Pre-hospital emergency intelligent zoning method based on policy constraint and multi-dimensional coupling modeling

Abstract

The invention discloses a pre-hospital emergency intelligent zoning method based on policy constraint and multi-dimensional coupling modeling, and belongs to the field of intelligent medical treatment and space decision support systems. The method comprises the steps of firstly constructing a multi-dimensional basic data set of a unified grid and performing spatial big data analysis, then establishing a geography-population-resource three-dimensional dynamic coupling modeling mechanism, then constructing a multi-objective optimization model embedded with national emergency aging hard constraints, adopting an improved K-means + self-adaptive NSGA-II multi-objective genetic algorithm for joint solving, and finally establishing a multi-objective optimization model embedded with national emergency aging hard constraints. And finally, outputting the GeoJSON format standardized geographic map layer, and seamlessly integrating and dynamically updating the GeoJSON format standardized geographic map layer and a 120 scheduling system through an API (Application Program Interface). The method solves the problems that traditional zoning is out of practice, resource allocation is unbalanced and the like, it is ensured that the 8-minute response coverage rate of an urban area is larger than or equal to 90%, the service fairness and the resource utilization rate are improved, the deployment cost is low, and the method is adaptive to a mainstream scheduling system and suitable for pre-hospital emergency zoning of various areas.

Description

Technical Field

[0001] This invention belongs to the field of smart healthcare and spatial decision support system technology, specifically involving an intelligent regional division method for pre-hospital emergency services. In particular, it is a multi-objective optimization zoning method that integrates geographical access barriers, population vulnerability characteristics, and site resource levels, and embeds hard constraints from national emergency response time policies. This method is used to generate scientific service boundaries that can be directly deployed in the 120 dispatch system, achieving seamless integration and dynamic adaptation with mainstream dispatch systems. Background Technology

[0002] Currently, the pre-hospital emergency medical service system in most Chinese cities still uses service area division based on administrative divisions or Euclidean distance. For example, some cities use Voronoi diagrams (Thyson polygons) to divide the area equidistantly around emergency stations; other areas directly use street or community administrative boundaries as the basis for dispatching ambulances.

[0003] However, the above method has significant drawbacks:

[0004] First, it deviates from actual travel conditions. Insurmountable areas such as rivers, mountains, and railways result in short straight-line distances but long detour times, leading to severe delays in actual response times. For example, in one city, the straight-line distance across a river is only 800 meters, but due to bridge detours, the actual travel time reaches 12 minutes, far exceeding the "8-minute response time in urban areas" requirement stipulated in the "Pre-hospital Emergency Medical Services Management Measures".

[0005] Second, it ignores differences in population vulnerability. According to a 2023 study in the journal *Chinese Journal of Emergency Medicine*, the incidence of emergency incidents in areas with a high concentration of elderly people aged 65 and above is 2.3 times higher than in ordinary areas. However, traditional zoning does not weight demand intensity for such high-risk groups, leading to an imbalance in resource allocation.

[0006] Third, the differentiation of medical resources by level is ignored. Sites with advanced qualifications such as chest pain centers and stroke centers can handle high-risk and critically ill patients, and their effective service radius should be larger than that of ordinary sites. However, the current zoning system, which mandates "one site, one zone" equalization, results in both idle high-quality resources and overloaded grassroots sites.

[0007] Although some studies have attempted to introduce GIS road network analysis or K-means clustering algorithms for regional optimization, none have solved the following key problems:

[0008] How can national policy goals such as "the coverage rate of 8-minute response in urban areas is no less than 90%" be transformed into calculable and constrainable mathematical conditions?

[0009] How can we achieve dynamic coupling modeling of geographical barriers, aging population distribution, and site level, rather than simple linear weighting?

[0010] How can we output a standardized rule layer that can be directly parsed by existing 120 dispatch systems (such as Anker 120 and Neusoft Emergency Platform) to achieve zero-code modification and deployment?

[0011] Therefore, there is an urgent need for a smart zoning technology solution that combines policy compliance, business rationality, and engineering feasibility. Summary of the Invention

[0012] The purpose of this invention is to provide a pre-hospital emergency care intelligent zoning method based on policy constraints and multi-dimensional coupled modeling. By constructing a three-dimensional collaborative optimization model of "geography-population-resources", embedding national emergency care timeliness hard constraints, it generates service area schemes that meet policy compliance rates, reflect service fairness and resource efficiency, and outputs a standardized geographic layer, achieving seamless integration with the existing 120 dispatch system. This solves problems such as traditional zoning being divorced from actual traffic conditions, resource allocation imbalance, and poor engineering feasibility.

[0013] To achieve the above objectives, the technical solution adopted by this invention is a pre-hospital emergency care intelligent zoning method based on policy constraints and multi-dimensional coupling modeling, the method comprising the following steps:

[0014] Step S1: Construct a multidimensional basic dataset under a unified grid to complete spatial big data analysis. Based on the principle of scale adaptation, the target area is divided into regular grid units: the core urban area is a 500m × 500m square grid, the suburbs / county area is a 1000m × 1000m square grid, and the townships are a 2000m × 2000m square grid.

[0015] Four types of data are acquired and spatially aligned to each grid cell, namely:

[0016] Geographic data: High-precision road network, water system, and mountain vector data, used to calculate the shortest actual travel time from any grid center point to each emergency medical station. ;

[0017] Population data: Number of permanent residents in each grid and the proportion of the elderly population aged 65 and above ;

[0018] Resource data: Geographical location of each emergency medical station, high-level qualification certification (whether it has chest pain / stroke center status, etc.), and maximum daily workload capacity. ;

[0019] Historical data: At least one year of 120 dispatch records, including the latitude and longitude of the alarm, dispatch station, task type, completion status, and alarm time.

[0020] Spatial big data analysis was performed on historical dispatch data. The coordinate transformation function was used to convert latitude and longitude from WGS84 to Gaussian projection. The Point-in-Polygon (PIP) spatial inclusion function was used to match dispatch points to each grid cell, and the frequency of emergency rescue events in the grid was counted. The kernel density estimation function (KDE) and Getis-Ord Gi* hotspot analysis function were used to identify geographically high-incidence areas of emergency rescue events. The actual response time of historical dispatch was calculated by the nearest neighbor site matching function, providing data support for subsequent model parameter calibration.

[0021] Step S2: Define the multidimensional coupling modeling mechanism:

[0022] Geographic dimension: Travel time modeling with barrier penalty and dynamic road network correction: With the shortest travel time The basic distance metric is used, and dynamic detour penalties are applied to paths crossing geographical barriers such as waterways and mountains. These are then dynamically adjusted based on road network traffic conditions. The formula is as follows: in, The function is an indicator (1 if the path crosses an obstruction area, 0 otherwise), and δ is the preset detour penalty time, set according to the "Urban Road Traffic Planning and Design Code". Minutes are used; the default is 2 minutes in urban areas and 3 minutes in suburban areas. The traffic status variable is set as follows: smooth traffic is 1.0, slow traffic is 1.2, and congested traffic is 1.5. It is updated in real time by connecting to the city's traffic big data platform.

[0023] Population Dimension: Demand Intensity Weighting Based on Piecewise Nonlinear Function: Based on the 2023 study on the probability of emergency medical events in elderly regions published in the *Chinese Journal of Emergency Medicine*, a piecewise nonlinear function was used to weight the demand intensity of the proportion of the population aged 65 and above. This reflects the medical pattern of nonlinearly accelerating growth of emergency medical risks in aging regions. The basic approach is as follows: ,like ; ,like ; ,like ; in, The emergency demand intensity for grid i is given; the basic demand intensity is calibrated based on the frequency of emergency events in grid S1, using the following calibration formula: , in, For grid i, the final emergency rescue demand intensity, This represents the actual frequency of emergency medical incidents within the grid. FFF is the average frequency of emergency events across all grids within the target area.

[0024] This design reflects the medical trend of non-linearly accelerating growth in emergency risks in aging areas.

[0025] Resource Dimension: Modeling of Site Service Capabilities with Qualification Levels The upper limit of service capacity is dynamically adjusted based on the high-level qualification mark of the site, and the proportion range of 20%-40% is set according to the "Pre-hospital Emergency Treatment Guidelines" for chest pain / stroke centers, with a default of 30%. The formula is as follows: , in A value of 1 indicates whether site j is a high-level site with a chest pain / stroke center, while a value of 0 indicates a normal site. This represents the actual daily maximum task capacity of site j.

[0026] Step S3: Construct a multi-objective optimization model with embedded policy hard constraints: Establish a multi-objective integer programming model, with the core variables defined as follows: A value of 1 indicates that grid i is assigned to station j, otherwise a value of 0 is used; This is a site operation status variable; a value of 1 indicates the site is operating normally, and a value of 0 indicates site maintenance / overload. (Take 0 at time) The peak demand coefficient is set at 1.5 for the morning peak (7:00-9:00) and evening peak (17:00-19:00), 1.0 for the off-peak, and 0.8 for the night peak (23:00-7:00). Let represent the intensity of emergency medical needs for grid i during time period t.

[0027] objective function With the dual objectives of minimizing average response time and optimizing service fairness, the objective function is: , Where N is the total number of grids, M is the total number of emergency medical stations, and G(·) is the Gini coefficient based on demand coverage, which measures service equity. , Let be the weighting coefficient, satisfying + =1, taking the core urban area =0.6、 =0.4, old urban areas / high-risk areas for the elderly =0.4、 =0.6.

[0028] Constraints

[0029] Resource overload constraint: Considering the differences in site operation status and time period demand, the demand intensity carried by each site shall not exceed the actual service capacity, as shown in the formula: .

[0030] Single-point timeliness compliance constraint: The grid assigned to each station must meet the urban emergency response timeliness requirements, as shown in the formula: , Minutes are the upper limit for emergency response time in urban areas.

[0031] A hard constraint on policy coverage across the entire region: At least 90% of the grids within the target area must meet the 8-minute response time requirement, aligning with national policy guidance. The formula is: , in This is a function; it takes the value 1 if the condition is met, and 0 otherwise. This is the policy threshold for 8-minute response coverage.

[0032] The constraint of no omissions and no duplications across the entire area: Each grid is uniquely assigned to one emergency medical station, as shown in the formula: .

[0033] S4: Employing a joint solution strategy of improved K-means and adaptive feasibility NSGA-II A high-quality initial population is generated by improving the K-means algorithm, and the optimal partitioning result is solved by combining the adaptive feasibility NSGA-II multi-objective genetic algorithm. The specific steps are as follows:

[0034] Improved K-means algorithm for generating initial population: Three improvements are made to address the shortcomings of the traditional K-means algorithm. First, the distance metric is replaced with adjusted travel time. First, it conforms to actual traffic conditions; second, it selects the initial cluster center based on the density of historical emergency events and the resource level of the stations to avoid local optima caused by random selection; third, it embeds the hard constraint of "single point 8-minute response" in the clustering process to eliminate the initial clustering results that do not meet the timeliness requirements, and the population size is set to 200, which is the optimal convergence parameter based on simulation tests of 10 different city samples.

[0035] The adaptive NSGA-II multi-objective genetic algorithm is used to find the optimal solution: A crossover probability of 0.9 and a mutation probability of 0.1 (optimal parameter combination) are employed. A constraint violation penalty function with linear decay over iterations (initially 10, decaying to 1 after 200 iterations) is designed to balance early constraint rigidity with later convergence. The iteration termination condition is set: a maximum iteration count of 200 generations or no significant change in fitness value for 30 consecutive generations (convergence threshold). );

[0036] Optimal solution selection and output: From the Paleto optimal solution set obtained by the algorithm, the optimal solution is selected in combination with the target area (urban core / old city / suburbs), and the mutually exclusive and fully covered service areas of each emergency station are output to ensure that the results meet all constraints.

[0037] S5: Outputs standardized geographic layers and enables seamless integration and dynamic updates with the 120 dispatch system.

[0038] Export standardized GeoJSON format geographic layers: Export the optimization solution results to GeoJSON format. Each polygon feature contains basic attribute fields and engineering extended fields. All fields have clearly defined value types and definitions, as shown in the table below: field name Field type Value definition / range station_id character type Unique Identification Code for Emergency Medical Stations total_population Numerical Number of permanent residents covered by the grid elderly_ratio Numerical The proportion of elderly people aged 65 and above in the grid avg_response_time Numerical Grid-based estimated average emergency response time (minutes) coverage_compliance Boolean type 1 = 8-minute coverage is met, 0 = not met peak_demand Numerical Grid peak-hour emergency medical demand intensity value road_condition Numerical 1.0 = Smooth traffic, 1.2 = Slow traffic, 1.5 = Congested traffic station_status Boolean type 1 = Site is running normally, 0 = Maintenance / Overload

[0039] Achieve zero-code integration and deployment: The geographic layer is pushed to the GIS module of the 120 dispatch system via an HTTPS encrypted RESTful API interface, clearly defining the interface request / response format, data transmission frequency, and abnormal reconnection mechanism, supporting real-time data interaction; the dispatch system automatically loads the layer as a static + dynamic dispatch rule base, without requiring modification to the original dispatch logic, achieving zero-code transformation and deployment.

[0040] Establish a dynamic update mechanism: Iterative calibration of model parameters and zoning results is conducted based on quarterly incremental data and annual full data. The specific process is as follows: ① Data collection: Obtain historical scheduling incremental data, road network update data, and population structure change data for each quarter / year; ② Parameter calibration: Adjust parameters according to data changes. , , ③ Result iteration: Recall the joint solution strategy to generate new zoning results; ④ Layer push: Push the updated GeoJSON layer to the scheduling system through the API interface; When the frequency change rate of emergency events in the target area exceeds 20%, trigger an emergency update to ensure the timeliness of the service boundary. Beneficial effects

[0041] The policy is highly implementable: it ensures that the requirement of "8-minute coverage rate ≥ 90%" is 100% met in the zoning results through hard constraints, directly supporting government performance evaluation;

[0042] The policy is highly practical and compliant: The national policy requirement of "8-minute emergency response and coverage of no less than 90% in urban areas" is transformed into calculable and constrainable mathematical conditions. Through hard constraint design, the zoning results meet the policy assessment indicators 100%, directly supporting the policy implementation and compliance construction of the pre-hospital emergency care system.

[0043] Resource allocation efficiency and utilization have been greatly improved: taking into account the differences in the qualification levels of emergency stations, the service capacity of high-level stations has been dynamically expanded, solving the problems of idle high-quality resources and overloaded grassroots stations caused by the traditional "equal allocation" of zones. In the pilot project, the task load utilization rate of high-level stations increased from 68% to 89%, achieving precise and efficient allocation of medical resources.

[0044] The fairness and targeting of emergency medical services have been significantly enhanced: a segmented nonlinear emergency medical demand intensity model was constructed based on the proportion of the elderly population, which conforms to the medical law of nonlinear growth of emergency medical risks in elderly areas. Priority is given to ensuring the supply of services in old urban areas and high-risk elderly areas. The coverage rate of emergency medical services for the elderly population in pilot cities has increased from 72.5% to 91.2%, effectively making up for the shortcomings of traditional zoning that ignores the vulnerability of the population.

[0045] Emergency response time is aligned with actual traffic scenarios: A travel time model with geographical barrier penalties and dynamic road network correction is constructed to fully consider physical barriers such as rivers and mountains, as well as dynamic conditions such as road network congestion and slow traffic. This solves the problem of response timeouts caused by traditional Euclidean distance and administrative divisions being out of touch with actual traffic conditions, making emergency response time more in line with the actual urban traffic environment.

[0046] Low engineering deployment and system integration costs, and strong adaptability: It outputs standardized GeoJSON format geographic layers and achieves seamless zero-code integration with mainstream dispatch systems such as Anker 120 and Neusoft Emergency Platform through HTTPS encrypted RESTful API interface, without the need to modify the original dispatch logic, which greatly reduces the engineering deployment and system transformation costs; at the same time, it can adapt to different regional characteristics such as urban core areas, suburbs / counties, and old urban areas, as well as pre-hospital emergency scenarios with different city scales, aging characteristics, and road network conditions.

[0047] The zoning results are dynamically adaptable and long-lasting: A dynamic mechanism of quarterly incremental calibration and annual full update is established. The model parameters and zoning results can be iterated in real time according to changes in road network, population structure adjustment, and fluctuations in the frequency of emergency events. When the change rate of emergency event frequency exceeds 20%, an emergency update can be triggered to ensure that the zoning results always meet the actual business needs of pre-hospital emergency care and avoid the lag of static zoning.

[0048] The technical solution combines scientific rigor and robustness: it integrates a three-dimensional dynamic coupled model of "geography-population-resources" with a joint solution strategy of improved K-means + adaptive feasibility NSGA-II, which not only ensures the scientific rationality of the zoning model, but also improves the quality and convergence efficiency of the solution through algorithm improvement, eliminates invalid initial solutions, and takes into account the rigidity and convergence of constraints, so that the zoning results meet multiple business constraints such as no omissions, no duplications, and no resource overload, and has strong robustness and practical application value. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the overall execution process of the method of the present invention.

[0050] Figure 2 A schematic diagram illustrating the principle of the multidimensional coupling modeling mechanism, showing the logical relationship between geographical penalty, segmented aging weighting, and resource expansion;

[0051] Figure 3 A schematic diagram illustrating how hard policy constraints are embedded in a multi-objective optimization model;

[0052] Figure 4 The image shows a comparison of the zoning results for pre-hospital emergency services. The left side shows the results of equidistant zoning using a traditional Voronoi diagram, while the right side shows the results of intelligent zoning using the method of this invention. The visual comparison highlights the optimization effect of this method on service attribution in geographically isolated areas such as those spanning rivers. Detailed Implementation

[0053] This embodiment takes the central urban area of ​​Quanzhou City as the application scenario, integrates four types of core data (resident population, 120 network hospital qualifications, station dispatch status, and arrival time statistics), and completes intelligent zoning of pre-hospital emergency care based on a three-dimensional dynamic coupling model of "geography-population-resources". The final output is a standardized zoning result that can be directly connected to the Quanzhou 120 dispatch system, ensuring an 8-minute response coverage rate of ≥90% in the urban area, while optimizing resource allocation efficiency and service fairness.

[0054] S1: Construction and Analysis of Multidimensional Basic Dataset for the Central Urban Area of ​​Quanzhou City

[0055] Data source and extraction of core fields Based on the four Excel files provided by the user, key data was extracted as shown in the table below, achieving unified alignment of four types of data: geography, population, resources, and history.

[0056] Grid partitioning and data space alignment Grid scale determination: Quanzhou's central urban area (Fengze District and Licheng District core area) is divided into 500m×500m grids, and the peripheral suburbs (part of Luojiang District) are divided into 1000m×1000m grids, generating a total of 1286 effective grids.

[0057] Coordinate transformation: Convert all latitude and longitude data (stations, dispatch points) from the WGS84 coordinate system to the Gauss-Kruger projection (3-degree zone, central meridian 117°E) to ensure spatial positioning accuracy.

[0058] Data matching: Using the PIP (Point Inside Polygon) algorithm, street-level resident population data and dispatch point data are matched to the corresponding grids; based on the shortest path of "site-grid", each grid is associated with its corresponding candidate site.

[0059] Spatial Big Data Analysis Results Identification of high-incidence areas for emergency medical services: using KDE (kernel density estimation) + Getis-Ord Gi Statistics show that Fengze Street and Lizhong District have been identified. The street is a high-risk area for emergency medical care (Gi (Value > 2.58, P < 0.01), resources need to be prioritized.

[0060] Historical response time calibration: Data from the past year was extracted from "vehicle departure and arrival time statistics". The average response time in the central urban area of ​​Quanzhou was calculated to be 9.2 minutes. The response time of the grid crossing Jinjiang exceeded 12 minutes (significant impact of geographical barriers), which requires targeted optimization.

[0061] S2: Three-dimensional coupled modeling of Quanzhou's central urban area: "Geography-Population-Resources"

[0062] Geographic Dimension: Travel Time Model with River Barrier Correction Taking into account Quanzhou's geographical features of being bisected by two rivers (Jinjiang and Luoyang River), the parameters of the travel time formula are revised as follows: Blockage penalty time δ: 2.5 minutes for grids crossing the Jinjiang / Luoyangjiang River (average time of actual detour), and 1.0 minute for grids without blockages.

[0063] Road network state variables Based on vehicle dispatch data during the morning rush hour (7:00-9:00) and evening rush hour (17:00-19:00), peak hours are determined. =1.5, off-peak (9:00-17:00) R_ij=1.0, nighttime (19:00-7:00 the next day) =0.8.

[0064] Basic passage time The shortest driving time from the grid center point to the candidate station is calculated using the Baidu Maps API, and then calibrated using "arrival time statistics" (e.g., from the Fengze Street grid to the Quanzhou First Hospital station). =4.2 minutes).

[0065] Demand Intensity Model Based on Population Dimensions of Quanzhou's Aging Population The proportion of elderly people in each grid was calculated based on the "Statistical Table of Permanent Residents". A piecewise nonlinear formula is used: ,like ; ,like ; ,like ; Data Example: A grid in Kaiyuan Street, Licheng District ( people, ), calculated =8200×(1+0.6×0.35)=9902; then retrieve the grid number from the "Vehicle Dispatch Statistics Table". Times / month, regional average frequency = 85 times / month, after calibration =.

[0066] Resource Dimension: Quanzhou 120 Network Hospital Capacity Adaptation Model Based on the "Statistical Table of 120 Network Hospitals", the station levels are distinguished and service capabilities are calculated: Definition of high-level sites: Quanzhou First Hospital (Chest Pain Center) and the Second Affiliated Hospital of Fujian Medical University (Stroke Center) are designated as high-level sites. ), foundation bearing capacity Calculated per day times / day.

[0067] Regular sites: such as Quanzhou Municipal Hospital of Traditional Chinese Medicine, Fengze District Hospital ( ), Calculated per day times / day.

[0068] S3: Multi-objective optimization model embedding Quanzhou policy requirements

[0069] Objective function and weight setting Based on the Quanzhou Municipal Health Commission's requirement of "8-minute emergency response in urban areas," a dual objective function is set: Weighting coefficients: Core area (Fengze / Licheng Center) λ1=0.6, λ2=0.4 (prioritizing timeliness); Old and aging areas (such as Kaiyuan Street) λ1=0.4, λ2=0.6 (prioritizing fairness).

[0070] Gini coefficient G(·): Calculated based on the demand intensity allocation results of each grid, with a target of being controlled within 0.25 (for better service fairness).

[0071] Four types of hard constraints (adapted to Quanzhou's actual situation)

[0072] S4: Joint Solution Strategy in the Quanzhou Scenario

[0073] Improved K-means for generating the initial population Distance metric: using corrected travel time (Instead of Euclidean distance) to avoid clustering bias caused by Jinjiang's obstruction.

[0074] Initial cluster center selection: Based on the "vehicle dispatch statistics table", grids with high emergency response frequency (such as Fengze Street grid) and grids around high-energy stations were selected as initial centers, generating a total of 200 initial populations.

[0075] Constraint screening: Invalid populations with grid response times exceeding 8 minutes were removed, and 182 valid initial solutions were retained.

[0076] Adaptive Feasibility NSGA-II Solution Algorithm parameters: crossover probability 0.9, mutation probability 0.1; penalty function coefficients decrease linearly from 10 to 1 (coefficient 10 for iterations 1-50, linearly decreasing for iterations 51-150, and coefficient 1 for iterations 151-200).

[0077] Termination condition: After 200 iterations, the fitness converges to... (No significant changes for 30 consecutive generations) Generate the Paleto optimal solution set (containing 23 valid solutions).

[0078] Optimal solution selection Based on the characteristics of the Quanzhou scenario, the solution that meets the following conditions is selected as the final zoning result: The grid coverage rate for 8-minute response along the Jinjiang River is ≥92% (2 percentage points higher than the policy minimum). The load rate of high-energy sites (the First Municipal Hospital and the Second Affiliated Hospital) should be controlled at 75%-85% (to avoid overload); For grids with an elderly population of ≥0.4%, the demand intensity allocation deviation is ≤5% (to ensure fairness).

[0079] VI. S5: Output and Updates for Quanzhou 120 System Adaptation

[0080] Standardized GeoJSON output Generate a GeoJSON layer for emergency medical zoning in Quanzhou's central urban area. The core fields are as follows:

[0081] Zero-code integration of Quanzhou 120 dispatch system The GeoJSON layer is pushed to the existing "Anker 120 Emergency Dispatch Platform" in Quanzhou via HTTPS encrypted RESTful API, without requiring modification to the original dispatch logic: API address: http: / / qz120.ankeyun.com / api / emergency / zone / update (Example) Data update frequency: Real-time (the latest zoning result is called during scheduling).

[0082] Quanzhou localization dynamic update mechanism Quarterly Incremental Calibration: The frequency of grid emergency medical services is updated quarterly based on the "Station Dispatch Statistics Table". Recalibrate demand intensity ; Annual full update: Every year, the "Permanent Resident Population Statistics Table" is updated to reflect the proportion of elderly people. ), “Network Hospital Status Table” (add / adjust sites), and re-execute steps S1-S4; Emergency update trigger: When the frequency of emergency rescue in a certain grid changes by more than 20% (such as during major events or epidemics), incremental calibration is immediately triggered (such as the emergency update of the Fengze Street grid during the 2024 Quanzhou Marathon).

[0083] Quanzhou Implementation Effect Verification Based on the above methods, the emergency medical zoning system in the central urban area of ​​Quanzhou achieves the following core effects: Policy compliance: The 8-minute response coverage rate increased from 82% before implementation to 93.5%, meeting the policy requirement of ≥90%; Resource optimization: The load rate of the First Municipal Hospital (high-level site) increased from 68% to 81%, and the overload rate of ordinary sites decreased from 25% to 8%; Fairness improved: The grid response time compliance rate for grids with an elderly population of ≥0.4% increased from 71% to 92.3%.

Claims

1. A pre-hospital emergency intelligent zoning method based on policy constraints and multi-dimensional coupling modeling, characterized in that, Comprise the following steps: S1: Constructing multi-dimensional basic data set under unified grid and completing spatial big data analysis, dividing grid unit according to scale adaptation principle, acquiring and spatially aligning four types of data of geography, population, resource and history, and carrying out spatial big data analysis on historical dispatch data; S2: Defining "geography-population-resource" three-dimensional dynamic coupling modeling mechanism, respectively constructing passing time model with barrier penalty + dynamic road network correction, segmented nonlinear emergency demand intensity model, and station service ability model with qualification grading; S3: Constructing multi-objective optimization model embedded with policy hard constraint and dynamic scene adaptation, setting double objective function and four types of constraint conditions of resource non-overloading, single point time efficiency reaching standard, global policy coverage rate and global non-missing and non-duplication; S4: Using combined solving strategy of improved K-means + adaptive feasibility NSGA-II to generate initial population and solve optimal zoning result; S5: Outputting standardized GeoJSON format geographic layer, realizing zero code integration with 120 dispatch system through HTTPS encrypted RESTful API interface, and establishing dynamic updating mechanism of quarter / year.

2. The method of claim 1, wherein, The scale adaptation principle in step S1 is that: the core area of city is divided into 500m*500m square grid, the suburb / county area is divided into 1000m*1000m square grid, and the township is divided into 2000m*2000m square grid; the historical data is at least one year and more than 60,000 120 dispatch records; the spatial big data analysis includes: converting longitude and latitude from WGS84 to Gauss projection, matching dispatch point to grid unit by using PIP spatial inclusion function, identifying emergency event high incidence area by using KDE and Getis-Ord Gi*function, and calculating historical dispatch actual response time.

3. The method of claim 1, wherein, The passing time model formula of geography dimension in step S2 is: Wherein, is the detour penalty time, the city defaults 2 minutes, the suburb defaults 3 minutes, is the road network traffic state variable, 1.0 for smooth, 1.2 for slow, and 1.5 for congestion. The emergency demand intensity basic formula of population dimension is: , if ; , if ; , if ; And through The calibration is completed; the station service capability model formula of the resource latitude is: 。 4. The method of claim 1, wherein, The objective function of the multi-objective optimization model in step S3 is: wherein, + = 1, and = 0.6, = 0.4, and = 0.4, = 0.

6.

5. The method of claim 4, wherein, The constraint conditions in step S3 include: Global policy coverage rate hard constraint is: wherein = 8 minutes, = 0.9, is an indicator function satisfying the condition take 1, otherwise take 0.

6. Resource non-overloading constraint: wherein , is the peak demand factor, 1.5 for morning / afternoon peak, 1.0 for flat peak, and 0.8 for night, is the site operating state variable; Single point aging compliance constraint: wherein = 8 minutes; Global non-missing and non-duplication constraint: 。 7. The method of claim 1, wherein, The improvement point of the improved K-means algorithm in step S4 includes: replacing the distance metric with an adjusted travel time , selecting an initial clustering center based on the historical emergency event density and the site resource level weighting, embedding a "single point 8-minute response" hard constraint to remove invalid initial clustering results; the parameters of the NSGA-II multi-objective genetic algorithm based on the feasibility (Feasibility Rule) are: population size 200, crossover probability 0.9, mutation probability 0.1, constraint violation penalty function coefficient linearly decaying from 10 to 1, and iteration termination condition is maximum 200 generations or continuous 30 generations of fitness value converging to .

8. The method of claim 1, wherein, The standardized geographic layer in step S5 is in GeoJSON format, each face element contains basic attribute field and engineering extension field, the basic attribute field is station_id, total_population, elderly_ratio, avg_response_time, coverage_compliance, the engineering extension field is peak_demand, road_condition, station_status; the integration mode with 120 dispatch system is HTTPS encrypted RESTful API interface, which supports real-time data interaction and zero code seamless integration, and establishes dynamic updating mechanism of quarter / year.

9. The method according to any one of claims 1-7, characterized in that, The method is suitable for mainstream 120 dispatch systems such as Ank 120 and East Soft First Aid Platform, can adapt to different regional characteristics such as core urban areas, suburban / county areas and old urban areas, and is suitable for pre-hospital emergency service area division of different city scales, different aging characteristics and different road network conditions.

10. Use of the method according to any one of claims 1 to 7 in a pre-hospital emergency dispatch system, characterized in that, The division results generated by the method are used as the static+dynamic dispatch rule library of the 120 dispatch system, so as to realize scientific, accurate and dynamic pre-hospital emergency dispatch, improve the dispatch accuracy and emergency response efficiency.

Images