Measuring under-5 mortality and fertility through mobile phone surveys: an assessment of selection bias in 34 low-income and middle-income countries

• Published in: BMJ open Vol. 13; no. 11; p. e071791
• Main Authors: Sánchez-Páez, David A, Masquelier, Bruno, Menashe-Oren, Ashira, Baruwa, Ololade Julius, Reniers, Georges
• Format: Journal Article
• Language: English
• Published: England British Medical Journal Publishing Group 17.11.2023; BMJ Publishing Group LTD; BMJ Publishing Group
• Series: Original research
• Subjects: Adolescent; Adult; Bias; Cell Phone; Cellular telephones; Child mortality; Demography; Developing Countries; Epidemiology; Female; Fertility; Humans; LDCs; Low income groups; Middle Aged; Research Methods; Selection Bias; STATISTICS & RESEARCH METHODS; Surveys and Questionnaires; Womens health; Young Adult; STATISTICS & RESEARCH METHODS; Epidemiology; Demography
• ISSN: 2044-6055; 2044-6055
• DOI: 10.1136/bmjopen-2023-071791

ObjectivesThis study aims to assess sample selection bias in mobile phone survey estimates of fertility and under-5 mortality.DesignWith data from the Demographic and Health Surveys, we use logistic regressions to identify sociodemographic correlates of mobile phone ownership and access, and Poisson regressions to estimate the association between mobile phone ownership (or access) and fertility and under-5 mortality estimates. We evaluate the potential reasons why estimates by mobile phone ownership differ using a set of behavioural characteristics.Setting34 low-income and middle-income countries, mostly in sub-Saharan Africa.Participants534 536 women between the ages of 15 and 49.Outcome measuresUnder-5 mortality rate (U5MR) and total fertility rate (TFR).ResultsMobile phone ownership ranges from 23.6% in Burundi to 96.7% in Armenia. The median TFR ratio and U5MR ratio between the non-owners and the owners of a mobile phone are 1.48 and 1.29, respectively. Fertility and mortality rates would be biased downwards if estimates are only based on women who own or have access to mobile phones. Estimates of U5MR can be adjusted through poststratification using age, educational level, area of residence, wealth and marital status as weights. However, estimates of TFR remain biased even after adjusting for these covariates. This difference is associated with behavioural factors (eg, contraceptive use) that are not captured by the poststratification variables, but for which there are also differences between mobile phone owners and non-owners.ConclusionsMobile phone surveys need to collect data on sociodemographic background characteristics to be able to weight and adjust mortality estimates ex post facto. Fertility estimates from mobile phone surveys will be biased unless further research uncovers the mechanisms driving the bias.