Use of a genetic algorithm approach for improving precision and efficiency of tests for estimating premorbid cognitive ability

<p dir="ltr">The 50-item National Adult Reading Test (NART; Nelson, 1982; NART–R; Nelson & Willison, 1991), first published in the UK in 1982, remains a widely used method for estimating premorbid intelligence in neurological patients.</p><p dir="ltr">We recently restandardised the NART against the most recent revision of the Wechsler Adult Intelligence Scale (WAIS-IV; Bright et al., 2018) and have confirmed the utility of the NART and other tests based on oral reading in the estimation of premorbid full-scale IQ (FSIQ) in a range of neurological conditions (Bright et al., 2002; Bright et al., 2020).

Nevertheless, such tests may contain words that provide little or no diagnostic power (i.e., can be eliminated without negatively affecting prediction accuracy). We recently proposed a genetic algorithm (GA)-based method (see Box 1) that tackles the computationally intractable problem of identifying optimal subsets of words from premorbid IQ word-reading tests.</p><p dir="ltr">Using this method we identified a subset of 17 of the original 50 NART words that predict FSIQ as accurately as the entire test (van der Linde & Bright, 2018).