Identifying Fraudulent Responses in a Study Exploring Opinions on Optimal Delivery Options for Pregnancies Impacted by Gestational Diabetes: Data Analysis of a Web-Based Survey (Preprint)

BACKGROUND Current literature is unclear on the safety and optimal timing of delivery for pregnant individuals with gestational diabetes mellitus (GDM), inspiring the study team to conduct a web-based survey study exploring patient and provider opinions on delivery options. An incident of fraudulent activity with survey responses prompted a shift in focus of the research project. Unfortunately, despite the significant rise of web-based surveys used in medical research, there remains very limited evidence on the implications of, and optimal methods to handle, fraudulent web-based survey responses. OBJECTIVE The objective of this paper was to highlight our approach to identifying fraudulent responses in a web-based survey study, in the context of clinical perinatal research exploring patient and provider opinions on delivery options for GDM pregnancies. METHODS We conducted cross-sectional web-based surveys across Canada for pregnant patients and perinatal healthcare providers. Surveys were available through REDCap, and recruitment took place between March and October 2023. A change to recruitment introduced a $5.00 gift card incentive to increase survey engagement. In mid-October 2023, an incident of fraudulent activity was reported, upon which the surveys were deactivated. Systematic guidelines were developed by the study team in consultation with information technology services and the research ethics board to filter fraudulent from true responses. RESULTS Between October 14th and 16th, 2023, an influx of almost 400 patient and over 2000 provider responses were recorded in our online survey. Systematic filtering flagged numerous fraudulent responses. Fraudulent patient responses were most often due to records with identical timestamp and responses (45.8%), same timestamp and responses with slight variations in wording (32.1%), and fraudulent email addresses provided (11.7%). Fraudulent provider responses were most often due to fraudulent email addresses (43.4%), suspicious timestamp (34.3%), and responses with very similar timestamp and similar responses (8.94%). CONCLUSIONS This fraudulent incident highlights the importance of preserving research integrity by using methodologically sound practices to extract true data for research findings. These fraudulent events continue to threaten the credibility of research findings and future evidence-based practices.