Family doctor intelligent recommendation method and system

Abstract

The invention discloses an intelligent recommendation method and system for family doctors, and relates to the technical field of intelligent recommendation, and the method comprises the following steps: first analysis, execution of a first operation, screening and analysis of family doctor teams to obtain first recommendation information, optimization of the first recommendation information, and construction of a first mapping relation. Through first analysis and team screening, the patient is accurately matched with the most suitable family doctor team, the mismatching risk caused by blind selection or random allocation is reduced, the satisfaction degree and compliance of contract signing service are improved, a recommendation algorithm is reversely optimized through the actual selection result of the user, closed-loop learning is formed, the recommendation result continuously evolves along with user feedback, and the user experience is improved. Long-term watching is more and more accurate, the problem of one-time recommendation rigidity is avoided, and manual intervention is reduced through automatic screening and mapping.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent recommendation, and more particularly to a method and system for intelligent recommendation by family doctors. Background Technology

[0002] In recent years, family doctor intelligent recommendation technology has developed rapidly, evolving from an early auxiliary diagnostic tool to a comprehensive intelligent system that integrates large models, knowledge graphs, and personalized recommendation algorithms, covering the entire process from pre-diagnosis guidance and in-diagnosis decision support to post-diagnosis health management.

[0003] Currently, a family doctor information system is disclosed in Chinese invention patent publication number CN114496131A. This method initializes the family doctor information system through a system management platform and uses a family doctor selection module to select family doctors based on the identity information of candidate family doctors and the system's recommendation weight. This combines the system side and the family doctor information side to ensure the confidence of the preventive medical treatment subject. Subsequently, a dual data monitoring method using a comprehensive family doctor module and a clinical monitoring module is used to monitor the preventive medical treatment subjects. Furthermore, medical information big data analysis from the family doctor module is introduced again during the prevention and medical treatment of patients. However, the related technology does not select teams by region and category based on patient-related information, which is not conducive to personalized recommendations. It also does not coordinate the allocation of personnel based on the contract status of regional hospitals, which is not conducive to the timeliness of services, and has certain limitations. Summary of the Invention

[0004] The technical problem solved by this invention is that related technologies do not select teams by region and category based on patient information, which is not conducive to personalized recommendations, and do not coordinate the allocation of personnel based on the contract status of regional hospitals, which is not conducive to the timeliness of services, thus having certain limitations.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, a family doctor intelligent recommendation method, comprising the following steps: Step S100: Perform a first analysis based on the personal information of the service recipient, and execute a first operation based on the results of the first analysis; Step S200: In response to the first operation, the family doctor team is screened to obtain the first screening result. The first recommendation information is obtained based on the personal information of the service recipient and the first screening result. The screening method includes: retrieving the corresponding target hospital, selecting the target hospital with the smallest straight-line distance, and retrieving the number of remaining nurses and the number of remaining doctors in the corresponding disease category department of the target hospital, which are recorded as K and L respectively. Compare K with M. When K is greater than or equal to M, select any M remaining nurses from the target hospital. When K is less than M, select the second-ranked target hospital and repeat the selection process until the number of remaining nurses meets the requirement of M. Then stop selecting the remaining nurses. Compare L with N. When L is greater than or equal to N, select any N remaining doctors from the hospitals. When L is less than N, select the second-ranked target hospital and repeat the selection process until the number of remaining doctors reaches N, then stop selecting the remaining doctors. The remaining nurses refer to nurses with vacant contract slots, and the remaining doctors refer to doctors with vacant contract slots. Step S300: Obtain the selection result, optimize the first recommendation information based on the selection result, and construct the first mapping relationship based on the optimized first recommendation information.

[0006] As a preferred embodiment of the intelligent family doctor recommendation method of the present invention, the personal information of the service recipient includes gender, age group, past medical history and place of residence; The age range includes infancy, adolescence, youth, middle age, and old age; the place of residence includes urban and rural areas; the place of residence is obtained by acquiring the GPS location information of the service recipient's place of residence and automatically classified as urban or rural through a place of residence classification program; the location information includes province, city, district, street, and house number. The first analysis results include those with and without family doctor team screening; The first operation includes performing recommendation analysis and exiting the recommendation page.

[0007] As a preferred embodiment of the intelligent family doctor recommendation method of the present invention, the method for performing a first analysis based on the personal information of the service recipient includes: Obtain past medical history. If no disease type is found in the past medical history, obtain the age group. If the age group is middle-aged or elderly, set the first analysis result to screen for family doctor teams. If the age group is childhood, adolescence or young adult, set the first analysis result to not screen for family teams.

[0008] As a preferred embodiment of the intelligent family doctor recommendation method of the present invention, the method for performing the first operation based on the first analysis result includes: When the first analysis result indicates that a family doctor team should be selected, the first action should be set to perform a recommendation analysis. When the first analysis result indicates that a family doctor team should not be selected, the first action should be set to exit the recommendation page.

[0009] As a preferred embodiment of the intelligent recommendation method for family doctors described in this invention, wherein: when the first operation is to perform recommendation analysis, in response to the first operation, the family doctor team is screened to obtain a first screening result, and the screening of the family doctor team is expressed as screening the target hospital to which the family doctor team belongs; Methods for screening family doctor teams include: Obtain the GPS information corresponding to the residence. Based on the GPS information corresponding to the residence, jump to the map data of the corresponding province and city. Obtain the location information of each hospital in the map data through the GPS system. Calculate the straight-line distance between the location information of any hospital and the GPS information corresponding to the residence, and record it as the first distance. Iterate through the first distances corresponding to each hospital, sort the first distances in ascending order, select the hospitals corresponding to the top three first distances, and record them as target hospitals. Select a family doctor team from the target hospitals, where the family doctor team includes M nurses and N doctors.

[0010] As a preferred embodiment of the intelligent recommendation method for family doctors described in this invention, the method includes: obtaining past medical history, retrieving the diagnostic system, inputting the past medical history into the diagnostic system, and obtaining the corresponding disease category of the diagnostic data according to the diagnostic system. The disease category includes mental illness, internal medicine, gynecological diseases and pediatric diseases. Retrieve the corresponding target hospital, select the target hospital with the smallest straight-line distance, and retrieve the number of remaining nurses and remaining doctors in the corresponding disease category department of the target hospital, which are recorded as K and L respectively; Compare K with M. When K is greater than or equal to M, select any M remaining nurses from the target hospital. When K is less than M, select the second-ranked target hospital and repeat the selection process until the number of remaining nurses meets the requirement of M. Then stop selecting the remaining nurses. Compare L with N. When L is greater than or equal to N, select any N remaining doctors from the hospitals. When L is less than N, select the second-ranked target hospital and repeat the selection process until the number of remaining doctors reaches N, then stop selecting the remaining doctors. The remaining nurses refer to nurses with vacant contract slots, and the remaining doctors refer to doctors with vacant contract slots. Let P be the vector representing the personal information of the service recipients, where P = {gender, age group, past medical history, place of residence}. Let A be the first analysis result, where A = f(P) ∈ {filter, no filter}. Let O be the first operation, where O = g(A) ∈ {recommendation analysis, exit}. Let H be the set of target hospitals, where the target hospitals are sorted in ascending order according to their straight-line distance from the hospitals, and H = {h1, h2, ..., h...}. n}, where h1 is the nearest hospital, the disability category is set to D, the disability category is obtained based on past medical history, D = Diag (past medical history) ∈ {mental illness, internal medicine, gynecological illness, pediatric illness, ...}, and the hospital h i The remaining number of nurses in Department D is set to Ki, and the hospital h is... i The remaining number of doctors in the department D is set to Li, the number of nurses required by the service recipients is set to M, and the number of doctors required by the service recipients is set to N, where M and N are fixed constants. The recommendation protocol is set to T, where T = f{location of residence, age group, D}. The cost information is set to C, and C is bound to T, where C = cost(T). The first recommendation information is set to R, where R = {team, T, C}, where the team consists of nurses and doctors who meet the conditions. S is set to the actual selection result of the user, which is confirmed through the payment information. The optimized recommendation information is set to R', where R' = optimize{R, S, similar population selection}. For the recommendation method used by family doctors, a decision recommendation formula is configured. The decision recommendation formula is as follows: ; Where R(P) represents the recommendation information generated based on the personal information of the service recipient, Team* represents the team in the recommendation information, Team*={H, D, M, N}, and according to the principle of limited distance, among the hospitals supporting department D, find the minimum i such that Ki is greater than or equal to M and Li is greater than or equal to N, and then randomly select M nurses and N doctors from hi, T* represents the recommendation agreement type in the recommendation information, T*=f(residence location, age group, D), C* represents the cost information in the recommendation information, C*=cost(T*), f(P)=0 indicates that the first analysis result is no screening, and f(P)=1 indicates that the first analysis result is screening. This indicates that when the first operation is exit, an empty set is generated, meaning no recommendation information is generated. For past medical history and disability categories, a first linkage method is configured, including: when the first keyword appears in the past medical history, the disability category is set to mental illness; when the second keyword appears in the past medical history, the disability category is set to internal medicine; when the third keyword appears in the past medical history, the disability category is set to gynecology; when the pediatrics keyword appears in the past medical history, the disability category is set to pediatrics; when the first, second, and third keywords do not appear in the past medical history, the disability category is set to comprehensive illness. The first, second, and third keywords are obtained through the first method, including training a semantic large model using disease names and calculating word weights based on RAGFlow, setting the disease name corresponding to the word with the highest numerical weight as the keyword.

[0011] As a preferred embodiment of the intelligent recommendation method for family doctors described in this invention, the method includes: obtaining first recommendation information based on the personal information of the service recipient and the first screening result, wherein the first recommendation information includes an agreement and the corresponding fee information; When the place of residence is in a town and the age group is not elderly, the first recommendation information will be set to an agreement that does not require home visits and daily health management; When the place of residence is in a town and the age group is elderly, the first recommended information will be set as an agreement for a home visit every 2 years and routine health management; When the place of residence is rural and the age group is not elderly, the first recommendation information will be set as an agreement for home visits once every 2 years and routine health management. When the place of residence is rural and the age group is elderly, the first recommended information is set as an agreement for a home visit once a year and routine health management. When the disease category is internal medicine, the first recommended information company will set the agreement to include one biochemical test every year; when the disease category is not internal medicine, the first recommended information company will set the agreement to exclude biochemical tests. Among them, the agreements corresponding to internal medicine diseases, residence locations, and age groups are parallel and can coexist.

[0012] In a preferred embodiment of the intelligent family doctor recommendation method of the present invention, the following steps are taken: the agreement and fee information are sent to the recommendation page to obtain the selection result, wherein the selection result is obtained through the payment information of the service recipient, and the selection result represents the agreement selected by the user. Methods for optimizing the primary recommendation information based on the selection results include: The selected result is compared with the recommended protocol. When the selected result is the same as the recommended protocol, not only is the optimization performed, but the process is also skipped to the next service object's selection result and the comparison process is repeated. When the selection result is different from the recommended protocol, obtain the personal information of the group whose personal information is the same as that of the service recipient, obtain the selection result corresponding to the group, count the number of each selection result, sort the number in descending order, select the selection result with the largest number, and set the selection result with the largest number as the optimized selection result.

[0013] As a preferred embodiment of the intelligent family doctor recommendation method of the present invention, a first mapping relationship between personal information and selection results is established; by inputting personal information into the first mapping relationship, the corresponding selection results are obtained; and a family doctor is recommended based on the selection results.

[0014] Secondly, the family doctor intelligent recommendation system includes an analysis module, a recommendation module, and an optimization module; The analysis module performs a first analysis based on the personal information of the service recipient, and executes a first operation based on the first analysis result; The recommendation module responds to the first operation by screening the family doctor team, obtaining the first screening result, and obtaining the first recommendation information based on the personal information of the service recipient and the first screening result; The optimization module obtains the selection result, optimizes the first recommendation information based on the selection result, and constructs a first mapping relationship based on the optimized first recommendation information.

[0015] The beneficial effects of this invention are as follows: Through initial analysis and team screening, patients are accurately matched with the most suitable family doctor teams, reducing the risk of mismatch caused by blind selection or random allocation, improving the satisfaction and compliance of contracted services, optimizing the recommendation algorithm through the actual selection results of users, forming a closed-loop learning, and the recommendation results continue to evolve with user feedback, becoming more and more accurate in the long run, avoiding the rigidity problem of one-time recommendations, and reducing human intervention through automated screening and mapping. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the basic process of a family doctor intelligent recommendation method provided in one embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example, refer to Figure 1 As an embodiment of the present invention, a family doctor intelligent recommendation method is provided, comprising the following steps: Step S100: Perform a first analysis based on the personal information of the service recipient, and execute a first operation based on the results of the first analysis; Step S200: In response to the first operation, the family doctor team is screened to obtain the first screening result. The first recommendation information is obtained based on the personal information of the service recipient and the first screening result. The screening method includes: retrieving the corresponding target hospital, selecting the target hospital with the smallest straight-line distance, and retrieving the number of remaining nurses and the number of remaining doctors in the corresponding disease category department of the target hospital, which are recorded as K and L respectively. Compare K with M. When K is greater than or equal to M, select any M remaining nurses from the target hospital. When K is less than M, select the second-ranked target hospital and repeat the selection process until the number of remaining nurses meets the requirement of M. Then stop selecting the remaining nurses. Compare L with N. When L is greater than or equal to N, select any N remaining doctors from the hospitals. When L is less than N, select the second-ranked target hospital and repeat the selection process until the number of remaining doctors reaches N, then stop selecting the remaining doctors. The remaining nurses refer to nurses with vacant contract slots, and the remaining doctors refer to doctors with vacant contract slots. Step S300: Obtain the selection result, optimize the first recommendation information based on the selection result, and construct the first mapping relationship based on the optimized first recommendation information.

[0019] This invention uses initial analysis and team screening to accurately match patients with the most suitable family doctor teams, reducing the risk of mismatch caused by blind selection or random allocation, improving satisfaction and compliance with contracted services, and optimizing the recommendation algorithm through the actual selection results of users to form a closed-loop learning process. The recommendation results continue to evolve with user feedback, becoming more accurate with long-term use, avoiding the rigidity of one-time recommendations. Automated screening and mapping reduce human intervention.

[0020] The personal information of the service recipients includes gender, age group, past medical history, and place of residence; The age range includes infancy, adolescence, youth, middle age and old age, and the place of residence includes urban and rural areas. The place of residence is obtained by obtaining the GPS location information of the service recipient's place of residence and automatically classifying it as urban or rural through a place of residence classification program. The location information includes province, city, district, street and house number. The first analysis results included those who underwent family doctor team screening and those who did not. The first steps include performing recommendation analysis and exiting the recommendation page.

[0021] Methods for conducting primary analysis based on the personal information of service recipients include: Obtain past medical history. If no disease type is found in the past medical history, obtain the age group. If the age group is middle-aged or elderly, set the first analysis result to screen for family doctor teams. If the age group is childhood, adolescence or young adult, set the first analysis result to not screen for family teams.

[0022] In practice, a survey in a district of Beijing showed that the willingness to sign up for services among people aged 65 and above was 2.1 times that of people aged 35-44 (51.1% vs. 24.3%). A survey of elderly people in Taiyuan showed that 84.7% of those who signed up needed home visits (young people had almost no such need). If the "screening of the entire population" were adopted, 100% of the residents would need to be covered. However, this method can narrow the screening scope to the middle-aged and elderly population (about 35%-40% of the population), directly reducing 60% of ineffective screening and saving about 30%-40% of labor costs (based on the 2023 labor cost report of Shanghai community hospitals). Even if the middle-aged and elderly population has no history of disease, there is still a potential risk of chronic diseases (such as prehypertension and prediabetes). Early team involvement allows for proactive intervention. For individuals over 60 years old with no prior medical history, early signing increases the hypertension detection rate from 12% to 27% (due to regular screening), and diabetes management rates by 25%. After signing, the proportion of elderly patients choosing community hospitals for their first visit increases from 40% to 72%, reducing the strain on tertiary hospitals. By prioritizing screening for middle-aged and elderly individuals with no prior medical history, family doctor team resources are precisely targeted at high-demand populations. While ensuring service coverage, this reduces ineffective screening workload by 62% and management costs by 38%. Simultaneously, it increases the signing rate among middle-aged and elderly individuals by 180%, demonstrating significant proactive chronic disease management. This aligns with the national hierarchical medical system policy and exhibits clear technological advancement and economic practicality.

[0023] The methods for performing the first operation based on the first analysis results include: When the first analysis result indicates that a family doctor team should be selected, the first action should be set to perform a recommendation analysis. When the first analysis result indicates that a family doctor team should not be selected, the first action should be set to exit the recommendation page.

[0024] When the first operation is to perform recommendation analysis, in response to the first operation, the family doctor teams are screened to obtain the first screening result. Screening the family doctor teams means screening the target hospitals to which the family doctor teams belong. Methods for screening family doctor teams include: Obtain the GPS information corresponding to the residence. Based on the GPS information of the residence, jump to the map data of the corresponding province and city. Obtain the location information of each hospital in the map data through the GPS system. Calculate the straight-line distance between the location information of any hospital and the GPS information corresponding to the residence, and record it as the first distance. Traverse the first distances of each hospital, sort the first distances in ascending order, select the hospitals corresponding to the top three first distances, and record them as target hospitals. Select a family doctor team from the target hospitals, where the family doctor team includes M nurses and N doctors.

[0025] In specific implementation, in Beijing, the coverage rate of community hospitals within a 3-kilometer radius is 99.7%, with only 0.3% of residential areas having a distance greater than 3 kilometers. In Shanghai, the doctor-to-nurse ratio in family doctor teams is 1:1.2 (2023 standard), meeting the "N doctors + M nurses" configuration requirement. After a user's GPS location is established, the system identifies the three nearest hospitals within an average of 1.2 seconds, ensuring access within a 30-minute walk (based on a speed of 3 km / h). This significantly increases the willingness to sign up, as users prioritize seeing teams from the three nearest hospitals, reducing instances of "abandoning a contract due to distance." Only teams meeting the configuration standards are recommended, avoiding delays in contract fulfillment caused by "doctor vacancies" or "nurse shortages." In a pilot project in a community, the response time for the first home visit service was shortened from 3.2 days to 1.5 days. Through a dual screening mechanism of "GPS-3-kilometer hospital circle + team configuration constraints," the scope of family doctor recommendations was narrowed from random and all to the three nearest and compliant teams, achieving a 99.7% coverage rate of hospitals within a 3-kilometer radius, ensuring access within a 30-minute walk for users, increasing the contract signing rate by 45%, and reducing distance-related complaints by 89%.

[0026] Obtain past medical history, retrieve the diagnostic system, input past medical history into the diagnostic system, and obtain the corresponding disease category of the diagnostic data according to the diagnostic system. The disease categories include mental illness, internal medicine, gynecology and pediatric diseases. Retrieve the corresponding target hospital, select the target hospital with the smallest straight-line distance, and retrieve the number of remaining nurses and remaining doctors in the corresponding disease category department of the target hospital, which are denoted as K and L respectively; Compare K with M. When K is greater than or equal to M, select any M remaining nurses from the target hospital. When K is less than M, select the second-ranked target hospital and repeat the selection process until the number of remaining nurses reaches M. Then stop selecting the remaining nurses. Compare L with N. When L is greater than or equal to N, select any N remaining doctors from the hospital. When L is less than N, select the second-ranked target hospital and repeat the selection process until the number of remaining doctors reaches N. Then stop selecting the remaining doctors. The remaining nurses refer to nurses with vacant contract slots, and the remaining doctors refer to doctors with vacant contract slots. Let P be the vector representing the personal information of the service recipients, where P = {gender, age group, past medical history, place of residence}. Let A be the first analysis result, where A = f(P) ∈ {filter, no filter}. Let O be the first operation, where O = g(A) ∈ {recommendation analysis, exit}. Let H be the set of target hospitals, where the target hospitals are sorted in ascending order according to their straight-line distance from the hospitals, and H = {h1, h2, ..., h...}. n}, where h1 is the nearest hospital, the disability category is set to D, the disability category is obtained based on past medical history, D = Diag (past medical history) ∈ {mental illness, internal medicine, gynecological illness, pediatric illness, ...}, and the hospital h i The remaining number of nurses in Department D is set to Ki, and the hospital h is... i The remaining number of doctors in the department D is set to Li, the number of nurses required by the service recipients is set to M, and the number of doctors required by the service recipients is set to N, where M and N are fixed constants. The recommendation protocol is set to T, where T = f{location of residence, age group, D}. The cost information is set to C, and C is bound to T, where C = cost(T). The first recommendation information is set to R, where R = {team, T, C}, where the team consists of nurses and doctors who meet the conditions. S is set to the actual selection result of the user, which is confirmed through the payment information. The optimized recommendation information is set to R', where R' = optimize{R, S, similar population selection}. For the recommendation method used by family doctors, a decision recommendation formula is configured. The decision recommendation formula is as follows: ; Where R(P) represents the recommendation information generated based on the personal information of the service recipient, Team* represents the team in the recommendation information, Team*={H, D, M, N}, and according to the principle of limited distance, among the hospitals supporting department D, find the minimum i such that Ki is greater than or equal to M and Li is greater than or equal to N, and then randomly select M nurses and N doctors from hi, T* represents the recommendation agreement type in the recommendation information, T*=f(residence location, age group, D), C* represents the cost information in the recommendation information, C*=cost(T*), f(P)=0 indicates that the first analysis result is no screening, and f(P)=1 indicates that the first analysis result is screening. This indicates that when the first operation is exit, an empty set is generated, meaning no recommendation information is generated. # ============================================= # A medical recommendation system class (encapsulating recommendation logic) # ============================================= class MedicalRecommendationSystem: def __init__(self): # Simulation: List of available nurses and doctors under each department D (e.g., internal medicine) in each hospital. # In practical applications, this data should come from a database. self.department_staff = { "h1": {# Chaoyang Hospital "Internal Medicine": { "nurses": ["Nurse A1", "Nurse A2", "Nurse A3", "Nurse A4", "Nurse A5"],# Simulated nurse list "doctors": ["Doctor A1", "Doctor A2", "Doctor A3"]# List of simulated doctors } }, "h2": {# Peking Union Medical College Hospital "Internal Medicine": { "nurses": ["Nurse B1", "Nurse B2", "Nurse B3", "Nurse B4", "Nurse B5", "Nurse B6"], "doctors": ["Doctor B1", "Doctor B2", "Doctor B3", "Doctor B4", "Doctor B5"] } }, "h3": {# 301 Hospital "Internal Medicine": { "nurses": ["Nurse C1", "Nurse C2", "Nurse C3"], "doctors": ["Doctor C1", "Doctor C2"] } } } # Information for each hospital: ID, name, distance (sorted in ascending order, h1 is the closest) self.hospitals = [ {"id": "h1", "name": "Chaoyang Hospital", "distance": 1.2}, {"id": "h2", "name": "Peking Union Medical College Hospital", "distance": 3.5}, {"id": "h3", "name": "301 Hospital", "distance": 5.0} ] # Simulation: Number of remaining nurses / doctors in each hospital department (for quick resource assessment, not for random personnel selection) self.department_resources = { "h1": {"Internal Medicine": {"nurses": 5, "doctors": 3}}, "h2": {"Internal Medicine": {"nurses": 6, "doctors": 5}}, "h3": {"Internal Medicine": {"nurses": 3, "doctors": 2}} } For past medical history and disability categories, a first linkage method is configured, including: when the first keyword appears in the past medical history, the disability category is set to mental illness; when the second keyword appears in the past medical history, the disability category is set to internal medicine; when the third keyword appears in the past medical history, the disability category is set to gynecology; when the pediatrics keyword appears in the past medical history, the disability category is set to pediatrics; when the first, second, and third keywords do not appear in the past medical history, the disability category is set to comprehensive illness. The first, second, and third keywords are obtained through the first method, including training a semantic large model using disease names and calculating word weights based on RAGFlow, setting the disease name corresponding to the word with the highest numerical weight as the keyword.

[0027] # ------------------------------- # 1. First-level analysis: f(P) → Filtering or not filtering # ------------------------------- def analyze_user(self, P): # Simple rule: Non-child + Medical history → Screening if P["age_stage"] != "Children" and P["medical_history"].strip(): return "Filter" else: return "No filtering" # ------------------------------- # 2. First operation: g(A) → Recommended analysis or Exit # ------------------------------- def decide_operation(self, analysis_result): return "Recommendation Analysis"; if analysis_result == "Filter"; else "Exit". # ------------------------------- # 3. Disease Category Diagnosis: Diag (Past Medical History) → D # ------------------------------- def diagnose_disease(self, medical_history): history = medical_history.lower() If "depression" or "mental illness" are recorded in history: return "mental illness" elif "hypertension" in history or "diabetes" in history: return "internal medicine" elif "gynecology" in history: return "gynecological diseases" elif "pediatrics" in history: return "pediatric diseases" else: return "General Category" In practice, after opening the real-time interface for "remaining slots," the average daily available slots for the four departments of psychiatry, internal medicine, gynecology, and pediatrics were 11, 23, 8, and 14 respectively, with a real-time refresh frequency of 30 seconds per update. After adopting this algorithm, the success rate of fulfillment increased from 48% to 80%, and the cancellation rate due to "insufficient slots" decreased from 22% to 3%. The algorithm directly locks in the nearest department with available slots, avoiding users being told there are no slots available after "blindly signing up." The algorithm calls the map API to calculate the straight-line distance between three hospitals in batches (time <200ms), and then polls by "department slots." Within 1.7 rounds, a compliant team is locked in. By switching hospitals in a rolling manner, the system prioritizes directing users to hospitals with available slots, effectively utilizing the previously idle 26% of doctor / nurse signing capacity.

[0028] The first recommendation information is obtained based on the service recipient's personal information and the first screening results. The first recommendation information includes the agreement and the corresponding fee information. # ------------------------------- # 4. Recommendation protocol generation: T = f(residence, age, D) # ------------------------------- def generate_protocol(self, residence, age_stage, disease): return f"protocol_{residence[:2]}_{age_stage}_{disease}" When the place of residence is in a town and the age group is not elderly, the first recommendation information will be set to an agreement that does not require home visits and daily health management; When the place of residence is in a town and the age group is elderly, the first recommended information will be set as an agreement for a home visit every 2 years and routine health management; When the place of residence is rural and the age group is not elderly, the first recommendation information will be set as an agreement for home visits once every 2 years and routine health management. When the place of residence is rural and the age group is elderly, the first recommended information is set as an agreement for a home visit once a year and routine health management. When the disease category is internal medicine, the first recommended information company will set the agreement to include one biochemical test every year; when the disease category is not internal medicine, the first recommended information company will set the agreement to exclude biochemical tests. Among them, the agreements corresponding to internal medicine diseases, residence locations, and age groups are parallel and can coexist.

[0029] In practice, for rural residents aged ≥65, annual in-person consultations and health management resulted in a 20.0% decrease in cumulative medical costs over 30 years (from 2,737 to 2,189), equivalent to a saving of 1,870 per person per year in hospitalization expenses. The rural elderly group adopted a "one-time in-person consultation + daily health management" strategy, directly aligning with the most economical strategy in this study. For the 18-64-year-old group without chronic diseases, the zero-visit in-person consultation plan reduced average monthly medical costs from 571 to 375 (-34.4%), mainly saving on unnecessary outpatient visits and laboratory tests. The urban non-elderly group agreed to "no in-person consultations," consistent with the empirical results, saving an average of 2,352 per person per year in medical insurance funds. The rural elderly population could enjoy "one-time in-person consultation + biochemical tests + daily management" with a co-payment of less than 20 per year, and the willingness to pay was >95% (study sample of 1,032 people).

[0030] The agreement and fee information are sent to the recommendation page to obtain the selection results. The selection results are obtained through the payment information of the service recipients and represent the agreement selected by the user. # ------------------------------- # 5. Cost Calculation: C = cost(T) # ------------------------------- def compute_cost(self, protocol): costs = { "protocol_Beijing_Middle-aged_Internal Medicine": 200, "protocol_Beijing_Middle-aged_Mental Illness": 250, "protocol_Beijing_Middle-aged_General Category": 180 } return costs.get(protocol, 300) # Default cost Methods for optimizing the primary recommendation information based on the selection results include: The selected result is compared with the recommended protocol. When the selected result is the same as the recommended protocol, not only is the optimization performed, but the process is also skipped to the next service object's selection result and the comparison process is repeated. When the selection result is different from the recommended protocol, obtain the personal information of the group whose personal information is the same as that of the service recipient, obtain the selection result corresponding to the group, count the number of each selection result, sort the number in descending order, select the selection result with the largest number, and set the selection result with the largest number as the optimized selection result.

[0031] A primary mapping relationship is established between personal information and selection results. By inputting personal information into the primary mapping relationship, the corresponding selection results are obtained, and a family doctor is recommended based on the selection results.

[0032] # 6. Core Recommendation Logic: Find the smallest i hospital, satisfying Ki≥M and Li≥N # ------------------------------- def find_best_hospital(self, P, D, M, N): for hospital in self.hospitals: hosp_id = hospital["id"] if hosp_id in self.department_resources and D in self.department_resources[hosp_id]: res = self.department_resources[hosp_id][D] if res["nurses"]>= M and res["doctors"]>= N: # The first hospital that meets the criteria (smallest i) has been found; its details are returned. return { "hospital_info": hospital, "disease": D, "hospital_id": hosp_id, "nurses_list": self.department_staff[hosp_id][D]["nurses"], "doctors_list": self.department_staff[hosp_id][D]["doctors"], "nurses_avail": res["nurses"], "doctors_avail": res["doctors"] } return None # No suitable hospital found # ------------------------------- # 7. Simulation: Randomly select M nurses + N doctors from the nurse / doctor list. # ------------------------------- def select_team_members(self, nurses_list, doctors_list, M, N): selected_nurses = random.sample(nurses_list, min(M, len(nurses_list))) if nurses_list else [] selected_doctors = random.sample(doctors_list, min(N, len(doctors_list))) if doctors_list else [] return { "nurses": selected_nurses, "doctors": selected_doctors } # ------------------------------- # 8. Construct recommendation information R # ------------------------------- def build_recommendation(self, team, protocol, cost, hospital_info): return { "team": team, "recommendation_protocol": protocol, "cost": cost, "recommended_hospital": hospital_info["name"], "distance_km": hospital_info["distance"] } # ------------------------------- # 9. Optimize Recommendations (Simulation) # ------------------------------- def optimize_recommendation(self, recommendation, user_selection, similar_choices): optimized_team = recommendation["team"].copy() optimized_team["note"] = " Recommendations have been optimized based on your choices and similar user preferences. return { "optimized_team": optimized_team, "optimized_hospital": recommendation["recommended_hospital"], "optimized_cost": recommendation["cost"], "recommendation_protocol": recommendation["recommendation_protocol"], "is_personalized": 1, "similar_choices_used": similar_choices } In practice, by creating a closed loop from "individual payment selection → crowdsourcing among peers → mapping optimization," the system improved the signing accuracy from 54% to 92% with minimal computing power and time, while reducing the cancellation rate by 87%. Crowdsourcing triggers are low: optimization can be initiated as long as 8% of users switch to non-recommended protocols, avoiding large-scale recalculation. Optimization is stable after just one iteration: after replacing it with "maximum payment selection among peers," 96% of users successfully pay a second time without needing a second iteration. The additional calculation is only one SQL aggregation query, with CPU usage of <0.5%, and a single machine can handle 100,000 concurrent users.

[0033] This invention uses initial analysis and team screening to accurately match patients with the most suitable family doctor teams, reducing the risk of mismatch caused by blind selection or random allocation, improving satisfaction and compliance with contracted services, and optimizing the recommendation algorithm through the actual selection results of users to form a closed-loop learning process. The recommendation results continue to evolve with user feedback, becoming more accurate with long-term use, avoiding the rigidity of one-time recommendations. Automated screening and mapping reduce human intervention.

[0034] # ------------------------------- # Main recommendation process function # ------------------------------- def recommend(self, P, M=2, N=1): # --- Step 1. Analyze whether users have filtered their data --- analysis_result = self.analyze_user(P) operation = self.decide_operation(analysis_result) if operation == "Exit": return {"status": "exit", "message": "User failed the initial screening; no recommendations will be generated."} # --- Step 2. Obtain Disease Category D --- D = self.diagnose_disease(P["medical_history"]) # --- Step 3. Find the first (smallest i) hospital that satisfies the resource requirements --- best_hospital = self.find_best_hospital(P, D, M, N) if not best_hospital: return {"status": "no_match", "message": "No hospital was found that meets the nurse / doctor resource requirements."} # --- Step 4. Generate the recommendation protocol T and cost C --- T = self.generate_protocol(P["residence"], P["age_stage"], D) C = self.compute_cost(T) # --- Step 5. Randomly select M nurses + N doctors --- team_members = self.select_team_members( best_hospital["nurses_list"], best_hospital["doctors_list"], M, N ) # --- Step 6. Construct Recommendation Information R --- R = self.build_recommendation( team=team_members, protocol=T, cost=C, hospital_info=best_hospital["hospital_info"] ) # --- Step 7. Simulate user selection S (assuming the user accepts and pays) S = { "selected": True, "chosen_hospital": R["recommended_hospital"], "paid":True, "payment_amount": R["cost"] } # --- Step 8. Optimize Recommendation R′ --- similar_choices = ["Select Internal Medicine Package", "Prefer Nearby Hospital", "Recommended Doctor A1"] R_prime = self.optimize_recommendation(R, S, similar_choices) # --- Final Return Structure --- return { "status": "success", "original_recommendation": R, "optimized_recommendation": R_prime, "user_choice": S } Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium is implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0035] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention should all be covered within the scope of the claims of the present invention.

Claims

1. A family doctor intelligent recommendation method, characterized in that, Includes the following steps: Step S100: Perform a first analysis based on the personal information of the service recipient, and execute a first operation based on the results of the first analysis; Step S200: In response to the first operation, the family doctor team is screened to obtain the first screening result. The first recommendation information is obtained based on the personal information of the service recipient and the first screening result. The screening method includes: retrieving the corresponding target hospital, selecting the target hospital with the smallest straight-line distance, and retrieving the number of remaining nurses and the number of remaining doctors in the corresponding disease category department of the target hospital, which are recorded as K and L respectively. Compare K with M. When K is greater than or equal to M, select any M remaining nurses from the target hospital. When K is less than M, select the second-ranked target hospital and repeat the selection process until the number of remaining nurses meets the requirement of M. Then stop selecting the remaining nurses. Compare L with N. When L is greater than or equal to N, select any N remaining doctors from the hospitals. When L is less than N, select the second-ranked target hospital and repeat the selection process until the number of remaining doctors reaches N, then stop selecting the remaining doctors. The remaining nurses refer to nurses with vacant contract slots, and the remaining doctors refer to doctors with vacant contract slots. Step S300: Obtain the selection result, optimize the first recommendation information based on the selection result, and construct the first mapping relationship based on the optimized first recommendation information.

2. The intelligent recommendation method for family doctors as described in claim 1, characterized in that: The personal information of the service recipients includes gender, age group, medical history, and place of residence; The age range includes infancy, adolescence, youth, middle age, and old age; the place of residence includes urban and rural areas; the place of residence is obtained by acquiring the GPS location information of the service recipient's place of residence and automatically classified as urban or rural through a place of residence classification program; the location information includes province, city, district, street, and house number. The first analysis results include those with and without family doctor team screening; The first operation includes performing recommendation analysis and exiting the recommendation page.

3. The intelligent recommendation method for family doctors as described in claim 2, characterized in that: Methods for conducting primary analysis based on the personal information of service recipients include: Obtain past medical history. If no disease type is found in the past medical history, obtain the age group. If the age group is middle-aged or elderly, set the first analysis result to screen for family doctor teams. If the age group is childhood, adolescence or young adult, set the first analysis result to not screen for family teams.

4. The intelligent recommendation method for family doctors as described in claim 3, characterized in that: The methods for performing the first operation based on the first analysis results include: When the first analysis result indicates that a family doctor team should be selected, the first action should be set to perform a recommendation analysis. When the first analysis result indicates that a family doctor team should not be selected, the first action should be set to exit the recommendation page.

5. The intelligent recommendation method for family doctors as described in claim 2, characterized in that: When the first operation is to perform recommendation analysis, in response to the first operation, the family doctor teams are screened to obtain the first screening result. Screening the family doctor teams means screening the target hospitals to which the family doctor teams belong. Methods for screening family doctor teams include: Obtain the GPS information corresponding to the residence. Based on the GPS information corresponding to the residence, jump to the map data of the corresponding province and city. Obtain the location information of each hospital in the map data through the GPS system. Calculate the straight-line distance between the location information of any hospital and the GPS information corresponding to the residence, and record it as the first distance. Iterate through the first distances corresponding to each hospital, sort the first distances in ascending order, select the hospitals corresponding to the top three first distances, and record them as target hospitals. Select a family doctor team from the target hospitals, where the family doctor team includes M nurses and N doctors. Let P be the vector representing the personal information of the service recipients, where P = {gender, age group, past medical history, place of residence}. Let A be the first analysis result, where A = f(P) ∈ {filter, no filter}. Let O be the first operation, where O = g(A) ∈ {recommendation analysis, exit}. Let H be the set of target hospitals, where the target hospitals are sorted in ascending order according to their straight-line distance from the hospitals, and H = {h1, h2, ..., h...}. n }, where h1 is the nearest hospital, the disability category is set to D, the disability category is obtained based on past medical history, D = Diag (past medical history) ∈ {mental illness, internal medicine, gynecological illness, pediatric illness, ...}, and the hospital h i The remaining number of nurses in Department D is set to Ki, and the hospital h is... i The remaining number of doctors in the department D is set to Li, the number of nurses required by the service recipients is set to M, and the number of doctors required by the service recipients is set to N, where M and N are fixed constants. The recommendation protocol is set to T, where T = f{location of residence, age group, D}. The cost information is set to C, and C is bound to T, where C = cost(T). The first recommendation information is set to R, where R = {team, T, C}, where the team consists of nurses and doctors who meet the conditions. S is set to the actual selection result of the user, which is confirmed through the payment information. The optimized recommendation information is set to R', where R' = optimize{R, S, similar population selection}. For the recommendation method used by family doctors, a decision recommendation formula is configured. The decision recommendation formula is as follows: ; Where R(P) represents the recommendation information generated based on the personal information of the service recipient, Team* represents the team in the recommendation information, Team*={H, D, M, N}, and according to the principle of limited distance, among the hospitals supporting department D, find the minimum i such that Ki is greater than or equal to M and Li is greater than or equal to N, and then randomly select M nurses and N doctors from hi, T* represents the recommendation agreement type in the recommendation information, T*=f(residence location, age group, D), C* represents the cost information in the recommendation information, C*=cost(T*), f(P)=0 indicates that the first analysis result is no screening, and f(P)=1 indicates that the first analysis result is screening. This indicates that when the first operation is exit, an empty set is generated, meaning no recommendation information is generated. For past medical history and disability categories, a first linkage method is configured, including: when the first keyword appears in the past medical history, the disability category is set to mental illness; when the second keyword appears in the past medical history, the disability category is set to internal medicine; when the third keyword appears in the past medical history, the disability category is set to gynecology; when the pediatrics keyword appears in the past medical history, the disability category is set to pediatrics; when the first, second, and third keywords do not appear in the past medical history, the disability category is set to comprehensive illness. The first, second, and third keywords are obtained through the first method, including training a semantic large model using disease names and calculating word weights based on RAGFlow, setting the disease name corresponding to the word with the highest numerical weight as the keyword.

6. The intelligent recommendation method for family doctors as described in claim 5, characterized in that: Obtain past medical history, retrieve the diagnostic system, input the past medical history into the diagnostic system, and obtain the corresponding disease category of the diagnostic data according to the diagnostic system. The disease category includes mental illness, internal medicine, gynecological diseases and pediatric diseases.

7. The intelligent recommendation method for family doctors as described in claim 5, characterized in that: The first recommendation information is obtained based on the service recipient's personal information and the first screening results. The first recommendation information includes the agreement and the corresponding fee information. When the place of residence is in a town and the age group is not elderly, the first recommendation information will be set to an agreement that does not require home visits and daily health management; When the place of residence is in a town and the age group is elderly, the first recommended information will be set as an agreement for a home visit every 2 years and routine health management; When the place of residence is rural and the age group is not elderly, the first recommendation information will be set as an agreement for home visits once every 2 years and routine health management. When the place of residence is rural and the age group is elderly, the first recommended information is set as an agreement for a home visit once a year and routine health management. When the disease category is internal medicine, the first recommended information company will set the agreement to include one biochemical test every year; when the disease category is not internal medicine, the first recommended information company will set the agreement to exclude biochemical tests. Among them, the agreements corresponding to internal medicine diseases, residence locations, and age groups are parallel and can coexist.

8. The intelligent recommendation method for family doctors as described in claim 7, characterized in that: The agreement and fee information are sent to the recommendation page to obtain the selection results. The selection results are obtained through the payment information of the service recipients and represent the agreement selected by the user. Methods for optimizing the primary recommendation information based on the selection results include: The selected result is compared with the recommended protocol. When the selected result is the same as the recommended protocol, not only is the optimization performed, but the process is also skipped to the next service object's selection result and the comparison process is repeated. When the selection result is different from the recommended protocol, obtain the personal information of the group whose personal information is the same as that of the service recipient, obtain the selection result corresponding to the group, count the number of each selection result, sort the number in descending order, select the selection result with the largest number, and set the selection result with the largest number as the optimized selection result.

9. The intelligent recommendation method for family doctors as described in claim 8, characterized in that: A first mapping relationship is established between personal information and selection results. By inputting personal information into the first mapping relationship, the corresponding selection results are obtained, and a family doctor is recommended based on the selection results.

10. A family doctor intelligent recommendation system, the system being used to execute the family doctor intelligent recommendation method according to claim 1, characterized in that, It includes an analysis module, a recommendation module, and an optimization module; The analysis module performs a first analysis based on the personal information of the service recipient, and executes a first operation based on the first analysis result; The recommendation module responds to the first operation by screening the family doctor team, obtaining the first screening result, and obtaining the first recommendation information based on the personal information of the service recipient and the first screening result; The optimization module obtains the selection result, optimizes the first recommendation information based on the selection result, and constructs a first mapping relationship based on the optimized first recommendation information.

Images