Expanding the WAIS-III Estimate of Premorbid Ability for Canadians (EPAC)

Abstract

Since the release of the Canadian WAIS-III normative data in 2001 (Wechsler, 2001), the clinical application of these norms has been limited by the absence of a method to estimate premorbid functioning. However, Lange, Schoenberg, Woodward, and Brickell (2005) recently developed regression algorithms that estimate premorbid FSIQ, VIQ and PIQ scores for use with the Canadian WAIS-III norms. The purpose of this study was to expand work by Lange and colleagues by developing regression algorithms to estimate premorbid GAI (Saklofske et al., 2005), VCI, and POI scores. Participants were the Canadian WAIS-III standardization sample (n = 1,105). The sample was randomly divided into two groups (Development and Validation group). Using the Development group, a total of 14 regression algorithms were generated to estimate GAI, VCI, and POI scores by combining subtest performance (i.e., Vocabulary, Information, Matrix Reasoning, and Picture Completion) with demographic variables (i.e., age, education, ethnicity, region of the country, and gender). The algorithms accounted for a maximum of 77% of the variance in GAI, 78% of the variance in VCI, and 63% of the variance in POI. In the Validation Group, correlations between predicted and obtained scores were high (GAI = .70 to .88; VCI = .87 to .88; POI = .71 to .80). Evaluation of prediction errors revealed that the majority of estimated GAI, VCI, and POI scores fell within a 95% CI band (93.5% to 97.0%) and within 10 points of obtained index scores (72.3% to 85.6%) depending on the subtests used. These algorithms provide a promising means for estimating premorbid GAI, VCI, and POI scores using the Canadian WAIS-III norms.

We thank The Psychological Corporation, a Harcourt Assessment company, for permission to use the Canadian WAIS-III Standardization Data. Portions of these data were presented at the annual conference of the International Neuropsychological Society, February 2005, St. Louis, Missouri.

Notes

We thank The Psychological Corporation, a Harcourt Assessment company, for permission to use the Canadian WAIS-III Standardization Data. Portions of these data were presented at the annual conference of the International Neuropsychological Society, February 2005, St. Louis, Missouri.

¹The American WAIS-III norms were generated based on the cumulative frequency of raw score distributions for a subset of the data defined by an age cohort (i.e., the distribution of individuals within each age cohort is normalized, scaled scores for each raw score are calculated, and minor sampling fluctuations are smoothed). This is the traditional method of forced normalization, with the norms for each age group based on 200 subjects. The Canadian norms were generated using a more complex statistical technique known as continuous norming. In essence, continuous norming employs polynomial regression to develop a mathematical model of the distribution of scores for a scale. Normative tables are generated from descriptive statistics of the entire sample (i.e., mean, standard deviation, skewness, and kurtosis) rather than from data of a subset of individuals in a group.

²An individual’s ethnicity category is determined based on their self-reported description of their ethnic background. Ethnicity is categorized into the following four categories (a) British: includes those individuals who report their ethnic background to be from the British Isles. Countries include England, Scotland, Wales, and Ireland. Individuals who describe themselves as “Canadian” would be included in this category if their family background originated from the British Isles. Individuals whose family background includes the British Isles and another European country not from the British Isles should not be included in this category. These individuals would be included in the French/European category; (b) French/European: includes those individuals who report their ethnic background to be from France or other European countries not including the British Isles. Countries include, but are not limited to France, Germany, Spain, Greece, Italy, Bulgaria, Denmark, Poland, Romania, Norway, Netherlands, Belgium, Ukraine, and Austria. Individuals who describe themselves as “Canadian” would be included in this category if their family background originated from one or more European countries not including the British Isles. Individuals whose family background includes French/European and another country not covered by the British or French/European category should not be included here. These individuals would be included in the Multiple Origin category; (c) Single Origin: includes those individuals who report their ethnic background to be from a country not covered by the British or French/European category, whose parents are from the same country. For example, an individual with two parents that originated from South Africa. Countries and/or regions include, but are not limited to Asia (e.g., Japan, China, Malaysia), Africa (South Africa, Nigeria, Kenya), Australia, New Zealand, Papua New Guinea, India, and South America (Brazil, Peru, Argentina). Individuals who describe themselves as “First Nations” should also be included in this category; (d) Multiple Origin: includes those individuals who report their ethnic background to be from a country not covered by the British or French/European category, whose parents are from different countries. For example, an individual whose mother is East Indian and father is Malaysian. Countries included in this category include those specified in the single origin category.

³Prediction models for the Working Memory Index (WMI) and Processing Speed Index (PSI) were not developed here because of methodological issues. The four subtests used in this study to estimate WAIS-III scores were selected because of their demonstrated reliability (The Psychological Corporation, 1997) and resistance to neurological insult (Donders, Tulsky, & Zhu, 2001; Fisher et al., 2000; Kaufman, 1990; Schoenberg et al., 2003; The Psychological Corporation, 1997). These four subtests were not considered appropriate to predict WMI or PSI because of their lack of association with these indexes. Although prediction models to predict WMI and PSI scores using demographic variables could have been obtained, the use of demographics variables alone was not undertaken because of their poor ability to estimate WAIS-III scores demonstrated in previous research using this methodology (i.e, Lange, Schoenberg, Woodward, and Brickell (2005)). Rather, we decided to provide prediction models for VCI and POI because (a) the four subtests used to predict WAIS-III scores were directly relevant to VCI and POI, and (b) VCI and POI scores reflect measures of visual and verbal ability that combine to create the global GAI score.