Supply–demand governance of hierarchical healthcare systems: mobile big data unveils non-random patient flow patterns and the bypass premium in cities

Qi Guo, Hengna Ren, Xinmiao ShaoFrontiers in Public Health 2026 · DOI

TL;DR

By analyzing anonymized mobile signaling data from millions of users in Shanghai, researchers have deconstructed the "bypass premium"—the extra distance patients travel to skip local clinics for top-tier hospitals. They discovered that nearly 74% of this extra travel is "passive structural friction" caused by administrative boundaries and poor connectivity, rather than a purely active choice for better quality.

What Problem Does This Paper Address?

Urban planners often look at healthcare through a static lens: "How many beds are in this district?" However, as populations sprawl and transportation networks evolve, the way people actually move to seek care—patient flow—tells a different story.

There is a systemic mismatch between where hospitals are and where patients live. This leads to "bypassing," where patients skip primary and secondary facilities to crowd into elite tertiary hospitals. Traditional models (like gravity models) often fail to capture these micro-level behavioral nuances. The authors sought to quantify the hidden geographic and social costs of this behavior and determine how much of our travel is a choice for quality versus a struggle against a poorly designed system.

Methods

The researchers constructed a city-scale, weighted, directed patient flow network for Shanghai using several novel layers:

  • Thiessen Polygon Partitioning: Instead of using arbitrary grids, they created "demand units" based on the service area of each hospital, allowing for a precise "one-hospital-one-domain" analysis.
  • Null Model Testing: They introduced a stochastic (random) benchmark to prove that patient flows aren't just following the path of least resistance; they are highly polarized toward top-tier institutions in a way that deviates significantly from random chance.
  • Channel Decomposition: They categorized flows into three "channels": Local (within the immediate area), Intra-district, and Inter-district.
  • Bypass Premium Index: A new metric developed to quantify the additional spatial burden patients take on to reach a specific grade of hospital compared to the nearest available one of that same grade.

Datasets

  • Mobile Signaling Data: Anonymized records from a major carrier in Shanghai for March 2019, capturing a "steady state" of urban mobility with 1-hour temporal resolution.
  • Hospital Directory: Official data on 352 public general hospitals (242 primary, 62 secondary, 48 tertiary) from the Shanghai Municipal Health Commission.
  • Spatial Vector Data: Geographic boundaries and road network data from the National Catalogue Service for Geographic Information.

Key Insights and Ideas

  • The "Value Filter" of Boundaries: Administrative boundaries don't just act as barriers; they act as filters. While local care stays local, inter-district travel is almost exclusively "funneled" into backbone pathways leading to tertiary hospitals. If you're crossing a district line, it’s almost certainly for a top-tier specialist.
  • Functional Neutrality of the Middle: Secondary hospitals are in a "sandwich" dilemma. They fail to act as hubs or diversions. They don't have the proximity of primary clinics or the prestige of tertiary centers, leaving them in a state of "functional neutrality" where their market share remains stagnant regardless of distance.
  • Friction vs. Selection: This is the "aha" moment. The study deconstructs the cost of seeking high-quality care. Most of the "premium" we pay in travel time isn't because we want to go further for a specific doctor; it's because the system forces us to overcome "structural friction" (like poor transit or administrative silos) to get there.

Key Results

  • Concentration: Tertiary hospitals handle 35.7% of total patient flow, and the top 10 hospitals alone carry 12.43% of the city's total volume.
  • The Bypass Premium: To reach a tertiary hospital, the average patient pays a "bypass premium" of 10.24 km.
  • The Cost Breakdown: Of that 10.24 km, 73.54% (7.53 km) is attributed to passive structural friction (overcoming boundaries and system inefficiencies). Only 26.46% (2.71 km) represents the active selective premium—the distance a patient willingly adds to their trip specifically to choose a higher-quality facility.
  • Inequality: Inter-district flows are highly polarized, with a Theil index of 0.453, indicating extreme concentration in a few "backbone" routes.

Limitations and Future Work

The authors acknowledge that mobile signaling data lacks clinical context. We don't know if a patient is bypassing a local clinic for a common cold or a rare cancer, which significantly changes the "rationality" of the bypass. Furthermore, the study uses "corrected distance" as a proxy for cost, which doesn't account for real-time traffic congestion or the specific nuances of public transit transfers.

Future research aims to use "multi-source data fusion," linking mobility data with Electronic Health Records (EHR) and insurance data. This would allow researchers to see if financial incentives (like insurance reimbursement gaps) can effectively "reduce friction" and steer patients back toward the underutilized secondary hospitals.

Summary updated

Abstract

Background The uneven distribution of healthcare resources and jobs-housing spatial separation are reshaping the spatiotemporal patterns of urban patient flows. This structural mismatch exacerbates inequalities in service utilization and imposes hidden geographic and social costs. However, conventional static statistics and theoretical models often fail to capture authentic micro-level behavioral patterns, rendering them unable to precisely quantify or deconstruct the inequalities and burdens concealed within patient flows. Methods Taking Shanghai as a representative megacity case study, we utilized anonymized mobile signaling data (March 2019) to construct a weighted, directed “demand–supply” patient flows network. We introduced a null model as a random benchmark and employed the channel decomposition method to deconstruct pathway structures. We developed a “bypass premium” index to quantify the specific burden of quality-driven hospital seeking. Results (1) Resource siphoning: Patient flows are highly concentrated toward top-tier hospitals, yet their spatial footprint is widely dispersed across the city, a pattern that deviates significantly from the random benchmark. (2) Boundary filtering: Administrative boundaries act as a “value filter.” Inter-district flows do not diffuse uniformly but are funneled into backbone pathways leading exclusively to tertiary hospitals. (3) Functional neutrality: Secondary hospitals fail to perform their intended hub-and-diversion function within the hierarchical healthcare system, resulting in a state of functional neutrality. (4) Cost deconstruction: The average bypass premium for reaching a tertiary hospital is 10.24 km. Crucially, 73.54% (7.53 km) of this constitutes passive structural friction required to overcome boundary barriers, while only 26.46% (2.71 km) represents the active selective premium paid for quality-driven access. Conclusion This study confirms the non-random polarization of patient flows and the screening mechanism of administrative boundaries in Shanghai. Our findings reveal that the costs of inter-district hospital-seeking stem primarily from passive structural friction rather than active selective premiums, occurring alongside a critical functional deficit in secondary hospitals. Consequently, policy interventions must prioritize strategies of “reducing friction” and “strengthening the middle.” Specifically, optimizing transportation networks, insurance integration, and medical consortiums is essential to dismantle barriers and revitalize the hub capacity of the intermediate tier.
BibTeX
@inproceedings{guo2026supplydemand,
  author = {Qi Guo and Hengna Ren and Xinmiao Shao},
  title = {{Supply–demand governance of hierarchical healthcare systems: mobile big data unveils non-random patient flow patterns and the bypass premium in cities}},
  booktitle = {Frontiers in Public Health},
  year = {2026},
  doi = {10.3389/fpubh.2026.1797443},
}
Figure