Preoperative anesthesia risk assessment method and system based on electronic medical record information

Abstract

The present invention relates to the technical field of medical record data risk assessment, and specifically to a preoperative anesthesia risk assessment method and system based on electronic medical record information. The electronic medical record data and anesthesia risk assessment level of patients undergoing preoperative anesthesia in a database are obtained, and an anesthesia risk assessment model is trained using the GBDT algorithm. During the training process, the impact of medical history and medical history combination on the anesthesia risk assessment level is considered, so as to optimize the loss function in each iteration, and improve the model training accuracy by evaluating the importance value of medical record data and adaptively adjusting the loss function, so as to enhance the model's sensitivity to medical history information. After the training is completed, the model can accurately predict the anesthesia risk assessment level, thereby assisting anesthesiologists in formulating safer and more reasonable anesthesia plans.

Description

Technical Field

[0001] The present invention relates to the technical field of medical record data risk assessment, and in particular to a preoperative anesthesia risk assessment method and system based on electronic medical record information. Background Art

[0002] With the rapid development of medical information technology, the electronic medical record system has become an indispensable part of modern hospitals. It records in detail key information such as the patient's medical history, treatment process and results. With the continuous improvement of living standards, the number of people who need painless diagnosis and treatment and surgical operations is also increasing every year. However, since anesthesia is a high-risk operation, patients with different disease conditions have different tolerance to anesthesia. Therefore, it is necessary to analyze the patient's electronic medical record information and fully understand the patient's medical history, so as to conduct preoperative anesthesia risk assessment and assist doctors in formulating anesthesia plans.

[0003] When the existing technology classifies the electronic medical records for anesthesia risk assessment, it usually uses a trained model to directly assess the anesthesia risk assessment level. However, in the process of model training, the supervised ensemble learning algorithm GBDT is usually used to directly learn the historical electronic medical record data and its anesthesia risk assessment level, so as to achieve the training of the GBDT model. However, due to individual differences, and different medical histories and medical history combinations may have different effects on anesthesia risks, the traditional GBDT algorithm does not fully consider the importance of medical history and medical history combinations during the training process, which will result in limited predictive ability of the obtained training model, affecting the credibility of anesthesia risk assessment, and thus affecting the accuracy of doctors in formulating anesthesia plans. Summary of the invention

[0004] In order to solve the technical problem that the traditional GBDT algorithm does not fully consider the importance of medical history and medical history combination during the training process, which will lead to limited prediction ability of the obtained training model, affect the credibility of anesthesia risk assessment, and further affect the accuracy of doctors in formulating anesthesia plans, the purpose of the present invention is to provide a preoperative anesthesia risk assessment method and system based on electronic medical record information, and the technical scheme adopted is as follows:

[0005] Obtain the medical history and anesthesia risk assessment level corresponding to each historical electronic medical record data;

[0006] Based on the GBDT algorithm, the model is trained on all historical electronic medical record data; in the initial iteration process, the anesthesia risk assessment level of all historical electronic medical record data is classified based on the GBDT algorithm to obtain the classification results;

[0007] For each iteration process after the initial iteration process: in the classification results of the previous iteration process, analyze the medical history of the historical electronic medical record data in each category to determine the medical history combination that appears frequently in each category; comprehensively analyze the difference in the number of occurrences of each medical history in different categories, as well as the difference in the number of occurrences of the medical history combination in each category and the number of occurrences in other categories, to determine the importance value of each historical electronic medical record data in each iteration process;

[0008] For the current iteration process, the classification accuracy, importance value and number of iterations of each historical electronic medical record data in the classification results of all previous iteration processes are combined to correct the loss function in the current iteration process to obtain a corrected loss function; the GBDT algorithm is iterated based on the corrected loss function. If the classification accuracy of all historical electronic medical record data in the classification results obtained in the current iteration process meets the preset conditions, the iteration is stopped to obtain a trained model;

[0009] The electronic medical record data to be evaluated are analyzed based on the trained model to obtain the anesthesia risk assessment level of the electronic medical record data to be evaluated.

[0010] Furthermore, in the initial iteration process, all historical electronic medical record data are classified into anesthesia risk assessment levels based on the GBDT algorithm to obtain classification results, including:

[0011] In each historical electronic medical record data, when a certain medical history exists in the historical electronic medical record data, the element value at the corresponding position of the medical history in the feature vector is set to 1, otherwise it is set to 0, so as to obtain the feature vector corresponding to each historical electronic medical record data;

[0012] The feature vectors corresponding to all historical electronic medical record data are used as the input of the GBDT algorithm to obtain the classification results under the initial iterative process.

[0013] Furthermore, the method for obtaining the medical history combination includes:

[0014] In each category of the classification results, obtain the medical history sequence corresponding to each historical electronic medical record data, and obtain other frequent item sets except one item set corresponding to all medical history sequences based on the FP-Growth algorithm;

[0015] The medical history corresponding to each frequent item set is regarded as a medical history combination.

[0016] Furthermore, the method for obtaining the importance value includes:

[0017] Determine the important factors for each medical history based on the differences in the number of occurrences of each medical history in different categories;

[0018] For the category with the lowest anesthesia risk assessment level, analyze the changes in the number of occurrences of the medical history combination in this category in other categories, and determine the importance value of each historical electronic medical record data in combination with the important factors of the medical history in this category;

[0019] The categories other than the category with the lowest anesthesia risk assessment level are taken as target categories. In any target category, based on the difference between the number of occurrences of each medical history combination in the target category and the number of occurrences in other categories, and combined with the difference between the important factors of the medical history in each medical history combination in the target category and the important factors in other categories, the important parameters of each medical history combination are determined;

[0020] In each target category, the importance value of each historical electronic medical record data is calculated by comprehensively considering the important parameters of all medical history combinations and the important factors of the medical history in each historical electronic medical record data.

[0021] Furthermore, the method for obtaining the important factors includes:

[0022] In each category, the ratio of the number of occurrences of each medical history to the number of historical electronic medical record data in the category was taken as the occurrence frequency of each medical history;

[0023] For any medical history, the difference between the frequency of occurrence of the medical history in each category and the frequency of occurrence in the category with the lowest anesthesia risk assessment level is normalized to obtain the important factors of the medical history.

[0024] Furthermore, for the category with the lowest anesthesia risk assessment level, the change in the number of occurrences of the medical history combination in this category in other categories is analyzed, and the importance value of each historical electronic medical record data is determined in combination with the important factors of the medical history in this category, including:

[0025] For any medical history combination in this category, calculate the ratio of the number of times this medical history combination appears in each other category to the number of all historical electronic medical record data in each other category as the frequency of occurrence of this medical history combination, and use the standard deviation of the frequency of occurrence of this medical history combination in all other categories as the fluctuation factor corresponding to this medical history combination;

[0026] The normalized value of the cumulative sum of the fluctuation factors corresponding to all medical history combinations in the category is used as the first important sum value corresponding to the category;

[0027] For any historical electronic medical record data in this category, the product of the sum of the important factors of all medical histories in the historical electronic medical record data and the first important sum value corresponding to the category is used as the importance value corresponding to the historical electronic medical record data.

[0028] Furthermore, the method for obtaining the important parameters includes:

[0029] For any target category, all categories with anesthesia risk assessment levels lower than the level corresponding to the target category are used as comparison categories for the target category;

[0030] For each medical history combination in the target category, calculate the ratio of the number of times each medical history combination appears in the target category to the number of all historical electronic medical record data in the target category as the first frequency, calculate the ratio of the number of times each medical history combination in the target category appears in each comparison category to the number of historical electronic medical record data in each comparison category as the second frequency, and normalize the difference between the first frequency and the second frequency as the frequency difference value corresponding to each medical history combination in the target category in each comparison category;

[0031] The normalized value of the difference between the important factor of each medical history in each medical history combination in the target category and the important factor in each comparison category is used as the change amplitude value of each medical history in each medical history combination, and the sum of the change amplitude values ​​of all medical histories in each medical history combination is used as the change difference value of each medical history combination in the target category in each comparison category;

[0032] In this target category, the product of the frequency difference value and the change difference value of each medical history combination is taken as the important value of each medical history combination in each comparison category; the average of the important values ​​of each medical history combination in all comparison categories is taken as the important parameter of each medical history combination.

[0033] Furthermore, in each target category, the importance value of each historical electronic medical record data is calculated by comprehensively considering the important parameters of all medical history combinations and the important factors of the medical history in each historical electronic medical record data, including:

[0034] In each target category, the normalized value of the accumulated sum of the important parameters of all medical history combinations is used as the second most important sum value corresponding to each target category;

[0035] For any historical electronic medical record data in each target category, the product of the sum of the important factors of all medical histories in the historical electronic medical record data and the second important sum corresponding to each target category is used as the importance value of the historical electronic medical record data.

[0036] Furthermore, the method for obtaining the modified loss function includes:

[0037] All iteration processes before the current iteration process are used as comparison processes, and the difference between the number of iterations of the current iteration process and the number of iterations of each comparison process is used as the number difference;

[0038] Compare the anesthesia risk assessment level of each historical electronic medical record data with the anesthesia risk assessment level in the classification result of the iterative process before each comparison process, and determine the error rate of each historical electronic medical record data in each comparison process;

[0039] Based on the error rate of each historical electronic medical record data in each comparison process and the number difference corresponding to each comparison process, determine the attention of each historical electronic medical record data in each comparison process, wherein the attention is positively correlated with the error rate, and the attention is negatively correlated with the number difference;

[0040] The product of the accumulated sum of the attention of each historical electronic medical record data in all comparison processes and the importance value of each historical electronic medical record data in the current iteration process is used as the weight of each historical electronic medical record data in the current iteration process;

[0041] In the current iteration process, the loss function is weighted based on the weights of all historical electronic medical record data to obtain a corrected loss function in the current iteration process.

[0042] The present invention also proposes a preoperative anesthesia risk assessment system based on electronic medical record information, the system comprising a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, and at least one instruction, at least one program, code set or instruction set, when loaded and executed by the processor, implements any one of the steps of a preoperative anesthesia risk assessment method based on electronic medical record information.

[0043] The present invention has the following beneficial effects:

[0044] By comprehensively acquiring historical electronic medical record data and its corresponding anesthesia risk assessment level, a rich and real data foundation is provided for subsequent model training. Model training is performed on all historical electronic medical record data based on the GBDT algorithm, and the classification results of the first iteration process are obtained, which sets the starting point for model training. Since the GBDT algorithm uses a loss function to calculate the prediction error of the current model in each iteration, and since different medical histories and medical history combinations have different effects on the anesthesia risk assessment level, in the present invention, the loss function is optimized in each iteration process to improve the training accuracy of each iteration process, which is used to improve the overall prediction accuracy of the final model. First, for each iteration process after the initial iteration process, the classification results obtained in the previous iteration process are analyzed, and the medical history combinations that frequently appear in the historical electronic medical record data of each category are analyzed. This medical history combination may be a medical history combination that has a significant impact on the anesthesia risk assessment level. Similarly, since different medical histories also have different effects on the anesthesia risk assessment level, the differences in the number of occurrences of each medical history in different categories are comprehensively analyzed, and the importance value of each historical electronic medical record data in each iteration process is evaluated in combination with the differences in the number of occurrences of medical history combinations in different categories. This process enhances the model's sensitivity to medical history information, thereby improving the subsequent model's ability to predict the corresponding anesthesia risk assessment level by analyzing medical record data. Then, for the current iteration process, the classification accuracy of the historical iteration process, the importance value of the historical electronic medical record data, and the number of iterations are comprehensively considered to correct the loss function in the current iteration process and obtain a corrected loss function. This adaptive adjustment mechanism enables the model to continuously optimize its decisions during the training process and reduce misjudgments. Then, the GBDT algorithm is iterated based on the corrected loss function. If the classification result obtained by the iteration meets the preset conditions, the iteration can be stopped, and the model training is completed. Finally, based on the trained model, the electronic medical record data to be evaluated is analyzed to obtain an accurate anesthesia risk assessment level, which can assist doctors in formulating more reasonable and safe anesthesia plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A method flow chart of a method for preoperative anesthesia risk assessment based on electronic medical record information provided by one embodiment of the present invention;

[0047] Figure 2 A method flow chart of a method for obtaining an importance value provided by an embodiment of the present invention;

[0048] Figure 3 A method flow chart of a method for obtaining the importance value of each electronic medical record data in other categories except the category with the lowest anesthesia risk assessment level provided by one embodiment of the present invention;

[0049] Figure 4 A system block diagram of a preoperative anesthesia risk assessment system based on electronic medical record information provided by an embodiment of the present invention;

[0050] Figure 5 A schematic diagram of the system structure of a preoperative anesthesia risk assessment system based on electronic medical record information provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following is a detailed description of the specific implementation method, structure, features and effects of a preoperative anesthesia risk assessment method and system based on electronic medical record information proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0052] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0053] The following is a detailed description of a method and system for preoperative anesthesia risk assessment based on electronic medical record information provided by the present invention in conjunction with the accompanying drawings.

[0054] See also Figure 1 , which shows a method flow chart of a method for preoperative anesthesia risk assessment based on electronic medical record information provided by an embodiment of the present invention, the method comprising the following steps:

[0055] Step S1: Obtain the medical history and anesthesia risk assessment level corresponding to each historical electronic medical record data.

[0056] Before performing surgery on a patient, or before performing a painless examination or painless operation, the anesthesiologist needs to fully understand the patient's condition and physical condition, assess the patient's tolerance to anesthesia, and develop a targeted anesthesia plan. In order to improve the efficiency of the assessment, the electronic medical record data of patients undergoing preoperative anesthesia and the anesthesia risk assessment level in the database are usually studied to obtain a training model. In this way, when determining the anesthesia risk of the patient to be assessed, the risk level can be obtained more quickly to assist the anesthesiologist in formulating anesthesia plans.

[0057] In view of the different medical histories recorded in the electronic medical record data, such as hypertension, heart disease, diabetes, allergens, and drug use, which will affect the anesthesia risk, each historical electronic medical record data is first analyzed to extract the corresponding medical history, that is, the corresponding keywords. Based on the JIEBA algorithm and the prior knowledge of the medical history, the medical history of each historical electronic medical record data can be extracted. For example, a certain electronic medical record data records "the patient has a history of hypertension for 10 years and has been taking nifedipine controlled-release tablets for a long time to control blood pressure. Recently, he was hospitalized due to a heart attack, accompanied by mild heart failure." Then the word segmentation result may be: "The patient / has / had / a / history / of / hypertension / for / 10 / years / , / has / been / taking / nifedipine / controlled-release tablets / to / control / blood pressure / . / Recently / was / hospitalized / due / to / heart / attack / , / accompanied / by / mild / heart failure." For words such as stop words, they are not used as keywords and should be filtered out. According to prior knowledge, the final extracted medical history should be: "hypertension, heart disease, heart failure". For some specific medical terms or drug names, JIEBA may not be able to identify them completely accurately. In this case, you can consider using a custom dictionary to expand JIEBA's vocabulary to improve the accuracy of medical history extraction. For electronic medical record data in the database that requires preoperative anesthesia, it also has an anesthesia risk assessment level, which can be obtained by the anesthesiologist through the evaluation of its medical records. The anesthesia risk assessment level can be set to 6 levels, with labels 1-6 respectively, where label 1 indicates the lowest anesthesia risk, and label 6 indicates the highest anesthesia risk.

[0058] It should be noted that the JIEBA algorithm is a technical means well known to those skilled in the art, and the specific process will not be described here; the specific assessment process of the anesthesia risk assessment level can be evaluated according to the patient's overall physical health status grading standards promulgated by the American Society of Anesthesiologists, and the specific process will not be described here; in the embodiments of the present invention, the collection and acquisition of electronic medical record data are authorized by relevant users, and the process does not violate relevant laws and regulations, and does not violate public order and good customs.

[0059] Step S2: Perform model training on all historical electronic medical record data based on the GBDT algorithm; in the initial iteration process, classify all historical electronic medical record data according to anesthesia risk assessment levels based on the GBDT algorithm to obtain classification results.

[0060] The Gradient Boosting Decision Tree (GBDT) algorithm can handle complex nonlinear relationships and high-dimensional data. It improves the accuracy of prediction by constructing multiple decision trees and integrating their prediction results. Therefore, the GBDT algorithm is used to learn from historical electronic medical record data. In each iteration, it is adjusted according to the results of the previous iteration to minimize the classification error, thereby obtaining a more accurate classification model. The trained GBDT model can then classify the anesthesia risk assessment level of new electronic medical record data, and the classification results can help anesthesiologists more accurately assess the patient's anesthesia risk and thus develop a more appropriate anesthesia plan.

[0061] Therefore, in an embodiment of the present invention, model training is performed on all historical electronic medical record data based on the GBDT algorithm, and the iterative process is initialized at the same time. That is, in the initial iterative process, all historical electronic medical record data are classified into anesthesia risk assessment levels based on the GBDT algorithm to obtain the classification results in the initial iterative process.

[0062] Preferably, in one embodiment of the present invention, at the initial number of iterations, all historical electronic medical record data are classified according to anesthesia risk assessment levels based on the GBDT algorithm to obtain classification results, including:

[0063] Machine learning algorithms rely on the characteristics of data for prediction or classification, so it is necessary to convert text-based medical histories into digital data that the algorithm can process. Therefore, in each historical electronic medical record data, when there is a certain medical history in the historical electronic medical record data, the element value of the medical history at the corresponding position in the feature vector is set to 1, otherwise it is set to 0, so as to obtain the feature vector corresponding to each historical electronic medical record data. For example: the types of all medical histories included in all historical electronic medical record data are hypertension, diabetes, heart disease, asthma, heart failure, and allergies; and the medical history in a certain historical electronic medical record data is hypertension, heart disease, and heart failure, then the feature vector corresponding to the historical electronic medical record data is [1,0,1,0,1,0].

[0064] Finally, the feature vectors corresponding to all historical electronic medical record data are used as the input of the GBDT algorithm, that is, the data is fitted based on the initial weak learner (usually a decision tree), so as to obtain the classification results under the initial iterative process. Each category in the classification result represents a level of anesthesia risk assessment. Among them, the loss function can use an exponential loss function, thereby realizing the initialization of the iterative process.

[0065] The formula model of the exponential loss function includes: ;in, represents the exponential loss function, N represents the total number of samples, Represents the true value of the i-th sample; Represents the predicted value of the i-th sample; Represents an exponential function with the natural constant e as the base.

[0066] It should be noted that in the initial iteration process, the GBDT algorithm is the same as the original algorithm and is a well-known technology, and the specific process will not be repeated here.

[0067] Step S3: For each iteration process after the initial iteration process: analyze the medical history of the historical electronic medical record data in each category in the classification results of the previous iteration process to determine the medical history combinations that appear frequently in each category; comprehensively analyze the differences in the number of occurrences of each medical history in different categories, as well as the differences in the number of occurrences of the medical history combination in each category and the number of occurrences in other categories, to determine the importance value of each historical electronic medical record data in each iteration process.

[0068] Due to the different effects of different medical histories on preoperative anesthesia risk assessment, the importance of different medical histories in the process of preoperative anesthesia risk assessment will also vary. For example, the history of cardiovascular disease (hypertension, heart disease) and respiratory disease history (asthma, chronic obstructive pulmonary disease) are all factors that anesthesiologists need to focus on during evaluation. These medical histories may significantly increase the risk of complications during anesthesia; at the same time, there are some medical histories that may not work alone for preoperative anesthesia evaluation, but when they appear at the same time as other medical histories, they will seriously interfere with the assessment of preoperative anesthesia risk. For example, if hypertension and diabetes appear at the same time in the historical electronic medical record data, then the stability of the cardiovascular system in response to anesthetic drugs may be more complex and difficult to predict; at the same time, given that the GDBT algorithm iteratively adds weak learners (decision trees) and adjusts them according to the results of the previous iteration process to minimize the classification error, thereby gradually improving the performance of the model. Therefore, in order to improve the prediction accuracy of the model finally obtained, in the embodiment of the present invention, each iteration process after the initial iteration process is adjusted to improve the accuracy of the model: first, for each iteration process after the initial iteration process, the medical history of the historical electronic medical record data in each category is analyzed in the classification results of the previous iteration process to determine the medical history combination that appears frequently in each category. Then, a comprehensive analysis is made of the impact of the medical history alone and the medical history combination on the anesthesia risk assessment level of the electronic medical record data, so as to quantify the importance value of each historical electronic medical record data in each iteration process, which is helpful to correct the loss function in the iteration process based on the importance value of the historical electronic medical record data in the iteration process in the subsequent process, thereby improving the prediction accuracy of the model.

[0069] For each iteration process after the initial iteration process, the medical history of the historical electronic medical record data in each category is analyzed in the classification results of the previous iteration process to determine the frequent medical history combinations in each category. In this way, the correlation between these medical history combinations and the anesthesia risk assessment level can be captured. This correlation may reflect the co-occurrence pattern of certain medical history combinations, which is crucial for distinguishing different categories of anesthesia risks.

[0070] Preferably, in one embodiment of the present invention, the method for obtaining the medical history combination includes:

[0071] In each category of the classification results, the medical history sequence corresponding to each historical electronic medical record data is obtained, and the frequent item sets of all items except one item set corresponding to all medical history sequences are obtained based on the FP-Growth algorithm, such as medical history sequence 1 {hypertension, heart disease, heart failure}, medical history sequence 2 {hypertension, diabetes, heart disease}, medical history sequence 3 {heart disease, heart failure, recurrence of heart disease}. Assuming that the minimum support is 2 (indicating that the item set must appear in at least two medical history sequences), the frequent item sets finally extracted (except one item set) are: {hypertension, heart disease} and {heart disease, heart failure}.

[0072] Then the medical history corresponding to each frequent item set is taken as a medical history combination, and the obtained medical history combination represents the medical history that often appears together in each category.

[0073] It should be noted that obtaining frequent itemsets based on the FP-Growth algorithm is a well-known technology, and the minimum support threshold can be set to 0.6 (indicating the frequency ratio of the item set), and the specific process will not be repeated here.

[0074] After mining the medical history combinations within each category in the classification results, by comprehensively analyzing the differences in the number of occurrences of each medical history in different categories and the differences between the number of occurrences of the medical history combination in each category and the number of occurrences in other categories, the contribution of each medical history and each medical history combination to distinguishing the level of anesthesia risk assessment can be evaluated, thereby obtaining the importance value of each historical electronic medical record data, which helps to converge to the optimal solution faster during the iteration process and improve the prediction accuracy of the model.

[0075] Preferably, in one embodiment of the present invention, the method for obtaining the importance value includes:

[0076] See also Figure 2 , which shows a method flow chart of a method for obtaining an importance value in one embodiment of the present invention, the method comprising the following steps:

[0077] Step S301: Determine the important factors of each medical history based on the difference in the number of occurrences of each medical history in different categories.

[0078] If a certain medical history appears frequently in a certain category, it can be said that the medical history has a more obvious correlation with the anesthesia risk assessment level corresponding to the category, and thus the importance of the medical history in the category will be greater; therefore, in each category, the ratio of the number of occurrences of each medical history to the number of historical electronic medical record data in the category is used as the frequency of occurrence of each medical history. At this time, the greater the frequency of occurrence, the more frequently each medical history appears in the historical electronic medical record data in the category, and the higher its importance.

[0079] If the frequency of a certain medical history in a category is significantly higher than the frequency of a certain medical history in the category with the lowest level of anesthesia risk assessment, it can be considered that the change in the level of anesthesia risk assessment is affected by this medical history. Therefore, for any medical history, the difference between the frequency of the medical history in each category and the frequency of the medical history in the category with the lowest level of anesthesia risk assessment is normalized to obtain the important factor of the medical history. At this time, the difference is positive and the larger it is, the greater the important factor of the medical history, indicating that the frequency of the medical history in each category is higher than the frequency of the medical history in the category with the lowest level of anesthesia risk assessment. For categories other than the category with the lowest level of anesthesia risk assessment, the medical history is more likely to be the reason for the anesthesia risk assessment level being higher than the lowest level of anesthesia risk assessment. Since the difference here may be positive or negative, the normalization method uses Function. After the calculations in the above steps, the important factors of each medical history in each category can be obtained. For the category with the lowest level of anesthesia risk assessment, the important factors of each medical history are calculated to be 0.5.

[0080] Step S302: For the category with the lowest anesthesia risk assessment level, analyze the changes in the number of occurrences of the medical history combination in this category in other categories, and determine the importance value of each historical electronic medical record data in combination with the important factors of the medical history in this category.

[0081] For the category with the lowest anesthesia risk assessment level, it can be considered that the historical electronic medical record data in this category does not have anesthesia risk, so the medical history combination in this category will not actually have an impact on the assessment of the anesthesia risk assessment level, and the importance of the medical history combination in the historical electronic medical record data under this category will also be low. However, in order to more accurately measure the importance of the historical electronic medical record data in the category with the lowest anesthesia risk assessment level, it can be reflected by the frequency changes of the medical history combination in this category in categories of other anesthesia risk assessment levels. If the frequency changes are relatively consistent, then it can be considered that the medical history combination has a smaller impact on the assessment of the anesthesia risk assessment level.

[0082] Therefore, for any medical history combination in this category, the ratio of the number of times this medical history combination appears in each other category to the number of all historical electronic medical record data in each other category is calculated as the frequency of occurrence of this medical history combination, and the standard deviation of the frequency of occurrence of this medical history combination in all other categories is used as the fluctuation factor corresponding to this medical history combination. At this time, the smaller the fluctuation factor, the more similar the frequency of occurrence of this medical history combination in all other categories is, and the performance is more stable. Then the smaller the impact of this medical history combination on the anesthesia risk assessment level will be.

[0083] Finally, the normalized value of the cumulative sum of the fluctuation factors corresponding to all medical history combinations in the category is used as the first important sum value corresponding to the category. The smaller the first important sum value is, the lower the importance of the medical history combination in the category is, and the smaller the impact on the anesthesia risk assessment level is. At the same time, for any historical electronic medical record data in the category, the sum of the important factors of all medical histories in the historical electronic medical record data and the product of the first important sum value corresponding to the category are used as the importance value corresponding to the historical electronic medical record data; the smaller the sum of the important factors of all medical histories in the historical electronic medical record data is, the lower the importance of the historical electronic medical record data is. At the same time, if the first important sum value of the category is also smaller, it means that the importance of the medical history combination in the category is also lower. Combining the two can more comprehensively measure the importance value of each historical electronic medical record data in the category. The larger the importance value is, the more important the historical electronic medical record data is. On the contrary, the smaller the importance value is, the less important the historical electronic medical record data is.

[0084] It should be noted that the normalization method here can be used Normalization function.

[0085] Step S303: For categories other than the category with the lowest anesthesia risk assessment level, determine the importance value of each historical electronic medical record data based on the difference in the number of occurrences of each medical history in different categories, the difference in the number of occurrences of medical history combinations in different categories, and the important factors of the medical history.

[0086] See also Figure 3 , which shows a method flow chart of a method for obtaining the importance value of each historical electronic medical record data in other categories except the category with the lowest anesthesia risk assessment level in one embodiment of the present invention, the method comprises the following steps:

[0087] Step S3031: Take all categories except the category with the lowest anesthesia risk assessment level as target categories. In any target category, determine the important parameters of each medical history combination based on the difference between the number of occurrences of each medical history combination in the target category and the number of occurrences in other categories, and in combination with the difference between the important factors of the medical history in each medical history combination in the target category and the important factors in other categories.

[0088] For any target category, all categories with anesthesia risk assessment levels lower than the level corresponding to the target category are taken as comparison categories for the target category. If the frequency of occurrence of the medical history combination in the target category is greater than the frequency of occurrence of the medical history combination in the comparison category, it can be regarded as that the increase in the anesthesia risk assessment level of the target category is caused by this medical history combination.

[0089] Therefore, for each medical history combination in the target category, the ratio of the number of times each medical history combination appears in the target category to the number of all historical electronic medical record data in the target category is calculated as the first frequency, and the ratio of the number of times each medical history combination in the target category appears in each comparison category to the number of historical electronic medical record data in each comparison category is calculated as the second frequency. Then, the difference between the first frequency and the second frequency is normalized and used as the frequency difference value corresponding to each medical history combination in the target category in each comparison category. Since the difference here may be positive or negative, the normalization method uses Function, when the difference is positive and the larger it is, the larger the frequency difference value is, indicating that the medical history combination in the target category is more important than each comparison category, and the greater the impact on the anesthesia risk assessment level.

[0090] The frequency difference value analyzes the impact of the medical history combination on the anesthesia risk assessment level, and then the impact of a single medical history in the medical history combination on the anesthesia risk assessment level can be analyzed. Since the important factors of the medical history are calculated based on the frequency of occurrence of the medical history in the category to which it belongs, it can reflect its impact on the anesthesia risk assessment level. Therefore, the difference between the important factors of each medical history in each medical history combination in the target category and the important factors in each comparison category is normalized as the change amplitude value of each medical history in each medical history combination in the target category. The difference here may be positive or negative, so the normalization method is still used. Function, at this time, the difference is positive and the larger it is, the larger the change amplitude value is, indicating that the medical history in the medical history combination in the target category has a greater contribution to the anesthesia risk assessment level than that in the comparison category. Then the sum of the change amplitude values ​​of all medical histories in each medical history combination is used as the change difference value of each medical history combination in the target category in each comparison category. The change difference value combines the change amplitude values ​​of all medical histories in each medical history combination. The larger the change difference value is, the greater the contribution of the medical history combination as a whole in the target category to the anesthesia risk assessment level compared to that in the comparison category.

[0091] Next, in the target category, the product of the frequency difference value and the change difference value of each medical history combination is used as the corresponding importance value of each medical history combination in each comparison category. The importance value takes into account both the difference in the frequency of occurrence of the medical history combination in different categories and the change of each medical history important factor in the medical history combination in different categories. Therefore, the importance value can more comprehensively and accurately represent the impact of each medical history combination in each target category on anesthesia risk assessment compared with each comparison category. The larger the importance value, the greater the impact.

[0092] At this point, each medical history combination in the target category has a corresponding importance value in each comparison category. Finally, the average of the importance values ​​of each medical history combination in all comparison categories can be used as the important parameter of each medical history combination. The larger the important parameter, the greater the impact of the medical history combination in the target category on the anesthesia risk assessment level.

[0093] Step S3032: In each target category, the importance value of each historical electronic medical record data is calculated by comprehensively analyzing the important parameters of all medical history combinations and the important factors of the medical history in each historical electronic medical record data.

[0094] Based on step S3031, the important parameters of each medical history combination in each target category can be calculated, and then the accumulated and normalized values ​​of the important parameters of all medical history combinations in each target category can be used as the second important sum value corresponding to each target category. The larger the second important sum value, the higher the importance of the medical history combination in the target category, and the greater its contribution to the evaluation of the anesthesia risk assessment level. The normalization method here can be used Normalization function.

[0095] Then, for any historical electronic medical record data in each target category, the sum of the important factors of all medical histories in the historical electronic medical record data and the product of the second important sum corresponding to each target category are used as the importance value of the historical electronic medical record data. The smaller the sum of the important factors of all medical histories in the historical electronic medical record data, the lower the importance of the historical electronic medical record data. At the same time, if the second important sum of the target category is also smaller, it means that the importance of the medical history combination in the target category is also lower. Combining the two, the importance value of each historical electronic medical record data in the target category can be more comprehensively measured. The larger the importance value, the more important the historical electronic medical record data is. Conversely, the smaller the importance value, the less important the historical electronic medical record data is.

[0096] At this point, for each iteration process after the initial iteration process, the importance value of each historical electronic medical record data in each iteration process can be obtained.

[0097] Step S4: For the current iteration process, the classification accuracy, importance value and number of iterations of each historical electronic medical record data in the classification results of all previous iteration processes are comprehensively considered, and the loss function in the current iteration process is corrected to obtain a corrected loss function; the GBDT algorithm is iterated based on the corrected loss function. If the classification accuracy of all historical electronic medical record data in the classification results obtained in the current iteration process meets the preset conditions, the iteration is stopped to obtain a trained model.

[0098] In the iterative process of machine learning, the loss function is a key indicator to measure the difference between the model's predicted value and the true value. For the current iteration of the GBDT algorithm, by combining the classification accuracy of the classification results in all previous iterations, the importance value of the historical electronic medical record data, and the number of iterations, the loss function in the current iteration process is corrected, so that the model can focus more on samples that are difficult to classify or samples that have a greater impact on the final result. This dynamic adjustment helps to improve the robustness and accuracy of the model. Then, by continuously correcting the loss function and iterating based on the corrected loss function, the model can gradually learn the complex patterns and features in the data. Especially when processing electronic medical records, the diversity and complexity of the data are very high. Dynamically adjusting the loss function helps the model better adapt to this complexity, thereby improving the performance (i.e., generalization ability) on unseen data, that is, the electronic medical record data to be evaluated. Finally, set the preset condition, that is, when the classification accuracy reaches a certain threshold, stop the iteration to prevent the model training process from going on indefinitely, thereby obtaining a trained model.

[0099] Preferably, in one embodiment of the present invention, the method for obtaining the modified loss function includes:

[0100] All iterations before the current iteration are used as comparison processes, and the difference between the number of iterations of the current iteration and the number of iterations of each comparison process is used as the number difference. The size of the number difference can be used to evaluate the timeliness of the comparison process.

[0101] The anesthesia risk assessment level of each historical electronic medical record data is compared with the anesthesia risk assessment level of each historical electronic medical record data in the classification results of the iteration process before each comparison process, and the error rate of each historical electronic medical record data in each comparison iteration process is determined. An example of the error rate calculation method is given: If a comparison process is the third iteration process, it is determined whether the anesthesia risk assessment level of each historical electronic medical record data is consistent with the anesthesia risk assessment level of each historical electronic medical record data in the initial iteration process and the second iteration process. If it is consistent with the anesthesia risk assessment level in the initial iteration process and inconsistent with the anesthesia risk assessment level in the second iteration process, then the error rate is 0.5 or 50%. It should be noted that if the comparison process is the initial iteration process, the error rate of each historical electronic medical record data in the comparison process is set to 1, because it is the initial iteration and there is no iteration result for comparison.

[0102] Then, based on the error rate of each historical electronic medical record data in each comparison process and the number difference corresponding to each comparison process, the attention degree of each historical electronic medical record data in each comparison process is determined, and the attention degree is positively correlated with the error rate, and the attention degree is negatively correlated with the number difference. The formula model of attention degree can be specifically, for example:

[0103] ;

[0104] in, Indicates the attention paid to each historical electronic medical record data in each comparison process; Indicates the error rate of each historical electronic medical record data in each comparison process; Indicates the difference in the number of times each comparison process corresponds to.

[0105] In the formula model of attention, when the error rate of each historical electronic medical record data in each comparison process is greater, then the historical electronic medical record data should be given higher attention, so the error rate is placed on the numerator; when the difference in the number of times corresponding to the comparison process is smaller, it means that the comparison process is closer to the current iteration process, and the reference value will be higher, so the difference in the number of times is placed on the denominator, and then the resulting ratio is used as the attention of each historical electronic medical record data in each comparison process. The larger the value, the higher the attention that should be paid to the historical electronic medical record data.

[0106] Next, the product of the cumulative sum of the attention of each historical electronic medical record data in all comparison processes and the importance value of each historical electronic medical record data in the current iteration process is used as the weight of each historical electronic medical record data in the current iteration process. The larger the cumulative value of the attention corresponding to the historical electronic medical record data, the greater the importance, which means that the historical electronic medical record data is more representative, has a greater impact on the results of the anesthesia risk assessment level, and is easily misclassified into other anesthesia risk assessment levels, requiring a higher attention level, and thus a greater weight.

[0107] Finally, in the current iteration process, the loss function is weighted based on the weights of all historical electronic medical record data, thereby obtaining a corrected loss function in the current iteration process: In an embodiment of the present invention, for the iteration process after the initial iteration process, the loss function can use an exponential loss function, multiplying the weight of each historical electronic medical record data by the error corresponding to each historical medical record data under the loss function, obtaining the loss term of each historical electronic medical record data, and accumulating the loss terms of all historical electronic medical record data to obtain a corrected loss function. The formula model of the corrected loss function includes:

[0108] ;

[0109] in, represents the corrected loss function; represents the amount of historical electronic medical record data; represents the weight of the i-th historical electronic medical record data; It represents the anesthesia risk assessment level of the i-th historical electronic medical record data in the current iteration, that is, the predicted value; represents the anesthesia risk assessment level of the i-th historical electronic medical record data, that is, the true value; Represents an exponential function with the natural constant e as the base.

[0110] In the modified loss function, the error value corresponding to each historical electronic medical record data under the loss function is first calculated. Since the weight of each historical electronic medical record data represents its impact on the anesthesia risk assessment result, the error value is weighted using the weight of each historical electronic medical record data to obtain the loss term of each historical electronic medical record data. , and then the loss terms are accumulated to correct the original loss function, so that the gradient direction obtained by the corrected loss function can be adjusted in the direction represented by the important data points, which helps the model converge to the optimal solution faster during the training process; at the same time, it improves the accuracy of the model's anesthesia risk assessment level assessment of electronic medical record data.

[0111] At this point, the iteration of the GBDT algorithm in the current iteration process can be completed according to the modified loss function of the current iteration process, and the classification results of all historical electronic medical record data in the current iteration process can be obtained. If the accuracy of the classification results of the historical electronic medical record data at this time reaches the preset condition, and the preset condition is set to the accuracy greater than or equal to the preset accuracy threshold, then it is considered that the prediction ability of the model meets the requirements, and the iteration can be stopped to obtain a trained model. The method for obtaining the accuracy includes: determining whether the anesthesia risk assessment level to which each historical electronic medical record data belongs is consistent with the anesthesia risk assessment level to which it belongs in the classification result of the current iteration process, and taking the ratio of the number of consistent historical electronic medical record data to the total number of historical electronic medical record data as the accuracy, and the accuracy can be set in the form of a percentage. When the accuracy is greater than or equal to the preset accuracy threshold, the iteration is stopped to obtain a trained model.

[0112] It should be noted that the iterative process of the GBDT algorithm is a technical means well known to those skilled in the art and will not be elaborated here; the preset accuracy threshold is set to 98%, and the specific value can be adjusted according to the implementation scenario and is not limited here.

[0113] Step S5: Analyze the electronic medical record data to be evaluated based on the trained model to obtain the anesthesia risk assessment level of the electronic medical record data to be evaluated.

[0114] Based on the above steps, a trained model can be obtained, and then the feature vector of the electronic medical record data to be evaluated is obtained, and the feature vector is used as the input of the trained model to obtain the classification result of the electronic medical record data to be evaluated, that is, the anesthesia risk assessment level. The anesthesiologist can then make further judgments based on the anesthesia risk assessment level of the electronic medical record data to be evaluated, thereby formulating a targeted anesthesia plan.

[0115] In summary, by comprehensively acquiring historical electronic medical record data and its corresponding anesthesia risk assessment level, a rich and real data foundation is provided for subsequent model training. Model training is performed on all historical electronic medical record data based on the GBDT algorithm, and the classification results of the first iteration process are obtained, which sets the starting point for model training. Since the GBDT algorithm uses a loss function to calculate the prediction error of the current model in each iteration, and since different medical histories and medical history combinations have different effects on the anesthesia risk assessment level, in the present invention, the loss function is optimized in each iteration process, thereby improving the training accuracy of each iteration process, which is used to improve the overall prediction accuracy of the final model. First, for each iteration process after the initial iteration process, analyze the classification results obtained in the previous iteration process, and the medical history combinations that frequently appear in the historical electronic medical record data of each category. This medical history combination may be a medical history combination that has a significant impact on the anesthesia risk assessment level. Similarly, since different medical histories also have different effects on the anesthesia risk assessment level, the differences in the number of occurrences of each medical history in different categories are comprehensively analyzed, and the importance value of each historical electronic medical record data in each iteration is evaluated in combination with the differences in the number of occurrences of medical history combinations in different categories. This process enhances the model's sensitivity to medical history information, thereby improving the subsequent model's ability to predict the corresponding anesthesia risk assessment level by analyzing medical record data. Then, for the current iteration process, the loss function in the current iteration process is corrected based on the correct classification of the historical iteration process, the importance value of the historical electronic medical record data, and the number of iterations, and the corrected loss function is obtained. This adaptive adjustment mechanism enables the model to continuously optimize its decision-making during the training process and reduce misjudgments. Then, the GBDT algorithm is iterated based on the corrected loss function. If the classification result obtained by the iteration meets the preset conditions, the iteration can be stopped, and the model training is completed. Finally, based on the trained model, the electronic medical record data to be evaluated is analyzed to obtain an accurate anesthesia risk assessment level, which can assist doctors in formulating more reasonable and safe anesthesia plans.

[0116] An embodiment of the present invention also provides a preoperative anesthesia risk assessment system based on electronic medical record information, the system comprising a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, and at least one instruction, at least one program, code set or instruction set, when loaded and executed by the processor, implements the steps of any one of the preoperative anesthesia risk assessment methods based on electronic medical record information.

[0117] See also Figure 4 , which shows a system block diagram of a preoperative anesthesia risk assessment system based on electronic medical record information provided by an embodiment of the present invention, including a data acquisition module 401: used to implement step S1 in the above method; an iterative initialization module 402: used to implement step S2 in the above method; an importance analysis module 403: used to implement step S3 in the above method; a model correction and training module 404, used to implement step S4 in the above method; an anesthesia risk assessment module 405, used to implement step S5 in the above method.

[0118] It should be noted that the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the preoperative anesthesia risk assessment system based on electronic medical record information and the preoperative anesthesia risk assessment method based on electronic medical record information provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0119] See also Figure 5 , which shows a system structure diagram of a preoperative anesthesia risk assessment system based on electronic medical record information provided by an embodiment of the present invention, including a processor 500, a memory 501, a bus 502 and a communication interface 503, wherein the processor 500, the communication interface 503 and the memory 501 are connected via the bus 502; wherein the memory 501 may include a high-speed random access memory, the bus 502 may be an ISA bus, a PCI bus or an EISA bus, etc., and the processor 500 may be an integrated circuit chip having signal processing capabilities.

[0120] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0121] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A preoperative anesthesia risk assessment method based on electronic medical record information, characterized in that: The method comprises: Obtain the medical history and anesthesia risk assessment level corresponding to each historical electronic medical record data; Based on the GBDT algorithm, the model is trained on all historical electronic medical record data; in the initial iteration process, the anesthesia risk assessment level of all historical electronic medical record data is classified based on the GBDT algorithm to obtain the classification results; For each iteration process after the initial iteration process: in the classification results of the previous iteration process, analyze the medical history of the historical electronic medical record data in each category to determine the medical history combination that appears frequently in each category; comprehensively analyze the difference in the number of occurrences of each medical history in different categories, as well as the difference in the number of occurrences of the medical history combination in each category and the number of occurrences in other categories, to determine the importance value of each historical electronic medical record data in each iteration process; For the current iteration process, the classification accuracy, importance value and number of iterations of each historical electronic medical record data in the classification results of all previous iteration processes are combined to correct the loss function in the current iteration process to obtain a corrected loss function; the GBDT algorithm is iterated based on the corrected loss function. If the classification accuracy of all historical electronic medical record data in the classification results obtained in the current iteration process meets the preset conditions, the iteration is stopped to obtain a trained model; The electronic medical record data to be evaluated is analyzed based on the trained model to obtain the anesthesia risk assessment level of the electronic medical record data to be evaluated; The method for obtaining the modified loss function includes: All iteration processes before the current iteration process are used as comparison processes, and the difference between the number of iterations of the current iteration process and the number of iterations of each comparison process is used as the number difference; Compare the anesthesia risk assessment level of each historical electronic medical record data with the anesthesia risk assessment level in the classification result of the iterative process before each comparison process, and determine the error rate of each historical electronic medical record data in each comparison process; Based on the error rate of each historical electronic medical record data in each comparison process and the number difference corresponding to each comparison process, determine the attention of each historical electronic medical record data in each comparison process, wherein the attention is positively correlated with the error rate, and the attention is negatively correlated with the number difference; The product of the accumulated sum of the attention of each historical electronic medical record data in all comparison processes and the importance value of each historical electronic medical record data in the current iteration process is used as the weight of each historical electronic medical record data in the current iteration process; In the current iteration process, the loss function is weighted based on the weights of all historical electronic medical record data to obtain a corrected loss function in the current iteration process.

2. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 1, characterized in that: In the initial iteration process, all historical electronic medical record data are classified into anesthesia risk assessment levels based on the GBDT algorithm to obtain classification results, including: In each historical electronic medical record data, when a certain medical history exists in the historical electronic medical record data, the element value at the corresponding position of the medical history in the feature vector is set to 1, otherwise it is set to 0, thereby obtaining the feature vector corresponding to each historical electronic medical record data; The feature vectors corresponding to all historical electronic medical record data are used as the input of the GBDT algorithm to obtain the classification results under the initial iterative process.

3. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 1, characterized in that: The method for obtaining the medical history combination includes: In each category of the classification results, obtain the medical history sequence corresponding to each historical electronic medical record data, and obtain other frequent item sets except one item set corresponding to all medical history sequences based on the FP-Growth algorithm; The medical history corresponding to each frequent item set is regarded as a medical history combination.

4. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 1, characterized in that: The method for obtaining the importance value includes: Determine the important factors for each medical history based on the differences in the number of occurrences of each medical history in different categories; For the category with the lowest anesthesia risk assessment level, analyze the changes in the number of occurrences of the medical history combination in this category in other categories, and determine the importance value of each historical electronic medical record data in combination with the important factors of the medical history in this category; The categories other than the category with the lowest anesthesia risk assessment level are taken as target categories. In any target category, based on the difference between the number of occurrences of each medical history combination in the target category and the number of occurrences in other categories, and combined with the difference between the important factors of the medical history in each medical history combination in the target category and the important factors in other categories, the important parameters of each medical history combination are determined; In each target category, the importance value of each historical electronic medical record data is calculated by comprehensively considering the important parameters of all medical history combinations and the important factors of the medical history in each historical electronic medical record data.

5. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 4, characterized in that: The method for obtaining the important factors includes: In each category, the ratio of the number of occurrences of each medical history to the number of historical electronic medical record data in the category was taken as the occurrence frequency of each medical history; For any medical history, the difference between the frequency of occurrence of the medical history in each category and the frequency of occurrence in the category with the lowest anesthesia risk assessment level is normalized to obtain the important factors of the medical history.

6. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 4, characterized in that: For the category with the lowest anesthesia risk assessment level, the change in the number of occurrences of the medical history combination in this category in other categories is analyzed, and the importance value of each historical electronic medical record data is determined in combination with the important factors of the medical history in this category, including: For any medical history combination in this category, calculate the ratio of the number of times this medical history combination appears in each other category to the number of all historical electronic medical record data in each other category as the frequency of occurrence of this medical history combination, and use the standard deviation of the frequency of occurrence of this medical history combination in all other categories as the fluctuation factor corresponding to this medical history combination; The normalized value of the cumulative sum of the fluctuation factors corresponding to all medical history combinations in the category is used as the first important sum value corresponding to the category; For any historical electronic medical record data in this category, the product of the sum of the important factors of all medical histories in the historical electronic medical record data and the first important sum value corresponding to the category is used as the importance value corresponding to the historical electronic medical record data.

7. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 4, characterized in that: The method for obtaining the important parameters includes: For any target category, all categories with anesthesia risk assessment levels lower than the level corresponding to the target category are used as comparison categories for the target category; For each medical history combination in the target category, calculate the ratio of the number of times each medical history combination appears in the target category to the number of all historical electronic medical record data in the target category as the first frequency, calculate the ratio of the number of times each medical history combination in the target category appears in each comparison category to the number of historical electronic medical record data in each comparison category as the second frequency, and normalize the difference between the first frequency and the second frequency as the frequency difference value corresponding to each medical history combination in the target category in each comparison category; The normalized value of the difference between the important factor of each medical history in each medical history combination in the target category and the important factor in each comparison category is used as the change amplitude value of each medical history in each medical history combination, and the sum of the change amplitude values ​​of all medical histories in each medical history combination is used as the change difference value of each medical history combination in the target category in each comparison category; In this target category, the product of the frequency difference value and the change difference value of each medical history combination is taken as the important value of each medical history combination in each comparison category; the average of the important values ​​of each medical history combination in all comparison categories is taken as the important parameter of each medical history combination.

8. A method for preoperative anesthesia risk assessment based on electronic medical record information according to claim 4, characterized in that: In each target category, the important parameters of all medical history combinations and the important factors of the medical history in each historical electronic medical record data are integrated to calculate the importance value of each historical electronic medical record data, including: In each target category, the normalized value of the accumulated sum of the important parameters of all medical history combinations is used as the second most important sum value corresponding to each target category; For any historical electronic medical record data in each target category, the product of the sum of the important factors of all medical histories in the historical electronic medical record data and the second important sum corresponding to each target category is used as the importance value of the historical electronic medical record data.

9. A preoperative anesthesia risk assessment system based on electronic medical record information, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and when the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor, the steps of a preoperative anesthesia risk assessment method based on electronic medical record information as described in any one of claims 1 to 8 are implemented.

Images