Hospital Visits: Mobility Data Improves Prediction

Predicting patient flow to hospitals remains a significant challenge for healthcare systems, yet current approaches often fail to consider the combined influence of hospital characteristics, socioeconomic factors, and human movement. Researchers Binbin Lin, Lei Zou, and Hao Tian, all from the Department of Geography at Texas A&M University, alongside Heng Cai, Yifan Yang, and Bing Zhou et al. from the University of Tennessee, tackled this problem by developing innovative flow prediction models. Their study, focused on Houston, Texas, integrates hospital attributes , including capacity and reputation , with population socioeconomic status and detailed mobility data to reveal how and why people choose specific healthcare facilities. By demonstrating that a ‘Deep Gravity’ model outperforms others and utilising techniques like SHAP analysis, this research offers crucial insights into the drivers of healthcare visitation, highlighting the importance of convenience for short trips and hospital quality for longer journeys, as well as revealing disparities in access and need across different demographic groups.

Houston hospital flows predicted via mobility data

Scientists often adopt a fragmented approach to healthcare accessibility research, examining individual determinants in isolation, yet recent findings demonstrate that more integrative modeling strategies yield deeper and more actionable insights. This study shows that the Deep Gravity model outperformed alternative flow-prediction approaches, offering meaningful guidance for policymakers seeking to design targeted interventions that promote equitable healthcare access. Despite notable advances in healthcare infrastructure, substantial inequalities persist, driven by socioeconomic conditions, uneven geographic distribution, and imbalanced resource allocation, which disproportionately affect vulnerable populations and exacerbate health disparities (Chen et al., 2023; McMaughan et al., 2020). Healthcare accessibility encompasses both spatial ease and social opportunity to obtain medical services, including geographic proximity, travel burden, and individuals’ socioeconomic capacity to utilize care (Guagliardo, 2004), while healthcare inequality refers to uneven access and utilization across population groups arising from these spatial and social disparities. Early research on hospital accessibility focused primarily on hospital capacity and population within service areas, but later studies incorporated distance decay effects (Luo & Qi, 2009), recognizing that healthcare utilization declines as travel distance increases, as well as hospital quality (Tao et al., 2020), which significantly shapes patient preferences.

Demographic differences have also been introduced to reflect varying healthcare needs across populations; however, many prior studies remain one-dimensional, emphasizing either hospital attributes or community demographics alone. In reality, healthcare access emerges from the interaction of multiple factors, including hospital quality, capacity, proximity, and socioeconomic status (SES), where low-income populations often prioritize proximity due to financial constraints, while higher-income groups may travel farther for higher-quality care. This fragmented focus limits understanding of the complex dynamics governing accessibility and underscores the need for a comprehensive framework that integrates hospital and demographic factors to address healthcare inequalities more effectively. Advances in location-tracking technologies now enable fine-grained analysis of human mobility, with SafeGraph providing large-scale, anonymized origin–destination data that capture population movement patterns across space and time (Advan Research, 2025), including healthcare-related flows that reflect realized healthcare utilization shaped by actual health needs. Complementing mobility data, user-generated content from platforms such as Google Maps offers patient-centered perspectives on hospital quality, where ratings and reviews reflect experiential assessments and review volume serves as a proxy for hospital popularity. Integrating these novel data sources enables a more holistic understanding of how hospital attributes and SES jointly influence healthcare access, addressing key limitations of traditional accessibility research. Mobility pattern modeling has been widely applied to contexts such as tourism, transportation, and urban flows, yet its effectiveness in predicting healthcare visitation remains uncertain, particularly when accounting for the interplay between hospital quality, population demographics, and spatial mobility. Identifying the most suitable modeling framework is therefore critical for advancing healthcare accessibility research and mitigating inequities.

To examine these joint influences, the study analyzed four years (2020–2023) of SafeGraph mobility data in Houston, Texas, incorporating hospital factors such as capacity, occupancy, service quality, and popularity alongside community SES characteristics including income, race, and educational attainment, while also modeling distance decay through drive-time measurements between hospitals and census block groups. By comparing five commonly used flow-prediction models, the study addresses how healthcare system attributes and resident characteristics shape access patterns, grounding the analysis in Andersen’s behavioral model of health services use, which has evolved to incorporate societal, feedback, and contextual influences (Andersen & Aday, 1978; Andersen, 1995; Alkhawaldeh et al., 2023). Prior research aligned with this framework identifies hospital-related factors—such as quality, reputation, patient experience, insurance coverage, size, and proximity—as enabling resources whose relative importance varies by institutional and geographic context. Evidence from Europe highlights the primacy of spatial accessibility and service quality, while studies in the Netherlands and the United Kingdom show strong preferences for shorter travel distances even at the expense of marginal quality reductions (Varkevisser et al., 2012; Salampessy et al., 2022; Smith et al., 2018). In contrast, the United States’ fragmented insurance system and pronounced socioeconomic disparities produce distinct access patterns, with race, insurance type, and income significantly influencing travel distance for care, as higher-income and commercially insured patients are more likely to travel farther than Medicaid beneficiaries (Orringer et al., 2022).

Houston Hospital Visitation Prediction Using Mobility Data

Researchers developed a novel methodology to predict healthcare visitation flows by integrating detailed hospital data with population socioeconomic status and spatial mobility patterns. Experiments employed a rigorous comparative analysis, training and evaluating each of the five models to determine the most accurate predictor of visitation flows. These techniques revealed the combined impacts of hospital attributes, population characteristics, and spatial considerations on visitation patterns, providing nuanced insights into decision-making processes. This methodological approach enables a deeper understanding of healthcare inequalities and informs targeted interventions to promote equitable access.

Deep Gravity Model Predicts Hospital Visitation Accurately

The team measured hospital capacities, ranging from 0 to 1,310 for Staffed All Beds with a mean of 281.43, and 0 to 162 for Staffed ICU Beds, averaging 30.23. Licensed All Beds ranged from 4 to 1,403, with a mean of 368.97. Occupancy rates were also quantified, with the All Bed Occupancy Rate ranging from 0% to 86%, averaging 51.26%, and the ICU Bed Occupancy Rate ranging from 0% to 92%, with a mean of 43.24. These measurements confirm a strong correlation between hospital resource availability and patient visitation. Data shows a total of 1,657,488 individual hospital visitations were recorded across 35 General Acute Care Hospitals in Harris County during the study period.

Analysis of 2,830 Census Block Groups revealed flow volumes ranged from 4 to 2,774.75, with an average of 24.69 visits per CBG-hospital pair. High-volume flows concentrated around eight hospitals, particularly in the western, central, and eastern regions of Houston. The study’s findings highlight the importance of considering socioeconomic factors and spatial dynamics when predicting and managing healthcare demand. This breakthrough delivers a powerful tool for optimising resource allocation and improving healthcare accessibility within urban environments, potentially informing future public health strategies and hospital planning initiatives.

Deep Gravity Model Predicts Houston Hospital Visits

Researchers have demonstrated a novel approach to predicting healthcare visitation flows by integrating hospital characteristics, population socioeconomic status, and human mobility data. This study moves beyond fragmented analyses by simultaneously considering hospital capacities, occupancy rates, reputation, and population demographics alongside spatial mobility patterns to understand visitation determinants. The authors acknowledge limitations stemming from data licensing restrictions preventing public availability of the original datasets, though access can be requested from SafeGraph for qualified researchers. Future research could explore these patterns in other metropolitan areas and investigate the impact of specific healthcare policies on visitation flows, potentially refining strategies for resource allocation and equitable healthcare provision.