Practical Fraud Detection and Prevention in Incentivized Online Surveys: Secondary Analysis of the ADOPT Study.

Background: Digital surveys are increasingly integrated into clinical and public health research to capture patient-reported outcomes. However, concerns about fraudulent or duplicate responses threaten data integrity. Most of the literature on survey fraud focuses on open-access online recruitment, where bot-generated or anonymous entries are common, but far less is known about fraud patterns in clinic-linked, incentive-based surveys. Evaluations of the real-world implementation of fraud-deterrence strategies remain limited. Objective: This study evaluated whether implementing enhanced fraud-deterrence procedures in an incentive-based, clinic-linked, postprocedure survey reduced the prevalence of potentially fraudulent responses. Second, the study evaluated which indicators were most frequently triggered before and after implementation. Methods: This evaluation was conducted within the ADOPT (Alternatives to Dental Opioid Prescribing After Tooth Extraction) study. Eligible patients aged 12 to 25 years were recruited through QR-coded, clinic-distributed flyers and invitation cards. Participants completed a screening survey followed by an incentivized postprocedure survey between days 6 and 10 after tooth extraction. A midstudy protocol modification introduced enhanced fraud-deterrence measures in the screening process, including a phone number requirement, prohibition of email invitations, date of birth confirmation, and the use of a participant list with SMS text message invitations. For analysis, survey responses were categorized as "control" (before modification) or "intervention" (after modification). A 6-item scoring system assessing completion time outliers, submission time, repeated screeners, duplicated phone numbers, a blank recruitment source, and illogical response patterns was used to classify responses as potentially fraudulent (≥2 indicators). Sensitivity analyses evaluated thresholds from 1 to 3 indicators. Results: A total of 573 survey responses were included, with 122 in the control cohort and 451 in the intervention cohort. The overall prevalence of potentially fraudulent responses (50/573, 8.7%) was lower than the rates reported in open-access online survey research, and the difference between the control and intervention cohorts was not statistically significant (15/122, 12.3% vs 35/451, 7.8%; P=.12). Fewer surveys in the intervention cohort were flagged for a blank recruitment source (16/451, 3.5% vs 15/122, 12.3%; P<.001) and completion of multiple screeners (29/451, 6.4% vs 16/122, 13.1%; P=.02). The frequency of duplicated phone numbers was higher in the intervention survey (82/451, 18.2% vs 3/122, 2.5%; P<.001), although this difference was not statistically significant when restricted to individuals who provided a phone number (82/451, 18.2% vs 3/46, 6.5%; P=.06). Sensitivity analyses showed consistent patterns across alternative thresholds, and subgroup analyses did not show overall differences in fraud rates based on age, sex, or recruitment location. Conclusions: Enhanced fraud-deterrence procedures did not statistically significantly reduce overall fraud prevalence in a survey setting using clinic-linked, QR-based recruitment with modest incentives. A transparent scoring system provides a replicable approach for assessing survey integrity and may be preferable to reliance on eligibility gating alone.