Reliability and validity of the CogState computerized battery in patients with seizure disorders and healthy young adults: comparison with standard neuropsychological tests.
The aim was to examine the reliability and validity of two measures of learning and memory within the CogState Computerized Battery: the One Card Learning (OCL) and the Continuous Paired Associative Learning (CPAL). Comparison of various reliable change measures was also included to examine rate of correct cognitive change classification, in particular, when using the Within-Subject Standard Deviation (WSD).
The OCL and the CPAL tests were administered twice and compared to standard and experimental versions of conventional neuropsychological tests in patients with seizure disorders (Baseline n = 80, Follow-up = 54) and university students (Baseline n = 89, Follow-up n = 87). Calculations of Reliable Change Indices (RCIs) on individual performance were obtained to detect reliable change across time using published CogState and current study parameters.
Results showed low retest reliabilities in both OCL and CPAL tests (r’s = .49-.77). Small to medium convergent validity correlations with traditional tests of learning and working memory were also found. Discrepancies in RCIs methods on performance estimation were observed when different test parameters and reliabilities were used. In contrast to recent recommendations by publishers of the CogState tests, the WSD method was found to substantially increase the rate of Type-I error when tests reliabilities were low.
Unsatisfactory reliability and validity estimates suggest caution regarding the CogState computerized measures of learning and memory as opposed to conventional tests. Caution should be used when interpreting OCL and CPAL test scores, as confidence intervals may be wide and encompass much of the population range. In line with previous recommendations, use of the WSD to detect change in performance over time should be avoided in neuropsychological testing, as this change measure is prone to elevated misclassification rates compared to other methods. Further independent research is needed to improve the psychometrics of CogState OCL and CPAL measures.