AUTHOR=Jungnickel Tobias , von Jan Ute , Albrecht Urs-Vito 

TITLE=Implementation of Mobile Psychological Testing on Smart Devices: Evaluation of a ResearchKit-Based Design Approach for the Implicit Association Test

JOURNAL=Frontiers in Digital Health

VOLUME=4

YEAR=2022

URL=https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2022.785591

DOI=10.3389/fdgth.2022.785591

ISSN=2673-253X

ABSTRACT=<sec><title>Objective</title><p>To determine whether a framework-based approach for mobile apps is appropriate for the implementation of psychological testing, and equivalent to established methods.</p></sec><sec><title>Methods</title><p>Apple's ResearchKit was used for implementing native implicit association test methods (IAT), and an exemplary app was developed to examine users' implicit attitudes toward overweight or thin individuals. For comparison, a web-based IAT app, based on code provided by Project Implicit, was used. Adult volunteers were asked to test both versions on an iPad with touch as well as keyboard input (altogether four tests per participant, random order). Latency values were recorded and used to calculate parameters relevant to the implicit setting. Measurements were analyzed with respect to app type and input method, as well as test order (ANOVA and χ<sup>2</sup> tests).</p></sec><sec><title>Results</title><p>Fifty-one datasets were acquired (female, <italic>n</italic> = 21; male, <italic>n</italic> = 30, average age 35 ± 4.66 years). Test order and combination of app type and input method influenced the latency values significantly (both <italic>P</italic>&lt;0.001). This was not mirrored for the D scores or average number of errors vs. app type combined with input method (D scores: <italic>P</italic> = 0.66; number of errors: <italic>P</italic> = 0.733) or test order (D scores: <italic>P</italic> = 0.096; number of errors: <italic>P</italic> = 0.85). <italic>Post-hoc</italic> power analysis of the linear ANOVA showed 0.8 by <italic>f</italic><sup>2</sup>=0.25, with α = 0.05 and 4 predictors.</p></sec><sec><title>Conclusions</title><p>The results suggest that a native mobile implementation of the IAT may be comparable to established implementations. The validity of the acquired measurements seems to depend on the properties of the chosen test rather than the specifics of the chosen platform or input method.</p></sec>