Virtual reality measures in neuropsychological assessment: a meta-analytic review

Abstract

Objective: Virtual reality-based assessment is a new paradigm for neuropsychological evaluation, that might provide an ecological assessment, compared to paper-and-pencil or computerized neuropsychological assessment. Previous research has focused on the use of virtual reality in neuropsychological assessment, but no meta-analysis focused on the sensitivity of virtual reality-based measures of cognitive processes in measuring cognitive processes in various populations. Method: We found eighteen studies that compared the cognitive performance between clinical and healthy controls on virtual reality measures. Results: Based on a random effects model, the results indicated a large effect size in favor of healthy controls (g = .95). For executive functions, memory and visuospatial analysis, subgroup analysis revealed moderate to large effect sizes, with superior performance in the case of healthy controls. Participants’ mean age, type of clinical condition, type of exploration within virtual reality environments, and the presence of distractors were significant moderators. Conclusions: Our findings support the sensitivity of virtual reality-based measures in detecting cognitive impairment. They highlight the possibility of using virtual reality measures for neuropsychological assessment in research applications, as well as in clinical practice.

Keywords:

• Neuropsychological assessment
• virtual reality
• ecological validity

Acknowledgment

This work was possible due to the financial support of the Sectorial Operational Program for Human Resources Development 2007–2013, co-financed by the European Social Fund, under the project number POSDRU/159/1.5/S/132400 with the title Young successful researchers – professional development in an international and interdisciplinary environment.

Notes

1. When addressing the diagnostic validity of a neuropsychological test one important aspect is the classification accuracy of the measure (Lezak, 1995). Classification accuracy refers to the correct percentage of correctly identified cases as belonging to either clinical group or healthy control group (Lezak, 1995). This classification accuracy is expressed via indexes like sensitivity/specificity. One approach to establish the sensitivity of a neuropsychological test is by comparing two contrasted groups, usually coming from clinical and healthy populations on the performance obtained on specific cognitive functions (Lezak, 1995; Urbina, 2004; Wasserman & Bracken, 2003). After computing the magnitude of the difference between the performance of the two groups which is usually expressed in terms of effect size one can obtain estimates of test overlap and probability of superiority. It is considered that a diagnostic marker in neuropsychological assessment has an appropriate level of sensitivity if the test overlap is lower than 5% expressed in an effect size larger than 3.0 in magnitude according to Cohen’s d metrics (Zakzanis, 2001).

2. general measures of executive functioning, as well as impulsivity/inhibition and attention indexes/measures.

3. memory and learning processes (e.g. target recall, target recognition, total errors).

4. spatial rotation and measures of visuospatial neglect.

5. accordingly to the specifications from the literature regarding the need to carefully interpret the magnitude of Cohen’s d qualification of ‘small,’ ‘medium,’ and ‘large’ effects depending on the context (Zakzanis, 2001). Although, for instance, 1.0 is a large effect according to Cohen’s metrics, this value reflects approximately only 45% percent overlap. This indicates that approximately 50% participants from the clinical group obtain scores different from those obtained by healthy controls. Therefore, we will not only rely in our analyses on Cohen’s d metrics, but will also provide estimates of test overlap to investigate the sensitivity of virtual reality measures and interpret the result through this frame.

6. a value of Hedge’s g between .20 and .50 indicates a small effect, one between .50 and .80 indicates a medium effect, while a value larger than .80 indicates a large effect size (Cohen, 1988).