Addressing participant validity in a small internet health survey (the restore study): Protocol and recommendations for survey response validation

James Dewitt, Benjamin D Capistrant, Nidhi Kohli, B. R. Simon Rosser, Darryl Mitteldorf, Enyinnaya Merengwa, William G West

Research output: Contribution to journalArticlepeer-review

8 Scopus citations

Abstract

Background: While deduplication and cross-validation protocols have been recommended for large Web-based studies, protocols for survey response validation of smaller studies have not been published. Objective: This paper reports the challenges of survey validation inherent in a small Web-based health survey research. Methods: The subject population was North American, gay and bisexual, prostate cancer survivors, who represent an under-researched, hidden, difficult-to-recruit, minority-within-a-minority population. In 2015-2016, advertising on a large Web-based cancer survivor support network, using email and social media, yielded 478 completed surveys. Results: Our manual deduplication and cross-validation protocol identified 289 survey submissions (289/478, 60.4%) as likely spam, most stemming from advertising on social media. The basic components of this deduplication and validation protocol are detailed. An unexpected challenge encountered was invalid survey responses evolving across the study period. This necessitated the static detection protocol be augmented with a dynamic one. Conclusions: Five recommendations for validation of Web-based samples, especially with smaller difficult-to-recruit populations, are detailed.

Original languageEnglish (US)
Article numbere96
JournalJournal of medical Internet research
Volume20
Issue number4
DOIs
StatePublished - Apr 2018

Keywords

  • Data accuracy
  • Data analysis
  • Design
  • Fraudulent data
  • Research
  • Research activities

Fingerprint Dive into the research topics of 'Addressing participant validity in a small internet health survey (the restore study): Protocol and recommendations for survey response validation'. Together they form a unique fingerprint.

Cite this