Perceptualand Motor Skills, 1991, 73, 315-322.
O Perceptual and Motor Skills 1991
USE O F T H E MODIFIED WAIS-R WITH PSYCHIATRIC INPATIENTS: A CAUTION '" DANIEL BOONE Trenton Psychiatric Hopi&[ Summary.-Use of the Modified WAIS-R described by Cella in 1984 with psychiatric inpatients was criticized on several grounds. First, the modification did not accurately identify areas of relative cognitive strength and weakness. An accuracy rate of 49% was considered unacceptable for interpretation of individual profiles, the task Second, use of the WAIS-RM the WAIS-RM was originally designed to does not significantly reduce total administration time. For the majority of 70 psychiatric inpatients, at least 80% of the WAIS-R items would be administered, resulting in saving just fifteen minutes. Finally, an inpatient's confidence and motivation to perform well could be compromised by skipping earlier, easier items. Administration of the fuU WAIS-R battery is recommended if a discussion of relative strengths and weaknesses is desired.
To decrease the time required to obtain an accurate estimate of intellec(Wechsler, tual functioning, psychologists have shortened the WAIS-R 1981) by using either selected subtests or selected items (Watkins, 1986). Subtest-reduction short forms attempt to provide an estimate of over-all intellectual functioning by administering every item of a selected number of subtests (Brooker & Cyr, 1986; Ishikuma & Kaufman, 1989; Reynolds, Wilson, & Clark, 1983; Silverstein, 1982, 1990b), resulting in a 53 to 86% reduction in administration time (Ryan & Rosenberg, 1984; Thompson, Howard, & Anderson, 1986; Ward, Selby, & Clark, 1987). Using a subtest-reduction approach, the clinician's task is to provide an estimate of global intellectual functioning, with no attention given to interpretation of the intersubtest profile. If a discussion of relative cognitive strengths and weaknesses is desired, advocates of the subtest-reduction approach encourage the administration of the standard WAIS-R (Kaufman, 1990; Sattler, 1788; Silverstein, 1990a; Watkins, 1986). In contrast, proponents of the item-reduction approach contend that referrals rarely seek only an estimate of the Full Scale I Q . In most cases, the referral also seeks statements regarding areas of relative intellectual strength and weakness (Cella, 1984; Himelstein, 1983; Satz & Mogel, 1962; Vincent, 1979). Instead of administering every item from a selected number of subtests, item-reduction techniques call for the administration of selected items from every subtest. This approach allows retention of some informa-
'I thank all the psychologists and . psychology . -. interns at Trenton Psychiatric Hospital for providing access to i h e u testing files. 'All correspondence should be addressed to Daniel Boone, Ph.D., Drake Complex, Trenton Psychiatric Hospital, P O Box 7500, West Trenton, NJ 08628.
316
D. BOONE
tion from all subtests which can then be used to make statements regarding areas of relative cognitive strength and weakness. The Modified WAIS-R (WAIS-RM; Cella, 1984; Himelstein, 1983; Vincent, 1979) is a criterion-referenced item-reduction WAIS-R short form. Starting points for several subtests are shortened by the obtained scaled score on the Information subtest, which is given in its entirety. Using the WAIS-RM, it is assumed that initial successive items resulting in a scaled score equal to that obtained on Information would have been passed on all other subtests. Exceptions to this are the Digit Span, Object Assembly, and Digit Symbol subtests, which are given in their entirety. For example, if a patient obtained a scaled score of 5 on Information, the clinician would begin Vocabulary on Item 7, Arithmetic on Item 5, and Picture Completion on Item 8 (Cella, 1984), since the raw score total of all skipped items would result in a scaled score of 5 on each of these subtests. Cella (1984) estimated that this procedure could reduce the total administration time of the WAISR by up to 30%. TABLE 1 STARTING ITEMS FORTHE WAIS-RM ON SEVENWAIS-R SUEITESTS AND TOTAL ~ R C E N I A G E (YO) OF ITEMSADMINISTERED AS DETERMINED BY THE INFORMATION SCALED SCORES Subtest 4
5
Information Scaled Scores 6 7
8+
Vocabulary 6 7 10 13 13 Arithmetic 4 5 6 8 9 Comprehension 3 4 6 6 6 Similarities 3 4 6 6 6 Picture Completion 5 8 8 8 8 Picture Arrangement 2 4 4 4 4 Block Design 1 2 3 3 3 % Items 86 80 74 71 70 Note.-From "The Modified WAIS-R: an extension and revision" by D. F, Cella, 1984, Journal ofClinzca1 Psychology, 40, 802. Copyright 1984 by D. F. Cella. Adapted by Permission.
Prior attempts to assess the usefulness of the WAIS-RM have examined either the correlation between predicted and actual subtest scaled scores or the magnitude and direction of mean subtest differences in scaled scores. Correlation coefficients between actual WAIS-R and predicted WAIS-RM subtest scaled scores have routinely been large and significant (Caplan, 1983; Cella, 1984; Ehrfurth, Phelan, & Bigler, 1981; Himelstein, 1983; Jeffrey & Jeffrey, 1984; Vincent, 1979). This evaluative criterion, however, is not very useful since statistically significant correlations between WAIS-RM and WAIS-R subtest scores are to be expected given their shared error variance (Kaufman, 1977; Silverstein, 1985). Another approach in assessing the WAIS-RM's usefulness has been to compare the WAIS-R and WAIS-
USE OF MODIFIED WAIS-R
3 17
RM subtest means using the t test (Cargnello & Gurekas, 1987; Cella, Jacobsen, & Hymowitz, 1985). Studies employing this evaluative criterion have routinely shown the WAIS-RM scores significantly overpredict the original WAIS-R means (Cargnello & Gurekas, 1987; Cella, et a/., 1985). Overprediction of the original WAIS-R subtest scaled score has been dismissed by WAIS-RM advocates as either minor (Cella, et al., 1985) or correctable though the subtraction of a constant (Cargnello & Gurekas, 1987). Neither use of correlations nor mean differences, however, directly addresses the accuracy of the WAIS-RM in reproducing the original WAISR subtest pattern of relative cognitive strength and weakness. Surprisingly, no studies have directly examined the effectiveness of the WAIS-RM in performing this task, even though this was the original reason for its development (Cella, 1984; Himelstein, 1983; Vincent, 1979). The purpose of this study was to evaluate accuracy of the WAIS-RM in predicting areas of relative cognitive strength and weakness for a sample of psychiatric inpatients.
METHOD Subjects Fifty-five men (33 white, 20 black, 2 Hispanic) and 20 women (13 white, 5 black, 1 Indian, 1 Korean) participated. They ranged in age from 18 to 64 years (Mean age = 31.3 yr.) and were referred for intellectual assessment at a state-run psychiatric institution. Fifty percent of these patients were primarily diagnosed with schizophrenic disorder, while 27% received a diagnosis of affective disorder. The remaining patients were diagnosed with either adjustment reaction (7%), personality disorder (7%), or substance abuse ~ r o b l e m(9%). The mean Full Scale I Q of this sample was 86.7 (SD = 15.9). Procedure Each patient received the entire WAIS-R, which was scored in the customary way. The WAIS-R records were then rescored using the WAISRM procedure as described by Cella (1984). Prediction of the actual Full Scale I Q from the WAIS-RM was evaluated via regression analysis. Each of the subtest scaled scores predicted by the WAIS-RM were then correlated with the actual WAIS-R subtest scaled score. I n addition, predicted and actual subtest mean scaled scores were compared using the t test. Next, accuracy of prediction of the original pattern of intersubtest scatter from the WAIS-RM was examined. Areas of relative strength and weakness for the standard WAIS-R were listed for each scale (Verbal and Performance), using a procedure described by Sattler (1988) and critical values ( p s .05) reported by Boone (1991a). This procedure was also used to identify areas of relative strength and weakness using subtest scaled scores predicted by the WAIS-RM. The percentage of agreement between WAIS-R and WAIS-
318
D. BOONE
RM in assigning a particular subtest as an area of cognitive strength or weakness was then calculated. RESULTS The Pearson correlation between the Full Scale IQ estimate obtained using the WAIS-RM and the actual WAIS-R value was large and significant (r = .99). The standard error of estimate obtained when using the WAIS-RM IQ to predict the actual IQ was 2.54, so a 5 confidence interval was needed to achieve 90% clinical accuracy (Boone, 1990, 1991b). Correlation coefficients between WAIS-RM scaled scores and the actual WAIS-R scaled scores were also large and significant, ranging from .95 (on Comprehension) to .99 (on Block Design). Although mean differences were small (all less than one point), t tests showed that difference scores between WAIS-RM and WAIS-R subtest scaled scores were all statistically significant; see Table 2. In all cases, WAIS-RM scaled scores were significantly WAIS-R
TABLE 2 MODIFII-DWAIS-R (WAIS-RM) SUBTESTCORRELATIONS, MEANS(M, STANDARI) DCVLAITONS (SD), AND SUBTEST COMPARISONS (td)
AND
Subtest
rD
.98 .97 .95 .96 .98 .97 .99
WAIS-RM
WAIS-R M
7.56 7.19 7.59 7.61 7.72 7.69 7.75
SD
3 77 2 89 3 42 Similarities 3.11 Picture Completion 3.25 Picture Arrangement 3.47 Block Design 3.37 %All correlation coefficients were significant ( p < ,001). *p
O Perceptual and Motor Skills 1991
USE O F T H E MODIFIED WAIS-R WITH PSYCHIATRIC INPATIENTS: A CAUTION '" DANIEL BOONE Trenton Psychiatric Hopi&[ Summary.-Use of the Modified WAIS-R described by Cella in 1984 with psychiatric inpatients was criticized on several grounds. First, the modification did not accurately identify areas of relative cognitive strength and weakness. An accuracy rate of 49% was considered unacceptable for interpretation of individual profiles, the task Second, use of the WAIS-RM the WAIS-RM was originally designed to does not significantly reduce total administration time. For the majority of 70 psychiatric inpatients, at least 80% of the WAIS-R items would be administered, resulting in saving just fifteen minutes. Finally, an inpatient's confidence and motivation to perform well could be compromised by skipping earlier, easier items. Administration of the fuU WAIS-R battery is recommended if a discussion of relative strengths and weaknesses is desired.
To decrease the time required to obtain an accurate estimate of intellec(Wechsler, tual functioning, psychologists have shortened the WAIS-R 1981) by using either selected subtests or selected items (Watkins, 1986). Subtest-reduction short forms attempt to provide an estimate of over-all intellectual functioning by administering every item of a selected number of subtests (Brooker & Cyr, 1986; Ishikuma & Kaufman, 1989; Reynolds, Wilson, & Clark, 1983; Silverstein, 1982, 1990b), resulting in a 53 to 86% reduction in administration time (Ryan & Rosenberg, 1984; Thompson, Howard, & Anderson, 1986; Ward, Selby, & Clark, 1987). Using a subtest-reduction approach, the clinician's task is to provide an estimate of global intellectual functioning, with no attention given to interpretation of the intersubtest profile. If a discussion of relative cognitive strengths and weaknesses is desired, advocates of the subtest-reduction approach encourage the administration of the standard WAIS-R (Kaufman, 1990; Sattler, 1788; Silverstein, 1990a; Watkins, 1986). In contrast, proponents of the item-reduction approach contend that referrals rarely seek only an estimate of the Full Scale I Q . In most cases, the referral also seeks statements regarding areas of relative intellectual strength and weakness (Cella, 1984; Himelstein, 1983; Satz & Mogel, 1962; Vincent, 1979). Instead of administering every item from a selected number of subtests, item-reduction techniques call for the administration of selected items from every subtest. This approach allows retention of some informa-
'I thank all the psychologists and . psychology . -. interns at Trenton Psychiatric Hospital for providing access to i h e u testing files. 'All correspondence should be addressed to Daniel Boone, Ph.D., Drake Complex, Trenton Psychiatric Hospital, P O Box 7500, West Trenton, NJ 08628.
316
D. BOONE
tion from all subtests which can then be used to make statements regarding areas of relative cognitive strength and weakness. The Modified WAIS-R (WAIS-RM; Cella, 1984; Himelstein, 1983; Vincent, 1979) is a criterion-referenced item-reduction WAIS-R short form. Starting points for several subtests are shortened by the obtained scaled score on the Information subtest, which is given in its entirety. Using the WAIS-RM, it is assumed that initial successive items resulting in a scaled score equal to that obtained on Information would have been passed on all other subtests. Exceptions to this are the Digit Span, Object Assembly, and Digit Symbol subtests, which are given in their entirety. For example, if a patient obtained a scaled score of 5 on Information, the clinician would begin Vocabulary on Item 7, Arithmetic on Item 5, and Picture Completion on Item 8 (Cella, 1984), since the raw score total of all skipped items would result in a scaled score of 5 on each of these subtests. Cella (1984) estimated that this procedure could reduce the total administration time of the WAISR by up to 30%. TABLE 1 STARTING ITEMS FORTHE WAIS-RM ON SEVENWAIS-R SUEITESTS AND TOTAL ~ R C E N I A G E (YO) OF ITEMSADMINISTERED AS DETERMINED BY THE INFORMATION SCALED SCORES Subtest 4
5
Information Scaled Scores 6 7
8+
Vocabulary 6 7 10 13 13 Arithmetic 4 5 6 8 9 Comprehension 3 4 6 6 6 Similarities 3 4 6 6 6 Picture Completion 5 8 8 8 8 Picture Arrangement 2 4 4 4 4 Block Design 1 2 3 3 3 % Items 86 80 74 71 70 Note.-From "The Modified WAIS-R: an extension and revision" by D. F, Cella, 1984, Journal ofClinzca1 Psychology, 40, 802. Copyright 1984 by D. F. Cella. Adapted by Permission.
Prior attempts to assess the usefulness of the WAIS-RM have examined either the correlation between predicted and actual subtest scaled scores or the magnitude and direction of mean subtest differences in scaled scores. Correlation coefficients between actual WAIS-R and predicted WAIS-RM subtest scaled scores have routinely been large and significant (Caplan, 1983; Cella, 1984; Ehrfurth, Phelan, & Bigler, 1981; Himelstein, 1983; Jeffrey & Jeffrey, 1984; Vincent, 1979). This evaluative criterion, however, is not very useful since statistically significant correlations between WAIS-RM and WAIS-R subtest scores are to be expected given their shared error variance (Kaufman, 1977; Silverstein, 1985). Another approach in assessing the WAIS-RM's usefulness has been to compare the WAIS-R and WAIS-
USE OF MODIFIED WAIS-R
3 17
RM subtest means using the t test (Cargnello & Gurekas, 1987; Cella, Jacobsen, & Hymowitz, 1985). Studies employing this evaluative criterion have routinely shown the WAIS-RM scores significantly overpredict the original WAIS-R means (Cargnello & Gurekas, 1987; Cella, et a/., 1985). Overprediction of the original WAIS-R subtest scaled score has been dismissed by WAIS-RM advocates as either minor (Cella, et al., 1985) or correctable though the subtraction of a constant (Cargnello & Gurekas, 1987). Neither use of correlations nor mean differences, however, directly addresses the accuracy of the WAIS-RM in reproducing the original WAISR subtest pattern of relative cognitive strength and weakness. Surprisingly, no studies have directly examined the effectiveness of the WAIS-RM in performing this task, even though this was the original reason for its development (Cella, 1984; Himelstein, 1983; Vincent, 1979). The purpose of this study was to evaluate accuracy of the WAIS-RM in predicting areas of relative cognitive strength and weakness for a sample of psychiatric inpatients.
METHOD Subjects Fifty-five men (33 white, 20 black, 2 Hispanic) and 20 women (13 white, 5 black, 1 Indian, 1 Korean) participated. They ranged in age from 18 to 64 years (Mean age = 31.3 yr.) and were referred for intellectual assessment at a state-run psychiatric institution. Fifty percent of these patients were primarily diagnosed with schizophrenic disorder, while 27% received a diagnosis of affective disorder. The remaining patients were diagnosed with either adjustment reaction (7%), personality disorder (7%), or substance abuse ~ r o b l e m(9%). The mean Full Scale I Q of this sample was 86.7 (SD = 15.9). Procedure Each patient received the entire WAIS-R, which was scored in the customary way. The WAIS-R records were then rescored using the WAISRM procedure as described by Cella (1984). Prediction of the actual Full Scale I Q from the WAIS-RM was evaluated via regression analysis. Each of the subtest scaled scores predicted by the WAIS-RM were then correlated with the actual WAIS-R subtest scaled score. I n addition, predicted and actual subtest mean scaled scores were compared using the t test. Next, accuracy of prediction of the original pattern of intersubtest scatter from the WAIS-RM was examined. Areas of relative strength and weakness for the standard WAIS-R were listed for each scale (Verbal and Performance), using a procedure described by Sattler (1988) and critical values ( p s .05) reported by Boone (1991a). This procedure was also used to identify areas of relative strength and weakness using subtest scaled scores predicted by the WAIS-RM. The percentage of agreement between WAIS-R and WAIS-
318
D. BOONE
RM in assigning a particular subtest as an area of cognitive strength or weakness was then calculated. RESULTS The Pearson correlation between the Full Scale IQ estimate obtained using the WAIS-RM and the actual WAIS-R value was large and significant (r = .99). The standard error of estimate obtained when using the WAIS-RM IQ to predict the actual IQ was 2.54, so a 5 confidence interval was needed to achieve 90% clinical accuracy (Boone, 1990, 1991b). Correlation coefficients between WAIS-RM scaled scores and the actual WAIS-R scaled scores were also large and significant, ranging from .95 (on Comprehension) to .99 (on Block Design). Although mean differences were small (all less than one point), t tests showed that difference scores between WAIS-RM and WAIS-R subtest scaled scores were all statistically significant; see Table 2. In all cases, WAIS-RM scaled scores were significantly WAIS-R
TABLE 2 MODIFII-DWAIS-R (WAIS-RM) SUBTESTCORRELATIONS, MEANS(M, STANDARI) DCVLAITONS (SD), AND SUBTEST COMPARISONS (td)
AND
Subtest
rD
.98 .97 .95 .96 .98 .97 .99
WAIS-RM
WAIS-R M
7.56 7.19 7.59 7.61 7.72 7.69 7.75
SD
3 77 2 89 3 42 Similarities 3.11 Picture Completion 3.25 Picture Arrangement 3.47 Block Design 3.37 %All correlation coefficients were significant ( p < ,001). *p
