University of Social Sciences and Humanities, Vietnam National University, 10-12 Dinh Tien Hoang street, Ben Nghe Ward, District 1, Ho Chi Minh City, Vietnam.
Institute of Psychology, Eötvös Loránd University, Izabella utca 46, Budapest 1064, Hungary.
Psychiatry Res. 2017 Sep;255:173-185. doi: 10.1016/j.psychres.2017.05.019. Epub 2017 May 19.
To fully understand the dimensionality of an instrument in a certain population, rival bi-factor models should be routinely examined and tested against oblique first-order and higher-order structures. The present study is among the very few studies that have carried out such a comparison in relation to the Symptom Checklist-90-R. In doing so, it utilized a sample comprising 2593 patients with substance use and impulse control disorders. The study also included a test of a one-dimensional model of general psychological distress. Oblique first-order factors were based on the original a priori 9-dimensional model advanced by Derogatis (1977); and on an 8-dimensional model proposed by Arrindell and Ettema (2003)-Agoraphobia, Anxiety, Depression, Somatization, Cognitive-performance deficits, Interpersonal sensitivity and mistrust, Acting-out hostility, and Sleep difficulties. Taking individual symptoms as input, three higher-order models were tested with at the second-order levels either (1) General psychological distress; (2) 'Panic with agoraphobia', 'Depression' and 'Extra-punitive behavior'; or (3) 'Irritable-hostile depression' and 'Panic with agoraphobia'. In line with previous studies, no support was found for the one-factor model. Bi-factor models were found to fit the dataset best relative to the oblique first-order and higher-order models. However, oblique first-order and higher-order factor models also fit the data fairly well in absolute terms. Higher-order solution (2) provided support for R.F. Krueger's empirical model of psychopathology which distinguishes between fear, distress, and externalizing factors (Krueger, 1999). The higher-order model (3), which combines externalizing and distress factors (Irritable-hostile depression), fit the data numerically equally well. Overall, findings were interpreted as supporting the hypothesis that the prevalent forms of symptomatology addressed have both important common and unique features. Proposals were made to improve the Depression subscale as its scores represent more of a very common construct as is measured with the severity (total) scale than of a specific measure that purports to measure what it should assess-symptoms of depression.
要全面理解某一特定人群中工具的维度，应定期检查并比较斜交一阶和高阶结构的竞争双因素模型。本研究是为数不多的对《症状清单-90-R》进行此类比较的研究之一。在这样做的过程中，它利用了一个包含 2593 名物质使用和冲动控制障碍患者的样本。该研究还包括对一般心理困扰的一维模型的测试。斜交一阶因素基于 Derogatis（1977）提出的原始的 9 维先验模型；以及 Arrindell 和 Ettema（2003）提出的 8 维模型——广场恐怖症、焦虑、抑郁、躯体化、认知表现缺陷、人际敏感和不信任、行为敌意、睡眠困难。以个体症状为输入，在二阶水平上测试了三个高阶模型，分别为：（1）一般心理困扰；（2）“恐慌伴广场恐怖症”、“抑郁”和“过度惩罚行为”；或（3）“烦躁敌对抑郁”和“恐慌伴广场恐怖症”。与之前的研究一致，没有发现支持单因素模型的证据。与斜交一阶和高阶模型相比，双因素模型被发现最适合数据集。然而，从绝对意义上讲，斜交一阶和高阶因子模型也能很好地拟合数据。高阶解决方案（2）支持了 R.F. Krueger 区分恐惧、痛苦和外化因素的心理病理学实证模型（Krueger，1999）。将外化和痛苦因素（烦躁敌对抑郁）结合起来的高阶模型（3）在数值上同样很好地拟合数据。总的来说，研究结果被解释为支持这样一种假设，即所涉及的常见症状形式具有重要的共同和独特特征。有人提议改进抑郁分量表，因为其分数代表的是更常见的结构，而不是特定的衡量标准，即试图衡量它应该评估的内容——抑郁症状。
