networks of resilience factors (RFN) after adversity exposure (Fritz et al., 2018). However, it remains unknown how RFN might potentially differ with respect to resilience itself.
The network structure and connectivity of resilience factors might provide important information about resilience after stress exposure. In dynamic PSN, Cramer et al. (2016) stated that stronger network connectivity is associated with an increase in synchronized activation of symptoms resulting in a stable activation pattern that is less likely to reverse. This leads to mutual reinforcement of psychopathology symptoms. This mechanism might actually be beneficial in the case of resilience. Additionally, connectivity differences of cross-sectional RFN might contribute to an enhancement of resilience factors after childhood adversity (Fritz et al., 2018). Thus, increased connectivity of RFN might potentially play a role in the development and maintenance of resilience.

A promising way to quantify resilience as such has been applied for example by van Harmelen and colleagues (2016). They used principal components analysis to extract latent scores of both childhood adversity and psychosocial functioning and related them to each other using regression analysis. Resilience scores (“resilient functioning scores”) were the regression residuals.

This study aims at investigating properties of a network model of self-efficacy with respect to a novel resilience score indicating high versus low individual levels of resilience. Adapting the idea of connectivity differences in dynamic PSN (Cramer et al., 2016) we expect the self-efficacy network to differ in high resilient compared to low resilient individuals. We hypothesize that high resilient individuals show stronger positive connectivity between self-efficacy items than low resilient individuals.

To address this, we firstly employ a novel methodology to compute resilience scores in a large cross-sectional sample of approximately 850 adults. These resilience scores indicate whether one’s mental health is better or worse or exactly as expected given individual stress exposure. We use partial-least squares (PLS) regression (Seidlitz, et al, 2018; Vértes et al. 2016; Whitaker et al., 2016) to predict an expected level of mental health by stress exposure and extract resilience scores as regression residuals. An established measure of resilient functioning (van Harmelen et al., 2017) will be used to validate our PLS regression approach. All data are self-report measures.
As a second step, we split our sample and explore the structure of self-efficacy network models of high vs. low resilient subjects. For the network analysis of both networks, we focus on global connectivity indices - e.g. degree, strength, expected influence, shortest path length, global efficiency - and additionally explore the network structure.

Results might provide a useful way to quantify resilience and help to better understand the nature of self-efficacy with respect to resilience itself.

以下论文已发表于本研究项目：

Schueler, K., Fritz, J., Dorfschmidt, L., Van Harmelen, A. L., Strömer, E., & Wessa, M. (2021). 高弹性与低弹性健康成年人一般自我效能的心理网络分析. 前沿精神病学. https://doi.org/10.3389/fpsyt.2021.736147

心理网络分析已被广泛用于构建精神障碍症状网络模型，并取得了对心理病理学更深入了解的令人鼓舞的结果（Borsboom et al., 2017）。仅有少数研究将此方法应用于与压力相关心理病理学概念的弹性。弹性是指应激暴露后维持或快速恢复的适应性过程（Kalisch et al., 2017）。弹性因素，如一般自我效能，是促进弹性的心理（或其他）结构（Kalisch et al., 2017）。迄今为止，研究方法主要将弹性视为心理病理学症状网络（PSN）的动态属性（Borsboom et al., 2017）或在逆境暴露后的弹性因素网络（RFN）的动态属性（Fritz et al., 2018）。然而，RFN可能如何与弹性本身相区别尚不清楚。

弹性因素的网络结构和连通性可能提供了关于应激暴露后弹性的重要信息。在动态PSN中，Cramer等（2016）指出，更强的网络连通性与症状同步激活的增加有关，从而产生一种稳定的激活模式，这种模式不太可能逆转。这导致了心理病理学症状的相互强化。这种机制实际上可能在弹性的情况下是有益的。此外，横截面RFN的连通性差异可能有助于在童年逆境后增强弹性因素（Fritz et al., 2018）。因此，RFN的连通性增加可能在弹性的发展和维持中发挥作用。

量化弹性的一个有希望的方法已被van Harmelen及其同事（2016）应用。他们使用主成分分析提取童年逆境和心理健康功能的潜在得分，并使用回归分析将它们相互关联。弹性得分（“弹性功能得分”）是回归残差。

本研究旨在调查自我效能网络模型相对于一种新的弹性得分（表示高弹性与低弹性个体水平的差异）的性质。借鉴动态PSN中连通性差异的想法（Cramer et al., 2016），我们预计自我效能网络在高弹性个体与低弹性个体之间会有所不同。我们假设高弹性个体在自我效能项之间的正向连通性比低弹性个体更强。

为了解决这个问题，我们首先采用一种新颖的方法来计算大约850名成年人大型横断面样本中的弹性得分。这些弹性得分表示在个体应激暴露下，一个人的心理健康状况是更好、更差还是正如预期的那样。我们使用偏最小二乘回归（PLS）（Seidlitz, et al, 2018; Vértes et al. 2016; Whitaker et al. 2016）来预测应激暴露下的预期心理健康水平，并将弹性得分作为回归残差提取。我们将使用一个已建立的弹性功能衡量标准（van Harmelen et al., 2017）来验证我们的PLS回归方法。所有数据都是自我报告的。

作为第二步，我们将样本分割并探索高弹性与低弹性个体自我效能网络模型的构成。对于两个网络的网络分析，我们关注全局连通性指标，例如度、强度、预期影响、最短路径长度、全局效率，并进一步探索网络结构。

结果可能提供了一种量化弹性的有用方法，并有助于更好地理解自我效能与弹性本身的本质。