Recruitment to Retention in Remote Research: Learnings from a Large Decentralized Real-World Study (Preprint)

BACKGROUND Smartphones are increasingly used in health research to reach, recruit and assess the health of large and diverse populations. These devices provide a continuous connection between participants and researchers to monitor long-term health and behavior trajectories using multimodal data streams, ranging from health surveys to sensor data at a fraction of the cost of traditional research studies. Despite the potential of real-world data for assessing health and behavior, representative and equitable recruitment and retention of the target population remain key challenges. Remote research has unique challenges, particularly issues with data quality from participants, the choice of recruitment channels, and the most effective ways to retain large and representative populations over a long period. OBJECTIVE We explored the impact of different recruitment and incentive distribution approaches on cohort characteristics and long-term retention. Real-world factors that significantly impact active and passive data collection were also evaluated. METHODS This is a secondary data analysis of research engagement from a large-scale, remote clinical study of flu, cold and COVID detection. Recruitment was conducted in two phases, between March 15, 2020, and January 04, 2022. Over 10,000 smartphone owners in the U.S. were recruited to provide 12 weeks of daily surveys and smartphone-based passive sensing data. Using multivariate statistics, we investigated the impact of different recruitment and incentive distribution approaches on the cohort characteristics. Survival analysis was used to assess the effects of socio-demographics on participant retention across two recruitment phases. Logistic regression was used to evaluate associations between passive data sharing from multiple smartphone sensors and cohort demographics. RESULTS We analyzed data from over 10,000 participants with 269,037 days of active data along with 336,292 days of passive sensor data. Our key findings show that i.) Participants recruited from social media/ads (Phase 1) had significantly lower compliance with the baseline survey completion and participated for a shorter period compared to those recruited from crowdsourcing platforms (Prolific, mTurk; Phase 2) (p < .0001). ii) Socioeconomic and demographic factors such as race/ethnicity, age, income, etc. of the cohort impacted participant retention (p < .0001), however, these factors varied remarkably between two phases. iii) Participants are more likely to adhere to baseline surveys if administered immediately after consent/enrollment. iv) Passive data sharing patterns across Android and iOS showed that Black/African Americans were significantly less likely to share passive sensor data than non-Hispanic whites (OR Range = 0.15 - 0.54; p < .001). CONCLUSIONS Our findings provide valuable empirical evidence about potential best practices to recruit and retain target populations in remote clinical research in the conduct of remote clinical research. Future studies should consider and account for possible biases in participant enrollment and retention based on the choice of recruitment platforms and incentive distribution approaches as well as engaging people from underrepresented populations.