Evidence brief: Understanding retention in longitudinal surveys conducted by mobile phone in lower-resource contexts: A case study from Kenya

Abstract

Mobile phone-based surveys are increasingly used for data collection in global contexts – and although the strengths and limitations of this methodology in high-income countries are well-documented, less is known about using this approach (particularly for longitudinal data collection) in low- and middle-income countries, where mobile phone ownership has expanded rapidly. In this paper, we leverage data from a parent study about cervical cancer prevention to understand the features associated with retention and loss to follow-up in repeated random-digit dialed mobile phone-based surveys over time in Kenya. From a cohort of 1405 people surveyed in 2022, 1167 were found again in 2024 (83.1%) and 1106 completed the 2024 survey (78.7%). Survey re-completion was significantly associated with older age (versus 17−32 year olds: OR 1.63 [95% CI 1.21–2.20] for 33−38 year olds, OR 1.65 [95% CI 1.22–2.22] for 39−45 year olds, and OR 2.62 [95% CI 1.84–3.72] for ≥46 year olds), male gender (OR 1.44 [95% CI 1.14–1.83]), and household income. Only age was significantly associated with our secondary outcome of being retained in the sample for follow-up (regardless of survey re-completion). This study’s high follow-up rate was not substantially different across these aforementioned groups, suggesting that mobile phone-based data collection is a promising approach, even for longitudinal surveys, in diverse global settings. We encourage more research about, and using, this methodology so researchers can better understand, and improve, follow-up and its determinants.

Introduction

There is increasing use of remote data collection methods, like mobile phone-based surveys, in a variety of global contexts [1–4]. The strengths and limitations of the mobile phone survey methodology in high-income countries have been well-documented [3,5–9]. These surveys can improve efficiency and lower costs, as well as potentially reduce certain reporting and measurement errors; but they can also introduce potential sources of bias, including non-representativeness and non-response. There has, however, been less empirical data on this methodological issue from low- and middle-income countries (LMICs) [10–14], where mobile phone penetration has exploded [15] and the potential for mobile phone-based data collection is great. In addition, emergency circumstances (like pandemics or natural disasters) may necessitate remote data collection approaches [9,11,16,17], so it is important to understand the potential for these methodologies in diverse global settings [18,19]. Longitudinal studies – wherein the same respondent provides multiple data points over time – have enormous potential for providing meaningful data particularly for time-varying exposures and outcomes [11,20–23], and may be particularly advantaged by using mobile phone-based surveys as these would reduce costs and complexities related to multiple rounds of in-person data collection. These are especially promising in settings with frequent migration where retention in in-person surveys may therefore be low.