Open access tools for quality-assured and efficient data entry in a large, state-wide tobacco survey in India

• Published in: Global health action 2017-01, Vol.10 (1), p.1394763-1394763
• Main Authors: Shewade, Hemant Deepak, Vidhubala, E, Subramani, Divyaraj Prabhakar, Lal, Pranay, Bhatt, Neelam, Sundaramoorthi, C., Singh, Rana J., Kumar, Ajay M. V.
• Format: Article
• Language: English
• Subjects: Access; Adult; Aged; Aged, 80 and over; Computers, Handheld; Data Collection - methods; data documentation; Data entry; EpiData; Female; GATS; Global Adult Tobacco Survey; Households; Humans; India; Male; Methods Forum; Middle Aged; Open access; open access tools; Polls & surveys; Population Surveillance - methods; Quality control; Questionnaires; Shooting; Smoking; Surveys and Questionnaires - standards; Tamil Nadu; Tobacco; tobacco surveillance; Tobacco Use Disorder; Tobacco use prevalence; Validity
• ISSN: 1654-9716; 1654-9880
• DOI: 10.1080/16549716.2017.1394763

Background: A large state-wide tobacco survey was conducted using modified version of pretested, globally validated Global Adult Tobacco Survey (GATS) questionnaire in 2015-22016 in Tamil Nadu, India. Due to resource constrains, data collection was carrid out using paper-based questionnaires (unlike the GATS-India, 2009-2010, which used hand-held computer devices) while data entry was done using open access tools. The objective of this paper is to describe the process of data entry and assess its quality assurance and efficiency. Methods: In EpiData language, a variable is referred to as 'field' and a questionnaire (set of fields) as 'record'. EpiData software was used for double data entry with adequate checks followed by validation. Teamviewer was used for remote training and trouble shooting. The EpiData databases (one each for each district and each zone in Chennai city) were housed in shared Dropbox folders, which enabled secure sharing of files and automatic back-up. Each database for a district/zone had separate file for data entry of household level and individual level questionnaire. Results: Of 32,945 households, there were 111,363 individuals aged ≥15 years. The average proportion of records with data entry errors for a district/zone in household level and individual level file was 4% and 24%, respectively. These are the errors that would have gone unnoticed if single entry was used. The median (inter-quartile range) time taken for double data entry for a single household level and individual level questionnaire was 30 (24, 40) s and 86 (64, 126) s, respectively. Conclusion: Efficient and quality-assured near-real-time data entry in a large sub-national tobacco survey was performed using innovative, resource-efficient use of open access tools.