Reliability of working memory assessment in neurocognitive disorders: a study of the Digit Span and Corsi Block-Tapping tasks.

Working memory (WM) is a core aspect of executive function, related to temporary storage of information to be manipulated and used by other cognitive processes.1 Despite the lack of a consensus definition and theoretical model to explain this function, its assessment is a key aspect in clinical practice, both for diagnostic and intervention purposes. Furthermore, as a cognitive process, WM is key to understanding functional capacity in clinical samples.

The Digit Span and Corsi Block-Tapping tasks are frequently used in WM assessment.2 In both tests, the subject must repeat a series of stimuli in the presented order (forward) or in the inverse order (backward). The Digit Span uses escalating series of numbers from 1 to 9, presented in a randomized fashion, to assess verbal WM. The Corsi Block-Tapping task uses nine cubes placed on a wooden board to assess visuospatial WM. Both tests are commonly used in neuropsychological assessment of older adults with a diagnostic hypothesis of pathological aging (e.g., minor and major neurocognitive disorders). However, we are unaware of studies investigating the reliability of these tasks for WM assessment in older adults diagnosed with neurocognitive disorders. Test reliability is an important measure used to estimate test precision and to create reliable change coefficients. These are simple statistical procedures used to examine whether changes in test scores over time (e.g., pre-intervention vs. post-intervention, baseline vs. follow-up) are likely due to measurement error (imprecision of test measures) or are associated with an external factor (the intervention, a placebo, or other non-documented causes).3

To analyze the reliability of the Digit Span and Corsi Block-Tapping tasks, we assessed 25 older adults with low formal education referred for neuropsychological assessment due to cognitive-functional complaints (13 patients with mild and 12 with major neurocognitive disorder, irrespective of etiological diagnosis). Diagnosis involved cognitive and functional assessment, performed with the Brazilian versions of the Mattis Dementia Rating Scale4 and Functional Activities Questionnaire.5 All participants gave written consent. The study was approved by the Universidade Federal de Minas Gerais Ethics Committee and conducted in accordance with the Declaration of Helsinki.

The same examiner administered and scored both tasks in all participants. Two trials were administered for each span (sequence length ranging from 2 to 8), in forward and reverse order, in both tasks. Two errors in the same span prompted interruption of the task. Reliability was estimated by the split-half method, using the first trials from each span to compose the first half and the second trials to compose the second half. Reliable change indexes (RCIs) were calculated for all test measures.

Table 1 shows descriptive data, internal consistency, and reliable change coefficients for each task. Reliability was high for Digit Span forward, low for Digit Span backward, and moderate for the Corsi Block-Tapping task. The RCI for the tasks can be used to track significant changes in longitudinal assessment, although the high variability and moderate reliability of the tasks are likely to hinder assessment of mild or slight changes.

Table 1

Participant characteristics, internal-consistency reliability, and reliable change indexes for working memory tests

	Mean	SD	Reliability	RCI*
Sociodemographic and clinical aspects
Age	76.20	7.92	-	-
Education	4.88	3.85	-	-
Mattis Dementia Rating Scale	107.80	16.23	-	-
Functional Assessment Questionnaire	7.24	4.28	-	-
Digit Span, forward
Correct trials	5.28	2.09	0.891	±2
Span	3.84	0.90		±1
Correct trials vs. span	21.84	14.39		±13
Digit Span, backward
Correct trials	3.16	1.40	0.598	±2
Span	2.88	1.01		±2
Correct trials vs. span	10.36	7.28		±13
Corsi Block-Tapping task, forward
Correct trials	4.64	1.11	0.753	±2
Span	3.60	0.65		±1
Correct trials vs. span	17.24	6.21		±9
Corsi Block-Tapping task, backward
Correct trials	2.80	1.83	0.782	±2
Span	3.16	0.75		±1
Correct trials vs. span	10.04	9.21		±12

RCI = reliable change index; SD = standard deviation.

If the difference between two assessments (baseline vs. follow-up, pre vs. post, etc.) is higher or lower than the RCI value, there is only a small probability (p < 0.05) of the difference being due to measurement error; instead, it is likely associated with an external factor (e.g., the tested intervention, time-related cognitive decline, placebo, or other non-documented factors).

Disclosure

The authors report no conflicts of interest.