Medical service improvement method and system based on remote medical treatment data analysis

Abstract

The invention discloses a medical service improvement method and system based on remote medical seeking data analysis, belongs to the technical field of computers, and through the technical scheme, accurate mapping between individual demands of patients and medical resources is achieved, and the accuracy of remote medical seeking recommendation is effectively improved. According to the method, the problem of data dimension mismatching in a traditional recommendation system is solved through deep feature fusion, the reasonability of medical resource recommendation is improved through a semantic coherent description information generation mechanism, and a hospital selection scheme better meeting the actual requirements of a patient is provided for the patient.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method and system for improving medical services based on analysis of medical treatment data in other places. Background Art

[0002] With socioeconomic development and increasing population mobility, cross-regional medical treatment is becoming increasingly common. When seeking medical services in non-insured areas, patients often face challenges such as information asymmetry, unfamiliar procedures, and poor resource matching. Efficiently and accurately matching patients seeking medical services in different locations with appropriate medical resources to improve their experience and outcomes has become a pressing challenge in the healthcare sector.

[0003] Existing inter-province medical service systems or platforms typically focus on simplifying processes (such as filing and settlement) or providing basic hospital directory information. They have obvious limitations in addressing patients' individualized and complex medical needs.

[0004] Furthermore, medical information systems in different regions exhibit significant differences in data processing capabilities, data models, and service priorities. Currently, there is a lack of effective mechanisms to intelligently adapt and utilize these heterogeneous computing node resources distributed across medical destinations based on patients' specific circumstances and needs. As a result, out-of-town medical recommendation services often remain limited to information display, struggling to achieve deep mining and precise matching based on multi-dimensional data, and failing to fully meet patients' personalized, high-quality medical needs. Summary of the Invention

[0005] The present application provides a method and system for improving medical services based on remote medical treatment data analysis, which can improve the accuracy of medical resource matching, optimize the efficiency of remote medical treatment decision-making, and improve the personalized service experience of patients. The technical solution is as follows:

[0006] The present application provides a medical service improvement method based on out-of-town medical treatment data analysis, and the technical solution is as follows: in response to the out-of-town medical treatment request of the target object, determine the current physiological state, medical reference information and reason for out-of-town medical treatment of the target object; based on the current physiological state and reason for out-of-town medical treatment of the target object, determine the target node that matches the physiological state and reason for out-of-town medical treatment from multiple candidate nodes associated with the medical destination, different candidate nodes have different computing power and hospital recommendation modes, and the medical destination is obtained from the out-of-town medical treatment request; the target object's electronic medical record, medical reference information and reason for out-of-town medical treatment are input into the target node, and the target node determines the target hospital recommended to the target object from multiple candidate hospitals at the medical destination.

[0007] Furthermore, the present application also proposes that medical reference information includes medical preferences and medical needs, and the reasons for out-of-town medical treatment include subjective reasons for out-of-town medical treatment and reference reasons for out-of-town medical treatment. In response to the target object's request for out-of-town medical treatment, the target object's physiological state, medical reference information and reasons for out-of-town medical treatment are determined, including: in response to the target object's request for out-of-town medical treatment, obtaining the target object's physiological data, subjective reasons for out-of-town medical treatment, historical medical data and medical feedback data, the medical feedback data is used to represent the hospital's evaluation of the target object; based on the physiological data, the target object's physiological state is determined; based on the historical medical data, the medical preference is determined; based on the historical medical data, the reference reasons for out-of-town medical treatment are determined.

[0008] Furthermore, the present application also proposes that historical medical data includes historical medical hospitals and historical medical times, and medical preferences are determined based on the historical medical data, including: obtaining the first hospital information of the historical medical hospital, and the first hospital information is used to introduce the historical medical hospital; based on the historical medical time, the basic medical time pattern of the target object is determined; based on the first hospital information and the historical medical time, the reference medical time pattern of the target object is determined; based on the first hospital information, the basic medical time pattern and the reference medical time pattern, the medical preference is determined; based on the historical medical data and medical feedback data, the reference reason for medical treatment in a different place is determined, including: obtaining the second hospital information of the historical medical hospital, and the second hospital information is used to describe the medical situation of the target object in the historical medical hospital; based on the second hospital information and the medical feedback data, the reference reason for medical treatment in a different place is determined.

[0009] Furthermore, the present application also proposes to determine the medical preference based on the first hospital information, the basic medical time pattern and the reference medical time pattern, including: determining the hospital preference and department preference of the target object based on the first hospital information; determining the medical time preference of the target object based on the basic medical time pattern and the reference medical time pattern; splicing the target object's hospital preference, department preference and medical time preference to obtain the medical preference; the second hospital information includes the number of medical payment times, the amount of medical payment, the medical travel distance and the medical effect information, and based on the second hospital information and the medical feedback data, determining the reference reason for medical treatment in a different place, including: determining the medical fatigue description information based on the number of medical payment times and the medical travel distance; determining the medical effect description information based on the medical payment amount and the medical effect information; determining the reference reason for medical treatment in a different place based on the medical fatigue description information, the medical effect description information and the medical feedback data.

[0010] Furthermore, the present application also proposes that, based on the physiological state of the target object and the reason for seeking medical treatment elsewhere, a target node that matches the physiological state and the reason for seeking medical treatment elsewhere is determined from multiple candidate nodes associated with the medical destination, including: based on the physiological state, multiple reference nodes are determined from multiple candidate nodes, and the multiple reference nodes are candidate nodes that match the physiological state; based on the reason for seeking medical treatment elsewhere, the target node is determined from multiple reference nodes.

[0011] Furthermore, the present application also proposes that, based on the physiological state, multiple reference nodes are determined from multiple candidate nodes, and the multiple reference nodes are candidate nodes matching the physiological state, including: determining the first node description information of each candidate node, the first node description information is used to describe the associated physiological state of the corresponding candidate node; based on the physiological state and the first node description information of each candidate node, determining the first matching degree between the physiological state and each candidate node; determining the candidate nodes whose first matching degree between the multiple candidate nodes and the physiological state is greater than or equal to the first matching degree threshold as reference nodes, to obtain multiple reference nodes; based on the reason for medical treatment in other place, determining the target node from multiple reference nodes, including: determining the second node description information of each reference node, the second node description information is used to describe the associated medical reason of the reference node; based on the reason for medical treatment in other place and the associated medical reason of each described reference node, determining the second matching degree between the reason for medical treatment in other place and each reference node; determining the reference node with the highest second matching degree between the reason for medical treatment in other place among the multiple reference nodes as the target node.

[0012] Furthermore, the present application also proposes to determine a target hospital recommended to a target object from multiple hospitals at the medical destination through a target node, including: determining multiple candidate hospitals from multiple hospitals through the target node based on electronic medical records, medical reference information and hospital description information of each hospital; sending the target object's electronic medical records, medical reference information and reasons for out-of-town medical treatment to the hospital node of each candidate hospital, so that the hospital node of each candidate hospital determines the degree of reception matching, the willingness to receive treatment and the reasons for generating the degree of reception matching and the willingness to receive treatment based on the target object's electronic medical records, medical reference information and reasons for out-of-town medical treatment; obtaining the degree of reception matching, the willingness to receive treatment and the reasons for generating the degree of reception matching and the willingness to receive treatment of each candidate hospital through the target node; determining the target hospital from multiple candidate hospitals through the target node based on the degree of reception matching, the willingness to receive treatment, the reasons for generating the willingness to receive treatment and the medical reference information of each candidate hospital.

[0013] Furthermore, the present application also proposes to determine multiple candidate hospitals from multiple hospitals through a target node based on electronic medical records, medical reference information and hospital description information of each hospital, including: determining the third matching degree between the target object and each hospital through a target node based on electronic medical records, medical reference information and hospital description information of each hospital; determining hospitals among multiple hospitals whose third matching degree is greater than or equal to the second matching degree threshold as candidate hospitals, and obtaining multiple candidate hospitals; determining the target hospital from multiple candidate hospitals through a target node based on the reception matching degree, reception willingness, generation reason and medical reference information of each candidate hospital, including: determining the reception intention score of each candidate hospital for the target object through a target node based on the reception matching degree, reception willingness and generation reason of each candidate hospital; determining the fourth matching degree between the target object and each candidate hospital based on the reception willingness and medical reference information of each candidate hospital; determining the target hospital from multiple candidate hospitals based on the reception intention score and the fourth matching degree.

[0014] Furthermore, the present application also proposes to determine the third matching degree between the target object and each hospital based on the electronic medical record, medical reference information and hospital description information of each hospital through the target node, including: generating object description information of the target object based on the electronic medical record and medical reference information through the target node; determining the third matching degree between the target object and each hospital based on the object description information of the target object and the hospital description information of each hospital; determining the acceptance intention score of each candidate hospital for the target object based on the acceptance matching degree, acceptance willingness and generation reason of each candidate hospital through the target node, including: inputting the acceptance matching degree, acceptance willingness and generation reason of each candidate hospital into the acceptance intention score determination model, and determining the acceptance matching process through the acceptance intention score determination model The fourth matching degree between the target object and each candidate hospital is determined based on the acceptance intention of each candidate hospital and the medical reference information, including: inputting the acceptance intention of each candidate hospital and the medical reference information into the matching degree determination model, processing the acceptance intention and the medical reference information through the matching degree determination model, and outputting the fourth matching degree between the target object and each candidate hospital; determining the target hospital from multiple candidate hospitals based on the acceptance intention score and the fourth matching degree, including: fusing the acceptance intention score and the fourth matching degree to obtain the target matching score of each candidate hospital; and determining the candidate hospital with the highest target matching score among the multiple candidate hospitals as the target hospital.

[0015] Furthermore, the present application also proposes to generate object description information of the target object based on electronic medical records and medical reference information, including: extracting features of the electronic medical records and medical reference information to obtain electronic medical record features of the electronic medical record and reference information features of the medical reference information; fusing the electronic medical record features and the reference information features to obtain object features of the target object; performing multiple rounds of iterative decoding on the object features to obtain object description information of the target object; determining the third degree of matching between the target object and each hospital based on the object description information of the target object and the hospital description information of each hospital, including: extracting features of the object description information of the target object and the hospital description information of each hospital to obtain a first description information feature of the object description information and a second description information feature of the hospital description information of each hospital; determining the feature similarity between the first description information feature and the second description information feature corresponding to each hospital as the third degree of matching between the target object and each hospital.

[0016] On the one hand, a medical service improvement system based on remote medical treatment data analysis is provided, the system comprising:

[0017] A first determining module is configured to determine the target subject's current physiological state, medical reference information, and reason for seeking medical treatment in a different place in response to the target subject's request for seeking medical treatment in a different place;

[0018] A second determination module is configured to determine, based on the current physiological state of the target subject and the reason for the remote medical treatment, a target node that matches the physiological state and the reason for the remote medical treatment from a plurality of candidate nodes associated with the medical treatment destination, wherein different candidate nodes have different computing powers and hospital recommendation modes, and the medical treatment destination is obtained from the remote medical treatment request;

[0019] The third determination module is used to input the target object's electronic medical record, the medical reference information and the reason for the out-of-town medical treatment into the target node, and determine the target hospital recommended to the target object from multiple candidate hospitals at the medical destination through the target node.

[0020] On the one hand, a computer device is provided, which includes one or more processors and one or more memories, wherein at least one computer program is stored in the one or more memories, and the computer program is loaded and executed by the one or more processors to implement the medical service improvement method based on the analysis of medical treatment data in a different place.

[0021] On the one hand, a computer-readable storage medium is provided, in which at least one computer program is stored. The computer program is loaded and executed by a processor to implement the medical service improvement method based on the analysis of medical treatment data in a different place.

[0022] On the one hand, a computer program product or computer program is provided, which includes a program code, which is stored in a computer-readable storage medium. The processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device executes the above-mentioned medical service improvement method based on the analysis of medical treatment data in a different place.

[0023] From the above, it can be seen that the present application provides a medical service improvement method and system based on the analysis of medical treatment data in other places. By integrating the patient's physiological status, historical preferences and reasons for medical treatment, it dynamically matches node resources with different computing power and recommendation modes, and combines multi-dimensional data mining with a two-way matching mechanism to effectively solve the mismatch problem between medical resources and patient needs. It has the advantages of improving matching accuracy, optimizing decision-making efficiency, and improving service experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0025] Figure 1 Schematic diagram of an implementation environment of a medical service improvement method based on remote medical treatment data analysis provided in an embodiment of the present application;

[0026] Figure 2 This is a flow chart of a medical service improvement method based on remote medical treatment data analysis provided by an embodiment of the present application;

[0027] Figure 3 This is a flowchart for determining the reason for seeking medical treatment in a different place, provided by an embodiment of the present application;

[0028] Figure 4 This is a flowchart of determining a target node provided by an embodiment of the present application;

[0029] Figure 5 This is a flowchart for determining a target hospital provided in an embodiment of the present application;

[0030] Figure 6 This is a structural diagram of a medical service improvement system based on remote medical treatment data analysis provided by an embodiment of the present application;

[0031] Figure 7 This is a structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0033] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with substantially the same effects and functions. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", nor is there any limitation on the quantity and execution order.

[0034] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to achieve better results.

[0035] Machine learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge sub-models to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and learning through demonstration.

[0036] Out-of-town medical treatment: Out-of-town medical treatment refers to patients seeking medical treatment at medical institutions in other regions, leaving their place of medical insurance coverage (usually their registered residence or place of work). This phenomenon is often driven by the uneven distribution of medical resources, patients' demand for high-quality medical services, or population mobility (e.g., for work or retirement).

[0037] Medical Information System (MIS): A comprehensive system that uses information technology to integrate medical data, optimize medical processes, and improve service efficiency. Its core goals include supporting clinical decision-making, improving patient management, and promoting cross-institutional collaboration.

[0038] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0039] Figure 1This is a schematic diagram of the implementation environment of a medical service improvement method based on remote medical treatment data analysis provided by the embodiment of the present application, see Figure 1 , the implementation environment may include a node 110 and a server 140 .

[0040] The node 110 is connected to the server 140 via a wireless network or a wired network. Optionally, the node 110 is a laptop computer, a desktop computer, etc., but is not limited thereto. The node 110 is installed and runs an application program that supports the improvement of medical services based on the analysis of medical treatment data in different places.

[0041] Server 140 is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Server 140 can provide backend services for applications running on node 110.

[0042] In related technologies, patients seeking medical treatment across regions often face the problem of inefficient matching of medical resources. Existing service platforms usually only provide a list of basic hospitals or simplify the filing process, and lack the ability to deeply analyze individual patient needs. When a patient needs urgent cross-provincial medical treatment due to a sudden illness, the existing system has difficulty in quickly identifying the urgency of the patient's condition and cannot dynamically match it based on the real-time reception capabilities of the medical institution. For example, when a patient with myocardial infarction needs to quickly find a hospital with interventional treatment capabilities in a different location, the traditional system can only display a list of hospitals and cannot comprehensively evaluate the hospital's current bed situation, specialist strength and the compatibility of the patient's condition.

[0043] In order to solve the above problems, researchers found that the key to matching medical resources lies in building a dynamic adaptation mechanism. First, it was observed that there are differences in computing power among medical information systems in different regions. Some nodes are good at real-time data processing, while others have complex model computing capabilities. Secondly, it was found that the reason why patients seek medical treatment directly affects the resource selection strategy. For example, cancer patients focus on specialist rankings, while emergency patients need to consider response speed. By establishing a mapping relationship between patient characteristics and computing node capabilities, attempts are made to match disease data with node characteristics in multiple dimensions, and finally a phased screening mechanism is formed: first, computing nodes are filtered according to physiological status, then the optimal node is determined according to the reason for seeking medical treatment, and finally hospital matching is completed through the recommendation algorithm of the selected node.

[0044] Therefore, this application proposes a medical service improvement method based on the analysis of medical treatment data in different places, see Figure 2 , taking the execution subject as a server as an example, the method includes:

[0045] 201. In response to a target subject's request for remote medical treatment, the server determines the target subject's current physiological state, medical reference information, and reason for remote medical treatment;

[0046] 202. The server determines a target node that matches the target subject's current physiological state and the reason for seeking medical treatment in a different location from multiple candidate nodes associated with the medical treatment destination. Different candidate nodes have different computing power and hospital recommendation modes. The medical treatment destination is obtained from the medical treatment request in a different location.

[0047] 203. The server inputs the target subject's electronic medical record, medical reference information, and reason for medical treatment in a different place into the target node, and determines a target hospital recommended to the target subject from multiple candidate hospitals at the medical treatment destination through the target node.

[0048] Among them, the current physiological state refers to the real-time health indicators obtained through wearable devices or electronic medical record analysis. Specifically, it can be achieved by combining heart rate variability analysis with medical record keyword extraction to assess the urgency of medical needs. Medical reference information includes a preference model formed by historical medical records. Specifically, a collaborative filtering algorithm can be used to process historical medical data to reflect the personalized needs of patients. Candidate nodes refer to heterogeneous computing units deployed at the medical destination. Specifically, they can be implemented using a containerized microservice architecture, with each node configured with an independent resource evaluation algorithm and recommendation strategy. The hospital recommendation model refers to the medical institution evaluation system. Specifically, a multi-index decision model based on the hierarchical analysis method can be used to quantify the match between the hospital and patient needs.

[0049] Specifically, when a patient requests out-of-town medical treatment, the system first analyzes diagnostic records and real-time monitoring data in the electronic medical record to generate a physiological status assessment report with a disease classification indicator. It also extracts the patient's medical records from the insured area over the past three years, analyzing characteristics such as the proportion of patients choosing tertiary hospitals and the interval between follow-up visits. This information model then constructs a reference information model that includes department preferences and hospital grade preferences. Based on the patient's self-reported referral reason and the system's inferred potential needs, three candidate nodes are selected from multiple pre-deployed computing nodes at the medical destination: the first node is equipped with a real-time bed query interface for emergency use; the second node is equipped with a disease prognosis prediction model for follow-up visits for chronic diseases; and the third node integrates multi-specialty ranking data for elective surgery. By comparing the time sensitivity of the patient's physiological status with the processing capacity of each node, the node with the fastest response time is prioritized as the initial screening result. Finally, based on the keyword "seeking specialized treatment" in the referral reason, the target node equipped with the specialist ranking analysis model is selected. This node matches the patient's tumor staging data with the destination hospital's radiotherapy equipment configuration and expert consultation status, generates a recommendation list containing predicted treatment success rates, and selects the tertiary cancer hospital with the highest comprehensive score as the target hospital.

[0050] Compared with related technologies, traditional methods use fixed algorithms to process all medical requests and are unable to dynamically adjust computing resources according to the urgency of the disease. This solution builds a configurable node pool so that patients with cardiovascular emergencies are automatically assigned to nodes with real-time data processing capabilities, while patients with rare diseases are routed to deep computing nodes equipped with knowledge graphs, achieving precise adaptation of computing resources to medical needs. Existing systems usually analyze hospital static data separately. This solution innovatively associates patients' physiological status, historical preferences and the hospital's dynamic reception capacity in multiple dimensions. For example, during an influenza outbreak, the bed turnover rate weight of the fever clinic is automatically increased to make the recommendation results more in line with the actual situation.

[0051] Through the above technical solutions, this application realizes the intelligent scheduling of medical computing resources, reducing the node matching time of emergency patients to one-fifth of the traditional method. Through the dual screening mechanism of dynamically combining physiological status and reasons for medical treatment, the accuracy of hospital recommendations for complex cases is significantly improved, especially for cancer patients referred across provinces. The consistency between the recommended target hospital and the actual hospital visited reaches a high level. The on-demand calling mechanism of heterogeneous computing nodes effectively balances the system load, and can still maintain stable service quality when handling concentrated medical requests caused by large-scale sudden epidemics.

[0052] This application further proposes that the medical reference information includes medical preferences and medical needs, and the reasons for medical treatment in a different place include subjective reasons for medical treatment in a different place and reference reasons for medical treatment in a different place. In response to the target object's request for medical treatment in a different place, the target object's physiological state, medical reference information and reasons for medical treatment in a different place are determined. Figure 3 , taking the execution subject as the server as an example, the steps include:

[0053] 301. In response to a target subject's request for medical treatment in a different location, the server obtains the target subject's physiological data, subjective reasons for medical treatment in a different location, historical medical treatment data, and medical treatment feedback data, where the medical treatment feedback data is used to represent the hospital's evaluation of the target subject.

[0054] 302. The server determines the physiological state of the target object based on the physiological data;

[0055] 303. The server determines the medical preference based on the historical medical data; and determines the reason for seeking medical treatment in a different place based on the historical medical data and medical feedback data.

[0056] Among them, medical preference refers to the patient's long-term medical behavior tendency. Specifically, this can be achieved by using the primary hospital information of historical medical hospitals and historical medical treatment time for pattern analysis. By analyzing the hospital level, department type and the distribution of medical treatment time, the patient's selection tendency for medical institutions is formed. Medical demand refers to the specific demands in the current medical treatment scenario. Specifically, this can be achieved by extracting features from the diagnosis records and treatment plans in the electronic medical record to reflect the patient's functional requirements for medical services. Subjective reasons for seeking medical treatment in other places refer to the motivation for seeking medical treatment across regions actively stated by the patient. Specifically, this can be achieved by using natural language processing technology to extract keywords from the text information input by the patient, such as referral needs and expert resource preferences. Reference reasons for seeking medical treatment in other places refer to potential motivations derived from objective data analysis. Specifically, this can be achieved by using the secondary hospital information of historical medical hospitals and medical feedback data for correlation analysis. For example, it can identify patients' dissatisfaction with the service efficiency or cost structure of historical hospitals.

[0057] Specifically, when a request for out-of-town medical treatment is triggered, the patient's physiological data is collected in real time to assess the current health status level, and the subjective reasons for out-of-town medical treatment are stored in a structured manner. At the same time, the hospital attributes and medical treatment time series in the historical medical treatment data are extracted, and a basic medical treatment time pattern is generated through a time series pattern mining algorithm. Combined with the hospital grade characteristics, a reference medical treatment time pattern is formed, and finally a medical treatment preference including hospital type preference, department selection tendency and medical treatment time period is derived. Medical treatment feedback data and data such as payment frequency and travel distance in historical medical treatment records are jointly analyzed. By calculating the medical treatment fatigue index and effect deviation, objective reasons for out-of-town medical treatment that are not clearly expressed by patients are identified, such as the traffic burden caused by frequent follow-up visits or the treatment effect not meeting expectations.

[0058] Compared to related technologies, traditional methods rely solely on patient-reported reasons for seeking medical treatment and a single-dimensional historical record to make recommendations, failing to identify implicit needs. This solution, by establishing a cross-validation mechanism between subjective statements and objective behavioral data, correlates service quality evaluations from hospital feedback data with patients' actual medical trajectories, identifying unexpressed referral motivations. Furthermore, based on a dual analysis of time series characteristics and hospital attributes, it accurately captures cyclical patterns in patients' medical behavior and institutional preferences.

[0059] Through the above technical solutions, this application achieves a three-dimensional analysis of patients' individualized medical characteristics, solving the recommendation bias problem caused by the single data dimension of traditional systems. The medical preference analysis module can accurately identify patients' long-term preference for specific hospital grades and department types. Referring to the analysis module of reasons for out-of-town medical treatment, it can effectively explore the hidden referral needs caused by medical fatigue or poor treatment results, providing multi-dimensional data support for the subsequent precise matching of medical resources.

[0060] This application further proposes a technical solution for determining the medical preferences of the target subject based on the first hospital information and historical medical treatment time of the historical medical hospital, and for determining the reference reasons for medical treatment in a different place based on the second hospital information and medical treatment feedback data of the historical medical hospital.

[0061] First, hospital information refers to attribute data describing historical hospital visits. Specifically, this data can be implemented using data on hospital grade, departmental settings, and service types. This data is used to analyze patients' basic preferences for hospital type and department. Basic visit time patterns refer to preliminary time distribution patterns derived from historical visit time statistics. This data can be implemented using a time series clustering algorithm to analyze the periodicity of historical visit times. This data reflects patients' initial time preferences for medical treatment. Reference visit time patterns refer to time distribution patterns modified based on hospital service characteristics. This data can be implemented by cross-validating hospital operating hours data with historical visit times. This data is used to adjust for time periods in the basic time pattern that don't match the hospital's actual service capacity. Second, hospital information refers to data describing the patient's specific visit journey at a historical hospital. This data can be implemented using data on visit frequency, travel distance, and payment records. This data is used to quantify patient fatigue and differences in treatment outcomes across specific hospitals. Medical feedback data refers to hospital evaluations of the patient's medical experience. This data can be implemented using satisfaction scores or doctor-patient communication records from hospital systems. This data supplements subjective medical experience that objective data cannot capture.

[0062] Specifically, to determine medical preferences, we first extract historical hospital grade and department configuration information to identify the hospital types and department categories most frequently chosen by patients, thereby forming hospital and department preferences. We then conduct a cyclical analysis of historical medical visit times to identify patients' preferred visit times, which serve as a baseline time pattern. Then, by combining the operating hours and department rosters of each hospital, we correct for time periods in the baseline time pattern that do not reflect actual preferences due to hospital service limitations, generating a reference time pattern. Finally, we integrate hospital preferences, department preferences, and the corrected time preferences to form a three-dimensional medical preference model that encompasses institution selection, department preferences, and temporal patterns. To determine the reasons for seeking medical treatment outside of a specific location, we calculate the physical exertion index generated by medical behavior by counting the number of visits and distance traveled at historical hospitals. We then analyze the cost-effectiveness of medical care by combining the ratio of payment amounts to treatment outcome data, thereby forming an objective indicator for evaluating medical fatigue and effectiveness. At the same time, the doctor-patient communication records and satisfaction scores recorded in the hospital system are integrated to identify patients' subjective evaluation of the quality of medical services, cross-validate objective indicators with subjective evaluations, and distinguish the fundamental reasons why patients choose to seek medical treatment in other places due to objective conditions or subjective dissatisfaction.

[0063] Compared with related technologies, existing methods typically only count visit frequencies or simply categorize hospital types, failing to correlate hospital service characteristics with temporal distribution patterns, leading to distorted time preference analysis. Traditional techniques rely on single-dimensional data such as visit frequency or cost to infer the reasons for seeking medical treatment in a different location, ignoring the interplay between hospital service capabilities and patients' subjective experience. This solution uses a multi-dimensional cross-analysis of hospital attribute data and time data to accurately distinguish between patients' true time preferences and passive choices driven by hospital service constraints. It also incorporates a dual verification mechanism of objective behavioral data and subjective evaluation data to effectively identify the subjective and objective drivers of decision-making in different locations.

[0064] Through the above technical solution, this application solves the problem of traditional methods ignoring the impact of hospital service characteristics on time distribution in medical preference analysis, improves the accuracy of time preference extraction, and avoids preference misjudgments caused by hospital business hours restrictions. At the same time, it overcomes the limitations of analyzing the reasons for out-of-town medical treatment with a single data source. Through the collaborative verification of subjective and objective data, it accurately distinguishes patients' out-of-town medical needs caused by objective conditions or subjective dissatisfaction, providing a reliable data foundation for subsequent precise matching of target hospitals.

[0065] This application further proposes to determine the hospital preference and department preference of the target object based on the first hospital information, determine the medical time preference based on the basic medical time pattern and the reference medical time pattern, and splice the hospital preference, department preference and medical time preference to form the medical preference; the second hospital information includes the number of medical payments, the payment amount, the moving distance and the medical effect information, generate medical fatigue description information based on the number of payments and the moving distance, generate medical effect description information based on the payment amount and the medical effect information, and determine the reference reason for medical treatment in other places in combination with the medical feedback data.

[0066] Hospital preference refers to a patient's preference for a medical institution's grade, nature, or service type. This can be achieved by combining historical hospital classification data with a correlation analysis algorithm for the departments visited, reflecting patients' trust in medical institutions. Department preference refers to a patient's tendency to choose specialized medical treatment for a specific disease type. This can be achieved by combining historical department distribution statistics with a disease type matching model to identify patients' specialized medical treatment habits. Time preference for medical treatment refers to the regularity of patients' medical treatment choices during specific time periods. This can be achieved by combining a time series pattern mining algorithm with a cluster analysis of historical registration times to capture patients' time-sensitive needs. Medical fatigue descriptive information quantifies the physical and mental burden incurred by frequent trips to the hospital. This can be achieved through a weighted calculation model combining travel distance and number of visits, assessing patients' potential demand for nearby medical resources. Medical effectiveness descriptive information is a comprehensive evaluation indicator of medical expenditure and treatment effectiveness. This can be achieved by combining a cost-effectiveness ratio algorithm with a treatment effectiveness grading standard to quantify patients' satisfaction with medical quality.

[0067] Specifically, in the medical preference generation phase, by analyzing historical primary hospital information, structured data such as hospital grade and departmental settings are extracted. A hierarchical clustering algorithm is then used to identify patients' preferences for tertiary hospitals or specialized hospitals, thereby forming hospital preferences. The correlation between historical medical department visits and disease diagnoses is also analyzed, and a department selection probability model is established to generate department preferences. In the time preference analysis, a basic medical time pattern is derived by statistically analyzing the distribution of historical registration times. This basic pattern is then modified by incorporating dynamic data on hospital capacity into the reference medical time pattern. Ultimately, a time window matching algorithm is used to determine the optimal time for medical treatment. In the analysis of the reasons for out-of-town medical treatment, the number of medical payment payments and travel distance are normalized and then input into a fatigue calculation model to generate a quantitative indicator reflecting transportation burden. The amount of medical payment and treatment effectiveness are then processed using a cost-effectiveness analysis model to generate medical quality evaluation indicators. These two objective indicators are then integrated with subjective evaluations from medical feedback data in a multi-dimensional manner. A decision tree classification algorithm is then used to identify the core factors leading to out-of-town medical treatment.

[0068] Compared with related technologies, traditional methods rely solely on the reasons for seeking medical treatment actively reported by patients, lack in-depth mining of historical behavioral data, and are prone to information bias. Existing systems usually use single-dimensional time statistics or simple department matching, and are unable to dynamically adjust the time preference model. This solution constructs a composite preference model by integrating multi-dimensional data such as hospital grade, department association, and dynamic time adjustment, effectively solving the problem of traditional recommendation systems ignoring implicit medical habits. In terms of analyzing the reasons for seeking medical treatment in other places, related technologies mostly rely on manually filled-in superficial reasons. This solution, by quantitatively analyzing medical fatigue and cost-effectiveness, combined with cross-validation of subjective and objective data, can accurately identify the deep motivations for seeking medical treatment that patients have not clearly expressed.

[0069] Through the above technical solution, this application can automatically identify patients' complex demand preferences for hospital grades, specialty types, and consultation time periods, and solve the matching bias problem caused by traditional recommendation systems ignoring historical behavioral patterns. By quantitatively analyzing the traffic burden and medical quality cost ratio during the medical treatment process, combined with subjective and objective evaluation data, it accurately captures the potential factors that lead to medical treatment in other places, and improves the fit between medical resource recommendations and patients' actual needs. This technical solution effectively overcomes the limitations of single data source analysis, realizes the synergy of multi-dimensional heterogeneous data, and provides accurate data support for personalized medical recommendations.

[0070] This application further proposes to determine the target node that matches the physiological state and the reason for seeking medical treatment in a different place from multiple candidate nodes associated with the medical destination based on the target object's physiological state and the reason for seeking medical treatment in a different place, see Figure 4 , the method comprises the following steps:

[0071] 401. The server determines, based on the physiological state, a plurality of reference nodes from a plurality of candidate nodes, where the plurality of reference nodes are candidate nodes that match the physiological state;

[0072] 402. The server determines a target node from multiple reference nodes based on the reason for seeking medical treatment in a different location.

[0073] Physiological status refers to the target patient's current health indicators assessed through physiological data. Specifically, this can be achieved by combining sensor data such as heart rate, blood pressure, and blood oxygen saturation with diagnostic records in electronic medical records, and classifying them using machine learning models to select candidate nodes with the corresponding medical data processing capabilities. The reason for seeking medical treatment outside of a specific location refers to the subjective and objective factors that trigger the target patient to seek medical treatment outside of their location. Natural language processing techniques can be used to extract keywords and classify intent from the text information provided by the patient, such as insufficient medical resources, poor treatment outcomes, and transportation needs, to adapt the recommendation algorithm model to different nodes. Candidate nodes refer to heterogeneous computing resources distributed across the medical destination. Specifically, a node registration mechanism can be used to manage server clusters with different computing power configurations, data processing models, and recommendation strategies. For example, GPU-accelerated nodes are suitable for image data analysis, while CPU-intensive nodes are suitable for medical record text processing. Reference nodes refer to the subset of candidate nodes initially screened based on physiological status. Specifically, a matching threshold filtering mechanism can be used, for example, setting a matching threshold of 80% or greater to advance nodes to the next stage of screening, reducing subsequent computational overhead. Among them, the target node refers to the final selected computing node. Specifically, a weighted scoring mechanism can be adopted, combining the matching degree between the reasons for medical treatment in other places and the node recommendation mode. For example, a node that supports a multi-factor decision tree is selected to handle complex reasons for medical treatment.

[0074] Specifically, the set of candidate nodes with corresponding medical data processing capabilities is first screened out based on physiological status. For example, when the patient's physiological status is in the acute stage of cardiovascular disease, nodes that support real-time vital signs monitoring are prioritized. Then, a secondary match is performed based on the reason for seeking medical treatment in a different place. For example, if the reason for the patient's medical treatment in a different place is "lack of specialized equipment locally", nodes that support medical resource gap analysis are selected. The two-stage screening mechanism first excludes nodes that are not related to physiological status, and then selects nodes that are adapted to the recommended mode based on specific reasons, avoiding the waste of computing resources caused by traversing all nodes, while adapting to the differences in computing power distribution and data processing models among different nodes.

[0075] Compared to related technologies, traditional methods typically use a single-dimensional matching approach, such as filtering based solely on hospital rank or directly calling a fixed node to handle all requests. This solution, through a phased screening mechanism, for the first time collaboratively analyzes physiological status and the reason for seeking medical treatment elsewhere, dynamically adapting heterogeneous node resources. Related technologies are unable to distinguish between differences in node computing power, resulting in high latency or biased recommendations. This solution, through a two-stage screening process, optimizes resource utilization while also improving recommendation accuracy.

[0076] Through the above technical solution, this application solves the problem that distributed heterogeneous node resources cannot adapt to the individual needs of patients, and realizes computing power matching based on physiological status and recommendation mode adaptation based on medical reasons. For example, in the scenario where a patient chooses to seek medical treatment in a different place for postoperative review, the system can prioritize calling nodes that support image data comparison, and then select a recommendation mode that supports traffic condition analysis based on the reason of "review convenience", and finally generate hospital recommendation results that take into account both medical expertise and geographical accessibility.

[0077] The present application further proposes to determine multiple reference nodes from multiple candidate nodes based on physiological state, and the multiple reference nodes are candidate nodes matching the physiological state, including: determining first node description information of each candidate node, the first node description information is used to describe the associated physiological state of the corresponding candidate node; determining a first matching degree between the physiological state and each candidate node based on the physiological state and the first node description information of each candidate node; determining the candidate nodes whose first matching degree between the multiple candidate nodes and the physiological state is greater than or equal to the first matching degree threshold as reference nodes, to obtain multiple reference nodes; determining a target node from multiple reference nodes based on the reason for medical treatment in a different place, including: determining second node description information of each reference node, the second node description information is used to describe the associated medical reason of the reference node; determining a second matching degree between the reason for medical treatment in a different place and each reference node based on the reason for medical treatment in a different place and the associated medical reason of each described reference node; determining the reference node with the highest second matching degree between the reason for medical treatment in a different place and the reason for medical treatment in a different place as the target node.

[0078] Among them, the first node description information refers to the range of physiological states that describe the processing capabilities of the candidate nodes. Specifically, natural language processing technology can be used to perform semantic analysis on the node service logs to generate information for screening computing nodes capable of processing the current patient's condition. The second node description information refers to the type of medical reason that describes the candidate node's adaptation. Specifically, it can be achieved by building an association between the node and the medical scenario through a knowledge graph, which is used to match the core demands of patients seeking medical treatment in other places. The first matching degree threshold refers to the minimum adaptation standard for node screening. Specifically, a dynamic adjustment algorithm can be used to automatically set it according to the number of candidate nodes and the distribution of computing power, to balance computing efficiency and matching accuracy. The second highest matching degree refers to the ranking of the strength of the association between the node and the reason for medical treatment. Specifically, the cosine similarity algorithm can be used to calculate the semantic matching degree between the reason text and the node description information, which is used to select the node resources that best meet the patient's needs.

[0079] Specifically, when a patient submits a request for out-of-town medical treatment, a standardized description of the physiological state is first generated by parsing the electronic medical record, such as converting the diagnostic conclusion in the medical record into an ICD-10 disease code. The first node description information of the candidate node is extracted through a pre-trained medical entity recognition model, such as extracting high-frequency diagnostic keywords from the node's historical case processing. The first matching degree is calculated using a vector space model, mapping the patient's physiological state and the node description information into feature vectors and then calculating the cosine similarity. When the matching degree exceeds the dynamic threshold, the node enters the reference node set. Subsequently, the patient's reason for out-of-town medical treatment is processed by word segmentation and semantically matched with the second node description information of the reference node, such as using the BERT model to calculate the text similarity score, and finally the node with the highest score is selected as the target node.

[0080] Compared with related technologies, traditional methods only perform single-dimensional matching based on hospital location or department classification, and are unable to handle the differentiated service capabilities of heterogeneous nodes. This solution, by establishing a dual description system for nodes, dynamically adapts to the computing needs of different medical scenarios while ensuring the basic service capabilities of the nodes. The node resource waste caused by static threshold settings in related technologies is optimized in this solution through dynamic adjustment of threshold strategies. At the same time, the maximum matching optimization mechanism avoids resource conflicts caused by multi-node parallel computing.

[0081] Through the above technical solutions, this application realizes the precise dynamic allocation of medical computing node resources. In the scenario where a patient has an emergency and needs urgent referral, it can quickly screen out nodes that have experience in treating the disease and support green channel services; in the scenario where a patient chooses to seek medical treatment in another place due to insufficient medical resources, it can accurately match nodes that are good at the disease and support cross-regional collaboration. The dual screening mechanism of node resources effectively reduces the probability of distributing invalid computing tasks, while ensuring the priority service needs of critically ill patients.

[0082] This application further proposes a method for determining a target hospital recommended to a target subject from multiple hospitals at the medical destination based on the target node, see Figure 5 Taking the execution subject as a server as an example, the method includes the following steps:

[0083] 501. The server determines, through the target node, multiple candidate hospitals from multiple hospitals based on the electronic medical records, medical reference information, and hospital description information of each hospital;

[0084] 502. The server sends the target patient's electronic medical record, medical reference information, and reason for out-of-town medical treatment to the hospital node of each candidate hospital, so that the hospital node of each candidate hospital determines the degree of match, willingness to receive the treatment, and reasons for generating the degree of match and willingness to receive the treatment based on the target patient's electronic medical record, medical reference information, and reason for out-of-town medical treatment.

[0085] 503. The server obtains the acceptance matching degree, acceptance willingness, and generation reason of each candidate hospital through the target node;

[0086] 504. The server determines a target hospital from the plurality of candidate hospitals through the target node based on the acceptance matching degree, acceptance willingness, generation reason and medical reference information of each candidate hospital.

[0087] Among them, candidate hospitals refer to medical institutions that have potential matches with the medical needs of target subjects through preliminary screening. Specifically, this can be achieved by using a multi-dimensional matching algorithm between hospital description information and patient characteristics, which is used to narrow the scope of recommendations and improve subsequent processing efficiency. The degree of match between admissions refers to a quantitative indicator of the hospital's diagnostic and treatment adaptability to the patient's condition. Specifically, it can be calculated by analyzing the correlation between the disease characteristics in the electronic medical record and the hospital's specialty advantages, and is used to objectively evaluate whether the hospital's professional capabilities meet treatment needs. Willingness to admit patients refers to the hospital's intention to accept patients under current resource conditions. Specifically, it can be generated through real-time resource monitoring data of hospital nodes combined with historical admission strategy models to reflect the hospital's dynamic admission capabilities. The reason for generation refers to the decision-making basis text for the degree of match between admissions and willingness to admit patients. Specifically, natural language generation technology can be used to convert the matching calculation logic and resource assessment results into explainable explanations to provide a transparent basis for recommendation decisions.

[0088] Specifically, the target node first selects a set of candidate hospitals that meet basic admission requirements based on static attributes such as the disease type in the patient's electronic medical record, preference data from the medical reference information, and hospital descriptions, such as specialty settings and equipment configuration. For example, if a patient has cardiovascular disease, the target node prioritizes hospitals with cardiology departments and interventional therapy equipment. The patient's complete medical data package is then sent to the hospital node of each candidate hospital. This triggers the hospital to generate quantitative admission match and willingness-to-accept values ​​based on dynamic data such as real-time bed occupancy and physician scheduling, combined with historical data such as the hospital's specialty treatment success rate and experience with similar cases. It also generates a textual reasoning for the decision, including a resource load analysis and a comparison of specialty strengths. After receiving feedback from each hospital, the target node comprehensively evaluates the professional compatibility reflected by the admission match, the resource availability reflected by the willingness to accept, and the rationality of the decision, as revealed by the reasoning. Furthermore, the target node uses a weighted scoring model to determine the final target hospital recommendation, taking into account personalized needs such as time preference and distance sensitivity from the patient's medical reference information.

[0089] Compared with related technologies, existing out-of-town medical recommendation systems usually only perform one-way matching based on public hospital information and basic patient needs, lacking real-time resource status feedback on the hospital side and dynamic evaluation of professional adaptability. This solution, by establishing a dynamic response mechanism for hospital nodes, triggers the hospital side to generate quantitative feedback based on real-time operational data and professional capability models after preliminary screening, so that recommendation decisions not only include static matching results, but also integrate dynamic evaluation data on the hospital's current reception capacity and professional advantages. For example, when a hospital has treatment capabilities but is currently at full capacity, its willingness to receive patients indicator will automatically decrease to avoid recommending medical institutions that are already overloaded.

[0090] Through the above technical solution, this application solves the problem of disconnection between recommendation results and actual reception capabilities caused by missing data on the hospital side. Through a two-way data interaction mechanism, it ensures that the recommendation results not only meet the patient's personalized needs, but also truly reflect the hospital's real-time reception status. At the same time, the transparent output of the generation reason enables patients to understand the recommendation logic and enhance their trust in the recommendation results. Furthermore, through a phased processing mechanism, the complexity of data processing is effectively reduced while ensuring the accuracy of recommendations, avoiding the waste of system resources caused by directly processing all hospital data.

[0091] This application further proposes a technical solution for determining the third matching degree between the target object and the hospital based on the electronic medical record, medical reference information and hospital description information of each hospital through the target node, and determining the hospital with the third matching degree greater than or equal to the second matching degree threshold as the candidate hospital; determining the acceptance intention score based on the candidate hospital's acceptance matching degree, acceptance willingness and generation reason, determining the fourth matching degree based on the acceptance willingness and medical reference information, and finally integrating the acceptance intention score and the fourth matching degree to determine the target hospital.

[0092] Among them, the third matching degree refers to the static matching degree between the target object and the hospital calculated based on the electronic medical record, medical reference information and hospital description information. Specifically, it can be achieved by feature extraction and similarity calculation, which is used to screen out the basic candidate hospital set and solve the limitation of traditional recommendation relying on a single indicator. The intention to receive treatment score refers to the quantitative indicator of the candidate hospital's willingness to receive the target object. Specifically, the model can be used to integrate the matching degree of reception, willingness to receive treatment and the reasons for generation to dynamically reflect the hospital's reception capacity and subjective willingness. The fourth matching degree refers to the dynamic matching degree between the target object and the candidate hospital calculated based on the willingness to receive treatment and medical reference information. Specifically, the model can be used to perform secondary calibration of the needs of both parties to balance patient preferences and hospital service capabilities. The target matching score refers to the fusion result of the intention to receive treatment score and the fourth matching degree. Specifically, it can be achieved by weighted summation or feature splicing to establish a two-way selection mechanism.

[0093] Specifically, the target node first extracts features from electronic medical records and medical reference information to generate object description information. It then calculates the third match degree based on the hospital description information, screening candidate hospitals that meet the threshold. Subsequently, the candidate hospital nodes provide feedback on the acceptance match degree, willingness to accept the patient, and the reasons for the match. The target node then processes the data using a model to generate an acceptance intention score. Simultaneously, a fourth match degree is calculated based on the medical reference information and the acceptance willingness, reflecting the dynamic adaptability between patient preferences and hospital service capabilities. Ultimately, the acceptance intention score and the fourth match degree are combined to form a target match score, and the candidate hospital with the highest score is selected as the target hospital.

[0094] Compared to related technologies, which typically make one-way recommendations based solely on static hospital information, lack dynamic assessment of hospital capacity and fail to integrate the two-way matching of patient preferences and hospital willingness, this solution introduces a fusion mechanism that combines the hospital acceptance intention score with the fourth degree of matching. This approach simultaneously considers the hospital's dynamic capacity and the patient's personalized needs when screening candidate hospitals, addressing the matching bias problem caused by data heterogeneity in traditional recommendation systems.

[0095] Through the above technical solution, this application achieves the following technical effects: First, candidate hospitals are screened through the third matching degree, the fit between the hospital's static attributes and the patient's characteristics is quantified, and the screening accuracy of the basic candidate set is improved; Second, the reception intention score is calculated based on the reception matching degree, reception willingness and generation reason, dynamically reflecting the hospital's reception capacity and subjective willingness, and avoiding recommendation failures caused by temporary changes in hospital resources; Third, patient preferences and hospital service capabilities are calibrated based on the fourth matching degree to ensure that the recommendation results meet the needs of both parties at the same time, and reduce the secondary referral rate caused by information asymmetry.

[0096] The present application further proposes generating object description information of the target object based on the electronic medical record and medical reference information through the target node, determining the third matching degree between the target object and each hospital based on the object description information and the hospital description information of each hospital, inputting the reception matching degree, reception willingness and generation reason of each candidate hospital into the reception intention score determination model to obtain the reception intention score, inputting the reception willingness and medical reference information into the matching degree determination model to output the fourth matching degree, fusing the reception intention score with the fourth matching degree to determine the target matching score, and selecting the candidate hospital with the highest score as the target hospital.

[0097] Among them, object description information refers to a structured set of patient features generated by extracting and fusing features from electronic medical records and medical reference information. Specifically, this can be achieved by using deep neural networks to extract and splice multi-dimensional features, and is used to comprehensively characterize the patient's medical history, real-time needs, and preferences. The third matching degree refers to a quantitative indicator of the similarity between patient characteristics and hospital service capabilities. Specifically, this can be achieved by calculating the feature vector similarity between object description information and hospital description information using the cosine similarity algorithm, and is used to screen candidate hospitals with matching basic attributes. The admission intention score determination model refers to a computational model used to comprehensively evaluate a hospital's admission capabilities and enthusiasm. Specifically, this can be achieved by using a multi-layer perceptron to perform nonlinear weighted calculations on the matching degree, admission willingness, and generation reasons, and is used to quantify the hospital's admission priority for patients. The matching degree determination model refers to an analytical model used to assess the degree of fit between patient preferences and hospital admission willingness. Specifically, this can be achieved by using an attention mechanism to calculate the correlation between medical reference information and admission willingness, and is used to enhance the objectivity of two-way matching.

[0098] Specifically, electronic medical records and medical reference information are input into the feature extraction module, and the temporal features of diagnostic records and medication history are extracted through a convolutional neural network. At the same time, the text features of medical preferences are analyzed through a word embedding model. The feature vectors of the fused object description information and hospital description information are input into the similarity calculation layer to generate a third matching degree score. The admission data of the candidate hospital are input into the admission intention score determination model, which weights the matching degree, admission intention and reason through a fully connected layer and outputs a 0-1 range admission intention score. At the same time, the matching degree determination model allocates attention weights to medical preferences and admission intention to generate a fourth matching degree score. Finally, the two scores are fused linearly weighted, and the candidate hospital with the highest total score is selected as the recommendation target.

[0099] In some specific embodiments, the feature extraction module can use a pre-trained BERT model to encode the electronic medical record text, the hospital description information can construct a hospital service capability vector through a knowledge graph, the admission intention score determination model can integrate a random forest algorithm to process discrete generated cause data, and the matching degree determination model can use a cross-attention mechanism to capture the potential correlation between preference and intention.

[0100] Compared to related technologies, traditional methods make one-way recommendations based solely on static hospital information, failing to consider the bidirectional matching of patients' dynamic needs and hospital capacity. This solution achieves a structured representation of patient characteristics by constructing object description information. It then employs a dual matching model to quantify the degree of match between hospital service capabilities and willingness to receive care, thus breaking through the limitations of single-dimensional recommendations. By dynamically integrating the willingness to receive care score and the degree of match, a bidirectional selection decision-making mechanism is established, effectively addressing the dynamic adaptation of patient needs and medical resources.

[0101] Through the above technical solutions, this application achieves a deep match between multidimensional patient data and hospital resources, improving the accuracy of recommendations. A hospital's admissions capacity is quantified through a patient intention score model, ensuring that recommendations align with the hospital's actual admissions requirements. The matching degree determination model effectively integrates patient preferences and hospital intentions, reducing the matching failure rate caused by information asymmetry.

[0102] This application further proposes a method for generating object description information of a target object based on electronic medical records and medical reference information, including extracting features from the electronic medical records and medical reference information to obtain electronic medical record features and reference information features, fusing the two types of features and generating object description information through multiple rounds of iterative decoding; a method for determining a third degree of matching based on object description information and hospital description information, including extracting features from the two to obtain first description information features and second description information features, and determining the degree of matching by calculating feature similarity.

[0103] Among them, feature extraction refers to the extraction of representative feature vectors from the original data, which can be implemented by convolutional neural networks or recurrent neural networks to capture disease characteristics in electronic medical records and preference characteristics in medical reference information. Feature fusion refers to the integration of feature vectors from different sources into a unified representation, which can be implemented by fully connected layers or attention mechanisms to construct a complete feature portrait of the patient. Multi-round iterative decoding refers to the generation of semantically coherent text descriptions through multiple feature reconstructions, which can be implemented by Transformer decoders to solve the semantic fault problem caused by discrete feature splicing. Feature similarity refers to the measurement of the degree of matching between two features through vector space distance, which can be calculated by cosine similarity or Euclidean distance to quantify the matching degree between patient needs and hospital resources.

[0104] Specifically, the diagnostic records and medication history in the electronic medical record are converted into structured feature vectors through a feature extraction network, and the preference data in the medical reference information is converted into preference feature vectors through a natural language processing model. The two types of features are integrated through a fusion layer and input into the decoder, and object description text containing complete semantic information is iteratively generated through multiple rounds of self-attention mechanism. Hospital description information is converted into hospital feature vectors through a feature extraction network with the same structure, and similarity is calculated with the patient feature vector. Hospitals with high matching degrees are included in the candidate list. This method maps individualized patient data and hospital resource descriptions into the same feature space through deep feature fusion and semantic reconstruction, solving the problem of insufficient matching accuracy caused by differences in feature dimensions in traditional methods.

[0105] Compared to related technologies, existing medical recommendation systems typically rely on keyword matching or simple feature concatenation, which cannot effectively handle the semantic differences in heterogeneous data. Traditional methods treat patient medical records and hospital information as independent features and perform linear matching, ignoring the deeper connections between the data. This solution builds a unified feature extraction framework to achieve a deep fusion of patient subjective preferences and objective medical data. Using multiple rounds of iterative decoding to generate contextually relevant descriptive information, this approach makes the matching process between patient needs and hospital resources semantically interpretable.

[0106] All of the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be described in detail here.

[0107] Through the above technical solution, this application achieves a precise mapping of individual patient needs with medical resources, effectively improving the accuracy of out-of-town medical recommendations. This method solves the data dimension mismatch problem in traditional recommendation systems through deep feature fusion, improves the rationality of medical resource recommendations through a semantically coherent description information generation mechanism, and provides patients with hospital selection options that better meet their actual needs.

[0108] Figure 6 This is a structural diagram of a medical service improvement system based on remote medical treatment data analysis provided by the embodiment of the present application, see Figure 6 , the system includes:

[0109] The first determining module 601 is configured to determine the target subject's current physiological state, medical reference information, and reason for seeking medical treatment in a different place in response to the target subject's request for seeking medical treatment in a different place.

[0110] The second determination module 602 is used to determine a target node that matches the current physiological state and the reason for out-of-town medical treatment from multiple candidate nodes associated with the medical destination based on the current physiological state of the target object and the reason for out-of-town medical treatment. Different candidate nodes have different computing powers and hospital recommendation modes. The medical destination is obtained from the out-of-town medical treatment request.

[0111] The third determination module 603 is used to input the target subject's electronic medical record, the medical reference information and the reason for the medical treatment in a different place into the target node, and determine the target hospital recommended to the target subject from multiple candidate hospitals at the medical destination through the target node.

[0112] It should be noted that: the medical service improvement system based on remote medical treatment data analysis provided in the above embodiment only uses the division of the above functional modules as an example to illustrate when improving medical services. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the medical service improvement system based on remote medical treatment data analysis provided in the above embodiment and the medical service improvement method embodiment based on remote medical treatment data analysis belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0113] Figure 7 This is a schematic diagram of the structure of a server provided in an embodiment of the present application. The server 700 may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 701 and one or more memories 702, wherein the one or more memories 702 store at least one computer program, and the at least one computer program is loaded and executed by the one or more processors 701 to implement the methods provided in the above-mentioned various method embodiments. Of course, the server 700 may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The server 700 may also include other components for implementing device functions, which will not be described in detail here.

[0114] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including a computer program. The computer program can be executed by a processor to implement the medical service improvement method based on remote medical treatment data analysis in the above embodiment. For example, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0115] In an exemplary embodiment, a computer program product or computer program is also provided, which includes a program code, which is stored in a computer-readable storage medium. The processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device executes the above-mentioned medical service improvement method based on the analysis of medical treatment data in a different place.

[0116] In some embodiments, the computer program involved in the embodiments of the present application may be deployed and executed on a computer device, or on multiple computer devices located at one location, or on multiple computer devices distributed at multiple locations and interconnected through a communication network. Multiple computer devices distributed at multiple locations and interconnected through a communication network may constitute a blockchain system.

[0117] Those skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware, or may be accomplished by a program to instruct the relevant hardware, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0118] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for improving medical services based on data analysis of medical treatment in other places, characterized in that: The method comprises: In response to a target subject's request for medical treatment in a different place, determining the target subject's current physiological state, medical reference information, and reason for medical treatment in a different place; Based on the current physiological state of the target subject and the reason for seeking medical treatment in a different place, determining a target node that matches the physiological state and the reason for seeking medical treatment in a different place from multiple candidate nodes associated with the medical treatment destination, where different candidate nodes have different computing power and hospital recommendation modes, and the medical treatment destination is obtained from the medical treatment request in a different place; The electronic medical record of the target subject, the medical reference information and the reason for the medical treatment in a different place are input into the target node, and the target hospital recommended to the target subject is determined from a plurality of candidate hospitals at the medical destination through the target node.

2. The method according to claim 1, characterized in that The medical reference information includes medical preferences and medical needs, the reasons for medical treatment in a different place include subjective reasons for medical treatment in a different place and reference reasons for medical treatment in a different place, and responding to the target subject's request for medical treatment in a different place, determining the target subject's physiological state, medical reference information, and reasons for medical treatment in a different place, includes: In response to a target subject's request for medical treatment in a different place, obtaining the target subject's physiological data, the subjective reason for medical treatment in a different place, historical medical treatment data, and medical treatment feedback data, wherein the medical treatment feedback data is used to represent the hospital's evaluation of the target subject; determining a physiological state of the target subject based on the physiological data; Determining the medical preference based on the historical medical data; Based on the historical medical treatment data and the medical treatment feedback data, the reference reason for seeking medical treatment in a different place is determined.

3. The method according to claim 2, characterized in that The historical medical treatment data includes historical medical treatment hospitals and historical medical treatment times. The determining of the medical treatment preference based on the historical medical treatment data includes: Obtaining first hospital information of the historical medical hospital, the first hospital information being used to introduce the historical medical hospital; determining a basic medical time pattern of the target subject based on the historical medical time; determining a reference medical time pattern of the target subject based on the first hospital information and the historical medical time; and determining the medical preference based on the first hospital information, the basic medical time pattern, and the reference medical time pattern; The determining of the reference reason for seeking medical treatment in a different place based on the historical medical treatment data and the medical treatment feedback data includes: Obtain the second hospital information of the historical hospital, where the second hospital information is used to describe the medical treatment situation of the target object in the historical hospital; based on the second hospital information and the medical feedback data, determine the reference reason for medical treatment in a different place.

4. The method according to claim 3, characterized in that The determining the medical treatment preference based on the first hospital information, the basic medical treatment time pattern, and the reference medical treatment time pattern includes: Determine the target subject's hospital preference and department preference based on the first hospital information; determine the target subject's medical time preference based on the basic medical time pattern and the reference medical time pattern; and concatenate the target subject's hospital preference, department preference, and medical time preference to obtain the medical preference; The second hospital information includes the number of medical payment times, the amount of medical payment, the medical travel distance, and medical treatment effect information. The determination of the reference reason for out-of-town medical treatment based on the second hospital information and the medical treatment feedback data includes: Based on the number of medical payment times and the medical travel distance, the medical fatigue description information is determined; based on the medical payment amount and the medical effect information, the medical effect description information is determined; based on the medical fatigue description information, the medical effect description information and the medical feedback data, the reference reason for medical treatment in a different place is determined.

5. The method according to claim 1, wherein The determining, based on the physiological state of the target subject and the reason for seeking medical treatment in a different place, a target node that matches the physiological state and the reason for seeking medical treatment in a different place from a plurality of candidate nodes associated with the medical destination, includes: Based on the physiological state, determining a plurality of reference nodes from the plurality of candidate nodes, the plurality of reference nodes being candidate nodes matching the physiological state; Based on the reason for seeking medical treatment in a different place, the target node is determined from the multiple reference nodes.

6. The method according to claim 5, characterized in that The determining, based on the physiological state, a plurality of reference nodes from the plurality of candidate nodes, the plurality of reference nodes being candidate nodes matching the physiological state, includes: determining first node description information for each candidate node, the first node description information being used to describe an associated physiological state of the corresponding candidate node; determining a first degree of match between the physiological state and each candidate node based on the physiological state and the first node description information of each candidate node; determining a candidate node among the multiple candidate nodes whose first degree of match with the physiological state is greater than or equal to a first matching degree threshold as a reference node, thereby obtaining the multiple reference nodes; The determining the target node from the multiple reference nodes based on the reason for seeking medical treatment in a different place includes: Determine the second node description information of each of the reference nodes, where the second node description information is used to describe the associated medical reasons of the reference nodes; determine the second matching degree between the reasons for medical treatment in a different place and each of the reference nodes based on the reasons for medical treatment in a different place; and determine the reference node with the highest second matching degree between the reasons for medical treatment in a different place among the multiple reference nodes as the target node.

7. The method according to claim 1, characterized in that The determining, through the target node, a target hospital recommended to the target subject from the multiple hospitals at the medical destination includes: Determining, through the target node, a plurality of candidate hospitals from the plurality of hospitals based on the electronic medical records, the medical reference information, and the hospital description information of each of the hospitals; Sending the target patient's electronic medical record, the medical reference information, and the reason for the out-of-town medical treatment to the hospital node of each candidate hospital, so that the hospital node of each candidate hospital determines the degree of match, willingness to receive the treatment, and the reasons for generating the degree of match and willingness to receive the treatment based on the target patient's electronic medical record, the medical reference information, and the reason for the out-of-town medical treatment; Obtaining the reception matching degree, reception willingness, and reasons for generating the reception matching degree and reception willingness of each candidate hospital through the target node; The target hospital is determined from the plurality of candidate hospitals through the target node based on the reception matching degree, the reception willingness, the generation reason and the medical reference information of each candidate hospital.

8. The method according to claim 7, characterized in that Determining, through the target node, a plurality of candidate hospitals from the plurality of hospitals based on the electronic medical record, the medical reference information, and the hospital description information of each of the hospitals includes: determining, through the target node, a third degree of match between the target object and each of the hospitals based on the electronic medical record, the medical reference information, and the hospital description information of each of the hospitals; determining a hospital among the multiple hospitals whose third degree of match is greater than or equal to a second degree of match threshold as a candidate hospital, thereby obtaining the multiple candidate hospitals; Determining the target hospital from the plurality of candidate hospitals through the target node based on the acceptance matching degree, the acceptance willingness, the generation reason, and the medical reference information of each candidate hospital includes: Through the target node, based on the reception matching degree, the reception willingness and the generation reason of each candidate hospital, the reception intention score of each candidate hospital for the target object is determined; based on the reception willingness and the medical reference information of each candidate hospital, the fourth matching degree between the target object and each candidate hospital is determined; based on the reception intention score and the fourth matching degree, the target hospital is determined from the multiple candidate hospitals.

9. The method according to claim 8, characterized in that Determining, by the target node, a third matching degree between the target object and each of the hospitals based on the electronic medical record, the medical reference information, and the hospital description information of each of the hospitals includes: generating, through the target node, object description information of the target object based on the electronic medical record and the medical reference information; and determining a third matching degree between the target object and each of the hospitals based on the object description information of the target object and the hospital description information of each of the hospitals; Determining, through the target node, the acceptance intention score of each candidate hospital for the target subject based on the acceptance matching degree, the acceptance willingness, and the generation reason of each candidate hospital, includes: Inputting the acceptance matching degree, the acceptance willingness and the generating reason of each candidate hospital into the acceptance intention score determination model, processing the acceptance matching degree, the acceptance willingness and the generating reason through the acceptance intention score determination model to obtain the acceptance intention score of each candidate hospital for the target patient; The determining, based on the acceptance willingness of each candidate hospital and the medical reference information, a fourth matching degree between the target object and each candidate hospital includes: Inputting the acceptance willingness and the medical reference information of each candidate hospital into a matching degree determination model, processing the acceptance willingness and the medical reference information through the matching degree determination model, and outputting a fourth matching degree between the target object and each candidate hospital; Determining the target hospital from the plurality of candidate hospitals based on the patient acceptance intention score and the fourth matching degree includes: The patient acceptance intention score and the fourth matching degree are integrated to obtain a target matching score for each candidate hospital; and the candidate hospital with the highest target matching score among the multiple candidate hospitals is determined as the target hospital.

10. A medical service improvement system based on remote medical treatment data analysis, characterized in that: The system comprises: A first determining module is configured to determine the target subject's current physiological state, medical reference information, and reason for seeking medical treatment in a different place in response to the target subject's request for seeking medical treatment in a different place; A second determination module is configured to determine, based on the current physiological state of the target subject and the reason for the remote medical treatment, a target node that matches the physiological state and the reason for the remote medical treatment from a plurality of candidate nodes associated with the medical treatment destination, wherein different candidate nodes have different computing powers and hospital recommendation modes, and the medical treatment destination is obtained from the remote medical treatment request; The third determination module is used to input the target object's electronic medical record, the medical reference information and the reason for the out-of-town medical treatment into the target node, and determine the target hospital recommended to the target object from multiple candidate hospitals at the medical destination through the target node.

Images