Reliability of a brief computerized attention test in older adults: a split-half and intraclass correlation analysis

Abstract

Background: Population aging is associated with an increased prevalence of cognitive impairment, making early detection of cognitive changes in older adults essential. Attention plays a central role in global cognitive functioning and is often affected early, being sensitive to different clinical conditions, including neurodegenerative diseases and systemic disorders. Reaction time–based measures have proven useful in the assessment of cognitive aging, as they reflect not only processing speed but also performance stability over time. In particular, reaction time variability has been described as a marker of cognitive instability.The Continuous Visual Attention Test allows the assessment of multiple attentional components, including omission errors, commission errors, reaction time, and its variability. Brief versions of the test have demonstrated applicability across different clinical contexts and the ability to discriminate levels of cognitive impairment in older adults. The analysis of attentional subdomains further suggests that deficits may occur heterogeneously, increasing the sensitivity of these measures in the assessment of cognitive aging. Objective: This study aims to evaluate the CVAT internal reliability in healthy elderly. Methods: Participants were recruited from a clinical setting. Inclusion criteria comprised older adults, while exclusion criteria werw severe neurological and psychiatric diseases and uncontrolled clinical conditions.The 90-second test consisted of 90 Go/No-Go stimuli, divided into two equal halves. Omission errors (OE), commission errors (CE), reaction time (RT), and reaction time variability (VRT) were calculated for each half.Internal consistency was assessed using Cronbach’s alpha, and test reliability was evaluated using the intraclass correlation coefficient. Results: Baseline performance showed mean commission errors of 0.219, omission errors of 0.122, reaction time of 464.31 ms, and reaction time variability of 133.97 ms. 173 participants were included. Mean age was 72,29 and 45,1% were females. Age varied from 61 to 91 years old. Split-half analysis demonstrated similar performance between the initial and final segments. Omission errors showed high internal consistency (alpha = 0.878) and good to excellent reliability (intraclass correlation = 0.780 to 0.876). Reaction time showed moderate internal consistency (alpha = 0.741) and reliability (intraclass correlation = 0.587 to 0.739). Commission errors showed moderate internal consistency (alpha = 0.705) and reliability (intraclass correlation = 0.546 to 0.706). Reaction time variability showed poor internal consistency (alpha = 0.489) and limited reliability (intraclass correlation = 0.323 to 0.488). Conclusion: Omission errors and reaction time demonstrated the most robust reliability and may represent the most suitable parameters for assessing attention in older adults using a brief computerized task. Reaction time variability showed limited reliability and should be interpreted with caution.