WAIS-IV Seven-Subtest Short Form: Validity and Clinical Use in Schizophrenia

Transcription

1 Archives of Clinical Neuropsychology 31 (2016) WAIS-IV Seven-Subtest Short Form: Validity and Clinical Use in Schizophrenia Ewa Bulzacka 1,2, *, John E. Meyers 3, Laurent Boyer 2,4, Tifenn Le Gloahec 1,2, Guillaume Fond 1,2,5,6, Andrei Szöke 1,2,5,6, Marion Leboyer 1,2,5,6, Franck Schürhoff 1,2,5,6 1 Department of Psychiatry, AP-HP, Henri Mondor-Albert Chenevier Group, Créteil F-94000, France 2 Fondation FondaMental (RTRS Santé Mentale), Créteil F-94000, France 3 Private Practice, Mililani 96789, Hawaii 4 Pôle psychiatrie universitaire, CHU Sainte-Marguerite, Marseille F-13274, France 5 INSERM U955, Translational Psychiatry Team, Créteil F-94000, France 6 Faculty of Medicine, University Paris Est-Créteil, Créteil F-94000, France *Corresponding author at: Centre Expert Schizophrénie, Pôle de Psychiatrie et d Addictologie des Hôpitaux Universitaires Henri Mondor, 40 rue de Mesly, CRETEIL, France. Tel.: (331) ; fax: (331) address: ewa.bulzacka@gmail.com (E. Bulzacka). Accepted 22 July 2016 Abstract Objective: This study assesses the psychometric properties of Ward s seven-subtest short form (SF) for WAIS-IV in a sample of adults with schizophrenia (SZ) and schizoaffective disorder. Method: Seventy patients diagnosed with schizophrenia or schizoaffective disorder were administered the full version of the WAIS-IV. Four different versions of the Ward s SF were then calculated. The subtests used were: Similarities, Digit Span, Arithmetic, Information, Coding, Picture Completion, and Block Design (BD version) or Matrix Reasoning (MR version). Prorated and regression-based formulae were assessed for each version. Results: The actual and estimated factorial indexes reflected the typical pattern observed in schizophrenia. The four SFs correlated significantly with their full-version counterparts, but the Perceptual Reasoning Index (PRI) correlated below the acceptance threshold for all four versions. The regression-derived estimates showed larger differences compared to the full form. The four forms revealed comparable but generally low clinical category agreement rates for factor indexes. All SFs showed an acceptable reliability, but they were not correlated with clinical outcomes. Conclusions: The WAIS-IV SF offers a good estimate of WAIS-IV intelligence quotient, which is consistent with previous results. Although the overall scores are comparable between the four versions, the prorated forms provided a better estimation of almost all indexes. MR can be used as an alternative for BD without substantially changing the psychometric properties of the SF. However, we recommend a cautious use of these abbreviated forms when it is necessary to estimate the factor index scores, especially PRI, and Processing Speed Index. Keywords: WAIS-IV; Wechsler Adult Intelligence Scale; Schizophrenia; Intelligence; Short form; IQ Introduction Mental, neurological, and substance-use (MNS) disorders constitute 13% of the global disease burden, surpassing both cardiovascular disease and cancer (Collins et al., 2011; Wittchen et al., 2011). MNS reduces the ability to adapt to a variety of real-life contexts. Wechsler defined intelligence as the aggregate or global capacity of the individual to act purposefully, to think rationally and to deal effectively with his environment (1939). A careful assessment of the abilities that enable the subject to adapt to his or her environment has been increasingly developed in the last few decades. As intellectual ability is no longer considered a single entity, but rather, a multifactorial construct covering multiple cognitive areas, the full assessment of cognitive functioning requires time and the selection of adapted tools. The Author Published by Oxford University Press. All rights reserved. For permissions, please journals.permissions@oup.com. doi: /arclin/acw063 Advance Access publication on 1 September 2016

2 916 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Cognition largely contributes to functional outcome in terms of professional insertion, community functioning, ability to acquire new skills and to manage everyday activities (Green & Harvey, 2014). In patients with schizophrenia (SZ), the Intellectual Quotient (IQ) has been shown a more sensitive and reliable predictor of functional outcome than specific cognitive measures (Leeson, Barnes, Hutton, Ron, & Joyce, 2009). Full-Scale IQ (FSIQ) is strongly associated with educational attainment, occupational level, and school achievement (Wechsler, 2008). An association between premorbid impaired IQ and the subsequent onset of psychosis has also been described (David, Malmberg, Brandt, Allebeck, & Lewis, 1997) which supports the neurodevelopmental contribution to schizophrenia. Cognitive functioning was suggested to be more accurate than neurobiology in differentiating SZ patients from healthy controls (Heinrichs, 2005). Systematic assessment of IQ has, therefore, been recommended in SZ by some authors (Harrison-Read, 2008). The Wechsler Adult Intelligence Scale (WAIS) is one of the most commonly used cognitive batteries designed to assess subject s IQ, a proxy of general cognitive functioning. Beyond the traditional dichotomic verbal versus performance intelligence distinction (Wechsler, 2008), the current version of the WAIS (WAIS-IV) is based on four major components: the Verbal Comprehension Index (VCI), the Perceptual Reasoning Index (PRI), the Working Memory Index (WMI) and the Processing Speed Index (PSI). FSIQ and General Ability Index (GAI) can be generated from these scores. Overall, the WAIS-IV includes 10 core subtests and 5 supplemental subtests. A typical pattern of performance among patients with SZ can be observed across the WAIS subscales. Tasks involving processing speed display the largest group differences, followed by working memory, and perceptual organization. The least impaired factor is verbal comprehension (Michel et al., 2013). Although in-depth assessment can certainly increase the understanding a of patient s cognitive profile and is a prerequisite for cognitive rehabilitation, in the context of efforts to decrease reimbursed assessment time for clinicians and researchers, the use of well-established abbreviated forms of the Wechsler scale may be helpful. The time savings could leave room for a more specific neuropsychological assessment. The mean administration time to obtain the FSIQ with the 4th edition (Wechsler, 2008) is estimated around 67 min, which is 13 min shorter than that of the WAIS-III (Sattler & Ryan, 2009). However, according to Ryan et al. (1998), the administration time may be increased by 50% 80% in patients suffering from neurological conditions. Schizophrenic patients may experience similar difficulties due to broad cognitive impairment (Heinrichs & Zakzanis, 1998), motivational problems, and low processing speed leading to fatigability. To this end, different short forms (SFs) of the Wechsler Scales have been developed. One of the most widely used was published by Ward (1990). This seven-subtest SF initially conceived for the Wechsler Adult Intelligence Scale-Revised (WAIS-R) includes Block Design (BD), Similarities (SI), Digit Span (DS), Arithmetic (AR), Information (IN), Coding (CD), and Picture Completion (PC) subtests. Each test is administered in original form according to the manual s instructions. The scaled scores are then weighted: Verbal IQ (VIQ) raw = 2 (IN + SI) + (DS + AR); Performance IQ (PIQ) raw = 2 (PC + BD) + (CD); FSIQ raw is the sum of the VIQ raw and the PIQ raw. IQ scores are obtained from the usual conversion tables from the manual. The administration time for this abbreviated form is estimated to be 46 min (Ryan et al., 1998). This SF offered a good estimation of FSIQ, VIQ and PIQ composite factors, as well as an excellent reliability (Axelrod, Woodard, Schretlen, & Benedict, 1996). Pilgrim, Meyers, Bayless, and Whetstone (1999) reported equivalent psychometric properties of the SF of the WAIS-III. The correlations found were 0.95 for PIQ, 0.97 for VIQ, and 0.98 for FSIQ. FSIQ and VIQ maintained good reliability with the full-version scores in both 2nd and 3rd editions (Axelrod & Paolo, 1998; Axelrod, Ryan, & Ward, 2001). The psychometric properties of this SF are conserved in other populations, such as traumatic brain injury patients (Guilmette, Dabrowski, Kennedy, & Gnys, 1999). These authors underlined the superiority of this version over other SF based on an absolute mean of differences between predicted and effective scores. Ryan and Ward (1999) published an alternative form that substitutes MR for BD. The MR form (7MRsf, as opposed to 7BDsf for BD version) was less time consuming, had a good reliability, and showed a high correlation with the initial form. Moreover, the 7MRsf seemed to slightly improve the accuracy of PIQ (Axelrod et al., 2001). Despite brief administration time and independent standardization of shorter instruments, such as WASI (Wechsler, 1999), the estimation of intellectual functioning reduced to very few scores provides less information about specific cognitive abilities and may lead to a lower validity of the estimation (Wymer, Rayls, & Wagner, 2003). More detailed information about underlying cognitive domains is often relevant in the context of the neuropsychological evaluation. Ward s SF was studied in a sample of 304 psychiatric inpatients (Benedict, Schretlen, & Bobholz, 1992). The results showed smaller errors in predicted IQ compared to two other SF and a better predictive power for the Ward s SF. Another study conducted in a mixed psychiatric sample confirmed the superiority of the Ward SF over shorter instruments in terms of reliability and validity (Abraham, Axelrod, & Paolo, 1997). Allen et al. (1997) assessed nine different SF in patients diagnosed with SZ. Compared to other SFs (Kaufman, Ishikuma, & Kaufman-Packer, 1991), the Ward s seven-subtest SF showed the lowest misclassification rate, lower difference from the

3 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); full-version IQ and the best correlation with the FSIQ. The authors advocate the use of Ward SF in schizophrenia when assessment time is not limited. Later, Ryan, Weilage, and Spaulding (1999) examined the accuracy of the SF for predicting IQs of 73 SZ inpatients. Results indicated that 93% of the estimated FSIQs were within ±5 points of their actual scores. The authors suggested using the SF with schizophrenic patients when only general estimates of intellectual functioning are required which is consistent with previous findings (Iverson, Guirguis, & Green, 1998). More recently, based on WAIS-IV, Meyers, Zellinger, Kockler, Wagner, and Miller (2013) examined 7BDsf properties in a sample of 102 subjects with different neurological, psychiatric, and somatic conditions. The authors compared two different methods to estimate FSIQ and index scores based on Ward s seven subtests: (i) the original method based on prorated scores and (ii) a regression-based method. Both methods were stable and adequately predictive of their full-version counterparts. The FSIQ was best predicted by the prorated scores method which is also easier to use in clinical practice. If index scores are also needed, the prorated score was more accurate for the VCI and nearly as accurate for the PRI. The PSI prediction showed better accuracy with the weighted regression. To date, no studies have been conducted to assess Ward s SF of the latest edition of the Wechsler scale in schizophrenia. The objective of this study was, therefore, to assess the properties of Ward s seven-subtest SF for WAIS-IV in a sample of patients diagnosed with SZ. We examined four different SF calculation algorithms to determine the most suitable one. Methods Study Sample Seventy outpatients with a diagnosis of schizophrenia or schizoaffective disorder, according to DSM-5 (American Psychiatric Association, 2013) criteria were consecutively assessed in a university-affiliated hospital (Cognitive and Social Rehabilitation Centre, Psychiatry Department, Hôpitaux Universitaires Henri Mondor, Assistance Publique des Hôpitaux de Paris (AP-HP)). The collected data were a part of a regular care process. A psychiatric interview was conducted with every participant by two experienced psychiatrists to confirm the diagnosis of schizophrenia or schizoaffective disorder and to ensure that patients were in clinical remission according to the remission criteria of Andreasen et al. (2014). All subjects were French-native speakers. Patients with a history of central nervous system disorder (including epilepsy, brain injury, and neurological disorders), uncorrected perceptive disorders affecting the understanding of oral and written instructions, a substance or alcohol disorder in the 6 months prior to participation, an electroconvulsive therapy treatment during the 12 months prior to participation, or a current depressive or manic episode were not included in the study. Socio-demographic data were collected. The sample characteristics are described in Table 1 and show no significant differences between patients with SZ and schizoaffective disorder. Table 1. Social, demographic, and cognitive characteristics of the study sample Whole sample Schizophrenia Schizoaffective disorder Sig. (2-sided) N (70%) 21 (30%) Sex, % male 56 (80%) 38 (77.6%) 18 (85.7%) p =.43 Age, mean (SD) 31.0 (9.6) 31.0 (10.0) 30.9 (9.0) p =.89 Handedness, % right handed 63 (90%) 46 (93.9%) 17 (81.0%) p =.23 Education level p =.44 Education class 1 17 (24%) 14 (28.6%) 3 (14.3%) Education class 2 21 (30%) 14 (28.6%) 7 (33.3%) Education class 3 32 (46%) 21 (42.9%) 11 (52.4%) Illness duration, years, mean (SD) 7.5 (5.9) 7.43 (6.7) 7.76 (3.4) p =.14 Functional outcome*, mean (SD) (16.4) (17.1) (13.8) p =.15 WAIS-IV, mean (SD) FSIQ 81.1 (15.5) 83.0 (15.3) 76.7 (15.4) p =.12 VCI 88.9 (17.1) 90.0 (15.4) 86.3 (20.8) p =.48 PRI 84.3 (15.4) 86.5 (15.7) 79.1 (13.8) p =.07 WMI 83.9 (14.1) 85.7 (14.4) 79.7 (12.6) p =.10 PSI 80.9 (14.3) 81.9 (14.8) 78.5 (13.0) p =.37 * Total score on Social Autonomy Scale. WAIS-IV, Wechsler Adult Intelligence Scale-IV; FSIQ, Full Scale Intellectual Quotient; VCI, Verbal Comprehension Index; PRI, Perceptive Reasoning Index; WMI, Working Memory Index; PSI, Processing Speed Index.

4 918 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Ethical Agreement This study involved data from regular care, no written consent was therefore required. Each patient was informed about a possible use of their data for research purpose and their ability to oppose this use at any time in which case patient s data were excluded from the analysis. A non-opposition form was signed by all participants. All data were collected and made anonymous. The study was carried out in accordance with ethical principles for medical research involving humans (WMA, Declaration of Helsinki). Measurements The neuropsychological assessment was conducted by a trained neuropsychologist and included WAIS 4th edition. All participants received a full version of WAIS-IV including all core subtests and the optional picture completion subtest that is necessary to calculate the SF. Functional outcome was assessed by a trained psychiatric nurse using the Social Autonomy Scale (SAS) (Leguay et al., 1998). This hetero-assessment instrument investigates the self-sufficiency level of the persons who suffer from chronic psychiatric conditions and measures subjects abilities to face up to tangible problems in everyday life. Data Analysis Based on the full WAIS-IV, four index scores (VCI, PRI, WMI, and PSI) and FSIQ composite score were generated. Descriptive statistics were computed for the demographic and neuropsychological measures (Table 1). Educational level was defined as the highest completed school grade and recorded as a trichotomous variable: (i) no formal education or uncompleted primary, completed primary, and uncompleted lower secondary; (ii) uncompleted secondary or any other case between (i) and (iii); (iii) at least some secondary education completed (Pichot, Lebeaux, Penhouët, & Simon, 1993). The data were examined for normal distribution (skewness, kurtosis, and outliers analysis). A preliminary analysis of the standard FSIQ and factor indexes was performed in order to compare the WAIS-IV profile in SZ patients with normative data issued from standardization subsample (Wechsler, 2008) using one sample t-tests. Ward s SF estimated indexes were calculated for each participant, using (i) prorated scores method and (ii) regressionbased method, for both BD (7BDsf) (Meyers et al., 2013) and Matrix Reasoning (7MRsf) (Meyers, personal communication) versions. Table 2 summarizes the formulae used to calculate the four SF. The prorated formulae were used to obtain the raw indexes that were then transformed according to conversion tables from the manual. The regression-based formulae directly provided the estimated values. The four SF formulae were evaluated following nine different criteria found in the literature (Donders & Axlerod, 2002; Iverson, Myers, & Adams, 1997; Meyers et al., 2013; Ryan et al., 1999): (i) correlation coefficients, (ii) mean comparison, Table 2. SF calculation algorithms using prorated scores and regression formulae* 7BDsf 7MRsf Prorated formulae FSIQ ((IN + SI + BD + PC + DS + AR + CD) 10)/7 ((IN + SI + MR + PC + DS + AR + CD) 10)/7 VCI ((IN + SI) 3)/2 ((IN + SI) 3)/2 PRI ((BD + PC) 3)/2 ((MR + PC) 3)/2 WMI DS + AR DS + AR PSI CD 2 CD 2 Regression-based formulae FSIQ 0.941(SI) (BD) (IN) (DS) (CD) (AR) (PC) (SI) (MR) (IN) (DS) (CD) (AR) (PC) VCI 2.937(SI) (IN) (SI) (IN) PRI 2.794(BD) (AR) (PC) (MR) (PC) WMI 2.835(DS) (AR) (AR) (DS) PSI 0.921(BD) (CD) (CD) * 7BDsf, Ward s seven-subtest SF (BD version); 7MRsf, Ward s seven-subtest SF (MR version); FSIQ, Full-Scale Intelligence Quotient; VCI, Verbal Comprehension Index; PRI, Perceptual Reasoning Index; WMI, Working Memory Index; PSI, Processing Speed Index; IN, Information; SI, Similarities; BD, Block Design; PC, Picture Completion; DS, Digit Span; AR, Arithmetic; CD, Coding; MR, Matrix Reasoning.

5 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); (iii) difference scores, (iv) category agreement, (v) band of error analysis, (vi) clinical accuracy, (vii) time savings, (viii) internal consistency reliability, and (ix) criterion validity. The SF scores were compared to the full-version scores through correlation analysis and comparison of means. As the comparisons between SF and full-form indexes were derived from the same administration of the WAIS-IV, which may lead to inflated Spearman correlations, we designated correlations above 0.90 as sufficiently accurate for clinical practice (Iverson et al., 1998). Dependent-sample t-tests were run to assess differences between SF and full-form scores. In case of normality violation, additional Wilcoxon tests were performed to confirm the result. The magnitude of the difference was estimated using Cohen s d in order to determine whether the difference between the actual and the estimated scores was clinically relevant. Cohen s d for within-subjects studies was corrected by the correlation between the means (Morris & DeShon, 2002). Mean score differences (Δ = Full version Short version) were provided. Next, each composite score (FSIQ, VCI, PRI, WMI, and PSI) was sorted on the basis of Wechsler s intelligence classification (manual) into seven groups: Extremely low, Borderline, Low Average, Average, High Average, Superior, Very superior, in order to perform a category agreement analysis (Iverson et al., 1997). We set an 80% agreement threshold to consider a SF clinically acceptable. A band of error analysis consisting of identification of cases within 2 SEM of the full WAIS-IV score was then conducted. According to the WAIS-IV technical manual, the SEM values are as follows: FSIQ = 2.56, VCI = 3.6, PRI = 3.48, WMI = 3.94, PSI = 5.26 (Manual, French edition, page 35, Table 4.3). We calculated the percentage of subjects scoring within 2 SEMs on the four SF compared to their actual scores. The criterion was met if at least 80% of cases fell in the same diagnostic category. We then determined the percentage of subjects who met either clinical classification or band of error criteria. If both criteria were violated at the same time, the SF was considered clinically inaccurate (Iverson et al., 1997). For information purpose, we estimated time savings for each SF, based on previous WAIS-III research (Ryan et al., 1998). To the best of our knowledge, no studies assessed the WAIS-IV separate subtests administration time in detail. Finally, reliability was measured to evaluate the internal consistency of each form. Along with Donders and Axlerod (2002), we set a minimum reliability threshold of Criterion validity assessed the ability of each SF to predict functional outcome (Social Autonomy Scale) and illness duration. The best methods were discussed following the above-mentioned decision criteria. All statistical analyses were performed with SPSS (IBM SPSS Statistics for Windows, Version Armonk, NY: IBM Corp.). All statistical tests were two-tailed. Given that multiple tests were computed, familywise error was adjusted by applying a Bonferroni correction to the alpha level (i.e., 4 forms 5 indexes: α = 0.05/(4 5)). As such, the revised alpha was set at Results Compared to the standardization sample (data issued from the manual), SZ patients obtained significantly worse scores on all measures (t-test, p ). The average FSIQ was 81.1 (SD = 15.5). VCI mean score was the highest factor score (88.9 ± 17.1), while PSI was the lowest (80.9 ± 14.3). The PRI and WMI scores were, respectively, 84.3 ± 15.4 and 83.9 ± The patterns of performance on the different composite scores are illustrated in Fig. 1. The mean comparison did not reveal sex-related differences of the WAIS-IV scores (independent samples t-test, p >.05) FSIQ VCI PRI WMI PSI 75 7BDsf PRO 7MRsf PRO 7BDsf REG 7MRsf REG Full Fig. 1. Composite scores estimated by the four Short forms and the full Wechsler Adult Intelligence Scale-IV in patients with schizophrenia.

6 920 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Table 3. Estimated IQ and Index scores (N = 70) * 7BDsf PRO 7MRsf PRO 7BDsf REG 7MRsf REG FSIQ (±15.00) (±14.97) (±14.38) (±14.49) VCI (±17.23) (±17.23) (±17.04) (±17.07) PRI (±16.21) (±15.18) (±14.21) (±13.29) WMI (±14.05) (±14.05) (±13.82) (±13.81) PSI (±14.98) (±14.98) (±12.29) (±12.56) * 7BDsf, Ward s seven-subtest SF (BD version); 7MRsf, Ward s seven-subtest SF (MR version); PRO, prorated scores version; REG, regression-based version; FSIQ, Full-Scale Intelligence Quotient; VCI, Verbal Comprehension Index; PRI, Perceptual Reasoning Index; WMI, Working Memory Index; PSI, Processing Speed Index. Means and SD for the estimated IQ and Index scores following the four SF formulae are presented in Table 3. The comparison of the four SF is detailed in Table 4. The Spearman correlation coefficients with the full versions were high and comparable across the four SF. Except PRI ranging from 0.84 to 0.89, all other coefficients were above the 0.90 threshold (all p , two-tailed). When analyzed separately in the two diagnostic groups, PRI correlation with the fullform counterparts was systematically lower in patients with schizoaffective disorder (Spearman s rho ranging from 0.67 to 0.76) compared to SZ patients (Spearman s rho ranging from 0.85 to 0.90). Other index correlations were comparable in both diagnostic groups. Analyses of the mean differences between SF and original scores showed that some SF scores, especially when derived from the regression-based versions, differed from the full-version scores (Cohen s d for dependent samples were negligible to small for prorated forms and mostly medium for the regression-based forms, see Table 4). Of note, WMI was not included in the mean comparison for the prorated versions, given the fact that all the subtests necessary to calculate the full form are included in the SF. The scores are thus identical to the full-version counterparts. Category agreement was comparable for all four SF, the percentage of cases falling into the same diagnostic category (compared to the full form) was systematically lower for PRI (54% 59%), PSI (61% 64%), and VCI (71 73%) indexes, all methods taken together. None of the SF showed an acceptable agreement on these dimensions. The general agreement ranged from 77% to 83% for the FSIQ. The band of error analysis showed generally higher misclassification rates for PRI index (all forms confounded). Only a half of the patients fell within the 2 SEM band for PRI when MR forms were used. 7MRsf REG was the least accurate of the four SF. The two BD forms showed slightly better accuracy for this criterion. Combining the two last decision criteria did not reveal any superiority among the four methods, but again PRI index showed the least accurate scores. When analyzed separately, PSI, and PRI indexes show noticeable differences in category agreement across diagnostic groups. Although preliminary and based on a small sample, the results show a very low category agreement (43% 48%) in patients with schizoaffective disorder. An additional analysis of agreement rates was performed. The sample was split into four quartiles following initial scores obtained with the full version of the WAIS-IV. FSIQ and WMI agreement rates did not differ across the groups. Instead, we observed differential agreement rates relative to the initial IQ on other indexes. PRI showed better agreement rates in subjects with higher IQ. PSI showed better agreement rates in subjects with lower IQ. VCI agreement rates were higher in persons within the percentiles 25 and 75; instead, more extreme values of VCI obtained slightly lower clinical agreement. The estimated time savings were slightly higher for the MR versions (54% vs. 49%). Criterion validity analysis did not reveal any significant correlation between WAIS-IV variables (both full and SF) and functional outcome and illness duration. Internal consistency estimates were assessed using Spearman Brown corrected splithalf or coefficient alpha methods for each form. The results were satisfactory and are presented in the Table 4. Cronbach s alpha/split-half reliability coefficients for subtests in the WAIS-IV ranged from 0.90 to Discussion We assessed seven-subtest SF of the WAIS-IV in a sample of SZ patients. We studied four different versions of the Ward s SF (Ward, 1990): BD (7BDsf) and MR (7MRsf) forms, both calculated following the prorated scores method (PRO) and regression-based method (REG).

7 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Table 4. Comparison of the four SF with the full-version counterparts in patients with schizophrenia (N = 70) * 7BDsf PRO 7MRsf PRO 7BDsf REG 7MRsf REG I Spearman correlation coefficients (all p <.001, 2-tailed) FSIQ VCI PRI WMI PSI II Mean comparison (paired samples t-tests, 2-tailed, Cohen s d) FSIQ t(69) = 0.662, p =.510 (d = 0.09) t(69) = 0.776, p =.440 (d = 0.10) t(69) = 4.650, p <.001 (d = 0.58) t(69) = 4.498, p <.001 (d = 0.57) VCI t(69) = 1.565, p =.122 (d = 0.20) t(69) = 1.565, p =.122 (d = 0.20) t(69) = 3.978, p <.001 (d = 0.51) t(69) = 3.453, p =.001 (d = 0.44) PRI t(69) = 1.456, p =.150 (d = 0.18) t(69) = 1.673, p =.099 (d = 0.20) t(69) = 2.621, p =.011 (d = 0.32) t(69) = 3.259, p =.002 (d = 0.40) WMI t(69) = 6.826, p <.001 (d = 0.39) t(69) = 6.002, p <.001 (d = 0.34) PSI t(69) = 4.904, p <.001 (d = 0.66) t(69) = 4.904, p <.001 (d = 0.66) t(69) = 0.921, p =.360 (d = 0.14) t(69) = 0.197, p =.844 (d = 0.02) III Difference scores compared to the full-version counterparts (mean, SD) FSIQ 0.26 (±3.25) 0.33 (±3.64) 1.80 (±3.24) 1.96 (±3.64) VCI 0.87 (±4.66) 0.87 (±4.66) 2.19 (±4.61) 1.90 (±4.61) PRI 1.53 (±8.78) 1.76 (±8.79) 2.26 (±7.21) 3.14 (±8.06) WMI 0.00 (±0.00) 0.00 (±0.00) 3.32 (±0.39) 0.29 (±0.40) PSI 3.79 (±6.46) 3.80 (±6.46) 0.64 (±5.83) 0.14 (±6.15) IV IQ category agreement (% cases falling in the same diagnostic category as in the full version) FSIQ 81% 83% 80% 77% VCI 73% 73% 71% 71% PRI 57% 54% 59% 54% WMI 100% 100% 100% 99% PSI 61% 63% 64% 64% V Band of error (% cases falling within 2 SEMs from the actual IQ) FSIQ 91% 86% 81% 76% VCI 89% 89% 81% 81% PRI 57% 51% 59% 50% WMI 100% 100% 100% 100% PSI 84% 84% 93% 90% VI Clinical accuracy (% cases with either 4 or 5 met) FSIQ 94% 91% 91% 89% VCI 90% 90% 86% 87% PRI 70% 69% 71% 64% WMI 100% 100% 100% 100% PSI 90% 91% 94% 96% VII Estimated time savings 49% 54% 49% 54% VIII Internal consistency reliability r = 0.90 r = 0.90 r = 0.93 r = 0.91 IX Criterion validity (Spearman s rho, p) Clinical remission 0.09 (p =.48) 0.09 (p =.48) 0.08 (p =.54) 0.07 (p =.60) Illness duration 0.11 (p =.37) 0.11 (p =.39) 0.10 (p =.43) 0.10 (p =.43) * 7BDsf, Ward s seven-subtest SF (BD version); 7MRsf, Ward s seven-subtest SF (MR version); PRO, prorated scores version; REG, regression-based version; FSIQ, Full-Scale Intelligence Quotient; VCI, Verbal Comprehension Index; PRI, Perceptual Reasoning Index; WMI, Working Memory Index; PSI, Processing Speed Index. WAIS-IV Score Pattern in Schizophrenia The full WAIS-IV factor index scores presented the typical pattern of performance in SZ (Michel et al., 2013). This pattern was maintained in each of the four SF. This confirms a relative preservation of verbal comprehension and perceptual abilities, along with a more pronounced working memory and processing speed impairment in SZ. As expected, all scores were found significantly lower compared to the standardization sample (Michel et al., 2013).

8 922 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Table 5. Regression equations for the FSIQ and index scores derived from the sample of patients with schizophrenia and schizoaffective disorder * 7BDsf 7MRsf FSIQ 1.071(SI) (BD) (IN) (DS) (CD) (AR) (PC) (SI) (MR) (IN) (DS) (CD) (AR) (PC) VCI 2.909(SI) (IN) (SI) (IN) PRI 3.343(BD) (DS) (PC) (MR) (PC) (AR) WMI 2.866(DS) (AR) (DS) (AR) PSI 1.02(DS) (CD) (DS) (CD) * 7BDsf, Ward s seven-subtest SF (BD version); 7MRsf, Ward s seven-subtest SF (MR version); FSIQ, Full-Scale Intelligence Quotient; VCI, Verbal Comprehension Index; PRI, Perceptual Reasoning Index; WMI, Working Memory Index; PSI, Processing Speed Index; IN, Information; SI, Similarities; BD, Block Design; PC, Picture Completion; DS, Digit Span; AR, Arithmetic; CD, Coding; MR, Matrix Reasoning. Clinical decision rules should be clearly stated to reliably predict SF accuracy and help clinicians choose the appropriate form. For this reason, we analyzed the four SF algorithms following nine decision criteria found in literature with a priori set acceptance thresholds. Criterion 1: Correlations In terms of correlation with the full version, the four SF were fairly comparable. All correlation coefficients ranged from 0.84 to 1.0 Only PRI did not meet the acceptance criterion (r 0.90) but the lower correlations were mainly observed in patients with schizoaffective disorder. While PRI scores showed no statistical difference between the diagnostic groups, there was a slight trend toward lower scores in schizoaffective patients (t-test, 2-tailed, p =.07). This is surprising in the light of previous studies showing more pronounced cognitive impairment in patients with SZ compared to schizoaffective disorder (Bora, Yucel, & Pantelis, 2009). Future studies are needed to explore this dimension in detail. For these reasons, we recommend a cautious use of the SF when estimating PRI. Criteria 2 & 3: Mean Scores Comparison Our results show that the regression-based SF tend to slightly overestimate the FSIQ, VCI, PRI, and WMI, while PSI is underestimated by the prorated methods (effect size ranged from small to medium) and thus led us to consider both regressionbased forms less suitable when used with patients with schizophrenia. Meyer s regression formulae were issued from a heterogeneous clinical sample and their use should be careful when assessing SZ patients. Although regression approach is widely used in literature, it is highly sample-related. Consequently, the regression equations issued from a previous study (Meyers et al., 2013, Meyers, personal communication) should be used with caution when assessing patients with SZ or schizoaffective disorder. To address this question, a linear regression was calculated using the age-scaled scores from the seven subtests to predict the FSIQ and the index scores within this diagnostic group (Table 5). Further research is therefore needed to assess the reliability of these algorithms in an independent sample of SZ patients. Criteria 4, 5 & 6: Category Agreement, Band of Error Analysis, and Clinical Accuracy The four forms revealed comparable but generally low category agreement rates. We found a noticeable change in clinical classification for PRI, PSI, and VCI scores across all methods. The band of error analysis showed inadequate estimates of PRI and more generally the MR forms. PRI indexes showed even lower agreement rates in schizoaffective patients, while PSI was more heterogeneous in SZ patients. One of the sources of PRI bias is certainly the fact that the latest edition of the WAIS does not require the Picture Completion subtest to calculate the original scores. If these were calculated using Picture Completion substitution (e.g., replacing the Visual Puzzle subtest, which is not included in either of the SF), the correlation coefficients and mean comparison would logically reveal a closer association between the full and SF, especially on the PRI index, which regularly shows lower rates of association with the full version on the different decision criteria. An additional analysis (using full-scale IQ recalculated with this optional subtest), might be conducted to confirm this point, but this is beyond the scope of this article. Future studies should investigate the utility of an alternative SF to replace this optional subtest with one of the core subtests. However, Michel et al. (2013) showed that WAIS-IV subtests that best differentiate SZ patients from controls (p =.001) include Picture Completion (along with Symbol search, CD, Cancellation, and Comprehension subtests).

9 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); However, in the three out of four forms, PSI was estimated based on one subtest only. This index generally tends to be the least reliable, for this reason using subtest dyads to provide estimates has been advocated (Christensen, Girard, & Bagby, 2007; Donders & Axlerod, 2002). This may explain the observed lower accuracy and led us to question the clinical advantage of these forms when estimating processing speed. Both category agreement and band of error analysis show a slightly better fit for prorated versions as far as FSIQ and VCI are concerned. PSI tends to be better predicted by the regression forms. On this point, our results are comparable with those published by Meyers et al., (2013). Finally, category agreement tends to be higher in subjects with higher IQ for PRI, lower IQ for PSI and the IQ between 25th and 75th percentiles for VCI. Overall, the low category agreement raises question of the ability of the Ward s SF to estimate the factor indexes and this point should be taken into account while interpreting clinical data. While WMI is measured in its full form, and VCI offers agreement rates in the low 70s, the only factor with at least 80% of clinical agreement was FSIQ. The results seem even more compromised in patients with schizoaffective disorder. Criterion 7: Time Savings We found no substantial difference between the MR and BD versions which is consistent with Ryan and Ward (1999). The clinician may thus want to select the appropriate form to match clinical hypotheses and requirements. The MR versions offer slightly better time savings and an additional abstract reasoning assessment. Visuo-construction will be better assessed with the BD version. Of note, when administered WAIS-IV, SZ patients tend to differ from the control group on BD, while this difference is not statistically significant on MR test (Michel et al., 2013). Criteria 8 & 9: Internal Consistency and Criterion Validity The four SF showed a good reliability. Instead, neither the full WAIS-IV nor any of the SF estimates was related to meaningful outcomes in terms of everyday autonomy or illness duration. Limits The limitations of our study include an unbalanced sex-ratio with a majority of male participants. This could be explained by the classical difference in severity of the disease between men and women (Aleman, Kahn, & Selten, 2003). Our study was conducted on a consecutive sample of patients recruited in a day care centre, where male patients are overrepresented due to a relatively higher severity of psychotic symptoms and a larger number of hospitalizations. Some studies suggest that patients with schizophrenia display gender differences on cognitive variables. Men tend to suffer from more pronounced cognitive impairments (Sota & Heinrichs, 2003). This was not supported by our findings. None of the WAIS-IV index scores differed according to sex and thus we considered this confound negligible. Second, low category agreement across the factor indexes (except WMI measured in its full form) questions the clinical utility of the Ward s SF when estimating cognitive performance in a more detailed way. To address this issue, we suggested alternative regression equations built on our sample. These require further assessment within an independent sample of SZ subjects. Third, another possible source of lower factor estimates and biased validity estimates may be related to suboptimal effort. Motivational factors can play a role in QI assessment, especially in patients with schizophrenia. Poor mental effort might be considered a core symptom of schizophrenia, representing an executive, monitoring, or motivational problem (Gorissen, Sanz, & Schmand, 2005). While SFs delivered independently can sometimes yield better estimates of performance as they are potentially less susceptible to fatigue, in the current case, with SF scores derived from the full administration, motivation may not explain differences between SF and full-scale scores, but it may contribute to error variance and decrease reliability, which in turn affects validity estimates. Fourth, this study raises concern about the ability of SF scales to predict outcome measures. Neither SF estimates nor full WAIS-IV IQ assessment did show any significant correlation with functional outcome and illness duration. This finding underlines the question of ecological validity of WAIS-IV as a complement to clinical assessment in schizophrenia. Finally, this study was carried out on a French speaking sample which may potentially affect the generalizability of our conclusions. Further research is needed to draw more general conclusions about the clinical advantages of the present SF.

10 924 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Strengths This study offers clear strengths: (i) a relatively large and homogenous sample of clinically stable SZ or schizoaffective patients, (ii) the parallel analysis of four different versions derived from the original SF, and (iii) the use of nine different comparison criteria. Conclusion In conclusion, our results provide strong support for using the Ward seven-subtest SF (Ward, 1990) of the WAIS-IV in patients with SZ or schizoaffective disorders when FSIQ estimate is needed. If factor indexes are required, the Ward s SF seems less appropriate and clinical conclusions should be drawn with caution, given the low clinical agreement rates. We recommend the use of the prorated forms in SZ given their better psychometric properties. Our study did not support the use of Ward s SF when estimating PRI in schizophrenia or schizoaffective disorder. Funding No funding sources had any role in study design, in the collection, analysis, and interpretation of data, in the writing of the report, or in the decision to submit the paper for publication. Acknowledgements We thank Mary Beth Spitznagel from Research and Editing Consulting Program (RECP) for her assistance in editing this manuscript. We thank Monik Graziani for her logistic assistance. References Abraham, E., Axelrod, B. N., & Paolo, A. M. (1997). Comparison of WAIS-R selected subtest short forms in a mixed clinical population. Assessment, 4 (4), Aleman, A., Kahn, R. S., & Selten, J.-P. (2003). Sex differences in the risk of schizophrenia: evidence from meta-analysis. Archives of General Psychiatry, 60 (6), Allen, D. N., Huegel, S. G., Gurklis, J. A., Kelley, M. E., Barry, E. J., & van Kammen, D. P. (1997). Utility of WAIS-R short forms in schizophrenia. Schizophrenia Research, 26 (2), American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (DSM-5 ). American Psychiatric Pub. Andreasen, N. C., Carpenter, W. T. Jr., Kane, J. M., Lasser, R. A., Marder, S. R., & Weinberger, D. R. (2014). Remission in schizophrenia: proposed criteria and rationale for consensus. American Journal of Psychiatry. Axelrod, B. N., & Paolo, A. M. (1998). Utility of WAIS R seven-subtest short form as applied to the standardization sample. Psychological Assessment, 10 (1), 33. Axelrod, B. N., Ryan, J. J., & Ward, L. C. (2001). Evaluation of seven-subtest short forms of the Wechsler Adult Intelligence Scale-III in a referred sample. Archives of Clinical Neuropsychology, 16 (1), 1 8. Axelrod, B. N., Woodard, J. L., Schretlen, D., & Benedict, R. H. (1996). Corrected estimates of WAIS-R short form reliability and standard error of measurement. Psychological Assessment, 8 (2), 222. Benedict, R. H., Schretlen, D., & Bobholz, J. H. (1992). Concurrent validity of three WAIS-R short forms in psychiatric inpatients. Psychological Assessment, 4 (3), 322. Bora, E., Yucel, M., & Pantelis, C. (2009). Cognitive functioning in schizophrenia, schizoaffective disorder and affective psychoses: meta-analytic study. The British Journal of Psychiatry, 195 (6), Christensen, B. K., Girard, T. A., & Bagby, R. M. (2007). Wechsler adult intelligence scale-short form for index and IQ scores in a psychiatric population. Psychological Assessment, 19 (2), 236. Collins, P. Y., Patel, V., Joestl, S. S., March, D., Insel, T. R., & Daar, A. S., et al. (2011). Grand challenges in global mental health. Nature, 475 (7354), David, A. S., Malmberg, A., Brandt, L., Allebeck, P., & Lewis, G. (1997). IQ and risk for schizophrenia: a population-based cohort study. Psychological Medicine, 27 (6), Donders, J., & Axlerod, B. N. (2002). Two-subtest estimations of WAIS-III factor index scores. Psychological Assessment, 14 (3), 360. Gorissen, M., Sanz, J. C., & Schmand, B. (2005). Effort and cognition in schizophrenia patients. Schizophrenia Research, 78 (2 3), /j.schres Green, M. F., & Harvey, P. D. (2014). Cognition in schizophrenia: past, present, and future. Schizophrenia Research: Cognition, 1 (1), e1 e9. Guilmette, T. J., Dabrowski, J., Kennedy, M. L., & Gnys, J. (1999). A comparison of nine WAIS-R short forms in individuals with mild to severe traumatic brain injury. Assessment, 6 (1), Harrison-Read, P. (2008). IQ tests as aids to diagnosis and management in early schizophrenia. Advances in Psychiatric Treatment, 14 (3),

11 E. Bulzacka et al. / Archives of Clinical Neuropsychology 31 (2016); Heinrichs, R. W. (2005). The primacy of cognition in schizophrenia. American Psychologist, 60 (3), 229. Heinrichs, R. W., & Zakzanis, K. K. (1998). Neurocognitive deficit in schizophrenia: a quantitative review of the evidence. Neuropsychology, 12 (3), 426. Iverson, G. L., Guirguis, M., & Green, P. (1998). Assessing intellectual functioning in persons with schizophrenia spectrum disorders using a seven subtest short form of the WAIS-R. Schizophrenia Research, 30 (2), Iverson, G. L., Myers, B., & Adams, R. L. (1997). Comparison of two computational formulas for a WAIS R seven subtest short form. Journal of Clinical Psychology, 53 (5), Kaufman, A. S., Ishikuma, T., & Kaufman-Packer, J. L. (1991). Amazingly short forms of the WAIS-R. Journal of Psychoeducational Assessment, 9 (1), Leeson, V. C., Barnes, T. R., Hutton, S. B., Ron, M. A., & Joyce, E. M. (2009). IQ as a predictor of functional outcome in schizophrenia: a longitudinal, four-year study of first-episode psychosis. Schizophrenia Research, 107 (1), Leguay, D., Cochet, A., Matignon, G., Hairy, A., Fortassin, O., & Marion, J. M. (1998). L Echelle d Autonomie Sociale: Premiers éléments de validation. L Encéphale, 24 (2), Meyers, J. E., Zellinger, M. M., Kockler, T., Wagner, M., & Miller, R. M. (2013). A validated seven-subtest short form for the WAIS-IV. Applied Neuropsychology: Adult, 20 (4), Michel, N. M., Goldberg, J. O., Heinrichs, R. W., Miles, A. A., Ammari, N., & Vaz, S. M. (2013). WAIS-IV profile of cognition in schizophrenia. Assessment, 20 (4), Morris, S. B., & DeShon, R. P. (2002). Combining effect size estimates in meta-analysis with repeated measures and independent-groups designs. Psychological Methods, 7 (1), 105. Pichot, P., Lebeaux, M.-O., Penhouët, C., & Simon, M. (1993). L estimation du niveau intellectuel prémorbide à la WAIS-R. Comparaison entre la méthode de Barona et l Index de détérioration de Wechsler. European Review of Applied Psychology, 43 (4), 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 (4), Ryan, J. J., & Ward, L. C. (1999). Validity, reliability, and standard errors of measurement for two seven-subtest short forms of the Wechsler Adult Intelligence Scale-III. Psychological Assessment, 11 (2), 207. 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 (4), Ryan, J. J., Weilage, M. E., & Spaulding, W. D. (1999). Accuracy of the seven subtest WAIS-R short form in chronic schizophrenia. Schizophrenia Research, 39 (1), Sattler, J. M., & Ryan, J. J. (2009). Assessment with the WAIS-IV. Jerome M Sattler Publisher. Sota, T. L., & Heinrichs, R. W. (2003). Sex differences in verbal memory in schizophrenia patients treated with «typical» neuroleptics. Schizophrenia Research, 62 (1), Ward, L. C. (1990). Prediction of verbal, performance, and full scale IQs from seven subtests of the WAIS-R. Journal of Clinical Psychology. Wechsler, D. (1939). The measurement of adult intelligence. Baltimore: William and Williams Co. Wechsler, D. (1999). WASI (Wechsler Adult Scale reduced). New York: The Psychological Corporation. Wechsler, D. (2008). Wechsler adult intelligence scale Fourth Edition (WAIS IV). San Antonio, TX: NCS Pearson. Wittchen, H. U., Jacobi, F., Rehm, J., Gustavsson, A., Svensson, M., & Jönsson, B., et al. (2011). The size and burden of mental disorders and other disorders of the brain in Europe European Neuropsychopharmacology: The Journal of the European College of Neuropsychopharmacology, 21 (9), Wymer, J. H., Rayls, K., & Wagner, M. T. (2003). Utility of a clinically derived abbreviated form of the WAIS-III. Archives of Clinical Neuropsychology, 18 (8),