It's not so much where people live, but where they frequently spend their time that can provide useful information for predicting community health measures. A team led by geographers in the Penn State College of Earth and Mineral Sciences found that adding place visitation data - geographical data points collected from millions of anonymous cell phone users with GPS-enabled devices - to a population health model increased the model's predictive performance by an average of 7.5%. This improvement in predicting public health factors, including the rates of depression and binge drinking, may assist public policymakers and health planners, researchers said.
They published their results in Computational Urban Science.
"In daily life, people visit many different places, such as restaurants, parks, gyms and bars," said Zhenlong Li, associate professor of geography, director of the Geoinformation and Big Data Research Lab and corresponding author of the paper. "Our results show that these place visitation patterns can provide useful additional information for estimating community health measures, complementing traditional demographic and social variables."
Traditionally, community health outcomes are determined based on static demographic and social statistics collected by the U.S. Census and State Departments of Health, such as race, age, socioeconomic status, access to healthcare and education levels. After creating a model containing just these measures, the researchers added aggregated cell phone data from the year 2019, which tracked visits to approximately seven million public points of interest, such as parks, restaurants, gyms, casinos, convenience stores, primary healthcare facilities and religious centers. The team then analyzed the data for the entire continental U.S., sorted into approximately 12,800 rural and 56,600 urban census tracts. The addition of place visitation data not only improved the model's predictive capabilities but also uncovered information about how the places people visit may shape community health.
"Our hypothesis was that the daily activity patterns at the neighborhood level would better predict neighborhood health, and by analyzing place visitation data on a supercomputer in the lab, we were able to verify that hypothesis," said Temitope Akinboyewa, doctoral student in geography and first author of the paper. "In particular, the model showed the biggest predictive gains for place visitations related to binge drinking, depression, routine medical checkups, obesity and asthma."
