Tracking Cell Phones to Predict Coronavirus - Will Big Brother Protect You?

Cell phone location data proves useful, but is it a slippery slope?

In your pocket right now, or on your desk, or maybe your bedside table, is a sophisticated GPS-connected tracking device that has, more likely than not, been logging your movements for years now. Of course, you know it as a cell phone, and most of us have a vague sense that it knows more about us than we might like. That’s right, big brother is watching you.

Is that ungood? It certainly irks my more libertarian tendencies, but such information can be put to good use, as demonstrated in this study, appearing in JAMA Internal Medicine, which shows how cell phone location data can predict new COVID cases.

A research team led by Josh Baker - full disclosure: Josh and I were residents together at UPenn and he is about the furthest thing from Big Brother you can get – accessed publicly available county-level cell phone data to determine what happened around the country when stay-at-home orders were put in place.

Overall, you can see dramatic reductions in cell phone pings at retail and workplace locations, a modest decline from grocery stores and parks, and an increase in residences. People were staying home.

But there was some variability around the country.

View fullsize

Behavior change after stay at home orders.

For instance, counties with higher poverty levels had lower declines in workplace activity – perhaps because the jobs in those areas weren’t the work-from-home type. Rural counties also had less of a dramatic behavior change after stay-at-home orders occurred. Counties with higher case-rates had more dramatic reductions in activity – people were taking things seriously. Concerningly, but not surprisingly, the further you got from the day of the stay-at-home order, the less impact it had. Retail activity increased by 0.5% every day after the stay-at-home order, for example.

That variation in behavior allowed the team to ask some interesting questions. Would counties that had greater reductions in retail and workplace activity see slower coronavirus growth rates?

View fullsize

Less behavior change —> more coronavirus

Indeed they did. Accounting for a 5-day lag between exposure and symptoms, the team showed that counties with the lowest levels of behavior change had the highest growth in cases. 

This held true even after adjustment for multiple county-level factors, including the amount of tests conducted. Put simply – counties where people didn’t heed the stay-at-home order as much (for whatever reason) had higher COVID growth rates thereafter.

Adding cell phone data to other epidemiologic information significantly improved a statistical model to predict new cases, suggesting that cell phone monitoring could be a useful tool to figure out where the next hotspots might be.

So… how do we feel about it?  Let’s assume this is a useful tool, let’s assume cell phone surveillance saves lives. And let’s remember this is county-level data, not individual cell phone location tracking. I think with those caveats this feels ok. But might there be a slippery slope between county-level cell phone data and individual-level cell phone data as some countries have reportedly been using?

I don’t see the U.S. going there, to be honest, we can’t even get people to wear masks. I am not particularly afraid that anytime soon we’ll be declaring victory over ourselves and loving Big Brother. But 2020 has been full of surprises.

This commentary first appeared on medscape.com.