By Martha Henry
Harvard AIDS Initiative (HAI) researchers are hoping to use data from cell phones to better understand sexual networks in Botswana villages.
The Botswana Combination Prevention Project (BCPP) will deliver a combination of HIV/AIDS prevention measures to 15 villages in Botswana to try to dramatically reduce new HIV infections. Villages receiving the intervention will be compared to 15 control villages that are similar, but receive standard HIV care. The goal of the BCPP is to significantly reduce new HIV infections at a community level. Success will be measured by how sharply the rate of new infections declines within entire villages.
Unlike a laboratory experiment in which conditions can be tightly controlled, the BCPP is a large clinical trial in the real world. And the real world is messy. Though people are more likely to have sexual partners within their own village, a certain fraction of relationships will likely occur across the villages, potentially leading to relationships that couple individuals from treatment and standard of care villages. If lots of people have relationships across the villages, the effect of the BCPP interventions will be diluted and become harder to measure. Researchers have to estimate how much social and sexual mixing takes place between villages and adjust their statistics accordingly.
In an innovative study, Dr. J.P. Onnela hopes to examine data from cell phone companies to measure how much mixing occurs between villages. Onnela has a PhD in Computational Science and did a postdoc at Harvard under Dr. Nicholas Christakis, a leader in social network research.
Botswana has one of the highest cell phone penetration rates in Africa—about 115%. The number of cell phones is greater than the number of citizens. Onnela, working with Drs. Vlad Novitsky and Rui Wang, plans to use phone data to investigate the social integrity of the villages in the BCPP. Two major cell phone operators have agreed to share data if the project gets funded.
The phone companies would turn over anonymous information about phone calls and text messages; no personal identifiers would be included. “We would know that person A lives within a certain geographical area and makes phone calls to 15 other people, 10 of whom reside within that area and 5 outside of it,” explained Onnela.
Though obviously people don’t have sex with everyone they call, phone usage is a good proxy for quantifying social interactions. “We’re not predicting sexual relationships at the level of individuals,” said Onnela. “Rather we want to understand to what extent two villages are socially intermixed.”
“We hope the project can give us the tools to understand how closely and how strongly different groups of individuals are connected and intermingled,” said Onnela. “If we can understand the extent of mixing, we can learn how effective different treatments really are.”
The idea of using cell phone data to study social networks in an HIV cluster-randomized trial is a new idea. Like most untried ideas, funding agencies consider it risky. “We’ve had a lot of enthusiastic feedback,” said Onnela, “but currently we don’t have funding.”
But he’s optimistic. “We can learn a lot from cell phone data that’s already being collected, and we’ve had reasonably good success with this in the past. This isn’t extra data that’s being collected for the BCPP. This is information that’s collected every day from your phone and mine and everyone else’s.”
“The beautiful thing is that we’re leveraging existing data,” he continues, enthusiastic about the methodology he’s developing. “We’d like to take the enormous data sets collected by cell phone operators and ask, ‘How can we use these data for public health?’”
Further reading:
Structure and Tie Strengths in Mobile Communication Networks
Geographic Constraints on Social Network Groups