Infectious disease risk prediction analysis method and system based on electronic cases

Abstract

The invention relates to the technical field of infectious disease risk prediction, in particular to an infectious disease risk prediction analysis method and system based on electronic cases. Comprising the following steps: S1, preprocessing a multi-modal electronic case, cleaning structured data, and performing structured conversion on unstructured data; S2, constructing dynamic risk characteristics, quantifying basic disease risks, evaluating a key symptom change rate, and calculating a dynamic risk comprehensive index; S3, predicting infectious disease risks, and evaluating dynamic relevance between a patient and a group infectious disease condition. S4, optimizing a risk judgment threshold value, calibrating the threshold value through a Kappa coefficient, iteratively optimizing a model and features in combination with clinical feedback, forming a standardized scheme, improving data quality through multi-modal data integration, achieving accurate risk assessment through two-dimensional layering, dynamically calibrating the threshold value, optimizing and enhancing model adaptability through clinical feedback, and improving the reliability of the model. A standardized scheme guarantees application consistency, and accuracy and clinical practicability of infectious disease risk prediction are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of infectious disease risk prediction, in particular to an infectious disease risk prediction analysis method and system based on electronic medical records. BACKGROUND

[0002] The current electronic medical record contains multi-modal heterogeneous data, and the existing infectious disease risk assessment has problems such as rough stratification, fixed threshold, weak model adaptability, and is difficult to match the dynamic prevention and control demand. In order to improve the prediction accuracy and clinical adaptability, it is urgent to integrate multi-source data and dynamically optimize the threshold of the systematic analysis scheme.

[0003] The prior art such as the patent for invention with publication number CN107220482B discloses a respiratory infectious disease risk assessment system and method, wherein the system comprises: a model building module for building a calculation model, a transmission source characteristic analysis module for simulating the process of exhaled droplets and their aerodynamic characteristics, a transmission substance diffusion transport analysis module for outputting transmission substance diffusion transport analysis results, a quantitative calculation module for obtaining the exposure level of susceptible population through a dose-response model, and a infectious disease risk assessment module for assessing the risk distribution in the space and outputting the risk distribution. It can solve the problem of accurate prediction of personnel infection level in infectious disease transmission risk evaluation, and can be applied to the design of susceptible population respiratory infectious disease protection scheme and emergency plan, and improve the accuracy of risk assessment and better ensure public safety.

[0004] For the above-mentioned scheme, at least the following technical problems exist: 1. The above-mentioned scheme lacks quantitative analysis of individual endogenous risk characteristics, and does not mention quantitative analysis of patient underlying disease risk, such as correlation analysis of underlying disease and severe incidence rate, disease severity level assessment, and does not involve dynamic assessment of patient key symptom change rate, such as quantitative analysis of symptom severity and change trend analysis, which will lead to the inability to distinguish the risk difference of different individuals due to their own health status, for example, it is impossible to identify the higher risk of severe illness of patients with severe underlying diseases under the same exposure condition, so that the risk assessment stays at the macro level of population exposure level, and lacks fine differentiation of individual risk.

[0005] 2. The above-mentioned scheme lacks dynamic correlation analysis of individual and group infectious disease, does not establish the time and space correlation between individual risk indicators and group infectious disease data, and does not quantify the correlation strength between individual and group epidemic trend through correlation coefficient, which will lead to the inability to adjust individual risk level in combination with the dynamic change of group epidemic, for example, in the peak stage of epidemic, the actual risk of individual should be higher than that in the trough period of epidemic, but the evaluation logic of the above-mentioned scheme cannot reflect this dynamic correlation, so that the risk assessment result is difficult to adapt to the stage change of epidemic. SUMMARY

[0006] In view of the above-mentioned defects of the prior art, the present application provides an infectious disease risk prediction analysis method and system based on electronic medical records, which can effectively solve the problems of lack of fine differentiation of individual risks and difficulty in adapting risk assessment results to the phased changes of the epidemic in the prior art.

[0007] To achieve the above object, the present application is implemented by the following technical solutions:

[0008] The present application provides an infectious disease risk prediction analysis method based on electronic medical records in the first aspect, comprising:

[0009] S1, multi-modal electronic medical record preprocessing: querying the electronic medical records of each patient in a set time period from a multi-source database, and preprocessing the electronic medical records corresponding to each patient.

[0010] S2, dynamic risk feature construction: based on the preprocessed electronic medical records corresponding to each patient, quantifying the basic disease risk corresponding to each patient, and evaluating the key symptom change rate of each patient in the set time period, and then calculating the dynamic risk comprehensive index corresponding to each patient.

[0011] S3, infectious disease risk prediction: evaluating the dynamic correlation of each patient and the group infectious disease, and then multi-dimensionally analyzing the infectious disease risk corresponding to each patient.

[0012] S4, risk judgment threshold optimization: by calibrating the infectious disease risk threshold, and then iteratively optimizing the model and features through clinical feedback, and forming a standardized scheme.

[0013] The present application provides an infectious disease risk prediction analysis system based on electronic medical records in the second aspect, comprising: a multi-modal electronic medical record preprocessing module for querying the electronic medical records of each patient in a set time period from a multi-source database, and preprocessing the electronic medical records corresponding to each patient.

[0014] A dynamic risk feature construction module is used to quantify the basic disease risk corresponding to each patient based on the preprocessed electronic medical records corresponding to each patient, and to evaluate the key symptom change rate of each patient in the set time period, and then to calculate the dynamic risk comprehensive index corresponding to each patient.

[0015] An infectious disease risk prediction module is used to evaluate the dynamic correlation of each patient and the group infectious disease, and then to multi-dimensionally analyze the infectious disease risk corresponding to each patient.

[0016] A risk judgment threshold optimization module is used to calibrate the infectious disease risk threshold, and then to iteratively optimize the model and features through clinical feedback, and to form a standardized scheme.

[0017] Compared with the known prior art, the technical scheme provided by the present application has the following beneficial effects:

[0018] 1、The electronic case-based infectious disease risk prediction analysis method and system provided by the embodiment of the present application, in the multi-modal electronic case preprocessing process, through the redundancy inspection sheet and prescription sheet record deletion of structured data, and the extraction of symptom description, image features and other information from unstructured data by using the BERT-based named entity recognition model and converting them into Boolean value labels, hierarchical labels and ICD-10 codes, it is beneficial to eliminate data noise and format differences, and to convert heterogeneous electronic case data into unified and usable structured information, and to provide a high-quality data basis for subsequent risk analysis.

[0019] 2、In the dynamic risk feature construction process of the embodiment of the present application, the basic disease information is extracted from the electronic case, the relative association coefficient is calculated by using the contingency table to count the proportion of each type of basic disease in severe and non-severe patients, and the key symptom change rate is obtained by labeling the first appearance time of symptoms, quantifying the severity and calculating the time interval, which is beneficial to comprehensively integrate individual health baseline and disease dynamic change information, accurately depict the endogenous risk characteristics of patients, and provide individualized core basis for risk assessment.

[0020] 3、In the infectious disease risk prediction process of the embodiment of the present application, the individual basic disease risk and the symptom change rate are aligned with the group infectious disease data on the time axis, the correlation window is determined, the Pearson time series correlation coefficient and the spatial overlap degree are used to calculate the correlation strength between the individual and the group epidemic, and then a two-dimensional hierarchical system is constructed by using the dynamic risk comprehensive index and the dynamic correlation index, and the risk level is divided according to the threshold, which is beneficial to analyze the individual risk in the group epidemic background, considering the patient's own condition and combining the group transmission trend, and realizing the multi-dimensional accurate stratification of risk.

[0021] 4、In the risk judgment threshold optimization process of the embodiment of the present application, the historical cases are retrieved from the database to construct a verification set, the Kappa coefficient is used to calculate the coincidence degree of different candidate thresholds to calibrate the risk threshold, the clinical application data is collected to optimize the features and weights of the dynamic risk comprehensive index and the calculation rules of the dynamic correlation index, and the standardization scheme is integrated, which is beneficial to realize the dynamic adaptation of the threshold, continuously correct the model bias through clinical feedback, form a unified operation specification, and ensure the consistency and practicality of risk assessment in different scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to make the technical solutions of the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0023] Figure 1 For the flowchart of the steps of the present application.

[0024] Figure 2 For the schematic diagram of the system structure connection of the present application. DETAILED DESCRIPTION

[0025] In order to make the technical solutions of the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0026] The present application will be further described below in combination with embodiments.

[0027] Please refer to Figure 1 As shown in the accompanying drawings, the present application provides an infectious disease risk prediction analysis method based on electronic medical records, which comprises the following steps: S1, multi-modal electronic medical record preprocessing: querying the electronic medical records of each patient in a set time period from a multi-source database, and preprocessing the electronic medical records corresponding to each patient.

[0028] It should be noted that the multi-source database refers to a plurality of database sets storing electronic medical record data of different types and different sources, such as the HIS system, LIS system and PACS system of a hospital, and the electronic medical record system for storing text data such as doctor's handwritten medical history records and diagnosis conclusions. These databases record patient information from different aspects of the treatment process, and together constitute the data source of multi-modal electronic medical records.

[0029] The preprocessing includes structured data cleaning and unstructured data structured conversion. The structured data cleaning process is as follows: the structured data in the electronic medical record of each patient includes test indicators, medication records and demographic information; the unique case number of each patient is matched with the visit time stamp of each patient; then redundant test records and prescription records with a repetition number greater than or equal to two are deleted; the missing test indicators of each patient are filled by using the average test indicators of the same disease patients in the corresponding department and within a set number of days from the onset time; and the birth date format, weight unit and blood pressure value of each patient are uniformly converted into a preset format and unit, so as to complete the cleaning of the structured data.

[0030] The unstructured data structured conversion process is as follows: the unstructured data includes medical history records, CT image reports and doctor's handwritten medical records; the text information in the medical history records and CT image reports is analyzed by using a BERT-based named entity recognition model; the symptom description, image feature and diagnosis conclusion are extracted; the symptom description is converted into a Boolean value label; the image feature is converted into a hierarchical label; and the diagnosis conclusion is mapped into an ICD-10 code, so as to realize the structured conversion of the unstructured data.

[0031] It should be noted that the test indicators include white blood cell count, body temperature and blood pressure; the medication records include drug name, dose and frequency; the demographic information includes birth date, gender and weight; the symptom description is extracted, such as continuous fever for 3 days, scattered patchy shadows in both lungs and new coronavirus infection; the symptom description is converted into a Boolean value label, such as "fever = yes" and "cough = no"; the image feature is converted into an ICD-10 code, such as "ground glass shadow = mild" and "consolidation shadow = none"; and the example is only illustrative and is not limited.

[0032] It should also be noted that in the unstructured data structured conversion process, BERT is a pre-training language model based on the Transformer architecture, which belongs to the existing mature technology. Here, the BERT-based named entity recognition model is used to accurately identify and extract key information such as symptom descriptions and image features such as "cough" and "lung nodules", and diagnosis conclusions such as "pneumonia", based on its semantic understanding ability of medical text, so as to provide accurate original extraction support for subsequent conversion of these information into structured data such as Boolean value labels and hierarchical labels; and ICD-10 code, i.e. International Classification of Diseases, 10th Revision, is a unified disease classification coding standard developed by the World Health Organization.

[0033] The method and system for infectious disease risk prediction based on electronic medical records provided by the embodiments of the present application are beneficial to eliminate data noise and format differences, convert heterogeneous electronic medical record data into unified and usable structured information, and provide a high-quality data basis for subsequent risk analysis.

[0034] S2, dynamic risk feature construction: based on the preprocessed electronic medical records corresponding to each patient, the risk of the underlying disease corresponding to each patient is quantified, and the change rate of the key symptoms corresponding to each patient in a set time period is evaluated, and then the dynamic risk comprehensive index corresponding to each patient is calculated.

[0035] In a specific embodiment, the quantification of the underlying disease risk corresponding to each patient is specifically as follows: extracting the underlying disease information of each patient from the preprocessed electronic medical records, recording each type of underlying disease in the underlying disease information as q, q = 1, 2, …, p, p is a positive integer, querying the underlying disease of the specified infectious disease patient in a set historical time period from the database, the specified infectious disease patient includes the specified infectious disease severe patient and the specified infectious disease non-severe patient, screening out the specified infectious disease patient belonging to the underlying disease contained in the underlying disease information, using a contingency table to count the proportion of each type of underlying disease in the specified infectious disease severe patient and the proportion in the specified infectious disease non-severe patient, dividing the proportion of each type of underlying disease in the specified infectious disease severe patient by the proportion of each type of underlying disease in the specified infectious disease non-severe patient to obtain the relative correlation coefficient of each type of underlying disease and the severe incidence rate, calculating the relative correlation coefficient of each type of underlying disease and the severe incidence rate according to a pre-set normalization formula, and then obtaining the differential risk coefficient.

[0036] Querying the disease severity level corresponding to each type of underlying disease from the clinical diagnosis and treatment guidelines corresponding to the underlying disease information, obtaining the nonlinear relationship between the disease course and the infectious disease severe risk, and then using a survival analysis model to fit the disease course year and the severe risk data of the historical cases to obtain a quantization curve, finally constructing a disease course correlation factor calculation method, calculating the disease course correlation factor corresponding to each type of underlying disease, and then integrating to calculate the underlying disease risk value corresponding to each patient by the underlying disease quantization formula.

[0037] It should be noted that the normalization formula is: b q R' q represents the differential risk coefficient of the qth type of underlying disease corresponding to the specified patient, R' q = 0, 1, 2, …, p, p is a positive integer, and b q represents the disease course correlation factor corresponding to the qth type of underlying disease, b q = 0, 1, 2, …, p, p is a positive integer. qThe relative correlation coefficient between the qth basic disease and the incidence of severe cases is denoted as max, and max denotes the maximum function.

[0038] The basic disease quantification formula is: wherein R is the risk value of the basic disease corresponding to the specified patient, c q and d q respectively represent the severity level and the course-related factor of the qth basic disease corresponding to the specified patient, and x is the adjustment coefficient corresponding to the set comorbidity superposition state.

[0039] It should be further noted that the nonlinear relationship between the course of a certain basic disease and the risk of severe cases after the patient is infected with a specified infectious disease is determined through clinical literature review and expert consensus. For example, the clinical regularity shows that for diabetes, the risk of severe cases increases rapidly with the number of years from 0 to 5 years, the risk increases at a slower rate from 5 to 10 years, and the risk tends to be stable or even slightly decreases after 10 years. Historical case data of patients with the basic disease are collected, including the number of years of the course, whether the patient is infected with a specified infectious disease, and whether the patient develops into a severe case. Survival analysis models such as the Cox proportional hazards model are used to fit the curve of the number of years of the course and the probability of severe cases. For example, the fitting curve shows that the risk of severe cases is 15% at 3 years, 30% at 8 years, and 28% at 12 years. Finally, the course is divided into several stages according to the clinical regularity and the fitting curve, and the course-related factor is set for each stage. The course-related factor value is positively correlated with the relative level of the risk of severe cases in the stage. For example, the course-related factor of diabetes is set to 1.5 for 0-5 years, 1.2 for 5-10 years, and 1.0 for more than 10 years. The course-related factors of each course stage of the basic disease are obtained in this way. The example is only illustrative and is not the only limitation.

[0040] In a specific embodiment, the change rate of the key symptoms of each patient in a set time period is evaluated. The specific process is as follows: the key symptom records in the set time period are extracted from the preprocessed electronic medical record, and the time point of the first occurrence of each symptom in the key symptoms and the severity of each record are marked. The severity is quantified as 0 to 3 points, wherein a severity quantification value of 0 indicates no symptoms, a severity quantification value of 1 indicates mild symptoms that do not affect daily activities, a severity quantification value of 2 indicates moderate symptoms that partially affect daily activities, and a severity quantification value of 3 indicates severe symptoms that completely restrict daily activities.

[0041] Let each symptom in the key symptoms be i, i = 1, 2, …, n, and n is a positive integer. The time interval of each record collected from the electronic medical record of each patient is obtained and denoted as Δt j , Δt jdenotes the time interval between the (j+1)th record and the jth record, j is the number corresponding to each record, j = 1, 2, …, m, m is a positive integer, and the key symptom change rate of each patient corresponding to each type of symptom at each record is calculated by the quantification formula of the single symptom dynamic change value;

[0042] Based on the key symptom change rate of each patient corresponding to the set time period, the key symptom change rate of each patient corresponding to the set time period is calculated by the key symptom change rate quantification formula.

[0043] It should be noted that the key symptoms include fever, cough and dyspnea, and the quantification formula corresponding to the single symptom dynamic change value is: α j，i = (S j+1，i -S j，i )*exp(-0.1*Δt j ), the single symptom dynamic change value α j，i of the ith type of symptom corresponding to the jth record of the specified patient is obtained, wherein S j+1，i and S j，i denote the severity quantification value of the ith type of symptom corresponding to the (j+1)th record and the jth record of the specified patient, respectively, and exp denotes the exponential function symbol, thereby obtaining the single symptom dynamic change value of each type of symptom in each record of each patient.

[0044] Symptom change rate quantification formula: The key symptom change rate β of the specified patient corresponding to the set time period is obtained, wherein is the weight factor corresponding to the ith type of symptom, and the key symptom change rate of each patient corresponding to the set time period is obtained.

[0045] In one specific embodiment, the dynamic risk comprehensive index corresponding to each patient is calculated, and the specific process is as follows: based on the basic disease risk corresponding to each patient and the key symptom change rate corresponding to the set time period, the basic disease risk and the key symptom change rate corresponding to each patient are calculated by weighting, and the dynamic risk comprehensive index corresponding to each patient is obtained.

[0046] It should be noted that the weighted calculation process is: multiplying the corresponding basic disease risk of each patient by the corresponding set basic disease risk weight factor and adding the corresponding key symptom change rate of each patient by the corresponding set key symptom change rate weight factor, and then obtaining the corresponding dynamic risk comprehensive index of each patient. The setting process of the basic disease risk weight factor and the key symptom change rate weight factor is the same as the setting process of the index weight in the disease risk scoring model in the prior art, which is based on the correlation strength of the basic disease risk, the key symptom change rate and the severe case outcome in the historical case data, and the weight distribution is determined by logistic regression coefficient quantification or clinical expert consensus calibration to ensure that the contribution of each factor matches the actual risk impact, so it is not described in detail here.

[0047] In the dynamic risk feature construction process of the embodiment of the application, the basic disease information is extracted from the electronic case, the relative association coefficient is calculated by using the contingency table to count the proportion of each type of basic disease in the severe and non-severe patients, and the key symptom change rate is obtained by labeling the first appearance time of the symptom, quantifying the severity and calculating the time interval, which is beneficial to comprehensively integrate the individual health baseline and the dynamic change information of the disease condition, accurately describe the endogenous risk characteristics of the patient, and provide individualized core basis for risk assessment.

[0048] S3, infectious disease risk prediction: evaluating the dynamic correlation of each patient and the group infectious disease condition, and then multi-dimensionally analyzing the infectious disease risk corresponding to each patient.

[0049] In a specific embodiment, the process of evaluating the dynamic correlation of each patient and the group infectious disease condition is as follows: querying the group specified infectious disease data in a set time period from the database, the group specified infectious disease data including the group daily average new case number, the virus variant proportion, the severe case incidence and the spatiotemporal transmission hotspot area distribution, aligning the basic disease risk index of each patient, the key symptom change rate and the group specified infectious disease data on the time axis, taking the time point of the first appearance of the symptom of each patient as the reference point, defining the correlation window of each set time period before and after the reference point, and establishing the time correlation basis between the individual risk index and the group specified infectious disease development trend.

[0050] The first correlation strength is calculated by a Pearson time-series correlation coefficient between the change rate of the key symptoms of each patient and the daily average number of new cases in the group within the correlation window, and the second correlation strength is calculated by a spatial overlap degree between the corresponding treatment area of each patient and the hotspot area of the specified infectious disease transmission within the set time range corresponding to the correlation window.

[0051] It should be noted that both the Pearson time-series correlation coefficient calculation and the spatial overlap degree calculation belong to the prior art. The Pearson time-series correlation coefficient is used to quantify the linear correlation degree between the fluctuation of the change rate of the key symptoms of the patient over time and the change trend of the daily average number of new cases in the group within the correlation window, and the spatial overlap degree is used to measure the spatial coverage overlap ratio between the treatment area of the patient and the hotspot area of the group infectious disease within the specified time range, so as to reflect the time and space correlation strength of the individual and the group epidemic situation respectively.

[0052] It should be further noted that the specified calculation method is to multiply the first correlation strength by the first weight based on the basic disease risk index, multiply the second correlation strength by the remaining weight, and then add the two product results to obtain the sum, which is the dynamic correlation index. For example, for a patient with a high basic disease risk index, the first weight can be set to 0.6, and the dynamic correlation index is equal to the first correlation strength multiplied by 0.6 plus the second correlation strength multiplied by 0.4. The example is only illustrative, and the actual first weight is not necessarily 0.6. This example is not the only limitation.

[0053] It should be further noted that the setting process of the preset dynamic correlation index threshold is the same as the setting process of the diagnostic test cutoff value in the prior art, such as the best critical point determined by the ROC curve. That is, based on the actual division results of the dynamic correlation close and not close in the historical cases, combined with the judgment demand of the correlation strength in the clinic, the initial threshold is determined by statistical analysis, and the verification set is calibrated and optimized subsequently, which will not be described in detail here.

[0054] In a specific embodiment, the multi-dimensional stratification analysis of the infectious disease risk corresponding to each patient is as follows: the dynamic risk comprehensive index corresponding to each patient is taken as the individual endogenous risk core index, and the dynamic correlation index corresponding to each patient is taken as the group exogenous risk index, so as to form a two-dimensional stratification system of endogenous and exogenous. The dynamic risk comprehensive index corresponding to each patient is divided into low, medium and high three levels according to the preset risk comprehensive index threshold, and the dynamic correlation index corresponding to each patient is divided into weak, medium and strong three levels according to the preset correlation index threshold.

[0055] When the level of the dynamic risk comprehensive index of a patient is high and the level of the dynamic correlation index of the patient is strong at this time, the patient faces extremely high risk of infectious diseases, and when the level of the dynamic risk comprehensive index of a patient is low and the level of the dynamic correlation index of the patient is weak at this time, the patient faces extremely low risk of infectious diseases, the dynamic risk comprehensive index and the dynamic correlation index of each patient corresponding to the remaining combinations are weighted and calculated, and then the result of the weighted calculation is divided into a medium-low risk, a medium risk and a medium-high risk according to a preset risk threshold, so as to obtain the risk of infectious diseases corresponding to each patient.

[0056] It should be noted that the remaining combinations include seven kinds, which are that the level of the dynamic risk comprehensive index is low and the level of the dynamic correlation index is medium, the level of the dynamic risk comprehensive index is low and the level of the dynamic correlation index is strong, the level of the dynamic risk comprehensive index is medium and the level of the dynamic correlation index is weak, the level of the dynamic risk comprehensive index is medium and the level of the dynamic correlation index is medium, the level of the dynamic risk comprehensive index is medium and the level of the dynamic correlation index is strong, the level of the dynamic risk comprehensive index is high and the level of the dynamic correlation index is weak, and the level of the dynamic risk comprehensive index is high and the level of the dynamic correlation index is medium.

[0057] It should be further noted that the process of weighting and calculating the dynamic risk comprehensive index and the dynamic correlation index of each patient corresponding to the remaining combinations is the same as the process of weighting and calculating the basic disease risk and the key symptom change rate corresponding to each patient, and will not be described in detail here.

[0058] In the process of infectious disease risk prediction, the individual basic disease risk, the symptom change rate and the group infectious disease data are aligned on the time axis, the correlation window is delimited, the correlation strength between the individual and the group epidemic situation is calculated by using the Pearson time sequence correlation coefficient and the spatial overlap degree, the two-dimensional hierarchical system is constructed by using the dynamic risk comprehensive index and the dynamic correlation index, the risk level is divided according to the threshold, the individual risk is analyzed in the group epidemic situation, the patient's own condition is considered, the group transmission trend is combined, and multi-dimensional accurate stratification of risk is realized.

[0059] S4, risk judgment threshold optimization: the infectious disease risk threshold is calibrated, and then the model and the features are iteratively optimized through clinical feedback, and a standardized scheme is formed.

[0060] In a specific embodiment, the calibration of the infectious disease risk threshold is performed as follows: the infectious disease risk threshold comprises a preset risk composite index threshold, a preset correlation index threshold, and a preset risk threshold; historical case data is queried from a database, and the queried historical case data is constructed into a verification set; the number of consistent cases of risk stratification and actual outcome, the number of actual severe cases but non-high-risk cases, and the number of actual non-infectious disease infections but high-risk cases in the historical case data are counted; and each group of candidate thresholds is set, each group of candidate thresholds comprising each group of candidate risk composite index thresholds, each group of candidate correlation index thresholds, and each group of candidate risk thresholds.

[0061] The degree of coincidence corresponding to each group of candidate thresholds is calculated by the Kappa coefficient calculation formula, the candidate risk composite index threshold corresponding to the maximum degree of coincidence is selected from each group of candidate risk composite index thresholds for the risk composite index threshold, the candidate correlation index threshold corresponding to the maximum degree of coincidence is selected from each group of candidate correlation index thresholds for the correlation index threshold, and the candidate risk threshold corresponding to the maximum degree of coincidence is selected from each group of candidate risk thresholds for the risk threshold. Thus, the candidate risk composite index threshold, the correlation index threshold, and the risk threshold corresponding to the maximum degree of coincidence are taken as the calibrated infectious disease risk threshold.

[0062] It should be noted that the setting process of the preset risk composite index threshold, the preset correlation index threshold, and the preset risk threshold is the same as the setting process of the dynamic correlation index threshold, and will not be described in detail here.

[0063] In a specific embodiment, the model and features are iteratively optimized through clinical feedback, and a standardized scheme is formed, which is performed as follows: clinical application data of risk stratification based on the calibrated infectious disease risk threshold in S3 is collected, the features and weights of the dynamic risk composite index in S2 and the calculation rules of the dynamic correlation index and the weighted score in S3 are optimized according to the clinical application data, and the optimized model parameters, feature rules, calibration thresholds, and clinical intervention paths are integrated into a standardized scheme.

[0064] It should be noted that, assuming that a hospital applies the calibrated threshold value to carry out S3 risk stratification during the influenza epidemic period, the clinical data shows that the actual severe rate of 18% of non-influenza vaccine recipients in the medium-risk patient group is significantly higher than the stratification expectation of 8%, and when the dynamic correlation index adopts a 14-day time window, the correlation strength of patients in high-prevalence areas is underestimated. The specific optimization process is: for the dynamic risk comprehensive index of S2, a new influenza vaccination history feature is added, with 0 points for non-vaccination and 1 point for vaccination, and the weight is set to 10%; for the dynamic correlation index of S3, the time window for calculating the spatial overlap degree is shortened from 14 days to 7 days, and in the weighted score formula, the weight of the dynamic risk comprehensive index is increased from 40% to 50%. The optimized standardized scheme formed by integration clearly states that the vaccination history needs to be extracted from the preventive vaccination record field of the electronic medical record, the time window parameter of the dynamic risk comprehensive index is fixed at 7 days, and the body temperature and respiratory symptoms of medium and high-risk patients need to be monitored daily. The calibrated infectious disease risk threshold table is embedded in the hospital HIS system together with the above rules. The example is only for illustrative purposes and is not the only limitation.

[0065] A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the steps of the method when executing the computer program.

[0066] Please refer to Figure 2 The electronic medical record-based infectious disease risk prediction analysis system comprises the following modules: a multi-modal electronic medical record preprocessing module, a dynamic risk feature construction module, an infectious disease risk prediction module, a risk judgment threshold optimization module, and a database.

[0067] The multi-modal electronic medical record preprocessing module is connected with the dynamic risk feature construction module and the database, the dynamic risk feature construction module is connected with the infectious disease risk prediction module and the database, the infectious disease risk prediction module is connected with the risk judgment threshold optimization module and the database, and the risk judgment threshold optimization module is connected with the database.

[0068] The multi-modal electronic medical record preprocessing module is used to query the electronic medical records of each patient within a set time period from a multi-source database, and preprocess the electronic medical records corresponding to each patient;

[0069] The dynamic risk feature construction module is used to quantify the basic disease risk of each patient based on the preprocessed electronic medical records corresponding to each patient, and to evaluate the key symptom change rate of each patient within a set time period, and then to calculate the dynamic risk comprehensive index of each patient;

[0070] The infectious disease risk prediction module is used to evaluate the dynamic correlation of each patient with the group infectious disease, and then to perform multi-dimensional stratification analysis on the infectious disease risk of each patient;

[0071] A risk judgment threshold optimization module is configured to calibrate the infectious disease risk threshold, and further optimize the model and features through clinical feedback iteration, and form a standardized scheme.

[0072] A database is configured to store the underlying diseases of the patients with the specified infectious disease in a historical time period, and further store the population data of the specified infectious disease in the set time period, and further store the historical case data.

[0073] In the risk judgment threshold optimization process, the historical cases are called from the database to construct a verification set, the Kappa coefficient is used to calculate the coincidence degree of different candidate thresholds to calibrate the risk threshold, the clinical application data is collected to optimize the features and weights of the dynamic risk comprehensive index and the calculation rules of the dynamic correlation index, and is integrated into a standardized scheme, which is beneficial to realize the dynamic adaptation of the threshold, continuously correct the model deviation through the clinical feedback, form a unified operation specification, and guarantee the consistency and practicality of the risk assessment in different scenarios.

[0074] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the protection scope of the technical solutions of the embodiments of the present application.

Claims

1. An electronic case-based infectious disease risk prediction analysis method, Comprising, characterized in that, further comprising: S1, multi-modal electronic medical record preprocessing: querying the electronic medical records of each patient in a set time period from a multi-source database, and preprocessing the electronic medical records corresponding to each patient; S2, dynamic risk feature construction: based on the preprocessed electronic medical records corresponding to each patient, quantifying the risk of the underlying diseases corresponding to each patient, and evaluating the change rate of the key symptoms corresponding to each patient in a set time period, and then calculating the dynamic risk comprehensive index corresponding to each patient; S3, infectious disease risk prediction: evaluating the dynamic correlation of each patient with the group infectious disease, and then multi-dimensionally analyzing the infectious disease risk corresponding to each patient; S4, risk judgment threshold optimization: by calibrating the infectious disease risk threshold, and then iteratively optimizing the model and features through clinical feedback, a standardized scheme is formed.

2. The electronic case-based infectious disease risk prediction analysis method of claim 1, wherein, The quantification of the risk of the underlying diseases corresponding to each patient is as follows: Extract the underlying disease information of each patient from the preprocessed electronic medical records, query the underlying diseases of the specified infectious disease patients in a set historical time period, and the specified infectious disease patients include specified infectious disease severe patients and specified infectious disease non-severe patients; Screen out the specified infectious disease patients belonging to the underlying diseases contained in the underlying disease information; Adopt the contingency table to count the proportion of each type of underlying disease in the specified infectious disease severe patients and the proportion in the specified infectious disease non-severe patients; Divide the proportion of each type of underlying disease in the specified infectious disease severe patients by the proportion of each type of underlying disease in the specified infectious disease non-severe patients to obtain the relative correlation coefficient of each type of underlying disease and the severe incidence rate; Calculate the relative correlation coefficient of each type of underlying disease and the severe incidence rate according to the pre-set normalization formula to obtain the differentiated risk coefficient; Query the disease severity level corresponding to each type of underlying disease; Obtain the non-linear relationship between the disease course and the infectious disease severe risk, and then use the survival analysis model to fit the disease course years and the severe risk data of the historical cases to obtain the quantification curve, and finally construct the disease course correlation factor calculation method to calculate the disease course correlation factor corresponding to each type of underlying disease, and then integrate to calculate the underlying disease risk value corresponding to each patient through the underlying disease quantification formula. 3.The electronic case-based infectious disease risk prediction analysis method of claim 2, wherein, The specific process of evaluating the change rate of the key symptoms corresponding to each patient in a set time period is as follows: Extract the key symptom records in the set time period from the preprocessed electronic medical records, and mark the time point of the first occurrence of each symptom in the key symptoms and the severity quantification value of each record; Obtain the time interval of each record collection from the electronic medical records of each patient, and calculate the key symptom change rate corresponding to each type of symptom of each patient at each record time through the quantification formula corresponding to the single symptom dynamic change value; Based on the key symptom change rate corresponding to each patient in a set time period, the key symptom change rate corresponding to each patient in a set time period is calculated through the key symptom change rate quantification formula. 4.The electronic case-based infectious disease risk prediction analysis method of claim 3, wherein, The specific process of calculating the dynamic risk comprehensive index corresponding to each patient is as follows: Based on the corresponding basic disease risk of each patient and the corresponding key symptom change rate in the set time period, the corresponding basic disease risk and the key symptom change rate of each patient are calculated by weighting, and then the dynamic risk comprehensive index corresponding to each patient is obtained. 5.The electronic case-based infectious disease risk prediction analysis method of claim 4, wherein, The dynamic relevance of each patient to the epidemic situation of the population is evaluated, and the specific process is as follows: Query the group specified infectious disease data in the set time period from the database; The group specified infectious disease data includes the average daily new case number of the group, the proportion of virus variant strains, the incidence of severe cases, and the distribution of spatiotemporal transmission hotspots; Align the basic disease risk index and the key symptom change rate of each patient with the group specified infectious disease data on the time axis, take the time point of the first appearance of symptoms of each patient as the reference point, and define the correlation window of each set time period before and after the reference point, thereby establishing the time correlation between the individual risk index and the development trend of the group specified infectious disease; Calculate the correlation strength of the key symptom change rate of each patient and the average daily new case number of the group in the correlation window by Pearson time series correlation coefficient, denoted as the first correlation strength; Calculate the correlation strength of the corresponding treatment area of each patient and the hot spot area of the group specified infectious disease transmission in the correlation window corresponding to the set time range by spatial overlap, denoted as the second correlation strength; Weight the first correlation strength and the second correlation strength to obtain the dynamic relevance index. When the dynamic relevance index is greater than or equal to the preset dynamic relevance index threshold, it indicates that the dynamic relevance of each patient to the epidemic situation of the population is close, otherwise, the dynamic relevance is not close. 6.The electronic case-based infectious disease risk prediction analysis method of claim 5, wherein, The multi-dimensional stratified analysis of the infectious disease risk corresponding to each patient is as follows: Take the dynamic risk comprehensive index corresponding to each patient as the endogenous risk core index, and take the dynamic relevance index corresponding to each patient as the exogenous risk index of the group, and construct a two-dimensional stratification system of endogenous and exogenous; Divide the dynamic risk comprehensive index corresponding to each patient into low, medium and high levels according to the preset risk comprehensive index threshold; Divide the dynamic relevance index corresponding to each patient into weak, medium and strong levels according to the preset relevance index threshold; When the level of the dynamic risk comprehensive index corresponding to a patient is high, and the level of the dynamic relevance index corresponding to the patient at this time is strong, the patient faces a very high risk of infectious diseases. When the level of the dynamic risk comprehensive index corresponding to a patient is low, and the level of the dynamic relevance index corresponding to the patient at this time is weak, the patient faces a very low risk of infectious diseases. Weight the dynamic risk comprehensive index and the dynamic relevance index corresponding to each patient in the remaining combination, and then divide the weighted calculation result into low risk, medium risk and high risk according to the preset risk threshold, to obtain the infectious disease risk corresponding to each patient.

7. The electronic case-based infectious disease risk prediction analysis method of claim 6, wherein, The specific process of calibrating the infectious disease risk threshold is as follows: The infectious disease risk threshold includes a preset risk composite index threshold, a preset correlation index threshold, and a preset risk threshold, each group of candidate thresholds includes a group of candidate risk composite index thresholds, a group of candidate correlation index thresholds, and a group of candidate risk thresholds; The Kappa coefficient calculation formula is used to calculate the degree of coincidence corresponding to each group of candidate thresholds. For the risk composite index threshold, the candidate risk composite index threshold corresponding to the maximum degree of coincidence is selected from the group of candidate risk composite index thresholds; For the correlation index threshold, the candidate correlation index threshold corresponding to the maximum degree of coincidence is selected from the group of candidate correlation index thresholds; For the risk threshold, the candidate risk threshold corresponding to the maximum degree of coincidence is selected from the group of candidate risk thresholds. Thus, the candidate risk composite index threshold, the correlation index threshold, and the risk threshold corresponding to the maximum degree of coincidence are used as the calibrated infectious disease risk threshold. 8.The electronic case-based infectious disease risk prediction analysis method of claim 7, wherein, The model and features are iteratively optimized through clinical feedback, and a standardized scheme is formed. The specific process is as follows: Clinical application data of risk stratification based on the calibrated infectious disease risk threshold is collected, and the features and weights of the dynamic risk composite index in S2 are optimized, and the calculation rules of the dynamic correlation index and the weighted score in S3 are optimized. The optimized model parameters, feature rules, calibrated thresholds, and clinical intervention paths are integrated into a standardized scheme. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8.

10. An infectious disease risk prediction analysis system for performing the infectious disease risk prediction analysis method according to any one of claims 1 to 8, characterized by, The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8.

Images