Effect of Interrupted Interview on Survey Responses and Data
Quality: an Implication for Validity of Demographic Health Survey
Davies, Joseph Hindowa Albert
Master's Thesis (84 pages)
Committee Chair / Thesis Adviser: Stephenson, Robert
Committee Members: McFarland, Deborah A
Research Fields: Health Sciences, Public Health
Partnering Agencies: Emory University schools, faculty or affiliated programs
Keywords: Interrupted Interview; Social Desirability Effect; Data Quality; Sensitive Question; Non Sensitive question
Program: Rollins School of Public Health, Hubert Department of Global Health
The Demographic Health Survey (DHS) is one of, if not the only, most widely used source of data for critical policy decision making, planning, and implementing, as well as monitoring and evaluating health and public health programs. Like most survey data collection, it is not devoid of data quality issues resulting from interrupted interview because, even though it reports the presence of a third party during interview, it fails to provide measures to correct for this effect on data quality, which can have implications for the validity of such data.
Aim of study
This study assesses the effect of Interrupted Interview on data quality of 2008 DHSs of two West African countries: Ghana and Nigeria. This study also proposes some implications for policy and public health programs and provides recommendations to improve future DHSs.
Bivariate and multivariate logistic regression analysis was done on/for an individual married women survey to assess the level and significant association of interrupted interview on response pattern in relation to both nature of question and socio-demographic characteristics of respondents.
Results show a relatively low proportion of interrupted interview among married women from both countries: Ghana, 7.8%; Nigeria, 6%. They also show significant association between interrupted interview and socio-demographic characteristics, such as education, place and region of residence, and household wealth status of respondents as follows: Ghana, place of residence p=0.002; region of residence, p=0.000; and wealth, p=0.003; Nigeria, age, p=0.000; education, p=0.000; religion, p=0.000; parity, p=0.001; and place/region of residence, wealth and ethnicity, p=0.000. The result, however, indicates no significant association between interrupted interview and nature of question except for >1 sex partners in the last 12 months among married women in Nigeria.
It is therefore concluded that the issue of interrupted interview should not be overlooked because it has an effect on the responses provided, which can have an implication for the quality of data generated by one of the most reliable data sources, the DHS. Measures should be instituted to correct for such effects in not only the data collection process, but also the design, collection, and analysis processes of future DHSs.
Table of Contents
Aim of study...4
Social Desirability effect and Data quality...6
Multivariate Logistic Regression...40
Multivariate Logistic regression for Ghana...40
Multivariate Logistic Analysis for Sensitive Outcome variables...40
Multivariate Logistic Analysis for Non Sensitive Outcome Variables (Ghana)...43
Multivariate Logistic Regression of Nigerian data...45
Multivariate Logistic Analysis for Sensitive Outcome Variables (Nigeria)...46
Multivariate Logistic Analysis for non sensitive Outcome Variables (Nigeria)...50
Discussion, Conclusion and