WAIS-R Performance Following Closed-Head Injury: A Comparison of the Clinical Utility of Summary IQs, Factor Scores, and Subtest Scatter Indices*

WAIS-R performance can be analyzed at the level of the IQs, factor scores, or subtests. This study examined the relative utility of such analyses in a closed-head-injured (CHI) sample (n = 233). Comparison of the sample with matched controls (n =117) indicated that factor scores may be the most appropriate level of analysis in clinical practice. This conclusion was supported by a series of discriminant function analyses performed on the WAIS-R scores of the CHI sample and 356 healthy controls drawn from the general population. Factor scores achieved a significantly higher classification accuracy than IQs or indices of subtest scatter (all /?s < .001). Combining the scatter indices with the composite measures (i.e., IQs or factor scores) did not produce a significant improvement in discrimination over that achieved by the composites alone. Counter to the assumptions underlying the clinical analysis of subtest profiles, CHI participants failed to exhibit significantly greater intraindividual subtest variability in their subtest profiles than the matched control sample. The Wechsler Adult Intelligence Scale-Revised (WAIS-R; Wechsler, 1981) is widely employed in clinical practice to assess the cognitive sequelae of closed-head injury (CHI) (Lezak, 1995; McKinlay & Gray, 1992; Richardson, 1990; Walsh, 1991). However, currently there is little published research on the characteristics of WAIS-R performance following a CHI (Crosson, Greene, Roth , Farr, & Adams, 1990). The studies conducted to date have predominantly employed small samples and have been concerned with the WAIS rather than WAIS-R. These WAIS studies have typically reported deficits in both Verbal and Performance IQ but with deficits being more pronounced on the latter scale (see Crosson et al., 1990; and Richardson, 1990 for reviews). Where performance on the individual subtests has been reported, Arithmetic and Digit Span have typically been the most impaired of the Verbal subtests with Digit Symbol exhibiting the largest Performance Scale deficit (Becker, 1975; Richardson, 1990). A number of authors have pointed out the dangers of assuming that the results of research obtained using the WAIS will be applicable to the WAIS-R (e.g. Bornstein, 1987; Reitan & Wolfson, 1990). Therefore, there is a need for new data obtained from CHI populations to aid clinicians in their interpretation of WAIS-R performance. Clinical interpretation of WAIS-R performance can be based on analysis at the level of the summary IQs, factor scores, or the subtests. Authorities on the Wechsler have divided opinions on the usefulness of the summary IQs. Matarazzo and Herman (1985) for example, have suggested that the discrepancy between Verbal and Performance IQs is the best Address correspondence to: J.R. Crawford, Department of Psychology, University of Aberdeen, Aberdeen AB92UB, UK. Accepted for publication: December 10, 1996. 346 J. R. CRAWFORD ET AL. validated Wechsler index of cerebral dysfunction, whereas Lezak (1995) is dismissive of its importance. Many clinicians base their interpretation of the WAIS-R primarily on the pattern of strengths and weaknesses in a client's subtest profile (e.g., Lezak, 1995; Walsh, 1991). Lezak (1988a) for example, notes that IQs can obscure clinically important strengths and weaknesses and suggests that "although the traditional scoring scheme for the Wechsler exemplifies the IQ problem, these tests also show us a way out of the problem by providing a profile of scores" (p. 359). However, the interpretation of subtest profiles is not without its own difficulties. The rationale underlying this approach is that, because individuals differ widely in terms of premorbid ability, the clinician needs to employ individual rather than normative comparison standards when attempting to detect impairment. It is assumed that some subtests will be relatively spared following neurological injury and thus will provide these comparison standards. Thus, in inferring impairment, the emphasis is placed more on the presence of large discrepancies among an individual's subtest scores than on comparison of these subtest scores with the relevant norms (e.g., Crawford, 1992; Lezak, 1995). There has been limited research on the utility of this approach in detecting and quantifying cognitive dysfunction. Furthermore, studies of healthy samples have revealed a surprisingly large degree of intraindividual variability; a relatively flat subtest profile is the exception rather than the rule in the general population (e.g., Crawford & Allan, 1996; Kaufman, 1990; Matarazzo, Daniel, Prifitera, & Herman, 1988). These studies suggest that there is a danger of over-inference when discrepancies between subtests are examined in clinical populations and underlines the need for statistical indices to supplement clinical judgment. Two such indices will be considered in the present study. McLean, Reynolds, and Kaufman (1990) have suggested using the variance of an individual's subtests as an index of subtest scatter: they termed this the Profile Variability Index (PVI). A recently proposed alternative approach to the analysis of subtest discrepancies uses the Mahalanobis Distance as an index (Burgess, 1991). The Mahalanobis Distance Index (MDI) is a multivariate indicator of the abnormality (i.e., rarity) of an individuals' scatter and has a chi-square distribution with 11 degrees of freedom when used with a full-length WAIS-R. In the present study it was hypothesized that the MDI would achieve better discrimination between healthy and CHI participants than the PVI as it factors in the estimated population correlations between subtests when evaluating the abnormality of discrepancies. For example, an extreme Digit Span score would be given less weight than an equally extreme Vocabulary score as the former subtest has a lower average correlation with the other subtests; that is, extreme Digit Span scores will occur with greater frequency in the healthy population. The greater sophistication of the MDI has a practical cost. In contrast to the PVI, which can be rapidly calculated by hand, the calculation of an individual's MDI requires a computer as it makes use of the subtest covariance matrix. Thus, the MDI would have to significantly outperform the PVI to justify its routine use in clinical practice. An alternative to the analysis of both IQs and subtest discrepancies is to examine scores obtained from factor analysis of the WAIS-R (Atkinson, 1991; Canavan, Dunn, & McMillan, 1986; Crawford, 1992; Kaufman, 1990). Although this suggestion has a long history (e.g., Maxwell, 1960), factor scores have not been widely employed in clinical practice or research. The present authors are unaware of any research which has examined WAIS or WAIS-R performance in CHI in terms of factor scores, despite the indications that the factor structure of the WAIS-R is preserved in CHI and other neurological disorders (e.g., Atkinson, Cyr, Doxey, & Vigna, 1989; Leckliter, Matarazzo, & Silverstein, 1986). This neglect of the factor score approach may be unfortunate as it has some advantages over the use of summary IQs or subtests; unlike IQs, factor scores are empirically derived composites rather than composites formed a priori, they therefore have superior construct validity. Because they are composites of correlated components, they necessarily possess superior WAIS-R INDICES IN CLOSED-HEAD INJURY 347 reliability to the subtests from which they are formed (e.g., see Guilford, 1954). Factor analyses of the WAIS and WAIS-R have commonly extracted three factors: a verbal (V) factor on which Information, Vocabulary, Comprehension, and Similarities have high loadings, a perceptual organization (PO) factor defined by high loadings from Block Design and Object Assembly, and a third factor termed attention/concentration (A/C) or freedom-fromdistractibility consisting of Arithmetic, Digit Span, and Digit Symbol (e.g., Crawford, Allan, Stephen, Parker, & Besson, 1989; Leckliter et al., 1986). The factor score approach may be particularly useful in the interpretation of WAIS-R performance following a CHI. Lezak (1988b) for example, in arguing the case against IQs and for examination of subtest profiles, notes that "Many persons who have suffered a mild concussion with loss of consciousness ... tend to perform relatively poorly on Digits Backward, Arithmetic and Digit Symbol" (p. 358). This clinical observation is in line with the present authors' experience and the research reported above, but it could be argued that it provides a justification for the use of factor scores as much as subtest profile analysis because these three subtests make up the attention/ concentration factor. In the present investigation the WAIS-R performance of a sample of CHI patients was compared with a matched sample of healthy controls to identify which components of WAIS-R performance typically reveal the largest deficits and which are relatively preserved. The aim was to update similar data obtained with the WAIS with the exception that factor scores were also examined. The second component of the study employed discriminant function analysis and was prompted by the fact that the principal concern for clinicians is how successful these indices are liable to be at detecting impairment when used with individual clients. Furthermore, in clinical work, healthy samples matched for age, sex, and education are obviously not readily available to provide comparison standards against which a client's performance can be compared. Because of this we employed a large sample drawn from the general adult population to serve as the healthy group in the discriminant function analyses. The foregoing discussion of WAIS-R indices focused on their relative merits. However, in clinical practice interpretation of the WAIS-R will be based on a consideration of these indices in combination (analysis of the qualitative features of performance also plays an important role, but is beyond the province of the present investigation). A final aim of the present study was therefore to examine whether the combination of composite scores (i.e., the IQs or factor scores) and indices of subtest scatter would achieve better discrimination between the performance of and healthy individuals and patients with CHI than either of these sets of measures alone.