Data Quality in HIV/Aids Web-Based Surveys

Handling Invalid and Suspicious Data

Datos Bibliográficos

ID2407413
AutoresJose A Bauermeister (University of Michigan), Jose Bauermeister (0000-0002-9276-2306, University of Michigan, autor de correspondencia), Emily Pingel (0000-0003-1053-1158, University of Michigan), Marc Zimmerman (0000-0001-8284-9910, University of Michigan), Marc A Zimmerman (0000-0001-5064-4259, University of Michigan), Mick P Couper (0000-0001-5006-9346, University of Michigan), Mick Couper (Institute for Social Research, University of Michigan, Ann Arbor, MI, USA), Alex Carballo-Diéguez (0000-0001-5888-031X, New York Psychoanalytic Society and Institute), Victor J Strecher (0000-0002-0587-2305, University of Michigan)
Año2012
Volumen24
Número3
Páginas272-291
Fecha de publicación2012-08-01
Peer ReviewedSí
Open AccessSí
TipoARTICLE
RevistaCAM (JOURNAL)
Identificadores de la revistaISSN: 1087-8513 • E-ISSN: 1552-3969
EditorialSAGE Publications Inc (PUBLISHER)
DOI10.1177/1525822x12443097
PMID23180978
PMCIDPMC3505140
OpenAlexW1622676990
IdiomaEN
Citas recibidas76
Referencias citadas14

Invalid data may compromise data quality. We examined how decisions made to handle these data may affect the relationship between Internet use and HIV risk behaviors in a sample of young men who have sex with men (YMSM). We recorded 548 entries during the 3-month period and created six analytic groups (i.e., full sample, entries initially tagged as valid, suspicious entries, valid cases mislabeled as suspicious, fraudulent data, and total valid cases) using data quality decisions. We compared these groups on the sample's composition and their bivariate relationships. Forty-one cases were marked as invalid, affecting the statistical precision of our estimates but not the relationships between variables. Sixty-two additional cases were flagged as suspicious entries and found to contribute to the sample's diversity and observed relationships. Using our final analytic sample, we found that very conservative criteria regarding data exclusion may prevent researchers from observing true associations. We discuss the implications of data quality decisions and its implications for the design of future HIV/AIDS web surveys

Bivariate analysis · Data quality · Family medicine · Human immunodeficiency virus (HIV) · Metric (unit) · Quality (philosophy) · Sample (material) · Statistics · Computer Science · Data-Driven Disease Surveillance · HIV/AIDS Research and Interventions · Mathematics · Medicine · Psychology · Social Psychology · Survey Methodology and Nonresponse