Boston naming test short forms: a comparison of previous forms with new item response theory based forms.

Two new short forms of the Boston Naming Test (BNT) were developed using item response theory (IRT) with data from 206 elderly outpatients. We evaluated the diagnostic ability of 12 short forms among the full sample and in a sub-sample of 69 patients diagnosed with mild Alzheimer's disease (AD) either alone or in combination with vascular dementia (VD). The full BNT (reliability alpha = .90) identified 44% of the AD/VD patients as abnormal on naming. Our 30 item short form (alpha = .90) also identified 44% of the AD/VD patients as abnormal, with 93% agreement with the full BNT on abnormal AD/VD patient classifications. Our 15 item short form (alpha = .84) identified 48% of the AD/VD patients as abnormal, with 90% agreement with the full BNT's abnormal classifications. An adaptive 30/15 item version equaled the performance of the full 30 item test while requiring only 15 items for 75% of the patients with normal naming ability. This study illustrates the utility of IRT for developing neuropsychological assessment tools.