Comparing Amazon Mechanical Turk with unpaid internet resources in online clinical trials

Abstract

Internet interventions face significant challenges in recruitment and attrition rates are typically high and problematic. Finding innovative yet scientifically valid avenues for attaining and retaining participants is therefore of considerable importance. The main goal of this study was to compare recruitment process and participants characteristics between two similar randomized control trials of mood management interventions. One of the trials (Bunge et al., 2016) was conducted with participants recruited from Amazon's Mechanical Turk (AMT), and the other trial recruited via Unpaid Internet Resources (UIR). MethodsThe AMT sample (Bunge et al., 2016) consisted of 765 adults, and the UIR sample (recruited specifically for this study) consisted of 329 adult US residents. Participants' levels of depression, anxiety, confidence, motivation, and perceived usefulness of the intervention were assessed. The AMT sample was financially compensated whereas the UIR was not. ResultsAMT yielded higher recruitment rates per month (p < .05). At baseline, the AMT sample reported significantly lower depression and anxiety scores (p < .001 and p < .005, respectively) and significantly higher mood, motivation, and confidence (all p < .001) compared to the UIR sample. AMT participants spent significantly less time on the site (p < .05) and were more likely to complete follow-ups than the UIR sample (p < .05). Both samples reported a significant increase in their level of confidence and motivation from pre- to post-intervention. AMT participants showed a significant increase in perceived usefulness of the intervention (p < .0001), whereas the UIR sample did not (p = .1642). ConclusionsBy using AMT, researchers can recruit very rapidly and obtain higher retention rates; however, these participants may not be representative of the general online population interested in clinical interventions. Considering that AMT and UIR participants differed in most baseline variables, data from clinical studies resulting from AMT samples should be interpreted with caution.