Updating psychometric tests is essential for neuropsychologists to correctly detect cognitive disorders in clinical practice. However, data concerning the usefulness of this update is often incomplete, and there is a lack of knowledge on the changes of psychometric qualities, especially with clinical populations.

The publication of the d2-R (Brickenkamp, 2015) by Hogrefe LTD incited us to compare this new normalization with the first standardization, still widely used, of this attention test (Brickenkamp, 1998).

We performed this comparison on a clinical population of 64 adult patients (median age 60.5 years [IQR: 49.5–69]) 4 months post ICU discharge, to whom we administered the d2 attention test.

We converted the results in standard scores according to both rating systems and normalizations. We then performed statistical analyzes with the R software, and compared the difference of scores distributions on two rating criteria, namely a quantitative criterion (GZ and CCT) and a criterion also taking into account the quality of answers (KL and CC).

We found good correlations for both criteria, confirming data provided by the publisher. But the Wilcoxon test shows a significant difference between GZ and CCT percentiles distributions (w=1256, P<0.001), not for KL and CC percentiles (w=1851, P=0.35).

The results analysis shows a divergence in the pathological zone of the curves: 51.4% of patients who are classified as “significantly slow” in the former norms, are now considered having a normal speed using the new norms (Fig. 1), thus showing great disagreement for slow processing diagnosis.

Test publishers should provide more information on the clinical implications of their updates of psychometric tests, correlations cannot be sufficient to discuss sensibility and specificity changes.