2008年1月27日 星期日

Factor Structure of the Yale–Brown Obsessive Compulsive Scale

Factor Structure of the Yale–Brown Obsessive Compulsive Scale
[Brief Reports]
Amir, Nader1,2; Foa, Edna B.1; Coles, Meredith E.1
1Center for the Treatment and Study of Anxiety, Department of Psychiatry, Allegheny University of the Health Sciences.
2Correspondence concerning this article should be addressed to Nader Amir, Department of Psychiatry, Allegheny University, 3200 Henry Avenue, Philadelphia, Pennsylvania 19129. Electronic mail may be sent to amir@ef.auhs.edu.
We thank Michael J. Kozak, Wayne K. Goodman, Eric Hollander, Michael A. Jenike, and Steven Rasmussen for helping to collect the data. This study was supported by National Institute of Mental Health Grant MH45404-05A1.
Received Date: December 9, 1996; Revised Date: January 6, 1997; Accepted Date: January 6, 1997
Abstract
The Yale–Brown Obsessive Compulsive Scale (Y-BOCS) is one of the most widely used measures of obsessive–compulsive disorder (OCD) symptoms (W. K. Goodman et al., 1989). The purpose of this study was to examine the dimensions underlying the Y-BOCS by performing a confirmatory factor analysis of the scale using responses from a large sample of patients. The results support a 2-factor model of OCD symptoms. The first factor reflected the degree of disturbance caused by OCD symptoms, and the second factor reflected the severity of OCD symptoms.
________________________________________


The Yale–Brown Obsessive Compulsive Scale (Y-BOCS) is one of the most widely used measures of severity of obsessive compulsive disorder (OCD) and has become the standard instrument for assessing OCD severity in both drug trials and studies of behavioral treatments (Steketee, 1993). This structured interview inquires about the presence of specific obsessions and compulsions as well as their severity (Goodman et al., 1989). The total severity score is the sum of the 10 item scores (range = 0–40) on a 5-point Likert scale (0 = none, 4 = extreme). Five items inquire about obsessions and five items inquire about compulsions. Each set of five items asks about the following aspects of OCD symptoms: distress, frequency, interference in daily life, resisting the symptoms, and controlling the symptoms.
An additional six items are included in the Y-BOCS but do not enter into the total severity score (Steketee, 1993). These additional items include insight into symptoms, avoidance, indecisiveness, pathologic responsibility, slowness, and pathologic doubting. The 10 core items of the Y-BOCS have excellent interrater reliability (intraclass correlation coefficients) as calculated across four raters of 40 patients with OCD (Obsession subtotal, r = .97, Compulsion subtotal, r = .96, and Total Score, r = .98; Goodman et al., 1989). The scale was also shown to be highly homogeneous (mean [alpha] across four raters = .89; Goodman et al., 1989). Although the division of items into obsessions and compulsions is theoretically derived, the factor structure of the Y-BOCS has received little attention.
Fals-Stewart (1992) conducted a principal-axis factor analysis that included the 10 core items of the Y-BOCS as well as the 6 investigational items that are not used in computing OCD severity (Goodman et al., 1989). Fals-Stewart examined the responses of 193 patients who met the criteria of the revised third edition of the Diagnostic and Statistical Manual of Mental Disorders for OCD(DSM–III–R; American Psychiatric Association, 1987). A single factor emerged, with all items loading .45 or higher on this factor. On the basis of this analysis, the author argued that the Obsession and the Compulsion subscales of the Y-BOCS comprise one construct and therefore only the total score should be considered in evaluating the severity of OCD. Because Fals-Stewart's analysis included the 6 investigational items of the Y-BOCS, this model, although fitting the data, may not be pertinent clinically. Furthermore, the inclusion of the 6 items that are not computed into the total score may have obscured the differences between obsessions and compulsions by emphasizing the overall severity of every symptom.
In a second study, McKay, Danyko, Neziroglu, and Yaryura-Tobias (1995) tested three models of the underlying factor structure of the Y-BOCS using a sample of 83 patients who met DSM–III–R criteria for OCD. In contrast to Fals-Stewart (1992), McKay et al.'s confirmatory factor analysis included only the 10 items that made up the total severity score. The first model tested included a single factor; the second model tested included two factors, with obsession items loading on the first factor and compulsion items on the second factor; the third model tested included two factors reflecting obsessions and compulsions and a higher order factor reflecting overall OCD symptoms. The single-factor model and the two-factor model with a higher order factor were not consistent with the data. The two-factor model without a higher order factor did fit the data. Because the two-factor model without a higher order factor was the only model to fit the data, McKay et al. concluded that clinical judgments should be based on the two subscales of the Y-BOCS rather than on the total score. Two methodological shortcomings render the McKay et al. (1995) results ambiguous. First, the sample size was relatively small for testing the models proposed. Second, the authors did not replicate their models in a second sample, and therefore, the generalizability of their findings to other samples cannot be ascertained.
Clinical observation suggests that OCD patients often do not clearly separate their OCD symptoms into obsessions and compulsions. Instead, they focus on disturbance caused by their symptoms and the severity of the symptoms. We therefore hypothesized that two factors underlying OCD symptoms would be disturbance (e.g., distress and interference) and symptom severity (e.g., time, control, and resistance). In the current study we used the 10 core items of the Y-BOCS to examine three models: a one-factor model composed of the first 10 items; a two-factor model, obsessions and compulsions, proposed by McKay et al. (1995); and a new two-factor model, with the first factor reflecting the degree of disturbance caused by OCD symptoms and the second factor reflecting the severity of symptoms. We did not include the 6 investigational items because their psychometric properties and clinical significance are not known (Goodman et al., 1989). The generalizability of the models was examined by replicating the results in a second split sample.
Method
Participants were 404 individuals who took part in the field trial of OCD (Foa et al., 1995) for the fourth edition of the DSM (DSM–IV; American Psychiatric Association, 1994), met DSM–III–R criteria for OCD, and had complete data. The total sample was randomly divided into two split samples (N = 202) to allow validation of the models proposed in this study. Each patient who had contacted one of the seven participating clinics for evaluation or treatment of OCD and was conversant in English entered the study. Additional participants included individuals who had had contact with one of the centers prior to the trial and remained symptomatic at the time of the study. A description of the sample is presented in Foa et al. (1995).
To test the three models described earlier, we used LISREL (Jöreskog & Sörbom, 1993). LISREL is a statistical technique for testing the fit between a set of relations among observed variables and a set of relations hypothesized by a conceptual model. Notably, a fit does not imply that the tested model is the only model that can account for the observed relations (Hoyle, 1995). Multiple models might fit one data set, and a particular model might fit one data set and not another. One strategy for choosing between competing models is to replicate all models that fit a given data set in an independent sample. The model that fits more than one sample is deemed generalizable and therefore preferable. This strategy was adopted to test the following three competing models. Model 1 predicted that one factor would account for OCD symptoms. Model 2 predicted that two factors, Obsessions and Compulsions, would account for OCD symptoms (McKay et al., 1995). Model 3 predicted that two factors, Disturbance and Symptom Severity, would account for OCD symptoms.
Results
An examination of the reliability of the Y-BOCS in our sample revealed that the scale has good internal consistency for the Obsessions and the Compulsions subscales as well as for the total scale: Obsessions subscale—Sample A, [alpha] = .83, Sample B, [alpha] = .87; Compulsions subscale—Sample A, [alpha] = .86, Sample B, [alpha] = .87; total scale, Sample A, [alpha] = .89, Sample B, [alpha] = .91. We used the covariance matrix among the 10 items to test the three models.1 Because there are no agreed on measures of fit between observed data and a proposed model, we used five commonly recommended indices of fit: (a) chi-square, (b) adjusted goodness of fit index (AGFI), (c) root mean square error of approximation (RMSEA) (Browne & Cudeck, 1993; Hoyle & Smith, 1994), (d) comparative fit index (CFI), and (e) the normed fit index (NFI; Mulaik et al., 1989). Of the five indices, chi-square is the most frequently used measure of fit. A significant chi-square value indicates that the data differ significantly from the predicted model, and that the model needs to be modified or rejected. A nonsignificant chi-square value indicates a good fit between the model and the data. AGFI is a global measure of fit between the data and the proposed model. It has a range of 0–1 with larger values indicating a better fit. Values of AGFI greater than .8 are considered a sign of good adjustment (Cole, 1987). RMSEA is a measure of residual variability not accounted for by the model. Browne and Cudeck (1993) suggested that RMSEA values of .05 indicate a close fit of the model to the data and values of up to .08 indicate reasonable fit. In this study we used RMSEA values of less than .05 to indicate a close fit between the model and the data. The CFI and the NFI measure the relative fit of a proposed model to a baseline model, usually the independence model. The CFI and NFI lie between 0 and 1, with larger values indicating better fit. Because the independence model often has a very large chi-square value, CFI and NFI often obtain values that are close to 1 (Jöreskog & Sörbom, 1993). Results of analyses of the three models are presented in Table 1.

________________________________________

[Help with image viewing]
[Email Jumpstart To Image]
Table 1 Relative Fit of Models
________________________________________
Model 1, proposing one factor, fit one sample with a nonsignificant chi-square value but had marginal fit to the second sample. This model had AGFIs of greater than .8 and satisfactory NFIs and CFIs in both samples. Similarly, the RMSEA value for this model was satisfactory in Sample A but did not meet our criteria of close fit in Sample B. The effect sizes for this model are presented in Figure 1.

________________________________________

[Help with image viewing]
[Email Jumpstart To Image]
Figure 1. One-factor model of Yale–Brown Obsessive Compulsive Scale.
________________________________________
Model 2, proposing two factors, Obsessions and Compulsions, did not fit either sample as indicated by significant chi-square values or RMSEAs. However, the AGFI, NFI, and CFI values were satisfactory for both samples. The effect sizes for this model are presented in Figure 2.

________________________________________

[Help with image viewing]
[Email Jumpstart To Image]
Figure 2. Replication of two-factor model of Yale–Brown Obsessive Compulsive Scale originally proposed by McKay et al. (1995).

________________________________________
Model 3, proposing two factors, Disturbance and Symptom Severity, fit both samples as indicated by nonsignificant chi-square values. This model also had good AGFIs, satisfactory RMSEAs, and good fit as indicated by NFI and CFI values. All indices supported this model in both samples. The effect sizes for this model are presented in Figure 3.

________________________________________

[Help with image viewing]
[Email Jumpstart To Image]
Figure 3. Two-factor model of the Yale–Brown Obsessive Compulsive Scale: Disturbance from and severity of obsessive–compulsive disorder symptoms.
________________________________________
Discussion
The present study provides support for a newly proposed two-factor model of the Y-BOCS symptom severity that represents disturbance and OCD symptom severity. The one-factor model fit one sample but only marginally fit the second sample. McKay et al.'s (1995) two-factor model, Obsessions and Compulsions, did not fit either of our subsamples. The finding that the one-factor model fit the first but not the second split sample lends support to the recommendation of LISREL experts to test a proposed model in more than one sample.
Our two-factor model is the only model that clearly fit both samples according to all indices. It suggests that the conventional division of OCD phenomenology into obsessions and compulsions may not reflect the structure underlying the different aspects of OCD symptoms. Instead, the new model suggests that the degree to which patients feel distressed by their OCD symptoms and the degree of interference these symptoms engender in the patient's functioning cannot be simply inferred from the severity of the OCD symptoms itself.
The present findings imply that clinicians should not focus exclusively on obvious severity-related dimensions, such as time spent on obsessing and ritualizing, when evaluating how disabled a patient is. Rather, the degree to which the patient's daily life is affected by the symptoms is an important dimension in clinical assessment and cannot simply be presumed to be a function of symptom severity. The difference between the symptoms targeted for therapy and general dysfunction has been recognized by treatment outcome researchers, who introduced the distinction between treatment efficacy and effectiveness: The former refers to reduction in symptom severity, and the latter to increase in general functioning.
In summary, the Y-BOCS appears to be viewed best as a two-factor scale, assessing impairment in functioning and symptom severity. Future research should examine the discriminant validity of the factors proposed in the current study by relating them to other measures of psychopathology (e.g., anxiety and depression) and treatment outcome.
References
American Psychiatric Association (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author. [Context Link]
American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. [Context Link]
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.). Testing structural equation models (pp. 136–162). Thousand Oaks, CA: Sage. [Context Link]
Cole, D. A. (1987). Utility of confirmatory factor analysis in test validation research. Journal of Consulting and Clinical Psychology, 55, 584–594. [Context Link]
Fals-Stewart, W. (1992). A dimensional analysis of the Yale–Brown Obsessive Compulsive Scale. Psychological Reports, 70, 239–240. Bibliographic Links [Context Link]
Foa, E. B., Kozak, M. J., Goodman, W. K., Hollander, E., Jenike, M. A., & Rasmussen, S. (1995). DSM-IV field trial: Obsessive–compulsive disorder. American Journal of Psychiatry, 152, 90–96. Bibliographic Links [Context Link]
Goodman, W. K., Price, L. H., Rasmussen, S. A., Mazure, C., Fleischmann, R. L., Hill, C. L., Heninger, G. R., & Charney, D. S. (1989). The Yale–Brown Obsessive Compulsive Scale: I. Development, use, and reliability. Archives of General Psychiatry, 46, 1006–1011. [Context Link]
Hoyle, R. H. (1995). Structural equation modeling: Concepts, issues, and applications. Thousand Oaks, CA: Sage. [Context Link]
Hoyle, R. H., & Smith, G. T. (1994). Formulating clinical research hypotheses as structural models: A conceptual overview. Journal of Consulting and Clinical Psychology, 62, 429–440. [Context Link]
Jöreskog, K. G., & Sörbom, D. (1993). LISREL VIII: User's reference guide. Chicago: Scientific Software. [Context Link]
McKay, D., Danyko, S., Neziroglu, F., & Yaryura-Tobias, J. A. (1995). Factor structure of the Yale–Brown Obsessive Compulsive Scale: A two dimensional measure. Behavior Research and Therapy, 33, 865–869. [Context Link]
Mulaik, S. A., James, L. R., VanAlstine, J., Bennett, N., Lind, S., & Stilwell, C. D. (1989). Evaluation of goodness-of-fit indices for structural equation models. Psychological Bulletin, 105, 430–445. Ovid Full Text Bibliographic Links [Context Link]
Steketee, G. S. (1993). Treatment of obsessive compulsive disorder. New York: Guilford Press. [Context Link]
1Copies of the covariance matrices used can be obtained from Nader Amir. [Context Link]
________________________________________

Accession Number: 00012030-199709000-00018
Copyright (c) 2000-2007 Ovid Technologies, Inc.

沒有留言: