Artificial intelligence-based post-hospital non-emergency transport platform recommendation method

Abstract

The application relates to the technical field of medical transportation, in particular to a post-hospital non-emergency transportation platform recommendation method based on artificial intelligence, which comprises the following steps: determining combination recommendation or direct recommendation according to the number of patients and a recommendation time length comparison value; in the combination recommendation, determining the illness change degree according to the examination parameter completeness, performing combination analysis based on the illness change degree and the medical record text correlation degree to determine the associated combination, and determining the recommendation priority coefficient corresponding to each associated combination based on the evaluation mean value; in the direct recommendation, determining the recommendation priority coefficient corresponding to each patient according to the comprehensive evaluation value; determining to perform generation analysis according to the number of patients in a single recommendation, or generating feature keywords according to the feature keyword emergence proportion; and determining the recommended hospital corresponding to each patient or each associated combination according to the keyword matching degree and the diagnosis waiting time length. The application can improve the accuracy of the recommended hospital.

Description

Technical Field

[0001] This invention relates to the field of medical transport technology, and in particular to an artificial intelligence-based method for recommending post-hospital non-emergency transport platforms. Background Technology

[0002] With the aging population and the continuous increase in the number of patients with chronic diseases, the demand for post-hospital non-emergency transport is increasing. Currently, although some hospitals have introduced artificial intelligence technology in the process of recommending transport destinations, most existing methods do not fully consider the individual characteristics of patients' conditions, resulting in inappropriate selection of transport destinations and delays in subsequent treatment. Therefore, how to improve the accuracy of transport hospital recommendations is an urgent problem to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN114861087A discloses an intelligent recommendation system for emergency transport based on proximity and capability, comprising: a condition parameter definition module for defining condition parameters that the intelligent recommendation system can call; a parameter acquisition module for acquiring the specific parameter content of the condition parameters and performing pre-calculation based on a data-driven emergency service mechanism to intelligently plan several emergency routes for treating patients; and a result display module for acquiring hospital recommendation results corresponding to each emergency route and displaying them on an emergency greenway APP in order of optimal recommendation logic. However, the above solution has the following problems: it only considers the emergency route planning and hospital recommendation for a single patient, lacks a flexible dynamic adjustment mechanism, cannot cope with complex scenarios, and does not fully consider the impact of individual patient characteristics on the selection of transport hospitals, resulting in recommendation results that are difficult to meet the needs of patients. Summary of the Invention

[0004] To address this, the present invention provides an AI-based recommendation method for post-hospital non-emergency transport platforms, which overcomes the problems of existing technologies that only consider emergency route planning and hospital recommendations for individual patients, lack flexible dynamic adjustment mechanisms, cannot cope with complex scenarios, and do not fully consider the impact of individual patient characteristics on the selection of transport hospitals, resulting in recommendation results that fail to meet patient needs.

[0005] To achieve the above objectives, the present invention provides an artificial intelligence-based method for recommending post-hospital non-emergency transport platforms, comprising:

[0006] The combination recommendation or direct recommendation is determined based on the number of patients and the comparison value of the recommended duration.

[0007] In the combination recommendation, the degree of variability of the condition is determined based on the completeness of the examination parameters, and the combination analysis is performed based on the degree of variability of the condition and the correlation of medical record text to determine the associated combination. The recommendation priority coefficient corresponding to each associated combination is determined based on the evaluation mean.

[0008] In direct recommendations, a comprehensive assessment value is determined based on the degree of change in the patient's condition and the number of deviation test parameters, and a recommendation priority coefficient is determined for each patient based on the comprehensive assessment value.

[0009] The generation analysis is determined based on the number of patients recommended in a single instance, or the characteristic keywords are generated based on the proportion of characteristic keywords appearing.

[0010] Recommended hospitals for each patient or combination of related information are determined based on keyword matching accuracy and diagnosis waiting time.

[0011] Furthermore, if the number of patients is greater than or equal to the preset number of patients or the comparison value of the recommended duration is greater than or equal to the preset comparison value of the recommended duration, then a combined recommendation is made.

[0012] Furthermore, if the number of patients is less than the preset number of patients and the recommended duration comparison value is less than the preset recommended duration comparison value, then a direct recommendation will be made.

[0013] Furthermore, the degree of change in the condition is determined based on the completeness of the examination parameters, including:

[0014] If the completeness of the examination parameters is greater than or equal to the preset completeness of the examination parameters, the degree of change in the condition is determined based on the mean of the examination fluctuation.

[0015] If the completeness of the examination parameters is less than the preset completeness of the examination parameters, the degree of change in the condition is determined based on the comparison value of the associated data.

[0016] Furthermore, a combined analysis was conducted based on the degree of change in the patient's condition and the correlation between medical record texts, including:

[0017] The initial association combination is determined based on the degree of change in the condition and the relevance of the medical record text, and whether to recombine the initial association combination is determined based on the degree of diagnostic difference.

[0018] For a single initial association combination, if the diagnostic difference corresponding to the initial association combination is greater than or equal to the preset diagnostic difference, then the patients in the initial association combination are recombined according to the difference in examination parameters.

[0019] Furthermore, if the number of patients recommended in a single instance is equal to the standard number, then a generative analysis is performed;

[0020] In the generation analysis, the number of recommended keywords is analyzed based on the completeness of the patient's examination parameters and the comparison value of the number of parameters to determine the total number of recommended keywords for each patient, and the feature keywords are determined based on the difference between the number of high-frequency keywords for the patient and the total number of recommended keywords.

[0021] Furthermore, based on the completeness of the patient's examination parameters and the number of deviation parameters, an analysis of the number of recommended keywords was conducted, including:

[0022] The number of initial recommended keywords for each patient is determined based on the completeness of the examination parameters, and the number of recommended keywords for pre-adjustment patients is increased based on the comparison value.

[0023] Among them, the number of recommended keywords for a single patient is positively correlated with the completeness of the examination parameters for that patient, and the increase in the number of recommended keywords for a single patient is positively correlated with the comparison value of the number of keywords for that patient.

[0024] The pre-adjusted patients are those whose number comparison value is less than the preset number comparison value.

[0025] Furthermore, feature keywords are determined based on the difference between the number of high-frequency keywords used by patients and the total number of recommended keywords, including:

[0026] For a single patient,

[0027] If the difference in quantity corresponding to the patient is greater than or equal to the standard quantity, then the characteristic keywords are determined based on the frequency of occurrence.

[0028] If the difference in the number corresponding to the patient is less than the standard number, then the characteristic keywords are determined based on the high-frequency keywords and the effective frequency.

[0029] Furthermore, if the number of patients recommended in a single instance exceeds the standard number, then feature keywords are generated based on the proportion of feature keywords appearing.

[0030] Furthermore, the recommended hospitals are those with a keyword matching degree greater than the preset keyword matching degree and a diagnosis waiting time less than the preset diagnosis waiting time.

[0031] Compared with the prior art, the beneficial effects of the present invention are that the technical solution of the present invention effectively reflects the current change in recommendation efficiency by comparing the number of patients and the recommendation duration, and then adaptively performs combined recommendations or direct recommendations based on the comparison of the number of patients and the recommendation duration, which is more in line with actual application scenarios. When the number of patients surges or the recommendation efficiency declines, it automatically switches to the combined recommendation mode to ensure that the recommendation task is completed on time, thereby improving recommendation efficiency.

[0032] Furthermore, this invention effectively reflects the comprehensiveness of a patient's diagnostic and treatment information by checking the completeness of the examination parameters. Then, different methods are adaptively selected to determine the degree of change in the patient's condition based on the completeness of the examination parameters. When the completeness of the examination parameters is high, the degree of change in the condition is determined based on the average fluctuation of the examination, which can accurately reflect the dynamic changes in the patient's individual physiological indicators. When the completeness of the examination parameters is low, the degree of change in the condition is determined based on the comparison value of associated data. By comparing the current patient with discharged patients with similar conditions, information from similar medical records is used to compensate for the lack of information in the current patient, thereby improving the accuracy of the assessment of the degree of change in the patient's condition.

[0033] Furthermore, the present invention determines the initial association combination based on the degree of change in the patient's condition and the correlation of the medical record text, which can achieve accurate clustering and preliminary grouping of the patient's condition. When the diagnostic difference is high, it can be reorganized based on the difference of the examination parameters, which can effectively identify the potential heterogeneity in the initial combination, avoid misclassifying patients with large diagnostic differences into one group, thereby improving the accuracy and consistency of patient grouping, and thus improving the accuracy of the recommendation.

[0034] Furthermore, by comprehensively analyzing the completeness of the parameters and the comparison of the number of parameters, this invention can dynamically adjust the number of recommended keywords to ensure that the number of recommended keywords for each patient matches the complexity of their condition and the completeness of the data. This avoids having too many or too few keywords, thereby improving the targeting and adaptability of keyword generation and thus enhancing the accuracy and practicality of the recommendations.

[0035] Furthermore, this invention uses keyword matching degree based on the overlap of characteristic keywords between patients (or associated combinations) and hospitals to ensure that recommended hospitals are highly matched with patients in terms of specialty capabilities and disease suitability. The diagnosis waiting time effectively reflects the hospital's reception efficiency, avoiding recommendations for congested or slow-responding hospitals. Thus, based on keyword matching degree and diagnosis waiting time, the recommended hospitals corresponding to each patient or associated combination are determined. This can optimize the patient's medical experience and improve the overall efficiency and satisfaction of medical services while ensuring the quality of recommendations. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the recommendation method for the post-hospital non-emergency transport platform based on artificial intelligence according to the present invention;

[0037] Figure 2 This is a flowchart illustrating the process of determining whether to make combined or direct recommendations based on the number of patients and the comparison value of the recommended duration in this invention.

[0038] Figure 3 This is a flowchart illustrating how the degree of change in a patient's condition is determined based on the completeness of examination parameters, as per the present invention.

[0039] Figure 4This is a flowchart illustrating the process of generating features based on the number of patients recommended in a single instance or based on the proportion of features appearing. Detailed Implementation

[0040] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0041] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0042] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0043] Please see Figures 1 to 4 As shown, this invention provides an artificial intelligence-based recommendation method for post-hospital non-emergency transport platforms, comprising:

[0044] The combination recommendation or direct recommendation is determined based on the number of patients and the comparison value of the recommended duration.

[0045] In the combination recommendation, the degree of variability of the condition is determined based on the completeness of the examination parameters, and the combination analysis is performed based on the degree of variability of the condition and the correlation of medical record text to determine the associated combination. The recommendation priority coefficient corresponding to each associated combination is determined based on the evaluation mean.

[0046] In direct recommendations, a comprehensive assessment value is determined based on the degree of change in the patient's condition and the number of deviation test parameters, and a recommendation priority coefficient is determined for each patient based on the comprehensive assessment value.

[0047] The generation analysis is determined based on the number of patients recommended in a single instance, or the characteristic keywords are generated based on the proportion of characteristic keywords appearing.

[0048] Recommended hospitals for each patient or combination of related information are determined based on keyword matching accuracy and diagnosis waiting time.

[0049] The application scenario of this invention is hospital recommendation during non-emergency transport of patients after hospitalization. This invention has several historical records, each of which records the recommendation duration, number of patients, recommendation duration comparison value, average examination fluctuation value, and related data comparison value of at least one hospital recommendation process during non-emergency transport of patients after hospitalization. Each historical record also has a corresponding qualified mark, which records whether the hospital recommendation process during non-emergency transport of patients after hospitalization meets the user's needs. The qualified mark can be recorded manually. It is understood that the user can determine whether the hospital recommendation process during non-emergency transport of patients after hospitalization meets the needs based on self-defined indicators. Self-defined indicators can be, but are not limited to, the number of errors, which will not be elaborated here. The number of errors is the number of times the recommended hospital is unsatisfactory to the patient.

[0050] The present invention sets a continuous cyclical monitoring cycle. At the end of each monitoring cycle, the number of patients and the recommended duration comparison value are determined. The duration of the monitoring cycle can be set according to the user's needs. The greater the user's need for monitoring accuracy, the shorter the monitoring cycle duration. A monitoring cycle value of 2 hours is provided, and the target monitoring cycle is the monitoring cycle that is currently being monitored.

[0051] The monitoring period that is adjacent to the target monitoring period and precedes the current monitoring period is recorded as the reference monitoring period.

[0052] The recommendation priority coefficient for each association combination is determined based on the evaluation mean. The recommendation priority coefficient for a single association combination is equal to the evaluation mean of that association combination × r0, where r0 is the priority weight coefficient. r0 > 0. Users can determine the value of r0 according to their actual needs. One possible value for r0 is 1.2.

[0053] The mean assessment for a single association combination is the average of the comprehensive assessment values ​​for each patient in that association combination.

[0054] Comprehensive assessment value = degree of change in condition / preset degree of change in condition + number of deviation test parameters / average number of deviation test parameters for each patient in the reference monitoring period;

[0055] The recommendation priority coefficient for each patient is determined based on the comprehensive assessment value. The recommendation priority coefficient for a single patient is equal to the comprehensive assessment value for that patient × r0. The higher the recommendation priority coefficient for a single patient (or related combination), the higher the priority of hospital recommendations for that patient (or related combination).

[0056] Specifically, if the number of patients is greater than or equal to the preset number of patients or the comparison value of the recommended duration is greater than or equal to the preset comparison value of the recommended duration, then a combined recommendation will be made.

[0057] The number of patients refers to the number of patients who require hospital recommendations at the end of the reference monitoring period.

[0058] Recommendation duration comparison value = average recommendation duration for each patient in the reference monitoring period - average recommendation duration for each patient in the historical records that can meet user needs. The recommendation duration for a single patient is the time required to start recommending hospitals for that patient until the corresponding hospital for that patient is determined.

[0059] The user can determine the preset number of patients and the preset recommendation duration comparison value according to the actual application scenario. The greater the user's need to improve recommendation efficiency, the smaller the preset number of patients and the preset recommendation duration comparison value should be. A method for determining the preset number of patients and the preset recommendation duration comparison value is provided by detecting the user's history of direct recommendations and recording the average number of patients and the average recommendation duration comparison value corresponding to the history that meets the user's needs as the preset number of patients and the preset recommendation duration comparison value, respectively.

[0060] It is understood that this invention effectively reflects the workload level through the number of patients and effectively reflects the recommendation efficiency through the comparison value of the recommendation duration. If the number of patients is greater than or equal to the preset number of patients or the comparison value of the recommendation duration is greater than or equal to the preset comparison value of the recommendation duration, it indicates that the workload is high or the recommendation efficiency is low. Using combined recommendation can improve processing efficiency. If the number of patients is less than the preset number of patients and the comparison value of the recommendation duration is less than the preset comparison value of the recommendation duration, it indicates that the workload is low and the recommendation efficiency is good. Using direct recommendation can ensure personalized accuracy.

[0061] Specifically, if the number of patients is less than the preset number of patients and the recommended duration comparison value is less than the preset recommended duration comparison value, then a direct recommendation will be made.

[0062] Specifically, the degree of change in the condition is determined based on the completeness of the examination parameters, including:

[0063] If the completeness of the examination parameters is greater than or equal to the preset completeness of the examination parameters, the degree of change in the condition is determined based on the mean of the examination fluctuation.

[0064] If the completeness of the examination parameters is less than the preset completeness of the examination parameters, the degree of change in the condition is determined based on the comparison value of the associated data.

[0065] In this invention, the individual medical record information includes examination information, disease description, medication records and treatment plans for each diagnosis and treatment of a single patient. The examination information includes textual descriptions of the examination results corresponding to several examination parameters, including but not limited to blood pressure, blood sugar, gallbladder wall thickness, heart rate and red blood cell count. This is content that is easily understood by those skilled in the art and will not be elaborated on in detail. This invention also includes several hospitals, and each hospital has medical record information corresponding to several discharged and transferred patients.

[0066] For a single patient, the medical records of discharged patients whose descriptions of the same condition have a higher percentage than the preset percentage of the same keywords are recorded as the matching medical records of that patient.

[0067] For the description of a patient's condition and the medical record information of a single discharged patient, the number of different keywords appearing in each description of the patient's condition is denoted as a1, and the number of different keywords appearing in each description of diagnosis and treatment in the medical record information of the discharged patient is denoted as a2. The percentage of identical keywords is the larger of the number of identical keywords in each description of the patient's condition and the description of diagnosis and treatment in the medical record information of the discharged patient and a2. The keywords appearing in the description of the condition are identified using NLP technology, which is a common technique used by those skilled in the art and will not be elaborated on in detail.

[0068] The preset percentage of identical keywords can be determined by the user based on the actual application scenario. The greater the user's need for improved recommendation accuracy, the higher the preset percentage of identical keywords will be. One preset percentage of identical keywords is provided, which is 80%.

[0069] Completeness of examination parameters = Number of examination parameters corresponding to this patient / Average number of examination parameters corresponding to each matched medical record information of this patient;

[0070] The number of examination parameters for a single patient is the total number of different examination parameters appearing in each examination information corresponding to that patient's medical record information;

[0071] Users can determine the preset value of the completeness of the examination parameters according to the actual application scenario. The smaller the preset value of the completeness of the examination parameters, the greater the user's need to determine the degree of change of the condition based on the average fluctuation of the examination. One preset value of the completeness of the examination parameters is provided, which is 40%.

[0072] The mean fluctuation of the examination for a single patient is the average value of the fluctuation reference value corresponding to each examination parameter, and the fluctuation reference value corresponding to a single examination parameter is the standard deviation of the value of that examination parameter in the examination information of each diagnosis and treatment of that patient.

[0073] When determining the degree of variability of a patient’s condition based on the mean fluctuation of examination results, the degree of variability of a patient’s condition is equal to the mean fluctuation of examination results for that patient, which is the average of the mean fluctuation of examination results for all patients in the historical records that can meet the user’s needs.

[0074] When determining the degree of disease change based on the correlation data comparison value, the degree of disease change for a single patient = the correlation data comparison value for that patient / the average correlation data comparison value for each patient in the historical records that can meet the user's needs;

[0075] The formula for calculating the correlation data comparison value for a single patient is as follows:

[0076]

[0077] Where ω is the correlation data comparison value corresponding to the patient, n is the total amount of matched medical record information corresponding to the patient, and i is 1, 2, ..., n. Let $\frac{i}{i}$ be the mean value of the examination fluctuations corresponding to the $i$-th matched medical record information for this patient. This represents the matching coefficient corresponding to the i-th matching medical record for this patient. The mean fluctuation of the examination for this patient is given. The matching coefficient for a single matched medical record is calculated as follows: (Number of second keywords in the matched medical record whose synchronous reference value is greater than the preset synchronous reference value + Total number of first keywords in the matched medical record) / Number of keywords in the matched medical record.

[0078] Each keyword appearing in each description of the patient's condition is recorded as the first keyword. Other keywords in a single matched medical record of the patient, excluding the first keyword, are recorded as the second keyword. For a single second keyword, the medical record of each discharged patient who has the second keyword is detected and recorded as reference information. The synchronous reference value corresponding to a single second keyword is the average number of first keywords corresponding to each reference information. The number of first keywords corresponding to a single reference information is the total number of first keywords appearing in each description of the condition of that reference information.

[0079] The preset synchronization reference value can be determined by the user based on the actual application scenario. The greater the user's need to improve the recommendation accuracy, the larger the preset synchronization reference value should be. One preset synchronization reference value is provided: the preset synchronization reference value = the average number of keywords in the disease description corresponding to the medical record information of each discharged and transferred patient × 50%. The number of keywords in the disease description corresponding to the medical record information of a single discharged and transferred patient is the total number of different keywords that appear in each disease description corresponding to that medical record information.

[0080] Understandably, the completeness of examination parameters effectively reflects the degree of completeness of the examination parameters. When the completeness of examination parameters is greater than or equal to the preset completeness of examination parameters, it indicates that the examination data is relatively sufficient. Based on the average fluctuation of the patient's own examination indicators, the degree of change in the condition can be directly assessed, thereby more accurately judging the stability of the condition and improving the accuracy of the judgment results. When the completeness of examination parameters is less than the preset completeness of examination parameters, it indicates that the examination data is insufficient. It is necessary to use historical medical records similar to the current patient for indirect inference to compensate for the impact of missing information, thereby improving the reliability of the condition assessment.

[0081] Specifically, a combined analysis based on the degree of change in the patient's condition and the correlation between medical record texts is conducted, including:

[0082] The initial association combination is determined based on the degree of change in the condition and the relevance of the medical record text, and whether to recombine the initial association combination is determined based on the degree of diagnostic difference.

[0083] For a single initial association combination, if the diagnostic difference corresponding to the initial association combination is greater than or equal to the preset diagnostic difference, then the patients in the initial association combination are recombined according to the difference in examination parameters.

[0084] For any two patients, the correlation degree of medical record text is the number of identical keywords appearing in the medical record information of the two patients;

[0085] The initial association combination is determined based on the degree of change in the patient's condition and the correlation of the medical record text. This includes: performing combination analysis for each patient; when performing combination analysis for a single patient, the patient is recorded as the target patient; other patients not recorded in the initial association combination are recorded as reference patients; each reference patient and the target patient whose absolute value of the difference between their degree of change in the patient's condition and the target patient's condition is less than a preset absolute value and whose correlation of the medical record text is greater than a preset correlation of the medical record text is recorded in an initial association combination; and the combination analysis continues for patients not recorded in the initial association combination until all patients are recorded in the corresponding initial association combination, at which point the combination analysis stops.

[0086] The values ​​of the preset absolute difference and the preset medical record text correlation can be determined by the user according to the actual application scenario. The greater the user's need for improved recommendation accuracy, the smaller the preset absolute difference and the larger the preset medical record text correlation. A method for determining the values ​​of the preset absolute difference and the preset medical record text correlation is provided. The method detects the historical records of initial correlation combinations determined based on the degree of disease change and the medical record text correlation. The average value of the reference absolute difference corresponding to each initial correlation combination in the historical records that can meet the user's needs and the average value of the reference medical record text correlation corresponding to each initial correlation combination are respectively recorded as the preset absolute difference and the preset medical record text correlation.

[0087] The absolute value of the reference difference is the absolute value of the difference in the degree of change of the condition of any two patients in a single initial association combination in the historical record; the reference medical record text association degree is the medical record text association degree of any two patients in a single initial association combination in the historical record.

[0088] It should be noted that for a single initial association combination, if the diagnostic difference corresponding to the initial association combination is less than the preset diagnostic difference, there is no need to recombine it, and the initial association combination is directly recorded as the association combination.

[0089] The diagnostic difference is the average of the outlier reference values ​​corresponding to each examination item. The outlier reference value for a single examination parameter is the standard deviation of the final reference value of that examination parameter for each patient corresponding to the medical record information that appears in the initial association combination.

[0090] The final reference value for a single examination parameter is the value of that examination parameter in the last diagnosis and treatment of a single patient.

[0091] When regrouping patients in the initial association group based on the difference in examination parameters, each patient in the association group is recorded as a patient to be recombined. Recombination analysis is performed on each patient to be recombined. When performing recombination analysis on a single patient to be recombined, the patient to be recombined is recorded as the target patient to be recombined. Other patients to be recombined who are not recorded in the association group except the target patient to be recombined are recorded as reference patients to be recombined. Reference patients to be recombined whose examination parameter difference with the target patient to be recombined is less than the preset examination parameter difference, as well as the target patient to be recombined, are recorded into an association group. Recombination analysis is continued for patients to be recombined who are not recorded into an association group until all patients to be recombined are recorded into the corresponding association group, at which point the recombination analysis is stopped.

[0092] The formula for calculating the difference in examination parameters between any two patients to be reconstituted is:

[0093]

[0094] in, Let be the difference in examination parameters between the two patients to be recombined, m be the number of examination parameters that appear in the examination information corresponding to the last diagnosis and treatment of both patients, and j be 1, 2, ..., m. This represents the absolute value of the difference between the final reference values ​​of the j-th examination parameter that appears in both patients' last diagnostic and treatment records. The j-th examination parameter is the average of the final reference values ​​for each patient to be recombined in this associated combination;

[0095] Users can determine the values ​​of the preset diagnostic difference degree and the preset examination parameter difference degree according to the actual application scenario. The greater the user's need for improved accuracy of recommendations, the smaller the values ​​of the preset diagnostic difference degree and the preset examination parameter difference degree should be. A method for determining the preset diagnostic difference degree and the preset examination parameter difference degree is provided, which is the average value of the diagnostic difference degree corresponding to the initial association combination that can meet the user's needs in the historical records that can be recombined as the preset diagnostic difference degree, and the average value of the reference examination parameter difference degree corresponding to each association combination in the historical records that can meet the user's needs as the preset examination parameter difference degree.

[0096] The reference examination parameter difference is the difference in examination parameters between any two patients in a single associated combination in the historical record.

[0097] Specifically, if the number of patients recommended in a single instance is equal to the standard number, then a generative analysis is performed;

[0098] In the generation analysis, the number of recommended keywords is analyzed based on the completeness of the patient's examination parameters and the comparison value of the number of parameters to determine the total number of recommended keywords for each patient, and the feature keywords are determined based on the difference between the number of high-frequency keywords for the patient and the total number of recommended keywords.

[0099] If a combination recommendation is made, the number of patients recommended in a single instance is the number of patients in a single associated combination; if a direct recommendation is made, the number of patients recommended in a single instance is 1; the standard number is 1.

[0100] Specifically, the number of recommended keywords is analyzed based on the completeness of the patient's examination parameters and the number of deviation parameters, including:

[0101] The number of initial recommended keywords for each patient is determined based on the completeness of the examination parameters, and the number of recommended keywords for pre-adjustment patients is increased based on the comparison value.

[0102] Among them, the number of recommended keywords for a single patient is positively correlated with the completeness of the examination parameters for that patient, and the increase in the number of recommended keywords for a single patient is positively correlated with the comparison value of the number of keywords for that patient.

[0103] The pre-adjusted patients are those whose number comparison value is less than the preset number comparison value.

[0104] The number of initial recommended keywords for a single patient is the smallest integer greater than or equal to q0, where q0 = (completeness of examination parameters for the patient / average completeness of examination parameters for each patient in the historical records that can meet the user's needs) × average number of recommended keywords for each patient in the historical records that can meet the user's needs.

[0105] The present invention includes a parameter range library, and each test parameter corresponds to a normal range. For example, the normal range of serum potassium is 3.5-5.0 mmol / L, and the normal range of fasting blood glucose is 3.9-6.1 mmol / L. Other contents that are easy for those skilled in the art to understand will not be elaborated further.

[0106] For a single inspection parameter, if the final reference value corresponding to the inspection parameter is not within the normal range corresponding to the inspection parameter, then the inspection parameter is recorded as a deviation parameter.

[0107] The number comparison value for a single patient = the number of initial recommended keywords for that patient / the number of deviation parameters for that patient - the average number of initial recommended keywords for each patient in the historical records that can meet the user's needs / the average number of deviation parameters for each patient in the historical records that can meet the user's needs.

[0108] The user can determine the preset number comparison value according to the actual application scenario. The greater the user's need to improve the recommendation accuracy, the smaller the preset number comparison value will be. A preset number comparison value is provided, which is the average of the number comparison values ​​corresponding to each pre-adjusted patient in the historical record that can meet the user's needs.

[0109] The number of deviation parameters for a single patient is the total number of deviation parameters in all examination parameters for that patient;

[0110] The increment value for the number of recommended keywords corresponding to a single pre-adjustment patient is the smallest integer greater than or equal to q1, where q1 = [(preset comparison value - the comparison value corresponding to this patient) / preset comparison value] × the initial number of recommended keywords corresponding to this pre-adjustment patient;

[0111] It should be noted that for a single patient, if the patient does not increase the number of recommended keywords, the initial number of recommended keywords for that patient will be recorded as the total number of recommended keywords.

[0112] If the patient adjusts the number of recommended keywords by increasing it, then the total number of recommended keywords = the initial number of recommended keywords for the patient + the increase in the number of recommended keywords for the patient.

[0113] Specifically, feature keywords are determined based on the difference between the number of high-frequency keywords used by patients and the total number of recommended keywords, including:

[0114] For a single patient,

[0115] If the difference in quantity corresponding to the patient is greater than or equal to the standard quantity, then the characteristic keywords are determined based on the frequency of occurrence.

[0116] If the difference in the number corresponding to the patient is less than the standard number, then the characteristic keywords are determined based on the high-frequency keywords and the effective frequency.

[0117] For a single patient, the quantity difference = the number of high-frequency keywords corresponding to that patient - the total number of recommended keywords corresponding to that patient;

[0118] The high-frequency keywords for a single patient are those keywords that appear more frequently than a preset frequency in the patient's medical records. The frequency of a single keyword is the number of times that keyword appears in the patient's medical records. The preset frequency can be determined by the user based on the actual application scenario. The greater the user's need to improve recommendation accuracy, the greater the preset frequency will be. A method for determining the preset frequency is provided, which is to record the average frequency of each high-frequency keyword in the historical records that can meet the user's needs as the preset frequency.

[0119] When determining feature keywords based on frequency of occurrence, select high-frequency keywords in descending order of frequency of occurrence until the total number of recommended keywords is reached.

[0120] When determining feature keywords based on high-frequency keywords and effective frequency, all high-frequency keywords are used as feature keywords, and low-frequency keywords are selected in descending order of effective frequency until the number of selected low-frequency keywords reaches k, where k = total number of recommended keywords - total number of high-frequency keywords.

[0121] Keywords other than high-frequency keywords are recorded as low-frequency keywords. Historical records that can meet user needs are recorded as reference historical records. The effective frequency corresponding to a single low-frequency keyword is the total number of reference historical records that record the low-frequency keyword as a feature keyword.

[0122] Specifically, if the number of patients recommended in a single instance exceeds the standard number, then feature keywords are generated based on the proportion of feature keywords appearing.

[0123] Among them, generating feature keywords based on the appearance ratio of feature keywords includes: for each patient in a single association combination, performing generation analysis for each patient in the association combination to determine the feature keywords corresponding to each patient, and taking each feature keyword corresponding to each patient as candidate keywords, and taking the candidate keywords whose appearance ratio of feature keywords is greater than the preset appearance ratio of feature keywords as the feature keywords corresponding to the association combination.

[0124] The percentage of appearance of a single feature keyword = the number of patients in the association combination who use that feature keyword / the total number of patients in the association combination;

[0125] The user can determine the preset feature keyword appearance ratio based on the actual application scenario. The greater the user's need to improve recommendation accuracy, the larger the preset feature keyword appearance ratio will be. One preset feature keyword appearance ratio is provided, which is 0.6.

[0126] Specifically, the recommended hospitals are those with a keyword matching degree greater than the preset keyword matching degree and a diagnosis waiting time less than the preset diagnosis waiting time.

[0127] It is understandable that the medical record information corresponding to each discharged and transferred patient also has several characteristic keywords. The method for generating characteristic keywords is also to use generation analysis, and each characteristic keyword in a single hospital is recorded as the analysis keyword.

[0128] For any patient (or associated combination) and any hospital, the keyword matching degree is the average of the reference values ​​of the occurrence frequency of each feature keyword corresponding to that patient (or associated combination), and the reference value of the occurrence frequency of a single feature keyword is the number of times that feature keyword appears in the analysis keywords corresponding to that hospital;

[0129] Users can determine the preset keyword matching degree and preset diagnosis waiting time values ​​according to the actual application scenario. The greater the user's need to improve recommendation accuracy, the larger the preset keyword matching degree value and the smaller the preset diagnosis waiting time value. A method for setting the preset keyword matching degree value is provided, which is the average keyword matching degree between each patient in the historical records that can meet the user's needs and the corresponding recommended hospital. A preset diagnosis waiting time value is provided, which is 6 hours.

[0130] The diagnostic waiting time for a single hospital is the average time from arrival at that hospital to the start of treatment for all patients who selected that hospital as the recommended hospital within the historical monitoring period. The historical monitoring period refers to each monitoring period prior to the target monitoring period.

[0131] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for recommending post-hospital non-emergency transport platforms based on artificial intelligence, characterized in that, include: The combination recommendation or direct recommendation is determined based on the number of patients and the comparison value of the recommended duration. In the combination recommendation, the degree of variability of the condition is determined based on the completeness of the examination parameters, and the combination analysis is performed based on the degree of variability of the condition and the correlation of medical record text to determine the associated combination. The recommendation priority coefficient corresponding to each associated combination is determined based on the evaluation mean. In direct recommendations, a comprehensive assessment value is determined based on the degree of change in the patient's condition and the number of deviation test parameters, and a recommendation priority coefficient is determined for each patient based on the comprehensive assessment value. The generation analysis is determined based on the number of patients recommended in a single instance, or the characteristic keywords are generated based on the proportion of characteristic keywords appearing. Recommended hospitals for each patient or combination of related information are determined based on keyword matching accuracy and diagnosis waiting time.

2. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 1, characterized in that, If the number of patients is greater than or equal to the preset number of patients, or the comparison value of the recommended duration is greater than or equal to the preset comparison value of the recommended duration, then a combined recommendation will be made.

3. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 2, characterized in that, If the number of patients is less than the preset number of patients and the recommended duration comparison value is less than the preset recommended duration comparison value, then a direct recommendation will be made.

4. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 1, characterized in that, The degree of change in the condition is determined based on the completeness of the examination parameters, including: If the completeness of the examination parameters is greater than or equal to the preset completeness of the examination parameters, the degree of change in the condition is determined based on the mean of the examination fluctuation. If the completeness of the examination parameters is less than the preset completeness of the examination parameters, the degree of change in the condition is determined based on the comparison value of the associated data.

5. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 1, characterized in that, A combined analysis based on the degree of change in the patient's condition and the correlation between medical record texts was conducted, including: The initial association combination is determined based on the degree of change in the condition and the relevance of the medical record text, and whether to recombine the initial association combination is determined based on the degree of diagnostic difference. For a single initial association combination, if the diagnostic difference corresponding to the initial association combination is greater than or equal to the preset diagnostic difference, then the patients in the initial association combination are recombined according to the difference in examination parameters.

6. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 1, characterized in that, If the number of patients recommended in a single instance is equal to the standard number, then a generation analysis is performed; In the generation analysis, the number of recommended keywords is analyzed based on the completeness of the patient's examination parameters and the comparison value of the number of parameters to determine the total number of recommended keywords for each patient, and the feature keywords are determined based on the difference between the number of high-frequency keywords for the patient and the total number of recommended keywords.

7. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 6, characterized in that, Based on the completeness of the patient's examination parameters and the number of deviation parameters, the number of recommended keywords was analyzed, including: The number of initial recommended keywords for each patient is determined based on the completeness of the examination parameters, and the number of recommended keywords for pre-adjustment patients is increased based on the comparison value. Among them, the number of recommended keywords for a single patient is positively correlated with the completeness of the examination parameters for that patient, and the increase in the number of recommended keywords for a single patient is positively correlated with the comparison value of the number of keywords for that patient. The pre-adjusted patients are those whose number comparison value is less than the preset number comparison value.

8. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 6, characterized in that, Feature keywords are determined based on the difference between the number of high-frequency keywords used by patients and the total number of recommended keywords, including: For a single patient, If the difference in quantity corresponding to the patient is greater than or equal to the standard quantity, then the characteristic keywords are determined based on the frequency of occurrence. If the difference in the number corresponding to the patient is less than the standard number, then the characteristic keywords are determined based on the high-frequency keywords and the effective frequency.

9. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 1, characterized in that, If the number of patients recommended in a single instance exceeds the standard number, then feature keywords are generated based on the proportion of feature keywords appearing.

10. The method for recommending post-hospital non-emergency transport platforms based on artificial intelligence according to claim 1, characterized in that, The recommended hospitals are those with a keyword matching degree greater than the preset keyword matching degree and a diagnosis waiting time less than the preset diagnosis waiting time.

Images