Hospital emergency triage method combined with medical history data

Abstract

The invention discloses a hospital emergency triage method in combination with medical history data, and belongs to the technical field of medical triage, and the method specifically comprises the steps: obtaining the actual illness state and historical treatment data of a current patient to form a pre-examination data set, inputting an illness state prediction model to obtain a pathological entity and the confidence probability thereof, and preferably evaluating a dominant illness in combination with a danger coefficient; obtaining a target department corresponding to the dominant disease, calculating the coverage overlap ratio of the non-dominant disease and the treatment capacity of the target department, if the overlap ratio is smaller than or equal to a preset threshold value, obtaining the department corresponding to the non-dominant disease, marking the department as a candidate department, and extracting the treatment data of the department to calculate a waiting pressure coefficient; calling cross-department cooperation data to calculate candidate department response efficiency coefficients; the comprehensive turnover efficiency scores corresponding to the candidate departments are calculated, and the candidate department with the highest score is selected as the first diagnosis department, so that the patient turnover link minimization and the medical resource utilization optimization are realized on the premise that the dominant threat is preferentially treated, and the emergency treatment efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of medical triage technology, specifically to a hospital emergency triage method that combines medical history data. Background Technology

[0002] Emergency departments need to triage patients based on the severity of their condition. Due to the large number of emergency patients, medical staff need to conduct pre-diagnosis of registered patients to triage them, distinguish the severity and urgency of their illnesses and the relevant specialty, and promptly arrange treatment procedures and guide them to the appropriate specialist, so that emergency patients can receive treatment as quickly as possible.

[0003] In current emergency triage technology, the department corresponding to the candidate condition with the highest risk and confidence level is usually selected as the triage department. This method does not take into account the real-time resource status of the triage department. When faced with dynamic load situations such as real-time fluctuations in the number of patients waiting in the department, dynamic changes in the occupancy rate of medical equipment, and phased adjustments in the number of medical staff on duty, it is difficult to achieve scientific and reasonable resource allocation.

[0004] When departments with high matching rates are experiencing peak patient volume, patient waiting times may exceed the safe threshold for disease progression, leading to mild cases turning into severe cases or even missing the optimal treatment window. Simultaneously, the lack of comprehensive evaluation of departmental processing efficiency and resource idleness can easily result in redundant equipment checks and wasted medical resources. This makes it impossible to minimize patient turnover while prioritizing the treatment of primary illnesses, and also makes it difficult to balance real-time departmental workload with efficient utilization of medical resources. Therefore, there is an urgent need for a method that can integrate dynamic resource data such as real-time departmental waiting data, equipment status information, and medical staff allocation to optimize triage decisions. Summary of the Invention

[0005] The purpose of this invention is to provide a hospital emergency triage method that combines medical history data, and to solve the following technical problems: Current emergency triage technology typically selects the department corresponding to the candidate condition with the highest risk and confidence level as the triage department, without considering the real-time resource status of the department. This leads to unreasonable resource allocation under dynamic load, increased patient risk, and waste of medical resources. There is an urgent need to integrate dynamic resource data to optimize triage decisions.

[0006] The objective of this invention can be achieved through the following technical solutions: A hospital emergency triage method that combines medical history data includes the following steps: S1, obtain the current patient's actual condition data and historical medical records to obtain the pre-screening dataset; the actual condition data includes condition description data and physiological characteristic data; S2, input the pre-examination dataset into the preset disease prediction model, obtain all pathological entities of the current patient and their confidence probabilities, and combine the risk coefficients corresponding to each pathological entity for optimization evaluation, select the pathological entity corresponding to the highest optimization score as the dominant disease, and label the remaining pathological entities as non-dominant diseases. S3, obtain the department corresponding to the dominant disease and mark it as the target department, calculate the coverage overlap between the set of non-dominant diseases and the treatment capacity of the target department, if the coverage overlap is less than or equal to the preset overlap threshold, then mark all departments corresponding to the non-dominant diseases as candidate departments, extract the current consultation data of each candidate department and calculate the waiting pressure coefficient of each candidate department, wherein the consultation data includes the average treatment rate, the number of patients waiting for consultation and the risk coefficient corresponding to each patient; S4. Obtain the collaboration data between the target department and any waiting department and calculate the response efficiency coefficient of the candidate department. The collaboration data includes physical path distance, historical average response delay and consultation success rate. S5. Calculate the comprehensive performance score of any candidate department based on the waiting pressure coefficient and response efficiency coefficient, and select the candidate department with the highest comprehensive performance score as the first-visit department and send it to the current patient.

[0007] As a further aspect of the present invention, the specific construction process of the disease prediction model is as follows: Acquire a number of historical medical records and divide them into training and testing sets according to a certain ratio. Each historical medical record is a dataset of a certain emergency patient, and the dataset includes disease description data and physiological characteristic data. Generate a backpropagation network model and set training parameters. Normalize the training set data and train the backpropagation network model using the training set to obtain a preliminary disease prediction model. Use the testing set to retrain the preliminary disease prediction model to obtain the final disease prediction model.

[0008] As a further aspect of the present invention: in step S2, the preferred evaluation process is as follows: A multidimensional risk coefficient assessment matrix is ​​constructed, which includes two orthogonal dimensions: timeliness coefficient and severity coefficient. The confidence probability value is weighted and fused with the timeliness coefficient and severity coefficient respectively to obtain the optimal score. The pathological entity corresponding to the highest optimal score is selected as the dominant disease.

[0009] As a further aspect of the present invention: the specific calculation process for the coverage overlap in S3 is as follows: A disease feature vector space is constructed based on a medical knowledge graph. Each disease in the non-dominant disease set is parsed into standard medical terminology codes and mapped to vector coordinates. At the same time, the historical electronic medical record database of the target department is obtained, and the disease spectrum and corresponding efficacy indicators are extracted through natural language processing to generate a multi-dimensional feature vector of the department's treatment capabilities. The cosine similarity algorithm is used to calculate the directional consistency between each disease feature vector and the department capability vector. After weighting the disease acuteness coefficient and the consequence severity coefficient, a single disease fitting value is obtained. All disease fitting values ​​in the set are normalized and the arithmetic mean is obtained. The arithmetic mean is labeled as the coverage overlap degree.

[0010] As a further aspect of the present invention: In S3, the specific calculation process of the waiting pressure coefficient is as follows: real-time acquisition of the waiting patient queue of the candidate department, extraction of the pathological entity and its risk coefficient corresponding to the waiting patient, screening out all waiting patients whose risk coefficient is greater than or equal to the current patient's risk coefficient and using them as the effective pressure load base; synchronously extracting the actual number of patients treated in the department within a preset time period as the dynamic treatment rate benchmark value; dividing the effective pressure load base by the dynamic treatment rate benchmark value to obtain the basic pressure ratio, synchronizing the real-time status database of the medical equipment of the candidate department, calculating the resource attenuation coefficient based on the degree of equipment shortage; calculating the product value of the basic pressure ratio and the resource attenuation coefficient, and labeling the product value as the waiting pressure coefficient.

[0011] As a further aspect of the present invention: the specific calculation process of the response performance coefficient in S4 is as follows: Obtain the measured physical path distance between the target department and the candidate department, and divide it by a preset reference distance to obtain the distance attenuation factor; simultaneously extract the historical average response delay and divide it by a preset reference response delay to obtain the delay attenuation factor; then calculate according to the formula... The response efficiency coefficient X is calculated, where P is the historical consultation success rate, D is the physical path distance, T is the historical average response delay, ε is the preset unit coefficient, Dref is the preset reference distance, and Tref is the preset reference response delay.

[0012] As a further aspect of the present invention: S5 further includes obtaining the comprehensive turnover efficiency score of the target department. If the comprehensive turnover efficiency scores of all candidate departments are less than or equal to the comprehensive turnover efficiency score of the target department, then the target department is selected as the first-visit department and sent to the current patient.

[0013] As a further aspect of the present invention: S3 further includes, if the coverage overlap is greater than a preset overlap threshold, then without repeating the subsequent steps, the target department is directly designated as the first-visit department and sent to the current patient.

[0014] The beneficial effects of this invention are: 1) During triage, this invention acquires current patient data such as the average treatment rate and the number of patients waiting in candidate departments to calculate the waiting pressure coefficient. It also retrieves historical cross-departmental collaboration databases, analyzes collaborative data such as physical path distances to calculate the response efficiency coefficient, and then generates a comprehensive turnover efficiency score through matrix weighted fusion operations to select the primary department. By considering dynamic load conditions such as real-time fluctuations in the number of patients waiting in departments, this invention overcomes the shortcomings of traditional methods that do not consider the real-time resource status of triage departments, achieving a scientific and rational allocation of resources.

[0015] 2) This invention determines the target department's ability to handle non-dominant diseases by calculating the coverage overlap. Other candidate departments will only be considered when the coverage overlap is less than or equal to a preset threshold. At the same time, the waiting pressure coefficient and response efficiency coefficient are combined to comprehensively evaluate the candidate departments. This avoids the situation where patients' waiting time exceeds the safe threshold for disease progression when the high-matching department is in peak hours. This reduces the risk of mild cases turning into severe cases or even missing the best treatment opportunity, and better protects the treatment effect of patients.

[0016] 3) This invention not only considers the primary disease, but also evaluates the departments corresponding to non-primary diseases. It selects the primary department by comprehensively evaluating the turnover efficiency score, avoiding problems such as repeated equipment examinations and waste of medical resources caused by not comprehensively evaluating indicators such as department treatment efficiency. It can optimize the patient turnover process while ensuring the priority treatment of the primary disease, and take into account dynamic resource data such as real-time consultation data of departments, thereby improving the utilization efficiency of medical resources. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a schematic diagram of a hospital emergency triage method that combines medical history data according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, this invention is a hospital emergency triage method that combines medical history data, including the following steps: S1, obtain the current patient's actual condition data and historical medical records to obtain the pre-screening dataset; the actual condition data includes condition description data and physiological characteristic data; S2, input the pre-examination dataset into the preset disease prediction model, obtain all pathological entities of the current patient and their confidence probabilities, and combine the risk coefficients corresponding to each pathological entity for optimization evaluation, select the pathological entity corresponding to the highest optimization score as the dominant disease, and label the remaining pathological entities as non-dominant diseases. S3, obtain the department corresponding to the dominant disease and mark it as the target department, calculate the coverage overlap between the set of non-dominant diseases and the treatment capacity of the target department, if the coverage overlap is less than or equal to the preset overlap threshold, then mark all departments corresponding to the non-dominant diseases as candidate departments, extract the current consultation data of each candidate department and calculate the waiting pressure coefficient of each candidate department, wherein the consultation data includes the average treatment rate, the number of patients waiting for consultation and the risk coefficient corresponding to each patient; S4. Obtain the collaboration data between the target department and any waiting department and calculate the response efficiency coefficient of the candidate department. The collaboration data includes physical path distance, historical average response delay and consultation success rate. S5. Calculate the comprehensive performance score of any candidate department based on the waiting pressure coefficient and response efficiency coefficient, and select the candidate department with the highest comprehensive performance score as the first-visit department and send it to the current patient.

[0021] 1) This invention, during triage, acquires current patient data such as the average treatment rate of candidate departments and the number of patients waiting to be seen to calculate the waiting pressure coefficient. It also retrieves historical cross-departmental collaboration databases, analyzes collaborative data such as physical path distances to calculate the response efficiency coefficient, and then generates a comprehensive turnover efficiency score through matrix weighted fusion calculations to select the primary department. This approach fully considers dynamic load conditions such as real-time fluctuations in the number of patients waiting in departments, overcoming the shortcomings of traditional methods that do not consider the real-time resource status of triage departments, and enabling scientific and rational resource allocation.

[0022] 2) This invention assesses a target department's ability to handle non-dominant conditions by calculating coverage overlap. Other candidate departments are considered only when the coverage overlap is less than or equal to a preset threshold. Simultaneously, it combines waiting pressure coefficients and response efficiency coefficients to comprehensively evaluate candidate departments. This avoids situations where patients' waiting times exceed safe thresholds for disease progression when highly matched departments are in peak hours, thereby reducing the risk of mild cases developing into severe cases or even missing the optimal treatment window, and better ensuring the patient's treatment outcome.

[0023] 3) This invention not only considers the primary disease, but also evaluates the departments corresponding to non-primary diseases. It selects the primary department by comprehensively evaluating the turnover efficiency score, avoiding problems such as repeated equipment examinations and waste of medical resources caused by not comprehensively evaluating indicators such as department treatment efficiency. It can optimize the patient turnover process while ensuring the priority treatment of the primary disease, and take into account dynamic resource data such as real-time consultation data of departments, thereby improving the utilization efficiency of medical resources.

[0024] It is worth noting that the specific formula for the comprehensive turnover efficiency score is as follows: Where b and d are preset coefficients, Y is the waiting pressure coefficient, and X is the response efficiency coefficient.

[0025] In a preferred embodiment, the specific construction process of the disease prediction model in step S2 is as follows: Acquire a number of historical medical records and divide them into training and testing sets according to a certain ratio. Each historical medical record is a dataset of a certain emergency patient, and the dataset includes disease description data and physiological characteristic data. Generate a backpropagation network model and set training parameters. Normalize the training set data and train the backpropagation network model using the training set to obtain a preliminary disease prediction model. Use the testing set to retrain the preliminary disease prediction model to obtain the final disease prediction model.

[0026] First, historical medical records of several emergency room patients were collected, including records from different patients over the past two years. Each patient's data included a description of their symptoms, such as "sudden chest pain and cough," as well as physiological characteristics, such as a body temperature of 38.5℃ and blood pressure of 140 / 90 mmHg. These data were then divided into training and testing sets in a 7:3 ratio. This division is used to allow the model to learn patterns, while the testing set is used to verify the model's learning performance. Next, a backpropagation network model was generated. This model acts like a machine that continuously adjusts its internal connection weights to learn. Training parameters, such as the learning rate and the number of iterations, were then set, determining the model's learning speed and extent. The training set data was then normalized, for example, scaling data from different ranges, such as blood pressure and body temperature, to between 0 and 1. This is because different data have different dimensions, and normalization allows the model to learn better. After this processing, the training set was used to train the backpropagation network model. During training, the model continuously adjusts its internal weights based on the input medical data and corresponding symptom labels, much like a student learning knowledge through practice problems, thus obtaining a preliminary disease prediction model. Finally, the initial model is retrained using the test set. During this process, the model's prediction results are observed, and the parameters are continuously adjusted to make the model more accurate until the final disease prediction model is obtained. This is done because the backpropagation network model can optimize the weights by continuously backpropagating the error, thereby improving the accuracy of the prediction.

[0027] Construct an accurate disease prediction model. Collecting historical data and dividing it into training and testing sets allows the model to learn patterns from emergency patient data during training, and verifies its learning during testing. Generating a backpropagation network model and setting its parameters leverages its ability to optimize weights through error backpropagation, allowing the model to continuously improve. Data normalization eliminates the influence of units, making model learning more efficient. Repeated training on the training and testing sets allows the model to continuously adjust and improve its predictive ability. The final accurate model can obtain the patient's pathological characteristics and confidence probabilities, providing a basis for determining the dominant disease, thus making triage more scientific and achieving the goal of optimizing emergency resource allocation and improving treatment efficiency.

[0028] In another preferred embodiment, the evaluation process in step S2 is preferably as follows: A multidimensional risk coefficient assessment matrix is ​​constructed, which includes two orthogonal dimensions: timeliness coefficient and severity coefficient. The confidence probability value is weighted and fused with the timeliness coefficient and severity coefficient respectively to obtain the optimal score. The pathological entity corresponding to the highest optimal score is selected as the dominant disease.

[0029] First, a multidimensional risk assessment matrix is ​​constructed. This matrix uses timeliness and severity as two independent dimensions, much like using the x-axis to represent the timeliness coefficient (e.g., the urgency of the condition requiring immediate treatment; for example, myocardial infarction has a high timeliness coefficient because delayed treatment can quickly endanger life) and the y-axis to represent the severity coefficient (e.g., the degree of harm caused to the patient's body; for example, extensive burns have a higher severity coefficient than the common cold). Then, for each pathological entity output by the disease prediction model and its corresponding confidence probability value, a weighted fusion operation is performed with the coefficients of these two dimensions. For example, if a pathological entity has a confidence probability of 0.8, a timeliness coefficient of 0.9, and a severity coefficient of 0.7, the weighted fusion can be understood as adjusting the confidence probability based on these two coefficients, highlighting the importance of different risk dimensions. This is like multiplying the confidence probability by different "weights" to reflect the combined impact of the condition's urgency and severity, thus obtaining an optimal score. This approach is taken because the confidence probability reflects the reliability of the model's judgment on the pathological entity, while the timeliness and severity coefficients supplement it from the perspective of actual clinical risk. Combining the two provides a more comprehensive assessment of the disease's risk. Finally, the pathological entity with the highest priority score is selected as the leading disease because this score comprehensively considers both the reliability of the model's prediction and the actual risk level of the disease, enabling a more reasonable determination of the diseases requiring priority treatment.

[0030] By constructing an orthogonal dimension matrix incorporating timeliness and severity, the risk level of pathological entities can be quantitatively analyzed from two key perspectives: time urgency and severity of harm, avoiding judgment based solely on a single indicator like confidence probability. Weighted fusion of confidence probability and two risk coefficients considers both the accuracy of model predictions and the risk characteristics of diseases in clinical practice, making the assessment results more aligned with the actual needs of emergency triage. The dominant disease identified in this way more accurately reflects the urgency and severity of the patient's condition, providing a more reliable basis for subsequent matching of target departments and comprehensive evaluation of other candidate departments. This ensures that emergency resources are prioritized for patients who need them most, improving the scientific and rational nature of triage, avoiding triage bias caused by single-indicator assessments, and enhancing patient treatment efficiency.

[0031] In another preferred embodiment, the specific calculation process for the coverage overlap in step S3 is as follows: A disease feature vector space is constructed based on a medical knowledge graph. Each disease in the non-dominant disease set is parsed into standard medical terminology codes and mapped to vector coordinates. At the same time, the historical electronic medical record database of the target department is obtained, and the disease spectrum and corresponding efficacy indicators are extracted through natural language processing to generate a multi-dimensional feature vector of the department's treatment capabilities. The cosine similarity algorithm is used to calculate the directional consistency between each disease feature vector and the department capability vector. After weighting the disease acuteness coefficient and the consequence severity coefficient, a single disease fitting value is obtained. All disease fitting values ​​in the set are normalized and the arithmetic mean is obtained. The arithmetic mean is labeled as the coverage overlap degree.

[0032] The specific calculation process for coverage overlap is as follows:

[0033] in, Let be the feature vector of the i-th non-dominant disease in the medical knowledge graph. The feature vectors of the dominant diseases in the medical knowledge graph. Let be the timeliness coefficient corresponding to the i-th non-dominant symptom. Let be the severity coefficient corresponding to the i-th non-dominant symptom, n be the total number of non-dominant symptom groups, and μ be the preset unit coefficient.

[0034] First, a disease feature vector space is constructed based on a medical knowledge graph. The medical knowledge graph is like a vast database containing various medical concepts and their relationships, such as recording the association between "pneumonia" and symptoms like "cough, fever, and lung shadows." Constructing the vector space is equivalent to defining a series of feature dimensions for each disease within this space, such as symptom type and location of the disease. Then, diseases in the non-dominant disease set, such as hypertension and diabetes coexisting in a patient, are parsed into standard medical terminology codes, like the International Classification of Diseases (ICD) code, thus giving each disease a unique standardized identifier. Next, these codes are mapped to vector coordinates. For example, in three-dimensional space, different coordinate axes represent different feature dimensions. Each disease determines its specific coordinates based on its features' performance in each dimension. This is like transforming each disease into a point in space, using coordinates to represent its characteristics. Simultaneously, the target department's historical electronic medical record database is loaded, such as all past patient medical records from the cardiology department. Natural language processing technology is used to extract the spectrum of diseases frequently treated by the department from these records, such as coronary heart disease and myocardial infarction, along with corresponding efficacy indicators, such as cure rate and improvement rate. This information is then integrated to generate a multi-dimensional feature vector representing the department's treatment capabilities. This vector acts as a "digital portrait" of the department's capabilities, with each dimension representing the department's level of ability in a particular type of disease or indicator. Next, a cosine similarity algorithm is used to calculate the directional consistency between each disease feature vector and the department's capability vector. The principle of the cosine similarity algorithm is to determine the degree of similarity by calculating the cosine value of the angle between two vectors. The smaller the angle, the closer the cosine value is to 1, indicating that the two vectors are more consistent in direction, meaning a better match between the disease features and the department's capabilities. After obtaining the similarity results, the acuteness coefficient and the severity coefficient of the consequences are weighted. For example, a disease with high acuteness and severe consequences is given a larger weight, thus highlighting the suitability of urgent and severe diseases, thereby obtaining a single disease suitability value. Finally, the fitting values ​​of all diseases in the set are normalized to the same range, such as between 0 and 1, for easy comparison. Then, their arithmetic mean is calculated. This mean is the coverage overlap, which reflects the degree of matching of the target department's overall ability to handle the set of non-dominant diseases.

[0035] By constructing a disease feature vector space based on a medical knowledge graph and mapping it to disease codes, complex disease information can be transformed into structured vector data, facilitating computer processing and analysis. Loading the target department's historical electronic medical records and extracting its treatment capability feature vectors is equivalent to summarizing the department's capability characteristics from actual clinical data, making the assessment more objective and realistic. Using a cosine similarity algorithm combined with the severity of the disease's acuteness and consequences to calculate the fit value considers both the degree of matching between the disease and the department's capabilities and the actual risk level of the disease, making the assessment results more in line with the actual needs of emergency triage. Calculating the coverage overlap quantifies the target department's overall coverage capability for non-dominant diseases. When the coverage overlap is less than a preset threshold, it indicates that the target department's ability to handle these diseases is limited, requiring further consideration of other candidate departments. This avoids the problem of incomplete patient treatment due to the target department's inability to effectively handle non-dominant diseases. In this way, all of the patient's conditions can be considered more comprehensively during the triage process, ensuring that the selected department can not only handle the primary symptoms, but also take into account non-primary symptoms to a certain extent. This improves the rationality and effectiveness of triage, provides patients with more comprehensive treatment plans, and ultimately optimizes the allocation of emergency resources and improves the treatment effect for patients.

[0036] In another preferred embodiment, the specific calculation process of the waiting pressure coefficient in S3 is as follows: The queue of patients waiting to be seen in the candidate department is acquired in real time; the pathological entities and their risk coefficients corresponding to the patients are extracted; all patients with risk coefficients greater than or equal to the current patient's risk coefficient are selected and used as the effective pressure load base; the actual number of patients treated in the department within a preset time period is extracted simultaneously as the dynamic treatment rate benchmark value; the effective pressure load base is divided by the dynamic treatment rate benchmark value to obtain the basic pressure ratio; the real-time status database of the medical equipment in the candidate department is synchronized, and the resource attenuation coefficient is calculated based on the degree of equipment shortage; the product of the basic pressure ratio and the resource attenuation coefficient is calculated, and the product value is calibrated as the waiting pressure coefficient.

[0037] The specific formula for calculating the waiting pressure coefficient is as follows: ; Where Y is the waiting pressure coefficient, R is the effective pressurization load base, V is the dynamic treatment rate benchmark value, K is the resource attenuation coefficient, and β is the preset unit coefficient.

[0038] First, the patient queues for candidate departments are acquired in real time, such as the current patient list for the emergency internal medicine department. Then, the pathological entity and its risk coefficient for each patient are extracted from this queue. The pathological entity can be a diagnosis like "acute gastritis" or "arrhythmia," while the risk coefficient is a value determined by the urgency and severity of the condition; for example, the risk coefficient for acute myocardial infarction is much higher than that for the common cold. Next, all patients with a risk coefficient greater than or equal to the current patient's risk coefficient are selected as the effective pressure load base. This is because only those with a risk level no lower than the current patient will actually put pressure on the current patient's waiting list. For example, if the current patient's risk coefficient is 8, then patients with a risk coefficient of 8 or higher will increase the current patient's waiting pressure. Simultaneously, the actual number of patients treated within a preset time period in the department is extracted, such as the number of patients actually treated in the past hour. This number is used as the dynamic treatment rate benchmark, reflecting the department's actual work efficiency within a certain time. The time period can be selected based on the department's daily work rhythm; for example, half an hour during busy periods and one hour during off-peak periods. Then, the effective pressurized load base is divided by the dynamic treatment rate benchmark to obtain the baseline pressure ratio. This ratio is calculated by dividing the number of patients requiring priority treatment by the department's treatment efficiency, thus measuring the department's workload under current conditions. For example, if the effective pressurized load base is 5 patients and the dynamic treatment rate benchmark is 2 patients per hour, the baseline pressure ratio is 2.5, meaning it would take 2.5 hours to process these more urgent patients at the current efficiency. Next, the real-time status database of the candidate department's medical equipment is synchronized to obtain equipment usage information. A resource attenuation coefficient is calculated based on the degree of equipment shortage. For example, when half of a critical piece of equipment, such as an electrocardiograph, is out of service, the equipment shortage is high, and the resource attenuation coefficient will decrease accordingly. This is because equipment shortages affect the department's treatment capacity, effectively reducing its work efficiency. Finally, the product of the baseline pressure ratio and the resource attenuation coefficient is calculated, and this product is designated as the waiting pressure coefficient. This comprehensively considers the pressure from waiting patients and the impact of equipment resources, resulting in a value that reflects the actual waiting pressure of the department.

[0039] By acquiring real-time data on the patient queue and selecting patients with higher risk factors as the effective pressure load baseline, the system accurately reflects the actual waiting load on current patients within the department, avoiding the overestimation of pressure by including patients with milder conditions. The actual number of patients treated within a preset time period is extracted as a dynamic treatment rate baseline, taking into account potential efficiency fluctuations in different time periods, making the assessment more realistic. The baseline pressure ratio is obtained by dividing the effective pressure load baseline by the treatment rate, quantifying waiting pressure from the perspectives of patient load and departmental efficiency. The resource attenuation coefficient is calculated by incorporating the real-time status of medical equipment because equipment conditions directly affect the department's treatment capacity; equipment shortages lead to a decrease in actual departmental efficiency, and this factor must be included in the pressure assessment. The final waiting pressure coefficient integrates multiple key factors such as patient load, departmental efficiency, and equipment resources, comprehensively and accurately reflecting the real-time workload of candidate departments. This is crucial for triage decisions because only by accurately understanding the waiting pressure of each department can we avoid assigning patients to departments with excessive pressure, leading to excessively long waiting times and affecting treatment outcomes. At the same time, it can also avoid wasting resources, optimize the allocation of emergency resources, and ensure that patients are assigned to departments with relatively low waiting pressure and appropriate treatment capabilities. This improves the efficiency and scientific nature of emergency triage, better guarantees the quality of patient treatment, optimizes emergency triage, and improves the efficiency of medical resource utilization.

[0040] In another preferred embodiment, the specific calculation process of the response performance coefficient in step S4 is as follows: Obtain the measured physical path distance between the target department and the candidate department, and divide it by a preset reference distance to obtain the distance attenuation factor; simultaneously extract the historical average response delay and divide it by a preset reference response delay to obtain the delay attenuation factor; then calculate according to the formula... The response efficiency coefficient X is calculated, where P is the historical consultation success rate, D is the physical path distance, T is the historical average response delay, ε is the preset unit coefficient, Dref is the preset reference distance, and Tref is the preset reference response delay; where X is the response efficiency coefficient, P is the historical consultation success rate, D is the physical path distance, T is the historical average response delay, ε is the preset unit coefficient, Dref is the preset reference distance, and Tref is the preset reference response delay.

[0041] In another preferred embodiment, step S5 further includes obtaining the comprehensive turnover efficiency score of the target department. If the comprehensive turnover efficiency scores of all candidate departments are less than or equal to the comprehensive turnover efficiency score of the target department, then the target department is selected as the first-visit department and sent to the current patient.

[0042] To ensure optimal triage, the most suitable department for each patient should be prioritized, especially when the target department already possesses strengths. By obtaining and comparing the comprehensive turnover efficiency scores of the target and candidate departments, a comprehensive assessment of each department's actual work efficiency and processing capacity can be achieved. When the scores of all candidate departments are no higher than those of the target department, it indicates that the target department performs better overall. It meets the professional treatment requirements of the primary illness and is not inferior to other departments in terms of turnover efficiency. In this case, selecting the target department as the initial consultation department ensures that the patient receives professional treatment, improves the efficiency of the consultation, avoids delays caused by changing departments, helps to achieve a rational allocation of emergency resources, allows patients to quickly find the most suitable department, reduces waiting time, improves treatment efficiency, protects patient health, and also ensures more effective use of medical resources, avoids waste, and improves the overall operational efficiency of the emergency system.

[0043] In another preferred embodiment, step S3 further includes, if the coverage overlap is greater than a preset overlap threshold, then without repeating the subsequent steps, the target department is directly designated as the first-visit department and sent to the current patient.

[0044] When the target department can effectively handle all patient conditions, the triage process is simplified, improving efficiency. When the overlap in coverage exceeds a preset threshold, it indicates that the target department is capable of handling both primary and non-primary conditions, eliminating the need to waste time and resources evaluating other candidate departments. Directly designating the target department as the first-line department allows patients to quickly enter the treatment process, reducing unnecessary waiting and procedural steps, and avoiding delays due to over-evaluation. Ultimately, this improves the efficiency of emergency triage, ensuring patients receive timely and appropriate treatment. It also optimizes the use of medical resources, preventing waste in unnecessary evaluations, allowing the emergency system to operate more efficiently and better serve patients.

[0045] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A hospital emergency triage method combining medical history data, characterized in that, Includes the following steps: S1, obtain the current patient's actual condition data and historical medical records to obtain the pre-screening dataset; the actual condition data includes condition description data and physiological characteristic data; S2, input the pre-examination dataset into the preset disease prediction model, obtain all pathological entities of the current patient and their confidence probabilities, and combine the risk coefficients corresponding to each pathological entity for optimization evaluation, select the pathological entity corresponding to the highest optimization score as the dominant disease, and label the remaining pathological entities as non-dominant diseases. S3, obtain the department corresponding to the dominant disease and mark it as the target department, calculate the coverage overlap between the set of non-dominant diseases and the treatment capacity of the target department, if the coverage overlap is less than or equal to the preset overlap threshold, then mark all departments corresponding to the non-dominant diseases as candidate departments, extract the current consultation data of each candidate department and calculate the waiting pressure coefficient of each candidate department, wherein the consultation data includes the average treatment rate, the number of patients waiting for consultation and the risk coefficient corresponding to each patient; S4. Obtain the collaboration data between the target department and any waiting department and calculate the response efficiency coefficient of the candidate department. The collaboration data includes physical path distance, historical average response delay and consultation success rate. S5. Calculate the comprehensive performance score of any candidate department based on the waiting pressure coefficient and response efficiency coefficient, and select the candidate department with the highest comprehensive performance score as the first-visit department and send it to the current patient.

2. The hospital emergency triage method combining medical history data according to claim 1, characterized in that, In S2, the specific construction process of the disease prediction model is as follows: Acquire a number of historical medical records and divide them into training and testing sets according to a certain ratio. Each historical medical record is a dataset of a certain emergency patient, and the dataset includes disease description data and physiological characteristic data. A backpropagation network model is generated and training parameters are set. The training set data is normalized, and the backpropagation network model is trained using the training set to obtain a preliminary disease prediction model. The initial disease prediction model was retrained using the test set to obtain the final disease prediction model.

3. The hospital emergency triage method combining medical history data according to claim 1, characterized in that, In S2, the specific process of the preferred evaluation is as follows: A multidimensional risk coefficient assessment matrix is ​​constructed, which includes two orthogonal dimensions: timeliness coefficient and severity coefficient. The confidence probability value is weighted and fused with the timeliness coefficient and severity coefficient respectively to obtain the optimal score. The pathological entity corresponding to the highest optimal score is selected as the dominant disease.

4. The hospital emergency triage method combining medical history data according to claim 1, characterized in that, In S3, the specific calculation process for the coverage overlap is as follows: A disease feature vector space is constructed based on a medical knowledge graph. Each disease in the non-dominant disease set is parsed into standard medical terminology codes and mapped to vector coordinates. At the same time, the historical electronic medical record database of the target department is obtained, and the disease spectrum and corresponding efficacy indicators are extracted through natural language processing to generate a multi-dimensional feature vector of the department's treatment capabilities. The cosine similarity algorithm is used to calculate the directional consistency between each disease feature vector and the department capability vector. After weighting the disease acuteness coefficient and the consequence severity coefficient, a single disease fitting value is obtained. All disease fitting values ​​in the set are normalized and the arithmetic mean is obtained. The arithmetic mean is labeled as the coverage overlap degree.

5. A hospital emergency triage method combining medical history data according to claim 1, characterized in that, In S3, the specific calculation process of the waiting pressure coefficient is as follows: real-time acquisition of the waiting patient queue of the candidate department, extraction of the pathological entity and its risk coefficient corresponding to the waiting patient, screening out all waiting patients whose risk coefficient is greater than or equal to the current patient's risk coefficient and using them as the effective pressure load base; synchronously extracting the actual number of patients treated in the department within a preset time period as the dynamic treatment rate benchmark value; dividing the effective pressure load base by the dynamic treatment rate benchmark value to obtain the basic pressure ratio, synchronizing the real-time status database of the medical equipment of the candidate department, and calculating the resource attenuation coefficient based on the degree of equipment shortage; Calculate the product of the basic pressure ratio and the resource attenuation coefficient, and calibrate the product as the waiting pressure coefficient.

6. A hospital emergency triage method combining medical history data according to claim 1, characterized in that, In S4, the specific calculation process for the response performance coefficient is as follows: Obtain the measured physical path distance between the target department and the candidate department, and divide it by a preset reference distance to obtain the distance attenuation factor; simultaneously extract the historical average response delay and divide it by a preset reference response delay to obtain the delay attenuation factor; then calculate according to the formula... The response efficiency coefficient X is calculated, where P is the historical consultation success rate, D is the physical path distance, T is the historical average response delay, ε is the preset unit coefficient, Dref is the preset reference distance, and Tref is the preset reference response delay.

7. A hospital emergency triage method combining medical history data according to claim 1, characterized in that, S5 also includes obtaining the comprehensive turnover efficiency score of the target department. If the comprehensive turnover efficiency scores of all candidate departments are less than or equal to the comprehensive turnover efficiency score of the target department, then the target department is selected as the first-visit department and sent to the current patient.

8. A hospital emergency triage method combining medical history data according to claim 1, characterized in that, S3 further includes that if the coverage overlap is greater than a preset overlap threshold, the subsequent steps are not repeated, and the target department is directly used as the first-visit department and sent to the current patient.

Images