Archives of Clinical Neuropsychology 30 (2015) 404–412
Dyadic Short Forms of the Wechsler Adult Intelligence Scale-IV David A. Denney1, Wendy K. Ringe1, Laura H. Lacritz 1,2,* 2
1 Department of Psychiatry, The University of Texas Southwestern Medical Center, Dallas, TX, USA Department of Neurology & Neurotherapeutics, The University of Texas Southwestern Medical Center, Dallas, TX, USA
*Corresponding author at: UT Southwestern Medical Center, Neuropsychology, 5323 Harry Hines Boulevard, Dallas, TX 75390-8846, USA. Tel.: +1 214 648 4646; fax: +1 214 648 4660. E-mail address: [email protected].
Abstract Full Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV) administration can be time-consuming and may not be necessary when intelligence quotient estimates will suffice. Estimated Full Scale Intelligence Quotient (FSIQ) and General Ability Index (GAI) scores were derived from nine dyadic short forms using individual regression equations based on data from a clinical sample (n ¼ 113) that was then cross validated in a separate clinical sample (n ¼ 50). Derived scores accounted for 70%– 83% of the variance in FSIQ and 77% –88% of the variance in GAI. Predicted FSIQs were strongly associated with actual FSIQ (rs ¼ .73–.88), as were predicted and actual GAIs (rs ¼ .80– .93). Each of the nine dyadic short forms of the WAIS-IV was a good predictor of FSIQ and GAI in the validation sample. These data support the validity of WAIS-IV short forms when time is limited or lengthier batteries cannot be tolerated by patients. Keywords: Intelligence; Assessment; Short form; Wechsler
Introduction The Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV; Wechsler, 2008) is a commonly employed measure of global intellectual and cognitive functioning. The WAIS-IV yields two measures of global functioning, the Full Scale Intelligence Quotient (FSIQ) and the General Ability Index (GAI). The FSIQ is a weighted index of verbal, perceptual reasoning, processing speed, and working memory abilities, whereas the GAI is a global measure that removes working memory and processing speed subtests. The GAI is new to the WAIS-IV and important in neuropsychological evaluations because it provides a summary score that is less sensitive than FSIQ to the influence of attention and/or processing speed deficits commonly seen in many neurologic populations. Short forms of previous WAIS versions (Wechsler, 1955, 1987, 1997) have been proposed to estimate intelligence quotient (IQ) in instances where full IQ testing is too lengthy for patients to tolerate or where an estimate of IQ will suffice. According to Thompson, LoBello, Atkinson, Chisholm, and Ryan (2004), the Wechsler intelligence scales are the most frequently applied measures used to derive short-form IQ estimations. Previous estimation methods include administering as few as one and as many as eight WAIS subtests to derive estimates (Axelrod, Dingel, Ryan, & Ward, 2000; Christensen, Girard, & Bagby, 2007; Engelhart, Eisenstein, Johnson, & Losonczy, 1999; Jeyakumar, Warriner, Raval, & Ahmad, 2004; Kaufman, Ishikuma, & Kaufman-Packer, 1991; Mendella, McFadden, Regan, & Medlock, 2000; Pilgrim, Meyers, Bayless, & Whetstone, 1999; Ringe, Saine, Lacritz, Hynan, & Cullum, 2002; Satterfield, Martin, & Leiker, 1994; Schoenberg, Duff, Dorfman, & Adams, 2004; Schoenberg, Duff, Scott, & Adams, 2002; Schoenberg, Scott, Ruwe, Patton, & Adams, 2004; Silverstein, 1982; Ward, 1990). These studies suggest that as short forms lengthen in number of subtests used to derive estimates, correlations between estimated and actual scores improve. However, as the number of subtests increase, the amount of time needed to attain the estimated score also increases. When time limitations factor into an evaluation, longer four- or seven-subtest short forms may not serve as an ideal solution. # The Author 2015. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected]. doi:10.1093/arclin/acv035 Advance Access publication on 8 June 2015
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Accepted 20 May 2015
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
405
Methods Participants Subjects included 163 consecutive individuals with known or suspected neurological disorders who underwent neuropsychological evaluation as part of their standard clinical care in the Neuropsychology Service at the University of Texas Southwestern Medical Center between September 2009 and January 2011. Subjects gave consent for their clinical data to be used for research purposes. The initial 113 consecutive individuals who presented for neuropsychological evaluation were included in the test group, and the following 50 consecutive patients comprised the cross-validation group. Procedures The core 10-subtest WAIS-IV was administered as part of a larger battery for clinical reasons, and all subjects gave written consent for their results to be used for research purposes. Short forms for FSIQ and GAI were derived by summing age-scaled scores from subtests administered from the Verbal Comprehension Index (VCI) (Vocabulary [V], Similarities [S], and Information [I]) to those from the Perceptual Reasoning Index (Block Design [BD], Matrix Reasoning [MR], and Visual Puzzles [VP]). This resulted in nine dyads: Vocabulary/Block Design (V/BD), Vocabulary/Matrix Reasoning (V/MR), Vocabulary/Visual Puzzles (V/VP), Similarities/Block Design (S/BD), Similarities/Matrix Reasoning (S/MR), Similarities/ Visual Puzzles (S/VP), Information/Block Design (I/BD), Information/Matrix Reasoning (I/MR), and Information/Visual Puzzles (I/VP). Within the test sample, the summations of the age-scaled scores for each dyad were entered into separate regression equations to predict FSIQ and GAI. Short form conversion tables were created from the regression equations, which were validated by comparing actual FSIQ and GAI scores of the test sample with estimated FSIQ and GAI scores acquired from the regression ′ ) between the short form predicted FSIQs and actual FSIQ, as well data. Composite reliability (rxx) and correlation coefficients (r sf as those between short form predicted GAIs and actual GAI, were calculated using the methods outlined by Levy (1967) to correct for inflated Pearson correlations. Results were cross-validated on a second mixed neurological sample in which corrected correlations (rsf′ ) between actual and estimated FSIQ and GAI were obtained. The percentages of estimated scores within 5 and 10 points of actual FSIQ and GAI scores were also calculated.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
There are several instruments designed to quickly estimate intelligence (e.g., Wechsler Abbreviated Scale of IntelligenceSecond Edition [Wechsler, 2011], Kaufman Brief Intelligence Test-Second Edition [Kaufman & Kaufman, 2004], Reynolds Intellectual Assessment Scales [Reynolds & Kamphaus, 2003], etc.). While these measures possess utility in estimating intelligence, using short forms of the Wechsler intelligence scales (e.g., WAIS-IV) designed to fully assess intelligence (i.e., FSIQ) holds several benefits. WAIS-IV short forms provide greater flexibility than brief stand-alone intelligence measures, as the core 10-subtest WAIS-IV allows various subtest combinations that can better meet the needs of individual circumstances. Subtests from the Wechsler intelligence scales are also used to assess domain-specific neuropsychological functioning. Utilizing short forms of the WAIS-IV to estimate intelligence eliminates the need for additional measures to assess intelligence, which saves time and resources. Multiple short forms have been examined for the WAIS-R and WAIS-III; however, information about WAIS-IV short forms is more limited. Sattler and Ryan (2009) derived tables using the Tellegen and Briggs (1967) procedure enabling the estimation of FSIQ from combinations of various selected subtests. These short forms were calculated using normative data and have yet to be validated in clinical populations. In a more recent study, Girard, Axelrod, Patel, and Crawford (2014) evaluated all possible dyad combinations using the 10 core subtests from the WAIS-IV. They used a clinical sample of 482 subjects (mostly male and lowaverage FSIQ) from a VA Medical Center to assess the reliability and validity of the dyads based on data from the standardization sample derived from linear scaling of composite subtest scaled scores. They note that no dyad consistently ranked among the five highest values of the psychometric properties they examined, though dyads that included processing speed or working memory subtests performed highest when examining a composite measure of validity. The authors encouraged readers to consider their goals for short-form use when evaluating or selecting a dyad. However, FSIQ estimation tables based on their approach were not provided. Intelligence estimates derived from a clinical population, therefore, remain absent for the WAIS-IV. Additionally, estimation tables or regression equations do not presently exist for estimating GAI. The present study used a regression-based approach, derived from clinical data of a mixed neurological sample, to generate dyadic short form estimates of WAIS-IV FSIQ and GAI.
406
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Results Test Sample Demographic data for the test sample (N ¼ 113) can be found in Table 1. Means and SDs of test sample FSIQ, GAI, and subtest performances can be seen in Table 2. The test sample WAIS-IV FSIQ scores ranged from 55 to 135, with a mean of 92.5 (SD ¼ 15.9). GAI scores ranged from 59 to 130, with a mean of 95.7 (SD ¼ 15.2). The regression equations for FSIQ and GAI short forms can be found in Table 3. All standard multiple regressions were significant (p , .001). The R 2 values ranged from 70% to 83% for Table 1. Test sample and validation sample demographic data Test sample (n ¼ 113)
Variable
M (SD)
Range
M (SD)
Range
59.9 (13.0) 14.2 (3.2)
25–86 1 –20
59.3 (14.2) 14.7 (2.4)
23–83 9 –19
48 52
62 38
11 0 3 86
10 4 10 76
Notes: Race was not equally distributed in the population, x2 (3, N ¼ 163) ¼ 8.82, p , .05. Other demographic variables were not significantly different between groups.
Table 2. Means and SDs of the Wechsler Adult Intelligence Scale-Fourth Edition variables Test sample (n ¼ 113)
Variable
Actual Full Scale IQ Actual GAI Vocabulary Information Similarities Block Design Matrix Reasoning Visual Puzzles
Validation sample (n ¼ 50)
M (SD)
Range
M (SD)
Range
92.5 (15.9) 95.7 (15.2) 10.4 (2.8) 9.4 (3.0) 9.7 (3.0) 8.7 (3.2) 9.2 (3.4) 8.8 (3.1)
55–135 59–130 4 –19 4 –17 1 –16 1 –18 3 –18 2 –17
98.4 (14.5) 101.2 (14.8) 11.6 (3.0) 10.1 (2.9) 10.7 (2.5) 9.3 (3.5) 10.0 (3.2) 9.5 (2.8)
51–131 55–137 4 –19 5 –17 3 –17 1 –19 1 –16 4 –17
Table 3. Regression analyses for dyads for predicting FSIQ and GAI Dyad
FSIQ (n ¼ 113) F
V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP
*
524.07 263.03 381.07 414.18 301.10 289.42 504.40 267.21 319.01
GAI (n ¼ 113) R
2
.83 .70 .77 .79 .73 .72 .82 .71 .74
Predicted FSIQ equation
F*
R2
Predicted GAI equation
(V + BD)2.75 + 39.94 (V + MR)2.43 + 45.03 (V + VP)2.90 + 36.95 (I + BD)2.61 + 45.13 (I + MR)2.50 + 45.94 (I + VP)2.69 + 43.64 (S + BD)2.67 + 43.37 (S + MR)2.38 + 47.61 (S + VP)2.70 + 42.57
848.04 419.93 638.98 587.46 497.50 419.78 613.77 361.37 401.35
.88 .79 .85 .84 .82 .79 .85 .77 .78
(V + BD)2.72 + 43.66 (V + MR)2.46 + 47.55 (V + VP)2.90 + 39.98 (I + BD)2.58 + 48.91 (I + MR)2.53 + 48.59 (I + VP)2.69 + 46.82 (S + BD)2.59 + 47.93 (S + MR)2.36 + 51.02 (S + VP)2.66 + 46.62
Notes: Dyads are sums of age-scaled scores; V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles. *df ¼ (1, 111); all significant, p , .001.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Age Education Gender (%) Female Male Race (%) African American Asian Hispanic White
Validation sample (n ¼ 50)
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
407
FSIQ and 77% to 88% for GAI. Means and SDs of dyadic FSIQ and GAI estimations derived from the test sample were similar across dyad estimates (FSIQ estimates range from 92.5 to 92.6; GAI estimates range from 95.6 to 95.8) and slightly higher for GAI than FSIQ, in keeping with subjects’ actual FSIQ and GAI scores (see Table 4). Estimated FSIQ scores were derived for each dyad, based on the regression equations in Table 3, and are presented in Table 5. The estimated GAI scores were also derived for each dyad and can be found in Table 6. Due to a lack of actual scores at the low and high ends of the spectrum, some estimated FSIQ and GAI scores presented in Tables 5 and 6 are theoretical and not based on actual data. Correlations between estimated and actual FSIQ in the test sample were highly significant (p , .001) for all dyads, with corrected correlations ranging from .76 to .86 (see Table 7). Likewise, correlations between estimated and actual GAI in the test sample were also significant (p , .001), ranging from .82 to .91 (see Table 7). Validation Sample
Table 4. Means and SDs of Wechsler Adult Intelligence Scale-Fourth Edition dyadic estimations Test sample
Estimated FSIQ V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP Estimated GAI V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP
Validation sample
M
SD
Range
92.5 92.6 92.5 92.5 92.5 92.6 92.6 92.6 92.5
14.4 13.4 14.0 14.1 13.6 13.5 14.4 13.4 13.7
62–128 67–123 60–133 61–126 66–128 62–130 54–129 62–121 59–126
95.6 95.7 95.5 95.7 95.7 95.7 95.7 95.7 95.8
14.3 13.5 14.0 13.9 13.7 13.5 14.0 13.3 13.5
65–131 70–126 63–136 64–129 69–132 66–133 58–131 65–124 63–129
M
SD
Range
97.5 97.6 98.1 95.9 96.3 96.4 98.4 97.0 97.1
15.7 12.6 14.1 14.6 12.1 12.3 14.5 11.6 12.3
54–136 57–128 60–133 61–139 61–121 68–135 54–131 57–117 61–126
100.6 100.7 101.1 99.1 99.6 99.6 99.8 100.0 100.4
15.6 12.7 14.1 14.4 12.3 12.3 13.8 11.5 12.1
57–139 60–131 63–136 64–142 64–124 71–138 58–133 60–119 65–129
Notes: V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Demographic data for the validation sample (N ¼ 50) can be found in Table 1. Validation sample WAIS-IV FSIQ scores ranged from 51 to 131, with a mean of 98.4 (SD ¼ 14.5) and validation sample GAI scores ranged from 55 to 137, with a mean of 101.2 (SD ¼ 14.8). Means and SDs of validation sample FSIQ, GAI, and subtest performance can be seen in Table 2. No significant differences were noted between the test and validation groups in terms of age [t(161) ¼ .264, p ¼ .792] and education [t(161) ¼ 2.977, p ¼ .330]. However, mean FSIQ [t(101.969) ¼ 22.308, p ¼ .023; Cohen’s d ¼ 0.46] and GAI [t(161) ¼ 22.163, p ¼ .032; Cohen’s d ¼ 0.34] were significantly higher in the validation sample than in the test sample. No gender differences were found between the test and validation samples (x2[1, N ¼ 163] ¼ 2.81, p ¼ .094). Both the test and validation samples were primarily white (86% and 76%, respectively) (x2 [3, N ¼ 163] ¼ 8.82, p ¼ .03). Using Tables 5 and 6, estimated FSIQ and GAI scores were, respectively, derived for the cross-validation sample. Means and SDs of dyadic FSIQ and GAI estimations derived from the validation sample can be found in Table 4 and have a slightly larger range across dyad estimates than seen in the test sample. Corrected correlations between estimated and actual FSIQ (rsf′ s ¼ .73 to .88, p , .001) and GAI (rsf′ s ¼ .80 to .93, p , .001) were highly significant in the cross-validation sample (see Table 7) and similar to correlations in the test sample. Paired t-test analyses between actual and estimated FSIQ scores significantly differed only for I/BD [t(49) ¼ 2.445, p ¼ .018; Cohen’s d ¼ 0.70] in the cross-validation sample. Comparisons between actual and estimated FSIQ scores for the remaining dyads in the cross-validation sample were not significant. Similarly, paired t-test analyses revealed a significant difference between the
408
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Table 5. Estimated full scale IQ scores from V/BD, V/MR, V/VP, I/BD, I/MR, I/VP, S/BD, S/MR, and S/VP dyad age-scaled score sums V/BD
V/MR
V/VP
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
45 48 51 54 56 59 62 65 67 70 73 76 78 81 84 87 89 92 95 98 100 103 106 109 111 114 117 120 122 125 128 131 133 136 139 142 144
50 52 55 57 60 62 64 67 69 72 74 77 79 81 84 86 89 91 94 96 98 101 103 106 108 111 113 116 118 120 123 125 128 130 133 135 137
43 46 49 51 54 57 60 63 66 69 72 75 78 80 83 86 89 92 95 98 101 104 107 109 112 115 118 121 124 127 130 133 136 138 141 144 147
50 53 56 58 61 63 66 69 71 74 76 79 82 84 87 90 92 95 97 100 103 105 108 110 113 116 118 121 123 126 129 131 134 136 139 142 144
51 53 56 58 61 63 66 68 71 73 76 78 81 83 86 88 91 93 96 98 101 103 106 108 111 113 116 118 121 123 126 128 131 133 136 138 141
49 52 54 57 60 62 65 68 71 73 76 79 81 84 87 89 92 95 97 100 103 106 108 111 114 116 119 122 124 127 130 132 135 138 140 143 146
49 51 54 57 59 62 65 67 70 73 75 78 81 83 86 89 91 94 97 99 102 105 107 110 113 115 118 121 123 126 129 131 134 137 139 142 145
52 55 57 60 62 64 67 69 71 74 76 79 81 83 86 88 90 93 95 98 100 102 105 107 109 112 114 117 119 121 124 126 129 131 133 136 138
48 51 53 56 59 61 64 67 70 72 75 78 80 83 86 88 91 94 97 99 102 105 107 110 113 115 118 121 124 126 129 132 134 137 140 142 145
Notes: The italicized scores in the table were not derived from actual data, and thus the Full Scale IQ estimates outside of these boundaries are suspect; V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/ Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles.
actual and estimated GAI scores derived from I/BD (t[49] ¼ 2.336, p ¼ .024; Cohen’s d ¼ .67) in the cross-validation sample, while all other dyad estimates did not significantly differ from actual GAI scores. Estimated FSIQ and GAI scores were compared with their actual counterparts to examine how often estimated scores fell within 5 and 10 points of actual FSIQ and GAI (see Table 8). For FSIQ, estimations were within five points of actual scores among 40% – 54% of cases. V/BD, I/BD, and S/BD were the most accurate dyads, each falling within five points of actual scores in 54% of cases. FSIQ estimations were within 10 points of actual scores in 80% – 94% of cases. V/BD was the most accurate dyad, as 94% of estimations were within 10 points of actual scores. For GAI, estimated scores were within five points of actual scores among 54% – 76% of cases. V/BD was the most accurate dyad, with 76% of estimations within five points of actual scores. GAI estimations were within 10 points of actual scores in 84% – 96% of cases. V/BD was again the most accurate dyad. Percentages of estimated scores that fell within 5 and 10 points of actual scores were also examined by IQ ranges (see Table 8). Discrepancies between actual and estimated scores were lower in the 90– 100 FSIQ/GAI ranges; though, it was not common to have greater than a 10-point discrepancy in any range. In order to compare the current study regression technique to the Tellegen and Briggs technique, estimated FSIQ scores based on the V/BD dyad were generated from the Sattler and Ryan tables (2009, p. 245) in our validation sample. Paired-sample t-tests were used to compare means of estimated FSIQ to actual FSIQ scores using both approaches. The mean Sattler and Ryan estimated
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Sum of age-scaled scores
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
409
Table 6. Estimated General Ability Index scores from V/BD, V/MR, V/VP, I/BD, I/MR, I/VP, S/BD, S/MR, and S/VP dyad age-scaled score sums V/BD
V/MR
V/VP
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
49 52 54 57 60 63 65 68 71 74 76 79 82 84 87 90 93 95 98 101 103 106 109 112 114 117 120 122 125 128 131 133 136 139 142 144 147
52 55 57 60 62 65 67 70 72 75 77 80 82 84 87 89 92 94 97 99 102 104 107 109 112 114 116 119 121 124 126 129 131 134 136 139 141
46 49 52 54 57 60 63 66 69 72 75 78 81 83 86 89 92 95 98 101 104 107 110 112 115 118 121 124 127 130 133 136 139 141 144 147 150
54 57 59 62 64 67 70 72 75 77 80 82 85 88 90 93 95 98 101 103 106 108 111 113 116 119 121 124 126 129 131 134 137 139 142 144 147
54 56 59 61 64 66 69 71 74 76 79 81 84 87 89 92 94 97 99 102 104 107 109 112 114 117 119 122 124 127 130 132 135 137 140 142 145
52 55 58 60 63 66 68 71 74 76 79 82 84 87 90 93 95 98 101 103 106 109 111 114 117 119 122 125 128 130 133 136 138 141 144 146 149
53 56 58 61 63 66 69 71 74 76 79 82 84 87 89 92 95 97 100 102 105 108 110 113 115 118 120 123 126 128 131 133 136 139 141 144 146
56 58 60 63 65 68 70 72 75 77 79 82 84 86 89 91 94 96 98 101 103 105 108 110 112 115 117 119 122 124 127 129 131 134 136 138 141
52 55 57 60 63 65 68 71 73 76 79 81 84 87 89 92 95 97 100 102 105 108 110 113 116 118 121 124 126 129 132 134 137 140 142 145 148
Notes: The italicized scores in the table were not derived from actual data, and thus the General Ability Index estimates outside of these boundaries are suspect; V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles.
Table 7. Corrected correlations between actual and predicted indices
FSIQ Test sample Validation sample GAI Test sample Validation sample
V/BD
V/MR
V/VP
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
.86 .88
.76 .82
.82 .83
.84 .86
.79 .79
.77 .80
.86 .82
.76 .73
.79 .74
.91 .93
.84 .87
.88 .89
.88 .86
.85 .85
.84 .85
.88 .90
.82 .80
.84 .82
Notes: Correlations corrected using procedure described by Levy (1967). All significant, p , .001.
FSIQ (M ¼ 102.68, SD ¼ 16.78) was significantly higher than actual FSIQ ([M ¼ 98.40, SD ¼ 14.55], t [49] ¼ 4.54, p , .001; Cohen’s d ¼ 1.30). In contrast, there was not a significant difference between the current study regression-based estimated FSIQ (M ¼ 97.5, SD ¼ 15.7) and actual FSIQ (M ¼ 98.40, SD ¼ 14.55).
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Sum of age-scaled scores
410
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Table 8. Percentage of estimated scores within 5 and 10 points of actual scores in the cross-validation sample by FSIQ/GAI range
Actual FSIQ ≤89 (n ¼ 8) 90–100 (n ¼ 22) ≥101 (n ¼ 20) Total (N ¼ 50) Actual GAI ≤89 (n ¼ 8) 90–100 (n ¼ 18)
Total (N ¼ 50)
V/MR
V/VP
+5 +10 +5 +10 +5 +10 +5 +10
63 88 55 96 50 95 54 94
50 88 59 96 45 85 52 90
25 88 59 100 45 70 48 86
+5 +10 +5 +10 +5 +10 +5 +10
88 100 83 100 67 92 76 96
88 100 72 94 63 83 70 90
63 88 61 100 67 92 64 94
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
38 88 59 86 55 90 54 86
50 100 55 86 25 70 42 82
63 88 59 86 40 90 52 88
50 75 50 82 60 80 54 80
25 88 64 86 35 70 46 80
25 88 55 96 30 65 40 82
50 100 61 94 54 79 56 88
75 100 83 94 33 71 58 84
75 88 78 89 54 88 66 88
63 88 72 94 75 92 72 92
63 100 61 89 71 83 66 88
38 100 72 94 46 79 54 88
Discussion Short forms of the Wechsler intelligence scales have been used in research and clinical settings to acquire estimates of FSIQ. Sattler and Ryan (2009) provided data for various dyads of WAIS-IV subtests, but these were based on data from the standardization sample and have not been validated in a clinical sample. The present study examined nine WAIS-IV dyadic short forms, derived from a clinical sample, by entering subtest scores into regression equations to predict FSIQ and GAI. Estimated scores accounted for a significant amount of variance in each of the nine short forms used to predict FSIQ (70% – 83%) and GAI (77% – 88%). Each of the nine WAIS-IV dyadic short forms examined was a good predictor of FSIQ and GAI in a mixed diagnostic sample. The V/BD dyad yielded the best estimates in that it accounted for the greatest amount of variance in FSIQ and GAI scores, had the highest percentage of predicted scores within 5 (54% in FSIQ; 76% in GAI) and 10 (94% in FSIQ; 96% in GAI) points, and ′ ′ ¼ .88) and GAI (r sf ¼ .93) in the validation sample. Overall, estimates correlated had the highest correlation with actual FSIQ (r sf better with actual GAI than actual FSIQ. This is not surprising as processing speed and working memory subtests were not included in the dyads. Dyads with BD had the three highest R 2 values among the nine dyads for FSIQ (V/BD ¼ .83; I/BD ¼ .79; S/BD ¼ .82) and three of the four highest R 2 values for GAI (V/BD ¼ .88; I/BD ¼ .84; S/BD ¼ .85). This suggests that regardless of the subtest used from the VCI, BD is the best single subtest for predicting global ability. However, when patients have motor limitations, BD may not be the optimal subtest because of the necessary physical manipulation of the blocks. Of the non-motor perceptual reasoning subtests, VP appears to be better than MR at predicting FSIQ and GAI with the V/VP dyad serving as a more accurate predictor (FSIQ R 2 ¼ .77; GAI R 2 ¼ .85) than I/VP, S/VP, V/MR, I/MR, or S/MR. Girard and colleagues’ (2014) examination of V/BD and V/VP ranked these dyads lower than others based on their composite validity measure due to their tendency to overestimate FSIQ in their sample. This discrepancy may be due to differences in subject populations and Girard and colleagues’s inclusion of working memory and processing speed subtests in their analyses. Sattler and Ryan (2009) also identified V/BD as being among the 10 best two-subtest short forms, using the Tellegen and Briggs (1967) approach. Comparisons of Sattler and Ryan’s estimated FSIQ tables (p. 245) using the V/BD dyad to actual FSIQ scores in our validation sample showed that mean Sattler and Ryan estimated FSIQ was significantly higher than actual FSIQ. In contrast, there was not a significant difference between the current study regression-based V/BD estimated FSIQ and actual FSIQ. These findings indicate that our V/BD regression-based estimates of FSIQ derived from a clinical/neurologic sample were more accurate in predicting actual FSIQ than the previously reported Tellegen and Briggs method that used WAIS-IV standardization data. While our dyadic regressions appeared to produce more reliable estimates when used within neurological populations, further investigation is needed to determine the generalizability of these findings in other populations. Despite the high correlations between predicted and actual FSIQ and GAI scores, caution is recommended in interpreting overall levels of intelligence based on these estimations. A comparison of agreement between actual and estimated classification ranges (based on the WAIS-IV technical manual) in the cross-validation sample showed only 56% – 70% agreement among the FSIQ dyads while the GAI dyads ranged between 64% and 86% agreement. Complementing this finding is the relatively low
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
≥101 (n ¼ 24)
V/BD
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
411
Conflict of Interest None declared.
References Axelrod, B. N., Dingel, J. D., Ryan, J. J., & Ward, L. C. (2000). Estimation of the Wechsler Adult Intelligence Sacle-III with the 7-subtest short form in a clinical sample. Assessment, 7, 157– 161. Christensen, B. K., Girard, T. A., & Bagby, M. R. (2007). Wechsler Adult Intelligence Scale—Third Edition short form for index and IQ scores in a psychiatric population. Psychological Assessment, 19, 236–240. Engelhart, C. I., Eisenstein, J., Johnson, V., & Losonczy, M. (1999). Comparison of linear equating and prorated short forms for estimating WAIS-R FSIQ in a neuropsychological population. The Clinical Neuropsychologist, 13, 95– 99. Girard, T. A., Axelrod, B. N., Patel, R., & Crawford, J. R. (2014). Wechsler Adult Intelligence Scale-IV dyads for estimating global intelligence. Assessment, 1– 8, doi: 10.1177/1073191114551551. Jeyakumar, S. L. E., Warriner, E. M., Raval, V. V., & Ahmad, S. A. (2004). Balancing the need for reliability and time efficiency: Short forms of the Wechsler Adult Intelligence Scale-III. Educational and Psychological Measurement, 64, 71–87. Kaufman, A. S., Ishikuma, T., & Kaufman-Packer, J. L. (1991). Amazingly short forms of the WAIS-R. Journal of Psychoeducational Assessment, 9, 4 –15. Kaufman, A. S., & Kaufman, N. L. (2004). Kaufman Brief Intelligence Test Second Edition. Circle Pines, MN: AGS Publishing. Levy, P. (1967). The correction for spurious correlation in the evaluation of short-form tests. Journal of Clinical Psychology, 1, 84– 86. Mendella, P. D., McFadden, L., Regan, J., & Medlock, L. (2000). Short-form prediction of WAIS-R scores in a sample of individuals diagnoses with multiple sclerosis. Applied Neuropsychology, 7, 102– 107. Pilgrim, B. M., Meyers, J. E., Bayless, J., & Whetstone, M. M. (1999). Validity of the Ward seven-subtest WAIS-III short form in a neuropsychological population. Applied Neuropsychology, 6, 243– 246. Reynolds, C. R., & Kamphaus, R. W. (2003). Reynolds Intellectual Assessment Scales and Reynolds Intellectual Screening Test: Professional Manual. Odessa, FL: PAR.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
number of cases whose estimations fell within five points of actual FSIQ (40% – 54%) and GAI (54% – 76%) across all nine dyads. However, all nine dyads provided estimated FSIQs within 10 points of actual FSIQ in at least 80% of cases while GAI estimations were within 10 points of actual GAI in at least 84% of cases. It may be advisable to report estimated FSIQ and GAI scores as a range or classification level (e.g., low average, average, etc.) rather than a specific score. An obvious advantage of using dyadic short forms to provide FSIQ or GAI estimates is the time savings that are inherent with abbreviated testing. While published data on WAIS-IV subtest administration times in clinical samples are lacking, such information is available for the WAIS-III, which suggest that among the subtests used in this study (V, S, I, BD, and MR), the V/BD dyad is the most time consuming (roughly averaging 26 min) but accounts for, on average, ,30% of the time required for WAIS-III FSIQ. The I/MR dyad is the quickest (roughly averaging 14 min), accounting for only 16% of the time necessary for attainment of FSIQ (Ryan, Lopez, & Werth, 1998). The WAIS-IV was developed to allow for quicker subtest administration than its predecessors, suggesting that short forms presented in this study are likely speedier than what has been previously reported (Wechsler, 2008). While no such information about VP is available, clinical experience as well as knowledge of the mechanics of the subtest indicates that, on average, VP administration time is less than BD or MR. The I/VP dyad may, therefore, be the most efficient option among those included in this study. Limitations to the generalizability of these findings may include a lack of broad diversity among some demographic factors. While gender was well balanced, the test sample was primarily white (86%), highly educated (mean ¼ 14.2 years; median ¼ 15), and at the upper end of middle age (mean ¼ 59.9 years; median ¼ 62). Also, it is less clear how accurate predicted scores are in individuals at the lower and higher ends of the IQ spectrum given the relatively few subjects with FSIQ scores ,90 (n ¼ 8) or .109 (n ¼ 8) in the validation sample. Processing speed and working memory subtests were not used, and it remains to be seen if inclusion of these subtests appreciatively adds to these results. However, Girard and colleagues (2014) found that overall, the Coding and Information dyad had the strongest measures of reliability and validity in their sample. Their results also showed that 9 of the top 10 dyads included either a processing speed or working memory subtest. Thus, inclusion of processing speed and working memory subtests for estimating intelligence may be beneficial in some populations. Overall, the nine dyadic short forms of the WAIS-IV examined in this investigation were highly correlated with actual FSIQ and GAI scores in test and cross-validation samples. This model also demonstrated greater accuracy at predicting FSIQ in a clinical sample than the Sattler and Ryan tables for the V/BD dyad. Despite these high correlations, caution is recommended when reporting predicted scores, as relative ranges may be more appropriate. While utilizing short forms to predict FSIQ and GAI has inherent limitations, these nine dyads appear to be appropriate for use in mixed neurological populations when an estimate of overall intellectual functioning will suffice.
412
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Ringe, W. K., Saine, K. C., Lacritz, L. H., Hynan, L. S., & Cullum, C. M. (2002). Dyadic short forms of the Wechsler Adult Intelligence Scale-III. Assessment, 9, 254–260. Ryan, J. J., Lopez, S. J., & Werth, T. R. (1998). Administration time estimates for WAIS-III subtests, scales, and short forms in a clinical sample. Journal of Psychoeducational Assessment, 16, 315 –323. Satterfield, W. A., Martin, C. W., & Leiker, M. (1994). A comparison of four WAIS-R short forms in patients referred for psychological/neuropsychological assessments. Journal of Psychoeducational Assessment, 12, 364– 371. Sattler, J. M., & Ryan, J. J. (2009). Assessment with the WAIS-IV. San Antonio, TX: Psychological Corporation. Schoenberg, M. R., Duff, K., Dorfman, K., & Adams, R. L. (2004). Differential estimation of verbal intelligence and performance intelligence scores from combined performance and demographic variables: The OPIE-3 verbal and performance algorithms. Clinical Neuropsychology, 16, 266– 276. Schoenberg, M. R., Duff, K., Scott, J. G., & Adams, R. L. (2002). An evaluation of the clinical utility of the OPIE-3 as an estimate of premorbid WAIS-III FSIQ. Clinical Neuropsychologist, 17, 308–321. Schoenberg, M. R., Scott, J. G., Ruwe, W., Patton, D., & Adams, R. L. (2004). Assumptions that underlie predicting premorbid IQ: A comment on the “Evaluation of the accuracy of two regression-based methods for estimated premorbid IQ.” Archives of Clinical Neuropsychology, 19, 1103– 1106. Silverstein, A. B. (1982). Two- and four-subtest short forms of the Wechsler Adult Intelligence Scale-Revised. Journal of Consulting and Clinical Psychology, 50, 415–418. Tellegen, A., & Briggs, P. F. (1967). Old wine in new skins: Grouping Wechsler subtests into new scales. Journal of Consulting Psychology, 31, 504. Thompson, A. P., LoBello, S. G., Atkinson, L., Chisholm, V., & Ryan, J. J. (2004). Brief intelligence testing in Australia, Canada, the United Kingdom, and the United States. Professional Psychology: Research and Practice, 35, 286– 290. Ward, L. C. (1990). Prediction of verbal, performance, and Full Scale IQs from seven subtests of the WAIS-R. Journal of Clinical Psychology, 46, 436–440. Wechsler, D. (1955). Wechsler Adult Intelligence Scale. New York: Psychological Corporation. Wechsler, D. (1987). Wechsler Adult Intelligence Scale-Revised. San Antonio, TX: Psychological Corporation. Wechsler, D. (1997). Wechsler Adult Intelligence Scale-Third Edition. San Antonio, TX: Psychological Corporation. Wechsler, D. (2008). Wechsler Adult Intelligence Scale-Fourth Edition administration and scoring manual. San Antonio, TX: Psychological Corporation. Wechsler, D. (2011). Wechsler Abbreviated Scale of Intelligence (2nd ed.). Bloomington, MN: Pearson.
Dyadic Short Forms of the Wechsler Adult Intelligence Scale-IV David A. Denney1, Wendy K. Ringe1, Laura H. Lacritz 1,2,* 2
1 Department of Psychiatry, The University of Texas Southwestern Medical Center, Dallas, TX, USA Department of Neurology & Neurotherapeutics, The University of Texas Southwestern Medical Center, Dallas, TX, USA
*Corresponding author at: UT Southwestern Medical Center, Neuropsychology, 5323 Harry Hines Boulevard, Dallas, TX 75390-8846, USA. Tel.: +1 214 648 4646; fax: +1 214 648 4660. E-mail address: [email protected].
Abstract Full Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV) administration can be time-consuming and may not be necessary when intelligence quotient estimates will suffice. Estimated Full Scale Intelligence Quotient (FSIQ) and General Ability Index (GAI) scores were derived from nine dyadic short forms using individual regression equations based on data from a clinical sample (n ¼ 113) that was then cross validated in a separate clinical sample (n ¼ 50). Derived scores accounted for 70%– 83% of the variance in FSIQ and 77% –88% of the variance in GAI. Predicted FSIQs were strongly associated with actual FSIQ (rs ¼ .73–.88), as were predicted and actual GAIs (rs ¼ .80– .93). Each of the nine dyadic short forms of the WAIS-IV was a good predictor of FSIQ and GAI in the validation sample. These data support the validity of WAIS-IV short forms when time is limited or lengthier batteries cannot be tolerated by patients. Keywords: Intelligence; Assessment; Short form; Wechsler
Introduction The Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV; Wechsler, 2008) is a commonly employed measure of global intellectual and cognitive functioning. The WAIS-IV yields two measures of global functioning, the Full Scale Intelligence Quotient (FSIQ) and the General Ability Index (GAI). The FSIQ is a weighted index of verbal, perceptual reasoning, processing speed, and working memory abilities, whereas the GAI is a global measure that removes working memory and processing speed subtests. The GAI is new to the WAIS-IV and important in neuropsychological evaluations because it provides a summary score that is less sensitive than FSIQ to the influence of attention and/or processing speed deficits commonly seen in many neurologic populations. Short forms of previous WAIS versions (Wechsler, 1955, 1987, 1997) have been proposed to estimate intelligence quotient (IQ) in instances where full IQ testing is too lengthy for patients to tolerate or where an estimate of IQ will suffice. According to Thompson, LoBello, Atkinson, Chisholm, and Ryan (2004), the Wechsler intelligence scales are the most frequently applied measures used to derive short-form IQ estimations. Previous estimation methods include administering as few as one and as many as eight WAIS subtests to derive estimates (Axelrod, Dingel, Ryan, & Ward, 2000; Christensen, Girard, & Bagby, 2007; Engelhart, Eisenstein, Johnson, & Losonczy, 1999; Jeyakumar, Warriner, Raval, & Ahmad, 2004; Kaufman, Ishikuma, & Kaufman-Packer, 1991; Mendella, McFadden, Regan, & Medlock, 2000; Pilgrim, Meyers, Bayless, & Whetstone, 1999; Ringe, Saine, Lacritz, Hynan, & Cullum, 2002; Satterfield, Martin, & Leiker, 1994; Schoenberg, Duff, Dorfman, & Adams, 2004; Schoenberg, Duff, Scott, & Adams, 2002; Schoenberg, Scott, Ruwe, Patton, & Adams, 2004; Silverstein, 1982; Ward, 1990). These studies suggest that as short forms lengthen in number of subtests used to derive estimates, correlations between estimated and actual scores improve. However, as the number of subtests increase, the amount of time needed to attain the estimated score also increases. When time limitations factor into an evaluation, longer four- or seven-subtest short forms may not serve as an ideal solution. # The Author 2015. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected]. doi:10.1093/arclin/acv035 Advance Access publication on 8 June 2015
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Accepted 20 May 2015
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
405
Methods Participants Subjects included 163 consecutive individuals with known or suspected neurological disorders who underwent neuropsychological evaluation as part of their standard clinical care in the Neuropsychology Service at the University of Texas Southwestern Medical Center between September 2009 and January 2011. Subjects gave consent for their clinical data to be used for research purposes. The initial 113 consecutive individuals who presented for neuropsychological evaluation were included in the test group, and the following 50 consecutive patients comprised the cross-validation group. Procedures The core 10-subtest WAIS-IV was administered as part of a larger battery for clinical reasons, and all subjects gave written consent for their results to be used for research purposes. Short forms for FSIQ and GAI were derived by summing age-scaled scores from subtests administered from the Verbal Comprehension Index (VCI) (Vocabulary [V], Similarities [S], and Information [I]) to those from the Perceptual Reasoning Index (Block Design [BD], Matrix Reasoning [MR], and Visual Puzzles [VP]). This resulted in nine dyads: Vocabulary/Block Design (V/BD), Vocabulary/Matrix Reasoning (V/MR), Vocabulary/Visual Puzzles (V/VP), Similarities/Block Design (S/BD), Similarities/Matrix Reasoning (S/MR), Similarities/ Visual Puzzles (S/VP), Information/Block Design (I/BD), Information/Matrix Reasoning (I/MR), and Information/Visual Puzzles (I/VP). Within the test sample, the summations of the age-scaled scores for each dyad were entered into separate regression equations to predict FSIQ and GAI. Short form conversion tables were created from the regression equations, which were validated by comparing actual FSIQ and GAI scores of the test sample with estimated FSIQ and GAI scores acquired from the regression ′ ) between the short form predicted FSIQs and actual FSIQ, as well data. Composite reliability (rxx) and correlation coefficients (r sf as those between short form predicted GAIs and actual GAI, were calculated using the methods outlined by Levy (1967) to correct for inflated Pearson correlations. Results were cross-validated on a second mixed neurological sample in which corrected correlations (rsf′ ) between actual and estimated FSIQ and GAI were obtained. The percentages of estimated scores within 5 and 10 points of actual FSIQ and GAI scores were also calculated.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
There are several instruments designed to quickly estimate intelligence (e.g., Wechsler Abbreviated Scale of IntelligenceSecond Edition [Wechsler, 2011], Kaufman Brief Intelligence Test-Second Edition [Kaufman & Kaufman, 2004], Reynolds Intellectual Assessment Scales [Reynolds & Kamphaus, 2003], etc.). While these measures possess utility in estimating intelligence, using short forms of the Wechsler intelligence scales (e.g., WAIS-IV) designed to fully assess intelligence (i.e., FSIQ) holds several benefits. WAIS-IV short forms provide greater flexibility than brief stand-alone intelligence measures, as the core 10-subtest WAIS-IV allows various subtest combinations that can better meet the needs of individual circumstances. Subtests from the Wechsler intelligence scales are also used to assess domain-specific neuropsychological functioning. Utilizing short forms of the WAIS-IV to estimate intelligence eliminates the need for additional measures to assess intelligence, which saves time and resources. Multiple short forms have been examined for the WAIS-R and WAIS-III; however, information about WAIS-IV short forms is more limited. Sattler and Ryan (2009) derived tables using the Tellegen and Briggs (1967) procedure enabling the estimation of FSIQ from combinations of various selected subtests. These short forms were calculated using normative data and have yet to be validated in clinical populations. In a more recent study, Girard, Axelrod, Patel, and Crawford (2014) evaluated all possible dyad combinations using the 10 core subtests from the WAIS-IV. They used a clinical sample of 482 subjects (mostly male and lowaverage FSIQ) from a VA Medical Center to assess the reliability and validity of the dyads based on data from the standardization sample derived from linear scaling of composite subtest scaled scores. They note that no dyad consistently ranked among the five highest values of the psychometric properties they examined, though dyads that included processing speed or working memory subtests performed highest when examining a composite measure of validity. The authors encouraged readers to consider their goals for short-form use when evaluating or selecting a dyad. However, FSIQ estimation tables based on their approach were not provided. Intelligence estimates derived from a clinical population, therefore, remain absent for the WAIS-IV. Additionally, estimation tables or regression equations do not presently exist for estimating GAI. The present study used a regression-based approach, derived from clinical data of a mixed neurological sample, to generate dyadic short form estimates of WAIS-IV FSIQ and GAI.
406
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Results Test Sample Demographic data for the test sample (N ¼ 113) can be found in Table 1. Means and SDs of test sample FSIQ, GAI, and subtest performances can be seen in Table 2. The test sample WAIS-IV FSIQ scores ranged from 55 to 135, with a mean of 92.5 (SD ¼ 15.9). GAI scores ranged from 59 to 130, with a mean of 95.7 (SD ¼ 15.2). The regression equations for FSIQ and GAI short forms can be found in Table 3. All standard multiple regressions were significant (p , .001). The R 2 values ranged from 70% to 83% for Table 1. Test sample and validation sample demographic data Test sample (n ¼ 113)
Variable
M (SD)
Range
M (SD)
Range
59.9 (13.0) 14.2 (3.2)
25–86 1 –20
59.3 (14.2) 14.7 (2.4)
23–83 9 –19
48 52
62 38
11 0 3 86
10 4 10 76
Notes: Race was not equally distributed in the population, x2 (3, N ¼ 163) ¼ 8.82, p , .05. Other demographic variables were not significantly different between groups.
Table 2. Means and SDs of the Wechsler Adult Intelligence Scale-Fourth Edition variables Test sample (n ¼ 113)
Variable
Actual Full Scale IQ Actual GAI Vocabulary Information Similarities Block Design Matrix Reasoning Visual Puzzles
Validation sample (n ¼ 50)
M (SD)
Range
M (SD)
Range
92.5 (15.9) 95.7 (15.2) 10.4 (2.8) 9.4 (3.0) 9.7 (3.0) 8.7 (3.2) 9.2 (3.4) 8.8 (3.1)
55–135 59–130 4 –19 4 –17 1 –16 1 –18 3 –18 2 –17
98.4 (14.5) 101.2 (14.8) 11.6 (3.0) 10.1 (2.9) 10.7 (2.5) 9.3 (3.5) 10.0 (3.2) 9.5 (2.8)
51–131 55–137 4 –19 5 –17 3 –17 1 –19 1 –16 4 –17
Table 3. Regression analyses for dyads for predicting FSIQ and GAI Dyad
FSIQ (n ¼ 113) F
V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP
*
524.07 263.03 381.07 414.18 301.10 289.42 504.40 267.21 319.01
GAI (n ¼ 113) R
2
.83 .70 .77 .79 .73 .72 .82 .71 .74
Predicted FSIQ equation
F*
R2
Predicted GAI equation
(V + BD)2.75 + 39.94 (V + MR)2.43 + 45.03 (V + VP)2.90 + 36.95 (I + BD)2.61 + 45.13 (I + MR)2.50 + 45.94 (I + VP)2.69 + 43.64 (S + BD)2.67 + 43.37 (S + MR)2.38 + 47.61 (S + VP)2.70 + 42.57
848.04 419.93 638.98 587.46 497.50 419.78 613.77 361.37 401.35
.88 .79 .85 .84 .82 .79 .85 .77 .78
(V + BD)2.72 + 43.66 (V + MR)2.46 + 47.55 (V + VP)2.90 + 39.98 (I + BD)2.58 + 48.91 (I + MR)2.53 + 48.59 (I + VP)2.69 + 46.82 (S + BD)2.59 + 47.93 (S + MR)2.36 + 51.02 (S + VP)2.66 + 46.62
Notes: Dyads are sums of age-scaled scores; V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles. *df ¼ (1, 111); all significant, p , .001.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Age Education Gender (%) Female Male Race (%) African American Asian Hispanic White
Validation sample (n ¼ 50)
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
407
FSIQ and 77% to 88% for GAI. Means and SDs of dyadic FSIQ and GAI estimations derived from the test sample were similar across dyad estimates (FSIQ estimates range from 92.5 to 92.6; GAI estimates range from 95.6 to 95.8) and slightly higher for GAI than FSIQ, in keeping with subjects’ actual FSIQ and GAI scores (see Table 4). Estimated FSIQ scores were derived for each dyad, based on the regression equations in Table 3, and are presented in Table 5. The estimated GAI scores were also derived for each dyad and can be found in Table 6. Due to a lack of actual scores at the low and high ends of the spectrum, some estimated FSIQ and GAI scores presented in Tables 5 and 6 are theoretical and not based on actual data. Correlations between estimated and actual FSIQ in the test sample were highly significant (p , .001) for all dyads, with corrected correlations ranging from .76 to .86 (see Table 7). Likewise, correlations between estimated and actual GAI in the test sample were also significant (p , .001), ranging from .82 to .91 (see Table 7). Validation Sample
Table 4. Means and SDs of Wechsler Adult Intelligence Scale-Fourth Edition dyadic estimations Test sample
Estimated FSIQ V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP Estimated GAI V/BD V/MR V/VP I/BD I/MR I/VP S/BD S/MR S/VP
Validation sample
M
SD
Range
92.5 92.6 92.5 92.5 92.5 92.6 92.6 92.6 92.5
14.4 13.4 14.0 14.1 13.6 13.5 14.4 13.4 13.7
62–128 67–123 60–133 61–126 66–128 62–130 54–129 62–121 59–126
95.6 95.7 95.5 95.7 95.7 95.7 95.7 95.7 95.8
14.3 13.5 14.0 13.9 13.7 13.5 14.0 13.3 13.5
65–131 70–126 63–136 64–129 69–132 66–133 58–131 65–124 63–129
M
SD
Range
97.5 97.6 98.1 95.9 96.3 96.4 98.4 97.0 97.1
15.7 12.6 14.1 14.6 12.1 12.3 14.5 11.6 12.3
54–136 57–128 60–133 61–139 61–121 68–135 54–131 57–117 61–126
100.6 100.7 101.1 99.1 99.6 99.6 99.8 100.0 100.4
15.6 12.7 14.1 14.4 12.3 12.3 13.8 11.5 12.1
57–139 60–131 63–136 64–142 64–124 71–138 58–133 60–119 65–129
Notes: V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Demographic data for the validation sample (N ¼ 50) can be found in Table 1. Validation sample WAIS-IV FSIQ scores ranged from 51 to 131, with a mean of 98.4 (SD ¼ 14.5) and validation sample GAI scores ranged from 55 to 137, with a mean of 101.2 (SD ¼ 14.8). Means and SDs of validation sample FSIQ, GAI, and subtest performance can be seen in Table 2. No significant differences were noted between the test and validation groups in terms of age [t(161) ¼ .264, p ¼ .792] and education [t(161) ¼ 2.977, p ¼ .330]. However, mean FSIQ [t(101.969) ¼ 22.308, p ¼ .023; Cohen’s d ¼ 0.46] and GAI [t(161) ¼ 22.163, p ¼ .032; Cohen’s d ¼ 0.34] were significantly higher in the validation sample than in the test sample. No gender differences were found between the test and validation samples (x2[1, N ¼ 163] ¼ 2.81, p ¼ .094). Both the test and validation samples were primarily white (86% and 76%, respectively) (x2 [3, N ¼ 163] ¼ 8.82, p ¼ .03). Using Tables 5 and 6, estimated FSIQ and GAI scores were, respectively, derived for the cross-validation sample. Means and SDs of dyadic FSIQ and GAI estimations derived from the validation sample can be found in Table 4 and have a slightly larger range across dyad estimates than seen in the test sample. Corrected correlations between estimated and actual FSIQ (rsf′ s ¼ .73 to .88, p , .001) and GAI (rsf′ s ¼ .80 to .93, p , .001) were highly significant in the cross-validation sample (see Table 7) and similar to correlations in the test sample. Paired t-test analyses between actual and estimated FSIQ scores significantly differed only for I/BD [t(49) ¼ 2.445, p ¼ .018; Cohen’s d ¼ 0.70] in the cross-validation sample. Comparisons between actual and estimated FSIQ scores for the remaining dyads in the cross-validation sample were not significant. Similarly, paired t-test analyses revealed a significant difference between the
408
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Table 5. Estimated full scale IQ scores from V/BD, V/MR, V/VP, I/BD, I/MR, I/VP, S/BD, S/MR, and S/VP dyad age-scaled score sums V/BD
V/MR
V/VP
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
45 48 51 54 56 59 62 65 67 70 73 76 78 81 84 87 89 92 95 98 100 103 106 109 111 114 117 120 122 125 128 131 133 136 139 142 144
50 52 55 57 60 62 64 67 69 72 74 77 79 81 84 86 89 91 94 96 98 101 103 106 108 111 113 116 118 120 123 125 128 130 133 135 137
43 46 49 51 54 57 60 63 66 69 72 75 78 80 83 86 89 92 95 98 101 104 107 109 112 115 118 121 124 127 130 133 136 138 141 144 147
50 53 56 58 61 63 66 69 71 74 76 79 82 84 87 90 92 95 97 100 103 105 108 110 113 116 118 121 123 126 129 131 134 136 139 142 144
51 53 56 58 61 63 66 68 71 73 76 78 81 83 86 88 91 93 96 98 101 103 106 108 111 113 116 118 121 123 126 128 131 133 136 138 141
49 52 54 57 60 62 65 68 71 73 76 79 81 84 87 89 92 95 97 100 103 106 108 111 114 116 119 122 124 127 130 132 135 138 140 143 146
49 51 54 57 59 62 65 67 70 73 75 78 81 83 86 89 91 94 97 99 102 105 107 110 113 115 118 121 123 126 129 131 134 137 139 142 145
52 55 57 60 62 64 67 69 71 74 76 79 81 83 86 88 90 93 95 98 100 102 105 107 109 112 114 117 119 121 124 126 129 131 133 136 138
48 51 53 56 59 61 64 67 70 72 75 78 80 83 86 88 91 94 97 99 102 105 107 110 113 115 118 121 124 126 129 132 134 137 140 142 145
Notes: The italicized scores in the table were not derived from actual data, and thus the Full Scale IQ estimates outside of these boundaries are suspect; V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/ Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles.
actual and estimated GAI scores derived from I/BD (t[49] ¼ 2.336, p ¼ .024; Cohen’s d ¼ .67) in the cross-validation sample, while all other dyad estimates did not significantly differ from actual GAI scores. Estimated FSIQ and GAI scores were compared with their actual counterparts to examine how often estimated scores fell within 5 and 10 points of actual FSIQ and GAI (see Table 8). For FSIQ, estimations were within five points of actual scores among 40% – 54% of cases. V/BD, I/BD, and S/BD were the most accurate dyads, each falling within five points of actual scores in 54% of cases. FSIQ estimations were within 10 points of actual scores in 80% – 94% of cases. V/BD was the most accurate dyad, as 94% of estimations were within 10 points of actual scores. For GAI, estimated scores were within five points of actual scores among 54% – 76% of cases. V/BD was the most accurate dyad, with 76% of estimations within five points of actual scores. GAI estimations were within 10 points of actual scores in 84% – 96% of cases. V/BD was again the most accurate dyad. Percentages of estimated scores that fell within 5 and 10 points of actual scores were also examined by IQ ranges (see Table 8). Discrepancies between actual and estimated scores were lower in the 90– 100 FSIQ/GAI ranges; though, it was not common to have greater than a 10-point discrepancy in any range. In order to compare the current study regression technique to the Tellegen and Briggs technique, estimated FSIQ scores based on the V/BD dyad were generated from the Sattler and Ryan tables (2009, p. 245) in our validation sample. Paired-sample t-tests were used to compare means of estimated FSIQ to actual FSIQ scores using both approaches. The mean Sattler and Ryan estimated
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Sum of age-scaled scores
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
409
Table 6. Estimated General Ability Index scores from V/BD, V/MR, V/VP, I/BD, I/MR, I/VP, S/BD, S/MR, and S/VP dyad age-scaled score sums V/BD
V/MR
V/VP
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
49 52 54 57 60 63 65 68 71 74 76 79 82 84 87 90 93 95 98 101 103 106 109 112 114 117 120 122 125 128 131 133 136 139 142 144 147
52 55 57 60 62 65 67 70 72 75 77 80 82 84 87 89 92 94 97 99 102 104 107 109 112 114 116 119 121 124 126 129 131 134 136 139 141
46 49 52 54 57 60 63 66 69 72 75 78 81 83 86 89 92 95 98 101 104 107 110 112 115 118 121 124 127 130 133 136 139 141 144 147 150
54 57 59 62 64 67 70 72 75 77 80 82 85 88 90 93 95 98 101 103 106 108 111 113 116 119 121 124 126 129 131 134 137 139 142 144 147
54 56 59 61 64 66 69 71 74 76 79 81 84 87 89 92 94 97 99 102 104 107 109 112 114 117 119 122 124 127 130 132 135 137 140 142 145
52 55 58 60 63 66 68 71 74 76 79 82 84 87 90 93 95 98 101 103 106 109 111 114 117 119 122 125 128 130 133 136 138 141 144 146 149
53 56 58 61 63 66 69 71 74 76 79 82 84 87 89 92 95 97 100 102 105 108 110 113 115 118 120 123 126 128 131 133 136 139 141 144 146
56 58 60 63 65 68 70 72 75 77 79 82 84 86 89 91 94 96 98 101 103 105 108 110 112 115 117 119 122 124 127 129 131 134 136 138 141
52 55 57 60 63 65 68 71 73 76 79 81 84 87 89 92 95 97 100 102 105 108 110 113 116 118 121 124 126 129 132 134 137 140 142 145 148
Notes: The italicized scores in the table were not derived from actual data, and thus the General Ability Index estimates outside of these boundaries are suspect; V/BD ¼ Vocabulary/Block Design; V/MR ¼ Vocabulary/Matrix Reasoning; V/VP ¼ Vocabulary/Visual Puzzles; I/BD ¼ Information/Block Design; I/MR ¼ Information/Matrix Reasoning; I/VP ¼ Information/Visual Puzzles; S/BD ¼ Similarities/Block Design; S/MR ¼ Similarities/Matrix Reasoning; S/VP ¼ Similarities/Visual Puzzles.
Table 7. Corrected correlations between actual and predicted indices
FSIQ Test sample Validation sample GAI Test sample Validation sample
V/BD
V/MR
V/VP
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
.86 .88
.76 .82
.82 .83
.84 .86
.79 .79
.77 .80
.86 .82
.76 .73
.79 .74
.91 .93
.84 .87
.88 .89
.88 .86
.85 .85
.84 .85
.88 .90
.82 .80
.84 .82
Notes: Correlations corrected using procedure described by Levy (1967). All significant, p , .001.
FSIQ (M ¼ 102.68, SD ¼ 16.78) was significantly higher than actual FSIQ ([M ¼ 98.40, SD ¼ 14.55], t [49] ¼ 4.54, p , .001; Cohen’s d ¼ 1.30). In contrast, there was not a significant difference between the current study regression-based estimated FSIQ (M ¼ 97.5, SD ¼ 15.7) and actual FSIQ (M ¼ 98.40, SD ¼ 14.55).
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Sum of age-scaled scores
410
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Table 8. Percentage of estimated scores within 5 and 10 points of actual scores in the cross-validation sample by FSIQ/GAI range
Actual FSIQ ≤89 (n ¼ 8) 90–100 (n ¼ 22) ≥101 (n ¼ 20) Total (N ¼ 50) Actual GAI ≤89 (n ¼ 8) 90–100 (n ¼ 18)
Total (N ¼ 50)
V/MR
V/VP
+5 +10 +5 +10 +5 +10 +5 +10
63 88 55 96 50 95 54 94
50 88 59 96 45 85 52 90
25 88 59 100 45 70 48 86
+5 +10 +5 +10 +5 +10 +5 +10
88 100 83 100 67 92 76 96
88 100 72 94 63 83 70 90
63 88 61 100 67 92 64 94
I/BD
I/MR
I/VP
S/BD
S/MR
S/VP
38 88 59 86 55 90 54 86
50 100 55 86 25 70 42 82
63 88 59 86 40 90 52 88
50 75 50 82 60 80 54 80
25 88 64 86 35 70 46 80
25 88 55 96 30 65 40 82
50 100 61 94 54 79 56 88
75 100 83 94 33 71 58 84
75 88 78 89 54 88 66 88
63 88 72 94 75 92 72 92
63 100 61 89 71 83 66 88
38 100 72 94 46 79 54 88
Discussion Short forms of the Wechsler intelligence scales have been used in research and clinical settings to acquire estimates of FSIQ. Sattler and Ryan (2009) provided data for various dyads of WAIS-IV subtests, but these were based on data from the standardization sample and have not been validated in a clinical sample. The present study examined nine WAIS-IV dyadic short forms, derived from a clinical sample, by entering subtest scores into regression equations to predict FSIQ and GAI. Estimated scores accounted for a significant amount of variance in each of the nine short forms used to predict FSIQ (70% – 83%) and GAI (77% – 88%). Each of the nine WAIS-IV dyadic short forms examined was a good predictor of FSIQ and GAI in a mixed diagnostic sample. The V/BD dyad yielded the best estimates in that it accounted for the greatest amount of variance in FSIQ and GAI scores, had the highest percentage of predicted scores within 5 (54% in FSIQ; 76% in GAI) and 10 (94% in FSIQ; 96% in GAI) points, and ′ ′ ¼ .88) and GAI (r sf ¼ .93) in the validation sample. Overall, estimates correlated had the highest correlation with actual FSIQ (r sf better with actual GAI than actual FSIQ. This is not surprising as processing speed and working memory subtests were not included in the dyads. Dyads with BD had the three highest R 2 values among the nine dyads for FSIQ (V/BD ¼ .83; I/BD ¼ .79; S/BD ¼ .82) and three of the four highest R 2 values for GAI (V/BD ¼ .88; I/BD ¼ .84; S/BD ¼ .85). This suggests that regardless of the subtest used from the VCI, BD is the best single subtest for predicting global ability. However, when patients have motor limitations, BD may not be the optimal subtest because of the necessary physical manipulation of the blocks. Of the non-motor perceptual reasoning subtests, VP appears to be better than MR at predicting FSIQ and GAI with the V/VP dyad serving as a more accurate predictor (FSIQ R 2 ¼ .77; GAI R 2 ¼ .85) than I/VP, S/VP, V/MR, I/MR, or S/MR. Girard and colleagues’ (2014) examination of V/BD and V/VP ranked these dyads lower than others based on their composite validity measure due to their tendency to overestimate FSIQ in their sample. This discrepancy may be due to differences in subject populations and Girard and colleagues’s inclusion of working memory and processing speed subtests in their analyses. Sattler and Ryan (2009) also identified V/BD as being among the 10 best two-subtest short forms, using the Tellegen and Briggs (1967) approach. Comparisons of Sattler and Ryan’s estimated FSIQ tables (p. 245) using the V/BD dyad to actual FSIQ scores in our validation sample showed that mean Sattler and Ryan estimated FSIQ was significantly higher than actual FSIQ. In contrast, there was not a significant difference between the current study regression-based V/BD estimated FSIQ and actual FSIQ. These findings indicate that our V/BD regression-based estimates of FSIQ derived from a clinical/neurologic sample were more accurate in predicting actual FSIQ than the previously reported Tellegen and Briggs method that used WAIS-IV standardization data. While our dyadic regressions appeared to produce more reliable estimates when used within neurological populations, further investigation is needed to determine the generalizability of these findings in other populations. Despite the high correlations between predicted and actual FSIQ and GAI scores, caution is recommended in interpreting overall levels of intelligence based on these estimations. A comparison of agreement between actual and estimated classification ranges (based on the WAIS-IV technical manual) in the cross-validation sample showed only 56% – 70% agreement among the FSIQ dyads while the GAI dyads ranged between 64% and 86% agreement. Complementing this finding is the relatively low
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
≥101 (n ¼ 24)
V/BD
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
411
Conflict of Interest None declared.
References Axelrod, B. N., Dingel, J. D., Ryan, J. J., & Ward, L. C. (2000). Estimation of the Wechsler Adult Intelligence Sacle-III with the 7-subtest short form in a clinical sample. Assessment, 7, 157– 161. Christensen, B. K., Girard, T. A., & Bagby, M. R. (2007). Wechsler Adult Intelligence Scale—Third Edition short form for index and IQ scores in a psychiatric population. Psychological Assessment, 19, 236–240. Engelhart, C. I., Eisenstein, J., Johnson, V., & Losonczy, M. (1999). Comparison of linear equating and prorated short forms for estimating WAIS-R FSIQ in a neuropsychological population. The Clinical Neuropsychologist, 13, 95– 99. Girard, T. A., Axelrod, B. N., Patel, R., & Crawford, J. R. (2014). Wechsler Adult Intelligence Scale-IV dyads for estimating global intelligence. Assessment, 1– 8, doi: 10.1177/1073191114551551. Jeyakumar, S. L. E., Warriner, E. M., Raval, V. V., & Ahmad, S. A. (2004). Balancing the need for reliability and time efficiency: Short forms of the Wechsler Adult Intelligence Scale-III. Educational and Psychological Measurement, 64, 71–87. Kaufman, A. S., Ishikuma, T., & Kaufman-Packer, J. L. (1991). Amazingly short forms of the WAIS-R. Journal of Psychoeducational Assessment, 9, 4 –15. Kaufman, A. S., & Kaufman, N. L. (2004). Kaufman Brief Intelligence Test Second Edition. Circle Pines, MN: AGS Publishing. Levy, P. (1967). The correction for spurious correlation in the evaluation of short-form tests. Journal of Clinical Psychology, 1, 84– 86. Mendella, P. D., McFadden, L., Regan, J., & Medlock, L. (2000). Short-form prediction of WAIS-R scores in a sample of individuals diagnoses with multiple sclerosis. Applied Neuropsychology, 7, 102– 107. Pilgrim, B. M., Meyers, J. E., Bayless, J., & Whetstone, M. M. (1999). Validity of the Ward seven-subtest WAIS-III short form in a neuropsychological population. Applied Neuropsychology, 6, 243– 246. Reynolds, C. R., & Kamphaus, R. W. (2003). Reynolds Intellectual Assessment Scales and Reynolds Intellectual Screening Test: Professional Manual. Odessa, FL: PAR.
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
number of cases whose estimations fell within five points of actual FSIQ (40% – 54%) and GAI (54% – 76%) across all nine dyads. However, all nine dyads provided estimated FSIQs within 10 points of actual FSIQ in at least 80% of cases while GAI estimations were within 10 points of actual GAI in at least 84% of cases. It may be advisable to report estimated FSIQ and GAI scores as a range or classification level (e.g., low average, average, etc.) rather than a specific score. An obvious advantage of using dyadic short forms to provide FSIQ or GAI estimates is the time savings that are inherent with abbreviated testing. While published data on WAIS-IV subtest administration times in clinical samples are lacking, such information is available for the WAIS-III, which suggest that among the subtests used in this study (V, S, I, BD, and MR), the V/BD dyad is the most time consuming (roughly averaging 26 min) but accounts for, on average, ,30% of the time required for WAIS-III FSIQ. The I/MR dyad is the quickest (roughly averaging 14 min), accounting for only 16% of the time necessary for attainment of FSIQ (Ryan, Lopez, & Werth, 1998). The WAIS-IV was developed to allow for quicker subtest administration than its predecessors, suggesting that short forms presented in this study are likely speedier than what has been previously reported (Wechsler, 2008). While no such information about VP is available, clinical experience as well as knowledge of the mechanics of the subtest indicates that, on average, VP administration time is less than BD or MR. The I/VP dyad may, therefore, be the most efficient option among those included in this study. Limitations to the generalizability of these findings may include a lack of broad diversity among some demographic factors. While gender was well balanced, the test sample was primarily white (86%), highly educated (mean ¼ 14.2 years; median ¼ 15), and at the upper end of middle age (mean ¼ 59.9 years; median ¼ 62). Also, it is less clear how accurate predicted scores are in individuals at the lower and higher ends of the IQ spectrum given the relatively few subjects with FSIQ scores ,90 (n ¼ 8) or .109 (n ¼ 8) in the validation sample. Processing speed and working memory subtests were not used, and it remains to be seen if inclusion of these subtests appreciatively adds to these results. However, Girard and colleagues (2014) found that overall, the Coding and Information dyad had the strongest measures of reliability and validity in their sample. Their results also showed that 9 of the top 10 dyads included either a processing speed or working memory subtest. Thus, inclusion of processing speed and working memory subtests for estimating intelligence may be beneficial in some populations. Overall, the nine dyadic short forms of the WAIS-IV examined in this investigation were highly correlated with actual FSIQ and GAI scores in test and cross-validation samples. This model also demonstrated greater accuracy at predicting FSIQ in a clinical sample than the Sattler and Ryan tables for the V/BD dyad. Despite these high correlations, caution is recommended when reporting predicted scores, as relative ranges may be more appropriate. While utilizing short forms to predict FSIQ and GAI has inherent limitations, these nine dyads appear to be appropriate for use in mixed neurological populations when an estimate of overall intellectual functioning will suffice.
412
D.A. Denney et al. / Archives of Clinical Neuropsychology 30 (2015); 404–412
Downloaded from http://acn.oxfordjournals.org/ at University of Sussex on July 30, 2015
Ringe, W. K., Saine, K. C., Lacritz, L. H., Hynan, L. S., & Cullum, C. M. (2002). Dyadic short forms of the Wechsler Adult Intelligence Scale-III. Assessment, 9, 254–260. Ryan, J. J., Lopez, S. J., & Werth, T. R. (1998). Administration time estimates for WAIS-III subtests, scales, and short forms in a clinical sample. Journal of Psychoeducational Assessment, 16, 315 –323. Satterfield, W. A., Martin, C. W., & Leiker, M. (1994). A comparison of four WAIS-R short forms in patients referred for psychological/neuropsychological assessments. Journal of Psychoeducational Assessment, 12, 364– 371. Sattler, J. M., & Ryan, J. J. (2009). Assessment with the WAIS-IV. San Antonio, TX: Psychological Corporation. Schoenberg, M. R., Duff, K., Dorfman, K., & Adams, R. L. (2004). Differential estimation of verbal intelligence and performance intelligence scores from combined performance and demographic variables: The OPIE-3 verbal and performance algorithms. Clinical Neuropsychology, 16, 266– 276. Schoenberg, M. R., Duff, K., Scott, J. G., & Adams, R. L. (2002). An evaluation of the clinical utility of the OPIE-3 as an estimate of premorbid WAIS-III FSIQ. Clinical Neuropsychologist, 17, 308–321. Schoenberg, M. R., Scott, J. G., Ruwe, W., Patton, D., & Adams, R. L. (2004). Assumptions that underlie predicting premorbid IQ: A comment on the “Evaluation of the accuracy of two regression-based methods for estimated premorbid IQ.” Archives of Clinical Neuropsychology, 19, 1103– 1106. Silverstein, A. B. (1982). Two- and four-subtest short forms of the Wechsler Adult Intelligence Scale-Revised. Journal of Consulting and Clinical Psychology, 50, 415–418. Tellegen, A., & Briggs, P. F. (1967). Old wine in new skins: Grouping Wechsler subtests into new scales. Journal of Consulting Psychology, 31, 504. Thompson, A. P., LoBello, S. G., Atkinson, L., Chisholm, V., & Ryan, J. J. (2004). Brief intelligence testing in Australia, Canada, the United Kingdom, and the United States. Professional Psychology: Research and Practice, 35, 286– 290. Ward, L. C. (1990). Prediction of verbal, performance, and Full Scale IQs from seven subtests of the WAIS-R. Journal of Clinical Psychology, 46, 436–440. Wechsler, D. (1955). Wechsler Adult Intelligence Scale. New York: Psychological Corporation. Wechsler, D. (1987). Wechsler Adult Intelligence Scale-Revised. San Antonio, TX: Psychological Corporation. Wechsler, D. (1997). Wechsler Adult Intelligence Scale-Third Edition. San Antonio, TX: Psychological Corporation. Wechsler, D. (2008). Wechsler Adult Intelligence Scale-Fourth Edition administration and scoring manual. San Antonio, TX: Psychological Corporation. Wechsler, D. (2011). Wechsler Abbreviated Scale of Intelligence (2nd ed.). Bloomington, MN: Pearson.
