Assessing Patient bypass Behavior Using Taxi Trip Origin-Destination (OD) Data

Gege Yang, Ci Song, Hua Shu, Jia Zhang, Tao Pei, Chenghu ZhouISPRS Int. J. Geo Inf 2016 · DOI

TL;DR

This paper uses taxi GPS data to analyze "hospital bypass behavior" – when patients travel to distant, high-quality hospitals instead of closer ones. By developing a "bypass behavior index" (BBI), the study reveals that factors like hospital reputation and accessibility significantly influence these choices, with Beijing showing more pronounced bypass phenomena and Suzhou exhibiting greater spatial heterogeneity in bypass trip distances.

What Problem Does This Paper Address?

In China, a significant number of patients opt to bypass nearby hospitals and travel to more distant, high-quality medical facilities, even for minor ailments. This "hospital bypass behavior" leads to overcrowding in top hospitals, inefficient resource allocation, and difficulties for patients seeking timely care. While previous studies have explored this phenomenon, they often relied on limited data and lacked spatial analysis, failing to capture the heterogeneity of bypass behavior across urban areas. This paper aims to address this gap by using large-scale taxi GPS data to quantify and analyze the spatial patterns and influencing factors of hospital bypass behavior in Beijing and Suzhou.

Methods

The core of this research lies in leveraging taxi GPS data to reconstruct patient travel patterns. The authors developed a multi-step process:

  1. Data Cleaning and Map Matching: Raw taxi GPS data was cleaned to remove erroneous records, and map matching was performed to accurately track taxi routes.
  2. Origin-Destination (OD) Pair Extraction: Taxi trip data was simplified into pick-up (origin) and drop-off (destination) points. A pick-up event was identified when a taxi's state changed from "without passenger" to "with passenger," and a drop-off event when it changed from "with passenger" to "without passenger."
  3. Hospital Target Point Identification: A buffer area around hospital entrances was defined as the "target point." GPS points falling within this buffer were considered as hospital visits.
  4. Source Point Recognition: For each hospital visit (target point), the corresponding pick-up point of the same taxi trip was identified as the "source point," representing the patient's origin.
  5. Bypass Behavior Index (BBI) Calculation: A novel "bypass behavior index" (BBI) was developed to quantify bypass behavior for each hospital. This index is an average of three standardized factors:
  6. Hospital bypass trip distance ranking: Ranks hospitals based on the average distance traveled by patients bypassing closer options.
  7. Hospital bypass share: Measures the proportion of a hospital's bypass source points relative to all bypass source points in the study area, indicating its attractiveness to bypass patients.
  8. Hospital bypass proportion: Represents the percentage of bypass patients for a single hospital relative to its total source points, reflecting the demand from non-local patients.

Datasets

The study utilized taxi GPS data from two Chinese cities:

  • Beijing: Collected from April 15-19, 2011, comprising records from over 10,000 taxis.
  • Suzhou: Collected from April 21-22, 2014, with data from approximately 4,300 taxis.

In addition to taxi data, hospital location data and road networks were used. The study focused on 20 tertiary hospitals in Beijing and 22 hospitals (10 tertiary, 12 secondary) in Suzhou.

Key Insights and Ideas

  • Taxi GPS as a Powerful Proxy: The research demonstrates the effectiveness of using taxi GPS data as a large-scale, real-time proxy for understanding complex human mobility patterns, specifically for healthcare-seeking behavior.
  • Quantifying "Bypass": The development of the BBI provides a standardized, quantitative measure for comparing bypass behavior across different hospitals and cities, moving beyond qualitative descriptions.
  • Spatial Heterogeneity Matters: The study highlights that bypass behavior is not uniform. Differences in hospital distribution, quality, and accessibility create distinct spatial patterns of bypass in urban environments.
  • Reputation Trumps Proximity (Often): The findings suggest that hospital reputation, specialty departments, and convenient transportation access are strong drivers for patients to bypass closer, potentially lower-tier, facilities.

Key Results

  • Bypass Trip Distances: Mean hospital bypass trip distances ranged from 5.988 km to 9.754 km in Beijing and 4.168 km to 10.283 km in Suzhou. Beijing generally showed longer bypass distances.
  • Top Bypass Hospitals: In Beijing, hospitals H301, ET, XH, BDDS, and ZL exhibited the highest bypass behavior. In Suzhou, SDDE, SZZY, SLBB, ET, and SA showed the highest bypass behavior.
  • Spatial Patterns: Beijing's bypass patterns were more dispersed, while Suzhou showed more significant spatial heterogeneity, with some hospitals in older downtown districts and others in surrounding areas.
  • Factors Influencing Bypass: Hospital reputation, transport accessibility, and spatial distribution were identified as key factors. Hospitals with strong reputations and specialized departments attracted more bypass patients.
  • City-Level Differences: While Beijing exhibited more pronounced bypass phenomena overall, Suzhou showed more significant differences in bypass trip distances between hospitals, suggesting greater spatial variation in its healthcare landscape.

Limitations and Future Work

The authors acknowledge several limitations:

  • Lack of Patient Attributes: The taxi OD data primarily reflects mobility and lacks crucial patient attributes like age, gender, income, and insurance status, which could further explain bypass behavior.
  • Limited Sample Size: While large for taxi data, the number of sampled hospitals (20 in Beijing, 22 in Suzhou) might not capture all meaningful patterns within the entire cities.
  • Exploratory Nature: The findings are considered preliminary and exploratory, suggesting the need for further data collection and sampling to refine the understanding of bypass behavior.

Future work could involve integrating patient demographic data, expanding the study to more cities, and exploring the impact of specific medical insurance policies on bypass behavior.

Summary updated

BibTeX
@article{yang2016assessing,
  author = {Gege Yang and Ci Song and Hua Shu and Jia Zhang and Tao Pei and Chenghu Zhou},
  title = {{Assessing Patient bypass Behavior Using Taxi Trip Origin-Destination (OD) Data}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2016},
  doi = {10.3390/ijgi5090157},
}
Figure