Feasibility, acceptability, and validity of crowdsourcing for collecting longitudinal alcohol use data

Challenges to conducting longitudinal research include financial, time, and geographic constraints. An emerging sampling method positioned to address these concerns is crowdsourcing. This study evaluated the feasibility, acceptability, and validity of collecting intensive longitudinal alcohol use data with the crowdsourcing platform, Amazon.com's Mechanical Turk (mTurk). Participants (N = 278) recruited from mTurk provided weekly recordings of daily alcohol and soda use over an 18-week period. Construct and external validity was evaluated using generalized linear mixed models describing associations of between-subject (e.g., alcohol use severity) and within-subject (e.g., day of week) variables with prospectively collected alcohol and soda use. High response rates were observed across the 18-week period demonstrating feasibility (64.1%-86.8%). The design was acceptable with 94% of participants indicating they were satisfied with the procedures. Multilevel models supported construct and external validity by replicating expected associations, such as more frequent and heavier drinking by individuals with higher AUDIT scores and on weekends. These effects were specific to alcohol use and did not extend to soda consumption. These data support the feasibility, acceptability, and validity of using mTurk for intensive longitudinal data collection. Future studies may leverage this platform to generate large, geographically diverse samples for prospective behavioral analytic research.