How can timely, local data on physical activity improve health?
Timely, actionable data on health behaviors and conditions is one of the cornerstones of public health. Data that capture physical activity behaviors within communities can inform action to improve health by answering two vital questions: How active is the community? Which parts of the community most need interventions to support physical activity? Answering these questions requires timely and geographically precise data that call tell us where people are and are not being active.
In the Division of Nutrition, Physical Activity, and Obesity at the Centers for Disease Control and Prevention, we have historically relied on national surveillance systems to measure physical activity behavior across the United States. Systems such as the National Health Interview Survey, the National Household Travel Survey, the American Time Use Survey, and the National Health and Nutrition Examination Survey have been invaluable sources of data. However, these national sources are not as useful for informing public health action in local communities because they:
- Rely on nationally representative samples of the population
- Usually rely on recall of past behaviors
- Have a considerable time lag between when data are collected and available to researchers
- Lack geographic granularity, rarely providing information below the state level.
Thus, these classic data sources would be less than ideal for informing public health action within communities. In search of physical activity data that are more comprehensive, accurate, timely, and geographically precise, we explored an emerging data source—the phone inside your pocket when you walk the dog or bike to work.
What is location-based services data?
We sought to tap into big data and location-based services (LBS) technologies. LBS data come from smartphone apps which anonymously collect users’ location either while the app is being used or in the background while the device is moving. For example, using this location data, weather apps provide the local forecast and navigation apps find nearby restaurants. LBS data can be used by private, nonprofit, and government entities for planning and research purposes. StreetLight Data, Inc., is one company that uses LBS for transportation planning and research. StreetLight feeds the data, comprising sequential “pings,” through an algorithm which classifies these ping sequences as transportation modes (including vehicle, biking, and walking).
What did we do?
To determine whether LBS can provide timely, local physical activity data, we first needed some evidence about its validity. We compared county-level LBS data purchased from StreetLight to a traditional surveillance data source, the American Community Survey (ACS). Our StreetLight dataset contained the average daily count of pedestrian and bicycle trips for 2019, from 298 sampled US counties, while our comparison dataset from ACS contained the proportion of workers who reported walking or bicycling to work (active transportation).
To validate the data and perhaps generate new research questions, we compared these two datasets using several pairs of metrics. For example, we conducted some analyses using only weekday or weekend trips in StreetLight, or restricted ACS data to workers who reported never teleworking. Looking at different combinations of the metrics may help us better understand which applications of the data are most valid. For the most strongly correlated metric pair, we also explored variations in the comparisons by subpopulations.
What did we find?
County-level correlations between the StreetLight and ACS measures were mostly moderate (Spearman’s rho of 0.11 to 0.53) for walking and mostly strong (0.49 to 0.61) for biking. For each activity, the most strongly correlated metric pair was between the number of trips per 1,000 county residents (from StreetLight) and the percent of non-teleworking individuals who actively commuted to work (from ACS). When we analyzed this metric pair within subpopulations, we found that the correlation was strongest in denser and more urban counties.
Why should you care?
These findings have several implications for physical activity practitioners and researchers. First and foremost, our study suggests that LBS data may be a valid, more timely, and actionable data source to measure physical activity behavior, particularly active transportation, and such data can complement traditional physical activity surveillance sources. Second, our study suggests that LBS have different utility based on the community context. LBS data may be particularly useful to understand physical activity patterns in denser and more urban areas. At least for now, it requires additional validation in sparsely populated rural communities.
Regardless of application, LBS data have some limitations that practitioners and researchers should consider. Because LBS data are generated by smartphones, they omit individuals who do not own or routinely use smartphones, or who have configured apps to block location-based permissions. If such individuals are unequally distributed across the population (e.g., more older persons than younger persons), the resulting LBS data may be less representative of the entire population in their communities. Another limitation is that neither LBS nor traditional physical activity data sources—though correlated with one another—are considered “gold standards” in and of themselves. In addition, physical activity practitioners and researchers should be aware that StreetLight provides repackaged LBS data, processed through its proprietary algorithm; therefore, we cannot comment from this study alone on the validity of metrics provided by other LBS data vendors.
Finally, LBS data raise ethical questions beyond those we typically confront in the field of physical activity surveillance. Are participants aware that they are generating LBS data? Do they understand the opt-in and opt-out features of apps installed on their phones? Is privacy protected? These questions and others demonstrate the importance of (1) ethics and review board oversight, and (2) using only aggregated and anonymized data, which we used in our study. We found that LBS data on walking and bicycling is correlated with other traditional measures of active transportation—and, by providing more timely and geographically precise information—may potentially complement our traditional surveillance systems.
Citation: Soto GW, Webber BJ, Fletcher K, Chen TJ, Garber MD, Smith A, Wilt G, Conn M, Whitfield GP. Association between passively collected walking and bicycling data and purposefully collected active commuting survey data—United States, 2019. Health & Place. Volume 81, 2023. https://doi.org/10.1016/j.healthplace.2023.103002.
Disclaimer: Use of trade names and commercial sources is for identification only and does not imply endorsement by the U.S. Department of Health and Human Services.