Health surveys using mobile phones in developing countries: Automated active strata monitoring and other statistical considerations for improving precision and reducing biases

Research output: Contribution to journalArticle

Abstract

In low-And middle-income countries (LMICs), historically, household surveys have been carried out by face-To-face interviews to collect survey data related to risk factors for noncommunicable diseases. The proliferation of mobile phone ownership and the access it provides in these countries offers a new opportunity to remotely conduct surveys with increased efficiency and reduced cost. However, the near-ubiquitous ownership of phones, high population mobility, and low cost require a re-examination of statistical recommendations for mobile phone surveys (MPS), especially when surveys are automated. As with landline surveys, random digit dialing remains the most appropriate approach to develop an ideal survey-sampling frame. Once the survey is complete, poststratification weights are generally applied to reduce estimate bias and to adjust for selectivity due to mobile ownership. Since weights increase design effects and reduce sampling efficiency, we introduce the concept of automated active strata monitoring to improve representativeness of the sample distribution to that of the source population. Although some statistical challenges remain, MPS represent a promising emerging means for population-level data collection in LMICs.

Original languageEnglish (US)
Article numbere121
JournalJournal of medical Internet research
Volume19
Issue number5
DOIs
StatePublished - May 2017

    Fingerprint

Keywords

  • Mobile health
  • Mobile phone
  • Research methodology
  • Sampling studies
  • Surveys and questionnaires

ASJC Scopus subject areas

  • Health Informatics

Cite this