Test-retest reliability and concurrent validity of the emotional stroop task (EST) in detecting anxiety and depression symptoms : an exploration of different summary measures

Abstract

In most clinical settings and practices, the measurement and diagnostic processes of common mental health disorders such as major depressive disorder and generalized anxiety disorder rely heavily on patients’ subjective self report with verified rating scales. With the vast development of computational methods and machine learning strategies, an increasing amount of research is interested in developing algorithms for psychopathology to inform severity of symptoms and preventive interventions. In this current study, a 300-item locally adapted Emotional Stroop Task (EST) in Cantonese with word stimuli of depressive, anxious and neutral emotional category was developed to explore the possibility of whether the reaction time task could serve as an objective measurement of mood disorder symptoms and severity in depression and anxiety among college students in Hong Kong. Test-retest reliability of the Cantonese EST with a 14-day time interval and its concurrent validity with currently established self-reported measures were investigated. A total of one hundred and fifty-nine students from the University of Hong Kong enrolled in the study, in which 90 participants were retained and have successfully completed all experimental procedures. Cronbach alpha was used to assess the test-retest reliability of the relevant measures, while Pearson correlations with scores of GAD-7 and PHQ-9 were used to assess its construct validity. By administering a repeated measures ANOVA analysis on accuracy and response latency between three emotional word stimuli categories, results indicated that the Cantonese EST used in the current study showed significant difference in mean response latency among the three emotional categories of word stimuli and a consistent pairwise difference in mean response latency in between depressive word stimuli and neutral word stimuli in both test and retest session. Other measures, such as the difference in mean response latency between anxious word stimuli and neutral word stimuli, and concurrent validity with self reported measures of GAD-7 and PHQ-9 only yielded significant difference in the first test session but not the retest session. Analyses from Drift Diffusion Model did not reveal significant impact from either compatible or incompatible trials on sequential evidence accumulation over time. The results provided evidence that the current Cantonese EST measurements are suggestive of objective measures towards a response delay reflective of psychopathology, but that further investigation on the task design and data analysis methodologies would need to be further investigated for more consistent and reliable measures.