Replication of factor structure of Wechsler Adult Intelligence Scale-III Chinese version in Chinese mainland non-clinical and schizophrenia samples

Size: px

Start display at page:

Download "Replication of factor structure of Wechsler Adult Intelligence Scale-III Chinese version in Chinese mainland non-clinical and schizophrenia samples"

Silas Armstrong
5 years ago
Views:

1 Psychiatry and Clinical Neurosciences (2007), 61, doi: /j x Regular Article Replication of factor structure of Wechsler Adult Intelligence Scale-III Chinese version in Chinese mainland non-clinical and schizophrenia samples SHUQIAO YAO, phd, 1 HUAN CHEN, md, 1 LI JIANG, phd 1 AND WAI-CHEONG CARL TAM, phd 2 1 Medical Psychological Research Center of the Second Xiangya Hospital, Central South University, Changsha, China and 2 Department of Psychology, Chung Yuan Christian University, Chung Li, Taiwan Abstract The aim of the present study was to evaluate the construct validity of the special four-factor structure of the Wechsler Adult Intelligence Scale (WAIS)-III Chinese version in a clinical sample of inpatients with schizophrenia as well as a sample of healthy adults. A sample of 114 inpatients with schizophrenia and a sample of 114 close-matched non-clinical adults on the Chinese mainland were studied with the measures of the WAIS-III Chinese version. Four competing hypothetical models of factors were tested for model fit and parsimony in both research samples, using maximum likelihood confirmatory factor analysis. Confirmatory factor analysis indicated that a four-factor model of Verbal Comprehension, Perceptual Organization, Working Memory, and Processing Speed best fit the data from both the schizophrenia sample and the healthy adult sample, and also fitted the data from both samples better than alternative models, which is similar to that reported by the developers and revisers of this scale. These empiric analyses and results support the construct validity of the WAIS-III Chinese version in patients with schizophrenia and healthy adults on the Chinese mainland. Key words adults, confirmatory factor analysis, neurocognitive ability, schizophrenia, WAIS-III Chinese version. INTRODUCTION The Wechsler Adult Intelligence Scale (WAIS) has been widely and frequently applied in testing adult intelligence and assessing those persons with schizophrenia for clinical and research purposes. Its psychometric capacities are acknowledged to be excellent. Compared with previous editions, the third edition of the WAIS (WAIS-III) with its subtests contains four discrete indices: Verbal Comprehension (VC, based on the Vocabulary, Similarities, and Information subtests), Perceptual Organization (PO, based on the Picture Completion, Block Design, and Matrix Reasoning subtests), Correspondence address: Shuqiao Yao, PhD, Medical Psychological Research Center, Second Xiangya Hospital, Central South University, Changsha, Hunan , China. shuqiaoyao@163.com Received 19 September 2006; revised 28 March 2007; accepted 30 March Working Memory (WM, based on the Arithmetic, Digit Span, and Letter Number Sequencing subtests), and Processing Speed (PS, based on the Digit Symbol- Coding and Symbol Search subtests). Confirmatory factor analyses (CFA) showed that the four-factor model best fitted the data for the total standardization sample in the USA and across the age ranges. 1 Replication of the WAIS-III factor structure was also examined by Gorsuch and Boyd. 2 The four-factor indices appear to provide a better way to conceptualize the test performance with 13 subtests (but this can also be explained by using only 11 subtests), and a better interpretation in both clinical and theoretical perspectives. 3 5 Saklofske et al. have replicated the four-factor structure in a larger independent Canadian sample across cultural boundaries, using CFA. 6 The examinations and analyses of the intelligence structure of the WAIS-III Chinese version with its Taiwanese standardization sample have been carried out

2 380 S. Yao et al. by its developers and revisors. 7 However, it is necessary to extend the prior analyses of the intelligence structure in Taiwan standardization sample to a Chinese mainland healthy sample so as to know whether the same applicability exists or not. There are no differences in ethnic group and language between the two regions. In addition, the possible cultural differences are decreasing in that both the opening policy or westernization of the Chinese mainland and the interactions of the two sides have been increasing enormously in the past 20 years. We therefore hypothesize that the same factor structure will be found in the Chinese mainland WAIS-III data even if there are some differences between the Chinese mainland and Taiwan in cultural background. For instance, both gross domestic product (GDP) per capita and average level of education in Taiwan are higher than those in the Chinese mainland. 8,9 Furthermore, it must be noted that the revised Letter Number Sequencing subtest in the Chinese version is to be excluded from the Working Memory index on account of cultural distinction. 7 There have been some previous studies testing whether the WAIS-III four-factor structure can generalize across clinical boundaries: Ryan and Paolo further supported the validity of the four-factor structure by using exploratory factor analysis (EFA) of data from a wide range of clinical samples such as substance use disorders, and psychiatric disorders; 10 and Heijden and Donders supported the construct validity of these four factors in patients with traumatic brain injury (TBI), using CFA. 11 Nevertheless, so far there is a paucity of research concerning the construct validity of WAIS-III in clinical samples of adult patients with schizophrenia. In the examination of factor structure of WAIS-III in a schizophrenia sample, Dickinson et al. used both EFA and CFA to indicate that the four-factor WAIS-III presented a significant improvement over other competing models but did not improve significantly over the fivefactor model in a schizophrenia sample. 12 However, whether the four-factor model of the WAIS-III Chinese version fits the data from schizophrenia and nonclinical samples in the Chinese mainland is unclear. For all of these reasons, the purpose of the present study was to evaluate the construct validity of the special four-factor structure of the WAIS-III Chinese version in a clinical sample of inpatients with schizophrenia as well as a sample of healthy adults by CFA. METHOD Participants We recruited 114 patients with schizophrenia (60 male, 54 female; age range, years; mean age, years; mean education level, years), and 114 healthy adults (53 male, 61 female; age range, years, mean age, years; mean education level, years). All the patients were, respectively, recruited from the Second Xiangya Hospital of Central South University (30%), the Second People s Hospital of Hunan Province (40%), and the Changsha Mental Hospital (30%) in Hunan province located in central south of the Chinese mainland, meeting the following inclusion criteria: (i) DSM-IV diagnostic criteria for schizophrenia or schizoaffective disorder (for schizophrenia, n = 111, including 56 undifferentiated, 34 paranoid, 10 residual, 7 disorganized, and 4 unspecified; for schizoaffective disorder, n = 3); (ii) right-handedness; (iii) age between 16 and 84 years; (iv) no visual (including color blindness) or auditory disabilities; (v) no motor system disability; (vi) no mental retardation; (vii) no history of substance-related disorders, delirium, dementia, amnestic and other cognitive disorders, and no history of any chronic general medical conditions; (viii) no history of electroconvulsive therapy; (ix) an ability and desire to cooperate with our experimental procedures; (x) written informed consent; and (xi) Global Assessment of Functioning (GAF) Scale score >61. All the healthy adults were recruited from the same district as the patients and had no personal or family history of mental illness or personal medical illness that might impair cognitive function such as stroke, seizure, or central nervous system infection. In addition, they also met the inclusion criteria (ii x). Approval was obtained from the ethical committee of the hospital. All participants gave written informed consent. We previously excluded 37 patients who had a diagnosis of a schizophrenia-spectrum disorder and also a diagnosis of substance abuse or dependence on alcohol or illicit drugs, and excluded 14 individuals from the healthy adult sample who had a history of mental retardation or physical disorder. The mean age of illness onset for the schizophrenia patients was years. All the patients were receiving antipsychotic medication during the time of testing, 46 inpatients were taking conventional antipsychotic drugs, 57 were taking atypical antipsychotic drugs, and 11 were drug naïve. WAIS-III Chinese version The data of the Taiwan WAIS-III standardization study initiated in 1999 were collected on a stratified random sample of 888 adults. The sampling plan followed the major descriptors used in the American WAIS-III standardization study: gender (50.97% male, 49.03% female), age (16 84 years collapsed into 11 age groups), education level (less than grade 6, grades 7 10,

3 Structure of WAIS-III Chinese version 381 high school diploma, some post-secondary training, or university degree). The measure consists of verbal IQ, performance IQ and Full IQ with four index scores. In the standardization study the WAIS-III Chinese version exhibited satisfactory reliability and validity similar to that of the WAIS-III. The same four-factor structure of the WAIS-III had been verified to best fit the standardization data underlying performance on the WAIS-III Chinese version. 7 Clinical assessment Every inpatient was assessed with the GAF to estimate their psychological, social and occupational functions. Those clinically stable inpatients with a GAF score >61, after combining information from interview with family informants, prior psychiatrists, and medical records to generate a final evaluation, were selected for administration of WAIS-III Chinese version (mean GAF score, ). Procedure Confirmatory factor analysis was used to compare four theoretical models derived from the technical manual and previous studies, with respect to the constructs measured by WAIS-III Chinese version of the patient and healthy adult samples, for parsimony and model fit. In addition, research by Gorsuch and others suggested that models must include a minimum of two subtests to define a stable factor because the underlying variables with fewer than two indicator variables often display some problems with identification and convergence. 1,13,14 Therefore, four models with 12 subtests were specified in the present study as follows. Model 1 was a one-factor model suggesting a single general intelligence factor grouping all 12 subtests. Model 2 was a traditional two-factor model that included six subtests representing a Verbal factor and six subtests representing a Performance factor. Model 3 was a three-factor model (WAIS-R model) that hypothesized four subtests representing a Verbal Comprehension factor, four subtests representing a Perceptual Organization factor, and four subtests representing an Attentional factor. Model 4 was the four-factor model having been noted as the best fit to the data from the WAIS-III Chinese version standardization sample, and being a refinement of the WAIS-R model with the Attentional factor divided to create the Working Memory (two subtests) and Processing Speed (two subtests) indices. Amos 5 was used to perform the CFA for each of the four models. 15 The Tucker-Lewis Index (TLI) was used; this index varies between 0 and 1, and values >0.9 indicate acceptable fit. Also, the comparative fit index (CFI) indicates an acceptable fit at >0.9. Both of them with values >0.95 indicate a close fit of the model. The adjusted goodness-of-fit index (AGFI) with values >0.8 indicates adequate model fit. The root mean square error of approximation (RMSEA) is used to test the degree of fit between objective data and each model of factors, with values <0.08 for a reasonable error of approximation and 0.05 for a close fit. The parsimonious normed fit index (PNFI) with values >0.6 indicates a good fit. RESULTS Table 1 lists the goodness of fit indices for each of these four models. It was obvious that model 4 (WAIS-III) was the best model in the sequence, meeting all the criteria for a best fit in both research samples, that is, AGFI > 0.8, RMSEA < 0.05, TLI and CFI > 0.95, and PNFI > 0.6. In addition, this WAIS-III model displayed an evident improvement over the WAIS-R model (model 3), which was the next best fit to data in both Table 1. Goodness-of-fit indices for four competing models Models d.f. c 2 AGFI RMSEA TLI CFI PNFI S H S H S H S H S H S H AGFI, adjusted goodness-of-fit index; CFI, comparative fit index; H, healthy adult sample (n = 114); PNFI, parsimonious normed fit index; RMSEA, root mean square error of approximation; S, schizophrenia sample (n = 114); TLI, Tucker Lewis Index. Model 1 is the one-factor model; model 2 is the traditional two-factor model; model 3 is the Wechsler Adult Intelligence Scale Revised (WAIS-R) model for three factors; model 4 is the four-factor model of WAIS-III.

4 382 S. Yao et al. Table 2. Confirmatory factor analysis correlation between factors in the WAIS-III model 4 Factor samples. The contrasts between one and two factors and between two and three factors and between three and four factors were statistically significant for each c 2 difference. The fit of factor structure for healthy adult sample presented a slight improvement over that for the schizophrenia sample. Although fit indices were better in the more complex models in Table 1, twofactor models were the most parsimonious gauged by the PNFI. Standardized structural coefficients for model 4 across study samples were used for the loading on factors for the WAIS-III model. Although WM was the least strong loading factor, all the loading >0.60 was significant (P < 0.001), which demonstrated and supported the predicted relationship between the subtests and the latent factors. For the healthy adult sample, four subtests of VC with coefficients were Vocabulary (0.82), Similarity (0.85), Information (0.88), and Comprehension (0.85); four subtests of PO with coefficients were Picture Completion (0.81), Block Design (0.77), Matrix Reasoning (0.77), and Picture Arrangement (0.78); two subtests of WM with coefficients were Arithmetic (0.65) and Digit Span (0.68); two subtests of PS with coefficients were Digit Symbol-Coding (0.70) and Symbol Searching (0.75). In addition, the factor loadings of VC, PO, WM, and PS were consistent between the schizophrenia sample and the healthy adult sample. Table 2 shows the intercorrelations among the factors for the WAIS-III model. The correlations were as expected and supported the appropriateness of scoring for general intelligence as the first-order factor for these four factors. DISCUSSION VC PO WM PS S H S H S H S H VC PO WM PS All correlations are significant (P < 0.001). H, healthy adult sample (n = 114); PO, Perceptual Organization; PS, Processing Speed; S, schizophrenia sample (n = 114); VC, Verbal Comprehension; WM, Working Memory. Recent studies have found that the four-factor structure of the WAIS-III was replicated perfectly in the UK, Canada, and Spain. 6,16,17 The results of the present study with two Chinese mainland samples provide further support for the WAIS-III Chinese version fourfactor structure. The present findings using CFA of various models of the factor structure based on WAIS- III Chinese version performance indicated that the same four-factor model proposed by the developers of the WAIS-III for the normative general population best fits the data underlying performance on the WAIS- III Chinese version in a sample of patients with schizophrenia and a healthy adult sample from the Chinese mainland, confirming the construct validity of VC, PO, WM, and PS indices. These findings are of empiric and clinical significance. The replication of the four-factor structure verifies the portability of WAIS-III across clinical boundaries, and demonstrates its psychometric integrity. The results of the current study are consistent with those suggested by Ryan and Paolo 10 and Dickinson et al. 12 Examination of fit indices indicates that in both samples the two-factor models produced a substantially better fit than the singe-factor model. Although the four-factor model best fitted the data, the twofactor model was the most parsimonious in terms of the PNFI in both research samples. These findings are similar to that reported in a prior study on CFA of the WAIS-III standardization data. 3 Whatever theoretical model can provide the best interpretation, the support for the construct validity of the PS factor is especially meaningful because it represents a conceptually independent processing speed ability as compared to the WAIS-R three-factor model. The PS index has been established empirically as a valuable clinical tool because of its sensitivity to brain pathology. 18 Furthermore, the Digit Symbol and Symbol Search may clearly appear as a separate factor only in samples of persons with impaired brain functioning. 3 This also appears to be the case for patients with TBI and schizophrenia. 11,12 This result was replicated in the present study in Chinese patients with schizophrenia with the WAIS-III Chinese version. Ward et al. suggested that the PS index should be loaded on the PO index or the WM index, and model 2B consisting of a Verbal Comprehension factor and a Performance factor afforded a substantially better fit in the younger groups than the traditional model (model 2A). 3 Dickinson et al. reported that Matrix Reasoning and Picture Arrangement emerged as possible candidates for revised specification in alternative models of WAIS-III performance in schizophrenia. 12 Whether the performance of Digit Span in the schizophrenia sample with WAIS-III Chinese version was a cue for an alternative model was considered as well in the present study.

5 Structure of WAIS-III Chinese version 383 There was strong evidence suggesting that demographic variables have a strong impact on WAIS-III scores. 19 Although the four WAIS-III factor indices provide invariant measurement across all ages, 20 differences between older and younger samples have occurred. 1 Similar structure consistency in the standardization sample across different age groups has been proved in the WAIS-III Chinese version. 7 However, we do not know whether the factor structure will remain consistent across different demographic groups in the patients with schizophrenia, which will be further explored in future research. In addition, because the data presented here are preliminary, this being the first available study on WAIS-III Chinese version factor structure in schizophrenia, only 12 WAIS-III Chinese version subtests were used to compute the factor index scores and were presented in hypothetical factor models for which the Letter Number Sequencing subtest is an alternative subtest and was excluded from the calculation of index scores. WAIS-III is characterized by assessing a broad range of abilities. Therefore the present study, providing a confirmatory measure, gives a meaningful way to compare the schizophrenia and non-clinical groups, and briefly compares schizophrenia groups from two different cultural backgrounds (China and USA) in some important domains of neurocognitive ability. Although the WAIS-III Chinese version four-factor structure seems to best fit the data from the schizophrenia sample, some evident distinctions in WAIS-III Chinese version performance across the four factors were found. Schizophrenia is heterogeneous, and some patients, especially those with prominent negative symptoms, will perform in an impaired fashion on a variety of neuropsychological measures. 21 It remains to be seen whether any factor or subtest performance presents problems in schizophrenia on the WAIS-III Chinese version. With regard to future research, the next logical step would be conducting studies in comparing performance on the four WAIS-III domains between patients with schizophrenia and normal controls. The present empiric analyses and results confirm that clinicians can use the WAIS-III Chinese version factor indices with confidence in diagnostic testing and assessment and can guide us when we search for ways to improve cognitive assessment. ACKNOWLEDGMENTS The authors thank Drs Yan-Ke Huang and Le Lv, Medical Psychological Research Center, the Second Xiangya Hospital for assistance in sampling work. REFERENCES 1. Wechsler D. Wechsler Adult Intelligence Scale, 3rd edn. Psychological Corporation, San Antonio, TA, Gorsuch RL, Boyd K. Factor replicability and invariance in the WAIS-III. In: Proceedings of the 105th Annual Convention of the American Psychological Association, Chicago, American Psychological Association, Washington, DC, Ward LC, Ryan JJ, Axelrod BN. Confirmatory factor analyses of the WAIS-III standardization data. Psychol. Assess. 2000; 12: Kaufman AS, Lichtenberger EO. Essentials of WAIS-III Assessment. Wiley, New York, Kaufman AS, Lichtenberger EO, Mclean JE. Two- and three-factor solutions of the WAIS-III. Assessment 2001; 8: Saklofske DH, Hildebrand DK, Gorsuch RL. Replication of the factor structure of the Wechsler Adult Intelligence Scale-Third Edition with a Canadian sample. Psychol. Assess. 2000; 12: Wechsler D, Chen YH, Chen XY. WAIS-III Chinese Version Technical Manual. Psychological Corporation, San Antonio, TA, National Bureau of Statistics of China. The China 2005 Statistical Yearbook. China Statistics Press, Beijing, Directorate-General of Budget, Executive Yuan (Taiwan). Taiwan 2005 Accounting and Statistics. [Cited 6 Dec 2006.] Available from URL: Attachment/ xls and public/data/dgbas03/bs2/socialindicator/ education-table. xls 10. Ryan JJ, Paolo AM. Exploratory factor analysis of the WAIS-III in a mixed patients sample. Arch. Clin. Neuropsychol. 2001; 16: Heijden P, Donders J. A confirmatory factor analysis of the WAIS-III in patients with traumatic brain injury. J. Clin. Exp. Neuropsychol. 2003; 25: Dickinson D, Iannone VN, Gold JM. Factor structure of the Wechsler Adult Intelligence Scale-III in schizophrenia. Assessment 2002; 9: Gorsuch RL. Factor Analysis, 2nd edn. Lawrence Erlbaum, Hillsdale, NJ, Hatcher L. A Step-by-Step Approach to Using the SAS System for Factor Analysis and Structural Equation Modeling. SAS Institute, Cary, NC, Arbuckle JL, Wothke W. AMOS, 4.0 User s Guide. Small- Waters Corporation, Chicago, Wycherley R, Lavender A, Holttum S, Crawford JR, Mockler D. WAIS-III UK: An extension of the UK comparability study. Br. J. Clin. Psychol. 2005; 44: Garcia LF, Ruiz MA, Abad FJ. Factor structure of the Spanish WAIS-III. Psicothema 2003; 15: Wechsler D. WAIS-III/WMS-III Technical Manual. Psychological Corporation, San Antonio, TX, Heaton RK, Taylor MJ, Manly J. Demographic effects and the use of demographically corrected norms with the

6 384 S. Yao et al. WAIS-III and WMS-III. In: Tulsky DS, Saklofske DH, Chelune GJ et al. (eds). Clinical Interpretation of the WAIS-III and WMS-III. Academic Press, San Diego, CA, 2003; Taub GE, McGrew KS, Witta EL. A confirmatory analysis of the factor structure and cross-age invariance of the Wechsler Adult Intelligence Scale, 3rd edn. Psychol. Assess. 2004; 16: Hawkins KA. Indicator of brain dysfunction derived from graphic representations of the WAIS-III/WMS-III technical manual samples: A preliminary approach to clinical utility. Clin. Neuropsychol. 1998; 12:

Applied Short-Form WAIS-III to Explore Global Cognitive Profile of the Patients with Schizophrenia

Applied Short-Form WAIS-III to Explore Global Cognitive Profile of the Patients with Schizophrenia Chia-Ju Lin 1,2, Chin-Chuen Lin 1, Yi-Yung Hung 1, Meng-Chang Tsai 1, Shih-Chun Ho 1, Ya-Ling Wang 1,