While the use of modern methods of contraception are now commonplace in many countries, one-third of women in developing countries who begin using a modern method of contraception quit within the first year and half quit within two years[i]. Most discontinuation occurs among women who want to avoid pregnancy putting them at risk for unwanted pregnancies, maternal morbidity and mortality[ii]. Traditional measures of contraceptive use are collected retrospectively from population representative surveys conducted only every five years which are not well-suited to measuring contemporary trends in contraceptive discontinuation. This is problematic because advocates and health ministries cannot address concerns in a reasonable amount of time to impact widespread change. “Big Data” can supplement these static sources by providing dynamic, real time tracking of the reasons women discontinue using contraceptives and open up possibilities to prevent discontinuation or help facilitate switching between methods.
So what exactly is “Big Data” and how can it supplement traditional reproductive health data?
Yes, big data means a lot of data (volume) but it is collected more frequently (velocity) than traditional reproductive data collected every five years through surveys like the Demographic and Health Surveys (DHS) or every 10 years like a census. Big data is also more than just numbers; it can be images, sounds, videos, satellite imagery and other sources (variety). Not all big data is created equal and researchers should not abandon their skepticism when evaluating big data sources (veracity). Internet-based samples are inherently biased by levels of internet coverage and selection into internet use; population-representative surveys remain the gold standard in tracking reproductive health indicators in developing countries.
Big data’s utility for reproductive health is that it may be able to enhance traditional data sources. The cost of collecting big data can be much lower than collecting population representative survey data. For example, Google search histories can be downloaded from the internet and analyzed using free software platforms like R and Python. Big data can be collected more quickly than survey data, as well. Big data on contraceptive discontinuation may also be more revealing than survey data because people may be more willing to ask Google questions they would be embarrassed to talk about on surveys or with their provider. Big data can allow health workers and program directors to monitor common myths and misconceptions about family planning methods, and address concerns with patients who may not ask them upfront, or tailor programmatic materials to current misconceptions or questions.
What if big data could help us understand why women discontinue contraceptive use and monitor these trends over time in real time? We may be able to keep more women engaged in health services if we can address specific reasons they are discontinuing use and improve public health outcomes for these women, their children and their families. This is an exciting new frontier in population health sciences and one that reproductive health researchers should embrace with creativity, innovation and skepticism.
At the Duke Global Health Institute, we have two projects underway to explore the use of big data for understanding contraceptive discontinuation. To find out more check out our Data+ team (Summer 2018) and Bass Connections team (2018-2019 academic year).
Amy Finnegan is a Postdoctoral Associate at the Duke Global Health Institute (DGHI). Her research broadly addresses health behavior change and communication in the domains of reproductive health and HIV.
[i] Jain, A.K., Obare, F., RamaRao, S. and Askew, I. (2013). Reducing unmet need by supporting women with met need. International Perspectives on Sexual and Reproductive Health, pp.133-141.
[ii] Castle, S. and Askew, I. (2015). Contraceptive Discontinuation: Reasons, Challenges, and Solutions. New York: Population Council.
[iii] Laney, D. (2014). Deja VVVu: Others claiming Gartner’s construct for big data. http://blogs.gartner.com/doug-laney/deja-vvvue-others-claiming-gartners-volume-velocity-variety-construct-for-big-data/
[iv] Hilbert, M. (2016). Big data for development: A review of promises and challenges. Development Policy Review, 34(1), 135-174.