Evaluating Biases in Cell-phone-based Healthcare Travel Data: A case study of Texas, USA

Authors

Sahasra Byreddy Academies of Loudoun, Leesburg, VA
Anastasia Soukhov Department of Geography and Environment, Western University, London, Ontario, Canada
Tao Hu Department of Geography, Oklahoma State University, Stillwater, Oklahoma
Armita Kar Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA

Healthcare travel behavior is an important determinant of health outcomes and is commonly evaluated by

measuring travel time between patients’ residences and healthcare facilities. Recently, anonymized cell-phone

mobility data have emerged as a valuable resource for estimating real-world travel behavior at large scales.

However, these cell-phone-based mobility datasets are not fully representative of the population, introducing

biases in the behavior that is observed¹ . This study characterizes these biases by comparing healthcare-related

origin-destination (OD) flows between cell-phone-based mobility data derived from SafeGraph and

administrative visitation records from the 2019 Texas Department of State Health Services (TDSHS) Public

Use Data Files, which serve as a benchmark independent of smartphone use. OD pairs were defined using

patients’ zip codes as origins and Texas hospitals as destinations. We performed comparative analysis of both

datasets using descriptive statistics, distance decay modeling, and map visualizations. Hospital-level weighted

network distances showed moderate agreement (Pearson r = 0.35) between the two datasets. Spatial analyses

revealed similar geographic patterns of hospital use with northeastern and eastern Texas being identified as

major hospital service areas. Distance decay patterns were similar between the two datasets with estimated

decay coefficients of beta = 0.94 for the cell-phone-based data and beta = 0.96 for the TDSHS data for network

distance. Overall, biases were more evident in the estimates of the travel magnitude than in broad spatial

patterns because the cell-phone-based mobility data consistently overestimated the travel distance seen in the

TDSHS data. These findings support that cell-phone-based mobility data may be more useful for characterizing

the geographic organization of healthcare utilization than precise hospital-level estimates.

Li, Z., Ning, H., Jing, F., & Lessani, M. N. (2024). Understanding the bias of mobile location data

across spatial scales and over time: A comprehensive analysis of SafeGraph data in the United States.