Smári J, Thor Olason D
Faculté des Sciences Sociales, Université d'Islande, Reykjavík, Islande.
Encephale. 2005 Mar-Apr;31(2):144-51. doi: 10.1016/s0013-7006(05)82381-4.
The Obsessive Thoughts Checklist (OTC) differs from several other measures of obsessive-compulsive symptoms in its focus on obsessive thoughts instead of compulsive behaviour. The OTC has been used in several studies in France and abroad and support for the discriminant and convergent validity of the instrument has been gathered. The authors of the OTC recently reported 3 underlying factors in this instrument: a perfectionism/verification factor, a contamination factor and a responsibility factor. In an earlier study of the OTC a 2 factor solution was however suggested. It therefore seems important to further elucidate the factorial structure of this instrument using confirmatory factor analysis. In this study data on the Icelandic translation of the OTC from three samples of Icelandic college students were submitted first to an exploratory and then to a confirmatory factor analysis. The total number of subjects was 614 college students, 254 men and 360 women with a mean age of 24.4 (sd = 5.0). The results of the exploratory factor analysis (PCA) submitted to a varimax rotation are presented in table I. The 3 expected factors were reproduced with few cross-loadings. In the confirmatory factor analysis the fit of three models to the data were evaluated: a 1 factor model, the 2 factor model of Bouvard et al. and the 3 factor model of Bouvard et al.. An initial examination of the data led to logarithmic transformation of 18 items to reduce skewness in their distributions. The data was subsequently subjected to a confirmatory factor analysis to compare the three-factor model with the two-factor and one-factor models for the OTC. The factors for the three- and two-factor models were allowed to correlate freely. The data were analysed using the EQS procedure, and the models tested were covariance structure models. Table II presents the goodness of fit indices for all three models. The results show that none of the three models provide an overall appropriate fit for the data. However, the fit indices for the three-factor model were considerably higher than found for the two or one factor models and the RMSEA index for the three-factor model suggested an acceptable fit for that model. Although the three-factor model suggests the best fit of all three models, the fit indices were still unacceptably low. Further examination of the data revealed a pattern of standardized residuals suggesting that this might in part be attributable to three items from the responsibility factor (items 26, 19 and 15) not being well specified within the model. When the residuals for these items were allowed to correlate, the fit of the model was substantially improved (CFI = 0.85; GFI = 0.87; AGFI = 0.85; RMSEA = 0.062). This indicates that a revision of the responsibility scale might be appropriate. Table III provides the means, standard deviations and the alpha coefficients for the 3 subscales of the OTC as well as for the total scale. In one subsample of the study (sample 1, n = 169) the OTC was administered together with the Padua Inventory-Washington State Revision (PI-WSUR) measuring obsessive-compulsive symptoms, the Penn State Worry Questionnaire (PSWQ) and the Community Epidemiological Scale-Depression (CES-D) measuring depression. In order to investigate the convergent validity and divergent of the OTC its correlation with the PI-WSUR was compared with its correlations with PSWQ and CES-D. These correlations shown in table IV support the convergent and divergent validity of the OTC. In another subsample of the study (sample 2, n = 296) the OTC was administered together with the Maudsley Obsessive Compulsive Inventory (MOCI). For samples 1 and 2, zero order and partial correlations were calculated between the subscales of the OTC and the subscales of the other instruments. As shown in table V the strongest correlations between the checking/perfectionism and the contamination subscales of the OTC were with corresponding subscales of the PI-WSUR and the MOCI. It is concluded that the factorial, the convergent and the divergent validities of the Icelandic translation of the OTC are supported in a student population even though the somewhat suboptimal fit of the three-factor model may indicate that a revision of the responsibility factor might be in order. This should however be further studied in a clinical population.
强迫观念检查表（OTC）与其他几种强迫症症状测量方法不同，它关注的是强迫观念而非强迫行为。OTC已在法国及其他国家的多项研究中使用，并收集到了该工具判别效度和聚合效度的支持证据。OTC的作者最近报告了该工具中的3个潜在因素：完美主义/核查因素、污染因素和责任因素。然而，在早期对OTC的一项研究中提出了一个双因素解决方案。因此，使用验证性因素分析进一步阐明该工具的因素结构似乎很重要。在本研究中，来自冰岛大学生三个样本的OTC冰岛语翻译数据首先进行探索性因素分析，然后进行验证性因素分析。受试者总数为614名大学生，其中254名男性和360名女性，平均年龄为24.4岁（标准差=5.0）。表I展示了进行方差最大化旋转的探索性因素分析（主成分分析）结果。3个预期因素得以重现，交叉负荷较少。在验证性因素分析中，评估了三个模型与数据的拟合度：单因素模型、Bouvard等人的双因素模型以及Bouvard等人的三因素模型。对数据的初步检查导致对18个项目进行对数转换，以减少其分布中的偏度。随后对数据进行验证性因素分析，以将三因素模型与OTC的双因素和单因素模型进行比较。三因素模型和双因素模型的因素被允许自由相关。使用EQS程序分析数据，测试的模型为协方差结构模型。表II展示了所有三个模型的拟合优度指标。结果表明，这三个模型均未对数据提供整体合适的拟合。然而，三因素模型的拟合指标远高于双因素模型或单因素模型，且三因素模型的RMSEA指标表明该模型拟合可接受。尽管三因素模型在所有三个模型中显示出最佳拟合，但拟合指标仍然低得令人无法接受。对数据的进一步检查揭示了标准化残差模式，表明这可能部分归因于责任因素中的三个项目（项目26、19和15）在模型中未得到很好的界定。当允许这些项目的残差相关时，模型的拟合得到了显著改善（比较拟合指数=0.85；拟合优度指数=0.87；调整拟合优度指数=0.85；近似误差均方根=0.062）。这表明责任量表的修订可能是合适的。表III提供了OTC三个子量表以及总量表的均值、标准差和阿尔法系数。在该研究的一个子样本（样本1，n = 169）中，OTC与测量强迫症症状的帕多瓦量表 - 华盛顿州修订版（PI - WSUR）、宾夕法尼亚州立大学忧虑问卷（PSWQ）以及测量抑郁的社区流行病学抑郁量表（CES - D）一起施测。为了研究OTC的聚合效度和区分效度，将其与PI - WSUR的相关性与其与PSWQ和CES - D的相关性进行了比较。表IV中的这些相关性支持了OTC的聚合效度和区分效度。在该研究的另一个子样本（样本2，n = 296）中，OTC与莫兹利强迫症量表（MOCI）一起施测。对于样本1和样本2，计算了OTC子量表与其他工具子量表之间的零阶相关和偏相关。如表V所示，OTC的检查/完美主义和污染子量表与PI - WSUR和MOCI的相应子量表之间的相关性最强。得出的结论是，即使三因素模型的拟合略显欠佳，这可能表明责任因素需要修订，但在学生群体中，OTC冰岛语翻译的因素效度、聚合效度和区分效度得到了支持。然而，这一点应在临床群体中进一步研究。
