The relationship between interviewer-respondent familiarity and family planning outcomes in the Democratic Republic of Congo: a repeat cross-sectional analysis. Issue 1 (21st January 2019)

• Record Type: Journal Article
• Authors: Anglewicz, Philip; Akilimali, Pierre; Eitmann, Linnea Perry; Hernandez, Julie; Kayembe, Patrick
• Is Part Of: BMJ open. Volume 9:Issue 1(2019)
• Publication Date: 2019-01-21
• Subjects: family planning -- survey methodology -- interviewer effects; Medicine -- Research -- Periodicals; 610.72
• DOI: 10.1136/bmjopen-2018-023069
• Languages: English
• ISSNs: 2044-6055

Abstract

Abstract : Objectives: The typical approach of survey data collection is to use interviewers who are not from the study site and do not know the participants, yet the implications of this approach on data quality have seldom been investigated. We examine the relationship between interviewer–respondent familiarity and selected family planning outcomes, and whether this relationship changes over time between 2015 and 2016. Setting: We use data from the Performance Monitoring and Accountability 2020 Project in Kongo Central Province, Democratic Republic of Congo. Participants: Participants include representative samples of women of reproductive ages (15 to 49), 1565 interviewed in 2015 and 1668 in 2016. The study used a two-stage cluster design: first randomly selecting enumeration areas (EAs), then randomly selecting households within each EA. Design: We first identify individual characteristics associated with familiarity between RE and respondent. Next, we examine the relationship between RE–respondent acquaintance and family planning outcomes. Finally, we use two waves of data to examine whether this relationship changes over time between 2015 and 2016. Results: In multivariate analysis, interviewer–respondent acquaintance is significantly associated with last birth unintended (OR 1.91, 95% CI 1.17 to 3.13) and reported infertility in 2015 (OR 2.26, 95% CI 1.03 to 4.95); and any contraceptive use (OR 1.51, 95% CI 1.01 to 2.28), traditional contraceptive use (OR 1.79, 95% CI 1.10 to 2.89), reported infidelity (OR 1.89, 95% CI 1.02 to 3.49) and age at first sex (coefficient −0.48, 95% CI −0.96 to −0.01) in 2016. The impact of acquaintance on survey responses changed over time for any contraceptive use (OR 2.09, 95% CI 1.33 to 3.30). Conclusions: The standard in many large-scale surveys is to use interviewers from outside the community. Our results show that interviewer–respondent acquaintance is associated with a range of family planning outcomes; therefore, we recommend that the approach to hiring interviewers be examined and reconsidered in survey data collection efforts.