Up to 94% of survey participants in online health research were excluded due to fraudulent responses, according to a recent scoping review.

Researchers conducted a review examining strategies to identify and mitigate fraudulent responses in online health research recruitment. Published in BMJ Evidence-Based Medicine, the study reviewed 23 peer-reviewed articles sourced from nine databases, including Medline and Scopus. Using the Joanna Briggs Institute methodology and PRISMA Extension for Scoping Reviews framework, the researchers evaluated fraud detection and mitigation strategies.

Among the studies, 18 (78%) reported fraudulent responses, with exclusion rates ranging from 3% to 94%. Fraud detection methods included survey completion time analysis (26%), where responses completed in under five minutes were flagged, and non-conforming responses (52%), such as inconsistent or implausible answers. Geographical incongruity (30%) was identified through IP address verification, ZIP code mismatches, or VPN usage. Temporal anomalies, such as responses submitted at odd hours, were noted in 17% of studies.

Higher incentives, ranging from $5 to $200, were linked to increased fraudulent responses in 73% of studies. Mitigation strategies included CAPTCHA systems (34%), manual verification (21%), and unique URL checks (8%). While CAPTCHA methods reduced automated responses, they were less effective against human fraud. Manual verification steps, including email and phone checks, were more labor-intensive but provided additional reliability.

The review highlighted the need for a multifaceted approach, combining detection and mitigation strategies to enhance data integrity. It also emphasized inconsistencies in evaluating antifraud measures and called for further exploration of advanced technologies like artificial intelligence to address evolving fraudulent activities. The authors concluded that no single method suffices and recommended integrating multiple strategies to safeguard online health research data.

Full disclosures and additional data are available in the published review.