TL;DR
Researchers used geospatial modeling to optimize the allocation of colorectal cancer (CRC) screening hospitals in Shanghai’s Pudong New District. By applying the Huff Model and Location-Allocation analysis, they identified significant accessibility gaps and proposed a data-driven hospital configuration that balances patient load and improves service coverage without requiring additional medical facilities.
What Problem Does This Paper Address?
Colorectal cancer is a leading cause of death in Shanghai, and while the city has established a community-based screening program (C-CRCSP), it faces a major bottleneck: follow-up compliance.
After an initial positive fecal occult blood test (FOBT), only about 28.2% of residents actually go for the necessary follow-up colonoscopy. The authors argue that this low rate is partially driven by poor spatial accessibility—hospitals aren't always where the patients are. Furthermore, the current selection of screening hospitals was made "subjectively" by local government without rigorous geographic analysis, leading to "ineffective screening" (wasteful resource allocation in over-served areas) and massive service gaps in rural regions.
Methods
The study leverages three primary geospatial modeling techniques to evaluate and then fix the screening network:
- Huff Model: Used to delineate Hospital Service Areas (HSAs). This model calculates the probability of a resident choosing a specific hospital based on the facility's capacity (number of colonoscopes) and the travel time (distance decay).
- Two-Step Floating Catchment Area (2SFCA): A standard method in health geography used here to measure spatial accessibility. It accounts for both the supply (hospitals) and the demand (target population) within a specific travel threshold (set at 22 minutes/1,300 seconds).
- Location-Allocation (L-A) Model: This is the optimization engine. The researchers used the "Maximize Coverage" and "Minimize Impedance" solvers in ArcGIS to determine which 13 hospitals (out of 18 available) would most efficiently serve the district's 74,608 high-risk residents.
Datasets
- C-CRCSP Screening Data: Provided by the Pudong CDC, covering 421,384 participants and identifying 77,473 residents who tested positive via FOBT.
- Hospital Infrastructure Data: Details on 18 hospitals equipped for colonoscopies, including their specific "service capacity" (calculated based on weekly consultation limits over a two-year period).
- Geographic Data: Administrative boundaries of 48 communities in Pudong New District (PND) and road network data sourced from OpenStreetMap.
- Coordinate Data: Hospital locations were geocoded using the Baidu Maps coordinate system.
Key Insights and Ideas
- The "Ineffective Screening" Paradox: The study found that urban centers like Lujiazui have the highest accessibility but also the highest rates of "ineffective screening" (where residents are rescreened too frequently despite previous negative results). This suggests that in high-access areas, the problem isn't a lack of resources, but a lack of targeted patient education.
- Subjectivity vs. Science: The current 13-hospital setup was chosen by administrative preference. The models showed that by simply swapping one hospital (Punan Hospital) for another (Hudong Hospital), the system could significantly better serve the population's needs.
- The Southeast Gap: There is a stark "accessibility desert" in the Southeast (Zhangjiang, Jichang, and Laogang). The models reveal that even with optimization, existing hospitals are too far away, suggesting that reallocating resources isn't enough—new service sites are likely required in these specific rural zones.
Key Results
- Low Compliance: Only 28.2% of residents with positive initial results followed through with a colonoscopy.
- Optimized Load Balancing: The proposed optimization reduced the average population served per colonoscope from 2360.10 to 2281.94.
- Equity Improvement: The standard deviation of the hospital load decreased from 1212.46 to 1160.45, indicating a more equitable distribution of patients across the medical system.
- Capacity Management: Before optimization, several hospitals were operating at over 100% of their theoretical capacity (leading to long wait times and drop-outs). The optimized model brought all hospitals back under 100% utilization.
Limitations and Future Work
The authors acknowledge a few "real-world" simplifications:
- Uniform Travel Speed: The model assumes a constant driving speed of 60 km/h. In a mega-city like Shanghai, traffic congestion varies wildly by time of day and road type, which could change accessibility scores.
- Centroid Simplification: Demand was calculated from the geometric center of each community rather than specific residential addresses.
- The "Hard" Gaps: While optimization helps, the authors note that for the extreme Southeast, no amount of "shuffling" existing hospitals will solve the problem. Future work should look at the feasibility of establishing entirely new colonoscopy centers in those underserved rural communities.
Summary updated