Comparison of Tetradic WAIS-III Short Forms in Predicting Full Scale IQ Scores in Neuropsychiatric Clinic Settings

Patients from two neuropsychology clinics (n = 283) were divided into model development (n = 189) and validation samples (n = 94) with proportional representation from each clinic. Three regression based models of four WAIS-III subtests were developed and cross-validated to evaluate their FSIQ estimation accuracy: (1) an optimized stepwise regression model derived from the development sample (Comprehension, Matrix Reasoning, Similarities, and Picture Arrangement), (2) an a priori WASI based model (Similarities, Matrix Reasoning, Vocabulary, and Block Design), and (3) an a priori model with relatively rapidly administered subtests from each Index Score factor of the WAIS-III (Information, Picture Completion, Arithmetic, and Digit Symbol-Coding). The models produced good (92 to 97%) FSIQ estimation accuracy within 10 points of FSIQ with none of these models performing significantly better than another. The utility of clinical use of short form FSIQ estimation is a function of accuracy needed versus administration time reduction and ease for the patient.