Multi-source spatiotemporal data-driven blood collection site efficiency diagnosis and network optimization
10.13303/j.cjbt.issn.1004-549x.2026.06.011
- VernacularTitle:基于多源时空数据的采血点效能诊断与网络优化研究
- Author:
Guiyun XIE
1
;
Jinyan CHEN
1
;
Jian OUYANG
1
;
Rongrong KE
1
;
Yi YANG
1
;
Zhenzhen CHEN
1
;
Ledong YANG
1
Author Information
1. Guangzhou Blood Center, The Key Medical Disciplines and Specialties Program of Guangzhou, Guangzhou 510095, China
- Publication Type:Journal Article
- Keywords:
blood collection site;
blood supply chain;
spatial analysis;
panel data;
ordinary least squares (OLS) model
- From:
Chinese Journal of Blood Transfusion
2026;39(6):776-783
- CountryChina
- Language:Chinese
-
Abstract:
Objective: To quantify the spatiotemporal factors and built environment effects on daily blood collection volumes at fixed sites using multi-source spatiotemporal data, explore the influence radii of different point of interest(POI) categories, and develop a classification framework for sites, thereby providing data references for efficiency diagnosis and network optimization of blood collection sites. Methods: We collected daily blood donation records from 16 fixed whole blood collection sites in Guangzhou in 2023, constructing a balanced “site-day” panel dataset (16 sites × 365 days = 5 840 observations). Multi-source geospatial data, including Amap POI, OpenStreetMap (OSM) road network, and meteorological data were integrated to establish 300 m, 500 m, 800 m, and 1 000 m buffer zones around each site. Variables extracted included counts of surrounding commercial facilities, distance to the nearest subway station, rainfall, and temperature. A two-way fixed effects panel model was employed with a first-order lag term of the dependent variable to capture the inertial effect of blood donation behavior. Based on the significant regression coefficients, a "Structure-Function Index (SFI)" was constructed for each site to identify its optimal buffer radius. Combining inherent site capacity (fixed effects) with the optimal SFI, a two-dimensional quadrant diagram was created to classify the sites. Results: The optimized model showed an adjusted R
of 0.89 and a Durbin-Watson statistic of 1.90, with the lag term coefficient stable at 0.36 (P<0.05). Temporal effects were significant: campaign days increased blood collection by 1 717.75 mL, holidays by 3 755.53 mL, and overlapping holidays with campaign days by 5 650.49 mL (P<0.05). Rainy days reduced collection by approximately 419 mL (P<0.05), and each 1℃ rise in daily average temperature decreased collection by approximately 31 mL (P<0.05). Distance to the nearest subway station was significantly positive (coefficient 0.95-1.39, P<0.05). POI effects showed scale-dependent differences: within 300 m, leisure and entertainment (+173.65 mL/unit) and shopping (+2.95 mL/unit) were significantly positive; within 500 m, tourist attractions (+72.55 mL/unit) and hotels (+40.09 mL/unit) were significantly positive; within 1 000 m, these effects diminished. Based on the two-dimensional classification of inherent collection capacity and optimal SFI, sites were categorized into four types: core, potential, adjustment, and resource-dependent. Conclusion: This study elucidates the multidimensional determinants of blood collection site efficiency, and provides quantitative evidence for performance diagnosis. It proposes a "concentric-zone planning" approach for site selection assessment alongside a" station-specific strategy" classification framework, offering data-driven references for network planning and resource allocation in blood collection and supply institutions. It should be noted that blood donation site selection also involves complex factors such as budgetary constraints, interdepartmental coordination, and venue conditions. This study provides a spatial econometric perspective for the quantitative optimization of blood collection networks.
