Weak index calculation method and device, electronic equipment and storage medium

Abstract

The invention provides a weakness index calculation method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a plurality of target weakness risk terms from the medical data of a user; according to the position of each target weakness risk term in the systematic clinical term system, constructing a semantic network of the plurality of target weakness risk terms; calculating the weight of each target weakness risk term according to the position of the target weakness risk term in the semantic network; the average value of the weights of all the target weakness risk terms is calculated, a weakness index of the user is obtained, the weakness index is used for predicting whether the user is in a weakness state or not, the weakness risk terms are obtained from the medical data of the user, the weights are given to the weakness risk terms through a semantic network, and the weakness risk terms are obtained. The prediction meanings of different weakness risk terms are evaluated, and the weakness risk terms with different granularities are brought into calculation through a semantic network, so that the prediction precision of the weakness index is improved.

Description

Technical Field

[0001] The present application belongs to the field of medical care technology, and specifically relates to a frailty index calculation method, device, electronic device and storage medium. Background Art

[0002] Frailty is a type of geriatric syndrome, which is related to the aging of skeletal muscles, and its core content is impaired physical function. Frailty is also a physiological decline state across multiple systems, which will have a significant impact on subsequent diagnosis and treatment activities under stress. It is very important to quickly and accurately assess the degree of frailty in elderly patients. For example, early detection of frailty can help to risk stratify and intervene in elderly patients.

[0003] It is generally believed that advanced age, falls, pain, malnutrition, sarcopenia, coexistence of multiple diseases, multiple medications, decreased activity function, sleep disorders, anxiety, depression and social support are all related to frailty. Therefore, there are many risk factors for frailty and the related factors are mixed. Although the use of frailty screening tools can better predict the degree of frailty and prognosis of the elderly, it requires additional manual data collection or subjective evaluation, which increases costs and limits large-scale implementation. Therefore, the electronic frailty index (eFI) based on the electronic medical record (EMR) has gradually developed, which can automatically obtain frailty characteristics and has good predictive efficiency.

[0004] However, the electronic weakness index in the prior art is calculated based on English disease words and has low prediction accuracy. Therefore, there is an urgent need in the art for a calculation method of the electronic weakness index with high prediction accuracy. Summary of the invention

[0005] In view of this, the purpose of the present application is to provide a frailty index calculation method, device, electronic device and storage medium to improve the prediction accuracy of the frailty index.

[0006] In a first aspect, an embodiment of the present application provides a method for calculating a frailty index, the method comprising: obtaining multiple target frailty risk terms from a user's medical data; constructing a semantic network of the multiple target frailty risk terms based on the position of each target frailty risk term in a systematic clinical terminology system; calculating the weight of each target frailty risk term based on the position of the target frailty risk term in the semantic network; calculating the average of the weights of all target frailty risk terms to obtain the frailty index of the user, wherein the frailty index is used to predict whether the user is in a frail state.

[0007] The above-mentioned frailty index calculation method obtains frailty risk terms in the user's medical data, assigns weights to the frailty risk terms through the semantic network of the frailty risk terms, evaluates the predictive significance of different frailty risk terms, and incorporates frailty risk terms of different granularities into the calculation through the semantic network, thereby improving the prediction accuracy of the frailty index.

[0008] In an optional embodiment of the present application, the semantic network of the multiple frailty risk terms is constructed based on the position of each target frailty risk term in the systematic clinical terminology, including: obtaining the standard term corresponding to each target frailty risk term, the standard term is the term corresponding to the frailty risk term in the systematic clinical terminology system; constructing the semantic network, the semantic network is the path for all standard terms to reach the general term, and the general term is the common parent term of all standard terms.

[0009] In the above embodiment, a semantic network is constructed through the semantic paths of frailty risk factors in a systematic clinical terminology system, which can accurately calculate the semantic similarities of different frailty risk factors, obtain the weights of frailty risk factors, and improve the prediction accuracy of the frailty index.

[0010] In an optional embodiment of the present application, the obtaining of multiple target frailty risk terms from the user's medical records includes: obtaining multiple frailty risk terms from the user's medical records; screening multiple target frailty risk terms from the multiple frailty risk terms based on prediction targets and / or frailty association relationships, wherein the frailty association relationships are associations between frailty risk terms and frail status, and the prediction targets are coping strategies to be taken when the user is determined to be in a frail status.

[0011] In the above embodiment, the frailty risk terms obtained from the medical records are screened to obtain frailty risk factors with better prediction effects, thereby optimizing the frailty assessment strategy and improving the prediction accuracy of the frailty index.

[0012] In an optional embodiment of the present application, the method further includes: matching a preset frailty risk term pool with the medical data to obtain a plurality of target frailty risk terms, wherein the frailty risk term pool includes a plurality of preset terms. In the above embodiment, matching based on the frailty risk term pool can focus on multiple sub-topics or elements in the information structure, thereby achieving more comprehensive acquisition of frailty risk terms.

[0013] In an optional embodiment of the present application, the method also includes: mapping each term in the preset term set to a systematic clinical term system to obtain a standard term set; mapping each term in the standard term set to a first set to obtain a first term set, the first set including a private term set, a Chinese tangram term set, and a Chinese systematic clinical term system; mapping each term in the standard term set to a second set to obtain a second term set, the second set including a unified medical language system; mapping each term in the second term set to a third set to obtain a third term set, the third set including a Chinese medical term system; mapping each term in the second term set to a fourth set to obtain a fourth term set, the fourth set including a human phenotype term set; mapping each term in the fourth term set to a fifth set to obtain a fifth term set, the fifth set including a Chinese human phenotype term set; using the preset term set, the standard term set, the first term set, the second term set, the third term set, the fourth term set and the fifth term set as the frailty risk term pool.

[0014] In the above embodiment, a synonym set of terms in the factor pool is obtained through cross-mapping, the conceptual scope of each term is expanded, and the term coverage of the frailty risk term pool is improved, so as to obtain frailty risk factors more comprehensively.

[0015] In an optional embodiment of the present application, if the frailty index characterizes that the user is in a frail state, the method further includes: extracting terms from the user's medical data to obtain new terms; and adding the new terms to the frailty risk term pool.

[0016] In the above embodiment, when a user is determined to be in a frail state, term mining is performed in the user's medical data to discover potential phenotypic terms and improve the term coverage of the frailty risk term pool, thereby obtaining frailty risk factors more comprehensively.

[0017] In an optional embodiment of the present application, based on the prediction target, multiple target frailty risk terms are screened out from the multiple frailty risk terms, including: obtaining a term combination corresponding to the prediction target, the term combination is obtained by performing term extraction on the medical data of all patients under the prediction target; obtaining the frailty risk terms in the term combination from the multiple frailty risk terms to obtain the multiple target frailty risk terms.

[0018] In the above embodiment, frailty term mining is performed from the medical records of all patients under a specific prediction target, and frailty risk terms with high correlation to the prediction target can be mined, thereby accurately screening out frailty terms with predictive significance.

[0019] In a second aspect, an embodiment of the present application provides a frailty index calculation device, comprising: an acquisition module for acquiring multiple target frailty risk terms from the user's medical data; a construction module for constructing a semantic network of the multiple target frailty risk terms based on the position of each target frailty risk term in the systematic clinical terminology system; a weight calculation module for calculating the weight of each target frailty risk term based on the position of the target frailty risk term in the semantic network; an index calculation module for calculating the average of the weights of all target frailty risk terms to obtain the user's frailty index, which is used to predict whether the user is in a frail state.

[0020] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory and a processor, the processor being connected to the memory; the memory being used to store programs; and the processor being used to call the programs stored in the memory to execute any of the methods described in the first aspect.

[0021] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in any one of the first aspects is executed.

[0022] Other features and advantages of the present application will be described in the following description. The purpose and other advantages of the present application can be realized and obtained through the structures specifically pointed out in the written description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The above and other purposes, features and advantages of the present application will be more clear as shown in the drawings.

[0024] Figure 1 A schematic diagram of a frailty index calculation method provided in an embodiment of the present application is shown; Figure 2 A schematic diagram showing a cross-mapping between term sets provided in an embodiment of the present application is shown; Figure 3 A schematic diagram of a process for screening frailty risk terms based on frailty association relationships provided in an embodiment of the present application is shown; Figure 4 A schematic diagram of a semantic network provided by an embodiment of the present application is shown; Figure 5A schematic block diagram of a frailty index calculation device provided in an embodiment of the present application is shown; Figure 6 A schematic block diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0025] Hereinafter, embodiments of the present application are described in detail with reference to the accompanying drawings. It should be noted that although the same elements are shown in different drawings, they will be represented by the same reference numerals. In the following description, specific details such as detailed configuration and components are provided to only help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications to the embodiments described herein can be made without departing from the scope of the present application. In addition, for clarity and conciseness, the description of well-known functions and configurations is omitted. The terms described below are defined in consideration of the functions in the present application, and may be different according to the user, the user's intention or custom. Therefore, the definition of the term should be determined based on the content throughout this specification.

[0026] The present application may have various modifications and various embodiments, in which the embodiments are described in detail below with reference to the accompanying drawings. However, it should be understood that the present application is not limited to the embodiments, but includes all modifications, equivalents and substitutes within the scope of the present application.

[0027] To facilitate understanding, the relevant terms and concepts involved in the embodiments of the present application are first introduced below.

[0028] The Frailty Index (FI) is a representative of the cumulative defect model, which defines frailty as the cumulative health defects determined in the Comprehensive Geriatric Assessment (CGA). It can be developed from most aging databases and mainly calculates the number of health defects, which can be symptoms, weight, comorbidities, disability, cognitive impairment, and laboratory, imaging or electrocardiographic abnormalities, generally involving dozens of items. The sum of the scores of each item divided by the actual number of items is the FI score. The more defects accumulate, the greater the probability of frailty. However, it takes more than an hour to complete all CGA item inquiries. Because of the time-consuming problem, the eFI in FI almost all selects assessment items that can be automatically obtained from the EMR.

[0029] The electronic frailty index (eFI) emphasizes that all frailty phenotype risk factors are automatically identified and scored from electronic medical records. The eFI can achieve better differentiation of frail populations. In addition to the frailty identification of the overall population, physicians can also use the eFI to predict postoperative adverse events, disease treatment and prognosis during hospitalization. Studies have shown that it has a certain discriminant effect. Of course, there are also literature proving that the discriminant effect of eFI is poor. However, this poor discriminant effect mainly involves treatments such as chemotherapy drugs and abdominal surgery. For this kind of prognosis discrimination with special treatments, there are generally risk factors that are more suitable for this field. Risk factors based on specialty fields may be more predictive than FRIs (Frailty Risk Items) in the general field.

[0030] There are many dimensions for conventional assessment of frailty, and the FRIs in the eFI are generally indicators that are easy to obtain directly in clinical practice. FRIs mainly include diagnosis and surgery codes, laboratory test items, and functional indicators such as cognitive deficits, visual and auditory deficits, and other deficits. At present, most studies advocate assessment from multiple dimensions of frailty, but the FRIs included in each study are quite different. All countries have borrowed or independently built electronic frailty assessment tools, verified them in their own medical systems, and then raised them to industry standards or national standards. However, there are currently no electronic frailty assessment tools in the Chinese environment, nor are there unified FRIs factor standards. Therefore, establishing FRIs factor standards suitable for the Chinese medical environment is one of the technical problems solved by the present invention.

[0031] To ensure that EMR data can effectively identify frailty states, EMR data elements need to be mapped to standardized terms that represent knowledge models or ontologies of frailty. Differences in clinical terminology standards affect both the transfer of knowledge in different countries and the automatic identification of relevant FRIs from EMRs. For example, compared to the 36-item eFI version in country A, the eFI in country B has not yet been standardized because frailty in that country is represented by clinical terms that are different from those in country A. Currently, there is little research on identifying standardized terminology in Chinese EMRs in the Chinese environment. Therefore, this application also establishes a standard for mapping frailty-related phenotypic features to standardized terminology in a Chinese environment.

[0032] Due to the reasons mentioned above, there is no feasible method for identifying frailty phenotypes in the Chinese environment.

[0033] like Figure 1 As shown, Figure 1 FIG. 2 is a flow chart showing a method for calculating a frailty index provided in an embodiment of the present application. Figure 1 As shown, the method includes steps 110 to 140: Step 110 , obtaining a plurality of target frailty risk terms from the user's medical data.

[0034] Obtain the user's medical data such as electronic medical records, doctor's orders, etc. Obtain multiple terms from the user's medical data. These terms are pre-specified, including clinical terms and non-clinical terms (such as symptoms described by patients, etc.). Use the acquired terms as target frailty risk terms. In this field, frailty risk terms are also called frailty risk factors (Frailty Risk Items, FRIs), which are indicators included in the statistics of eFI.

[0035] For example, a term library is pre-established, which contains pre-specified frailty-related terms, such as: not recognizing spouse and children, spatial orientation loss, temporal orientation loss, deterioration of ability to handle things, etc. Information is extracted from the user's electronic medical record, and multiple words are extracted from it. It is determined whether the extracted words include the terms in the term library. If an extracted word is in the term library, then the word is a target frailty risk term.

[0036] In one embodiment of the present application, the method further includes: matching a preset frailty risk term pool with the medical data to obtain a plurality of target frailty risk terms, wherein the frailty risk term pool includes a plurality of preset terms.

[0037] A matching method, such as a rule and terminology based algorithm (RTBA), is used to extract target frailty risk factors from medical data.

[0038] In the embodiment of the present application, a frailty risk term pool is pre-established, and the frailty risk term pool includes multiple frailty-related terms. The term information in the clinical text in the EMR is extracted, and the term recognition is realized based on the frailty risk term pool and the inverse maximum matching method.

[0039] Furthermore, the rule-based algorithm can also make negative judgments on terms in the text and determine whether the terms describe the patient himself or his family members.

[0040] In order to solve the problem that there is no FRIs factor standard applicable to the Chinese medical environment, this application establishes a frailty risk terminology pool through the following methods: setting frailty indicator dimensions through prior experience; selecting multiple frailty indicator dimensions and filling them with specific terms commonly used in medical data as FRIs under the frailty indicator dimensions.

[0041] Based on previous research, the inventors of this application set up multiple frailty indicator dimensions, as shown in Table 1 below. Table 1 is an example of a frailty indicator dimension shown in an embodiment of this application. It should be noted that the inclusion scope of FRIs is different according to different prediction purposes and outcome events. In other embodiments of this application, dimensions can also be set according to different situations.

[0042]

[0043] Table 1 Dimensions of frailty risk According to different usage scenarios, such as different prediction purposes or the difficulty of extracting terms from medical data, multiple frailty indicator dimensions are selected and FRIs within these dimensions are supplemented.

[0044] For example, dimensions such as "basic information", "comorbidity status", "symptoms and characteristics", "test indicators", and "assessment scales and questionnaires" are selected, and FRIs under each dimension are specified to generate a list of high-risk factors for frailty in the EMR, namely, a frailty risk term pool, as shown in Table 2. Table 2 shows an example of a frailty risk term pool.

[0045]

[0046] Table 2 Sources of each FRIs in the EMR and recommended methods of obtaining them The FRIs in Table 2 above are only used as examples. The embodiments of the present application can flexibly select specific FRIs, and various physiological system deficiency items available in EMR can be included. However, whether these deficiency items truly reflect the frail state can be judged and verified by the real-world diagnosis and treatment data driven by subsequent data.

[0047] This application integrates authoritative knowledge on frailty assessment in the elderly with the empirical research conclusions on frailty risk in elderly frail patients in the project implementation hospitals, to form an elderly frailty risk assessment project and tool that is more suitable for the Chinese environment.

[0048] Through the interoperability of clinical terminology, the assessment tool can more conveniently transfer knowledge and codes between hospitals, and can further sink the geriatric frailty assessment capabilities into the EMR used in clinical frontline, so that mainstream geriatric frailty risk assessment experience can be integrated into the clinical frontline workflow of grassroots hospitals.

[0049] In the above embodiment, matching based on the frailty risk term pool can focus on multiple sub-topics or elements in the information structure, thereby achieving more comprehensive acquisition of frailty risk factors.

[0050] Existing technologies focus on subtopics of a certain information structure, such as only focusing on medical ontology, knowledge management processes or text mining, etc.; or only focusing on elements of a certain information structure, such as time relationships, clinical concepts, vital signs, symptoms or diseases; or only focusing on a set of development technologies such as machine learning. In this solution, the scope of the task can be limited through matching solutions, focusing only on the identification of frailty-related phenotypic terms involved above, and whether these concepts express negation and family descriptions. At the same time, specifying the content of the frailty term pool can achieve multi-factor and multi-topic focus. For example, clinical terms can be added to the frailty risk term pool using a collection method, and term extraction through machine learning algorithms can add daily expressions to the frailty term pool.

[0051] In one embodiment of the present application, the method also includes: mapping each term in the preset term set to a systematic clinical term system to obtain a standard term set; mapping each term in the standard term set to a first set to obtain a first term set, the first set including a private term set, a Chinese tangram term set, and a Chinese systematic clinical term system; mapping each term in the standard term set to a second set to obtain a second term set, the second set including a unified medical language system; mapping each term in the second term set to a third set to obtain a third term set, the third set including a Chinese medical term system; mapping each term in the second term set to a fourth set to obtain a fourth term set, the fourth set including a human phenotype term set; mapping each term in the fourth term set to a fifth set to obtain a fifth term set, the fifth set including a Chinese human phenotype term set; using the preset term set, the standard term set, the first term set, the second term set, the third term set, the fourth term set and the fifth term set as the frailty risk term pool.

[0052] In order to more accurately identify frailty-related phenotypes, this application enriches the Chinese synonyms and hyponyms of the frailty phenotype terms involved in FRIs as a basis.

[0053] First, the initial terminology set, such as the initial frailty risk terms (FRIs) indicator pool, was mapped to SNOMED CT (Systematized Nomenclature of Medicine Clinical Terms) concepts.

[0054] The mapping relationship between FRIs and standardized terms can be obtained by conducting term mapping and SNOMED CT knowledge training for three clinical experts in the team, and then using the SNOMED CT browser to independently map FRIs. The mapping process uses a predefined FRIs mapping SCT standard to generate matching criteria with direct matching, indirect matching, or non-matching corresponding SNOMED CT items. The matching criteria for mapping are shown in Table 3.

[0055]

[0056] Table 3 Matching criteria between FRIs and SNOMED CT terminology concepts In Table 3, when a data element has only one SNOMED CT item, it is defined as a direct match; when a single data element has multiple SNOMED CT items, or the SNOMED CT item is semantically similar to the data element but not a direct match, it is defined as a partial match; when a data element cannot be matched to any SNOMED CT item, it is defined as a mismatch.

[0057] Two clinical experts independently mapped. After each round, a third clinical expert and medical record coder will be invited to discuss and analyze the causes and determine the final results of inconsistent mappings. After each round of mapping, the kappa consistency coefficient of the matching results of the two mappers will be calculated to continuously evaluate the reliability and consistency of the matching results of the two mappers.

[0058] Through the above consensus method, this application establishes a standard for mapping frailty-related phenotypic features to standardized terms in a Chinese environment. The concept of reliably mapping FRIs to the existing SNOMED CT can achieve many effects. For example, by mapping the data elements of FRIs to standardized terms in EMR (such as SNOMED CT), a more unified terminology identification method will be allowed to evaluate frailty risk factors in EMR. It is also possible to automatically calculate continuous frailty scores using assessment scale data that has been collected by doctors, thereby significantly improving the efficiency of frailty assessment.

[0059] Mapping these frailty phenotypes to standard medical ontology terminology sets not only facilitates the reuse of the frailty assessment knowledge system, but the standardized terminology also allows the electronic medical record data to be linked to natural language processing algorithms, thereby automatically monitoring and identifying frailty-related health problems from electronic medical records.

[0060] Each term in the initial frailty risk term index pool is mapped to a systematic clinical terminology system to obtain a standard terminology set. Based on SNOMED CT, the frailty-related phenotypes involved are mapped to the corresponding concepts of SNOMED CT by cross-mapping between terminology sets. The Chinese terms in the "Chinese SNOMED International Version 3.4" and "Chinese Tangram Terminology Set" are mapped to the corresponding concepts of SNOMED CT; the Chinese terms in the "Chinese Human Phenotype Ontology Consortium (CHPO)" are transitively cross-mapped to the corresponding concepts of SNOMED CT through the HPO (Human Phenotype Ontology Consortium) and UMLS (Unified Medical Language System) terminology sets. Figure 2 As shown, Figure 2 A schematic diagram showing a cross-mapping between term sets is shown.

[0061] Figure 2 In the above, each term in the standardized term set is mapped to the first set, and the first set includes multiple term sets, such as hospital private terminology, Chinese tangram term set, and Chinese systematic clinical term system. After mapping the standardized term set, the terms mapped in the first term set are obtained to obtain the first term set. The private terms of the hospital in this application are Chinese terms accumulated by the hospital in the medical process, such as clinical discovery terms and patient expression terms. This application does not limit the concept of private terms.

[0062] For example, there is a term called progressive cognitive deterioration in the initial frailty risk factor indicator pool, which is mapped to the systematic clinical terminology system to obtain the term cognitive impairment. Map cognitive impairment with the first set, assuming that cognitive impairment is mapped to inability to identify objects normally and mental retardation in private terms. In the Chinese tangram terminology set, it is mapped to reduced ability to handle things, and in the Chinese systematic clinical terminology system, it is mapped to cognitive impairment. Then obtain those terms in the first set that have a mapping relationship with cognitive impairment, and obtain the first term set: inability to identify daily objects, mental retardation, and reduced ability to handle things.

[0063] The standard terms are mapped to a second set such as the unified medical language system to obtain a second term set, and each term in the second term set is mapped to a third set such as the Chinese medical term system published by HITSA to obtain a third term set. For example, cognitive impairment is mapped to loss of language communication ability in the unified medical language system, and this term with a mapping relationship is obtained to obtain the second term set, and loss of language communication ability is mapped to the Chinese medical term system published by HITSA, such as the term is mapped to reduced ability to handle things and loss of time orientation, and these two terms are obtained as the third term set.

[0064] The terms in the second term set are mapped to a fourth set such as a human phenotype term set to obtain a fourth term set. Each term in the fourth term set is mapped to a fifth set such as a Chinese human phenotype term to obtain a fifth term set. Loss of language communication ability is mapped to a human phenotype term set such as loss of spatial orientation, and this term is used as the fourth term set. And loss of spatial orientation is mapped to the Chinese human phenotype term to obtain loss of spatial orientation and not recognizing spouse and children, to obtain a fifth term set.

[0065] The preset term set, standard term set, first term set, second term set, third term set, fourth term set and fifth term set are used as the final frailty risk term pool. For example, after the above enrichment of progressive cognitive function deterioration, the final frailty risk term pool includes: progressive cognitive function deterioration, inability to identify daily objects, cognitive impairment, mental retardation, decreased ability to handle things, loss of time orientation, loss of space orientation, and not recognizing spouse and children.

[0066] After cross-mapping, a set of Chinese synonyms for SNOMED CT frailty phenotype terms can be obtained.

[0067] Furthermore, when cross-mapping between Chinese and English, once the English target concept has no more subordinate concepts, the synonyms and hyponyms of the Chinese concept are incorporated into the current English target concept to increase the subsequent term detection rate.

[0068] In the above embodiment, a synonym set of terms in the factor pool is obtained through cross-mapping, the concept scope of each term is expanded, the term coverage of the frailty risk term pool is improved, and the frailty risk factors are obtained more comprehensively.

[0069] In one embodiment of the present application, the medical data includes electronic medical records, test data, imaging data, monitoring equipment data, and diagnosis and treatment data.

[0070] In the existing technology, frailty risk factors are usually extracted only from medical documents written by doctors, such as electronic medical records, but other data are still of great value. For example, structured data such as tests and examinations; imaging data such as head CT and MRI of the elderly may contain descriptions of brain sulci, cisterns and ventricles related to "brain atrophy"; some wearable devices in hospitals, as well as non-contact radio frequency imaging monitoring equipment, can monitor frailty phenotypes such as slow movement and low physical activity levels; various diagnosis and treatment information based on timeline changes, after discharge, patients often have multiple outpatient follow-up records, including tests and examinations of key outcome indicators.

[0071] Through these different modal data, the frailty phenotype can be identified in multimodal data. Through the frailty phenotype under multiple modalities, the frailty index can be calculated more comprehensively and accurately.

[0072] In one embodiment of the present application, the method of obtaining multiple target frailty risk terms from the user's medical records includes: obtaining multiple frailty risk terms from the user's medical records; screening multiple target frailty risk terms from the multiple frailty risk terms based on prediction targets and / or frailty association relationships, wherein the frailty association relationships are associations between frailty risk terms and frail status, and the prediction targets are coping strategies to be taken when the user is determined to be in a frail status.

[0073] Different FRIs have different predictive significance for frailty. For example, the inventors found that the granularity of most FRIs is a certain systemic disease, such as urinary system diseases, thyroid diseases, ischemic heart disease, etc. However, the conceptual granularity of these FRIs may have a better solution. For example, "left heart failure->heart failure->heart disease" are three medical concepts with an is-a relationship from fine to coarse concept granularity, that is, "left heart failure is a heart failure" and "heart failure is a heart disease". However, when "heart failure" itself is used as a FRI of an eFI, through feature selection based on real-world data, its parent concept (heart disease) or child concept (left heart failure) may have better predictive significance than its own concept (heart failure). Therefore, it is necessary to screen the frailty risk factors extracted from medical data. In this application, this kind of three is-a relationships from fine to coarse are called a bottom-to-top hierarchical relationship or a parent-child relationship.

[0074] For example, the frailty risk terms extracted from the medical data based on the frailty risk term pool are screened, and the remaining frailty risk terms after the screening are used as target frailty risk terms.

[0075] The screening methods in the present application include: screening based on predicted targets and screening based on frailty association relationships. Only one of the screening methods can be used, or both can be used at the same time. For example, a round of screening of frailty risk factors is first performed based on the predicted targets, and then a screening method based on frailty association relationships is used for the remaining frailty risk factors to finally obtain the target frailty risk factors.

[0076] Based on the prediction target, multiple target frailty risk terms are screened out from the multiple frailty risk terms, including: obtaining a term combination corresponding to the prediction target, the term combination being obtained by performing term extraction on the medical data of all patients under the prediction target; obtaining the frailty risk terms in the term combination from the multiple frailty risk terms to obtain the multiple target frailty risk terms.

[0077] The above-mentioned prediction targets include prediction purposes and outcome events in the medical field.

[0078] Frailty can often be defined by some adverse outcome events, and the mainstream frailty outcome events are falls, disability, dependence on life, need for long-term hospitalization and death. The outcome events of eFI can be set in the EMR environment during hospitalization, or in the community during the observation period of the corresponding retrospective study cohort. In the EMR, the entire hospitalization, or a specific perioperative period, or anesthesia safety may be the object of observation. Generally speaking, disability, long-term hospitalization, total hospital stay during hospitalization, discharge for higher-level care, falls during hospitalization, readmission rate and mortality rate are all mainstream indicators. In addition, frailty has become an important parameter to be considered in disciplines other than geriatrics (such as surgery, cardiac surgery and orthopedics), and these disciplines are increasingly in need of automatic frailty assessment tools to evaluate adverse events in various subspecialties. As shown in Table 4, Table 4 is a specific prediction purpose and outcome event provided in this application.

[0079]

[0080] Table 4 Prediction purpose and outcome events of eFI The optimal combination of FRIs may be different depending on the purpose of eFI prediction. In EMR, frailty outcome events that are easily and directly obtained include: death indicators (obtained from the medical record homepage or follow-up system), hospitalization duration and readmission indicators (obtained from the medical record homepage), postoperative complications (such as return to the emergency department or return to the critical care after surgery, which can be obtained from the postoperative medical record, nursing record and transfer record form, etc.), and adverse events (such as adverse events such as falls can be obtained from the adverse event registration form).

[0081] As shown in Table 4, eFI can also be verified for consistency with a variety of frailty assessment scales filled out by medical staff. It can be seen that eFI can form a lot of research directions and prediction purposes. The eFI constructed after feature selection for different prediction purposes has different model application scenarios and performance.

[0082] For specific prediction purposes, the candidate FRIs (or phenotypic terms) in the frailty indicator pool can be combined with knowledge-driven and data-driven methods to generate a subset of FRIs features suitable for specific purposes.

[0083] Knowledge-driven means selecting FRIs based on domain knowledge or literature.

[0084] Data-driven means obtaining high-correlation FRIs through feature selection of outcome event labels of large sample actual case data in EMR. In one embodiment of the present application, the method further includes: obtaining medical data and outcome events of multiple patients, wherein the outcome event is a label pre-set for each patient; performing term extraction on the medical data of all patients under the outcome event to obtain multiple terms under each outcome event; using the multiple terms under the outcome event as a term combination of the outcome event to obtain a term combination of each outcome event.

[0085] Obtain the medical data of multiple patients and the outcome events or prediction purposes of each patient. For multiple patients with the same outcome event / prediction purpose, identify the terms in the medical data of these patients. These terms are the FRI associated with the outcome event / prediction purpose. These terms are used as the term combination of the outcome event / prediction purpose.

[0086] For the frailty risk terms extracted by the user, the frailty risk factors within the term combination are obtained, for example, the frailty risk terms are matched with the term combination of the prediction target, the matched frailty risk terms are obtained, and multiple target frailty risk terms are obtained.

[0087] In the above embodiment, frailty term mining is performed from the medical records of all patients under a specific prediction target, and frailty risk terms with high correlation to the prediction target can be mined, thereby accurately screening out frailty terms with predictive significance.

[0088] Based on the frailty association relationship, a plurality of target frailty risk terms are screened out from the plurality of frailty risk terms, including: using a plurality of single-factor feature selection methods to screen the plurality of frailty risk terms in parallel to obtain a plurality of screening results, and screening the plurality of frailty risk terms according to the screening results to obtain a screening set; and using a multi-factor feature selection method to screen the screening set to obtain a plurality of target frailty risk terms.

[0089] For example, Figure 3 As shown, Figure 3 A flowchart of screening frailty risk terms based on frailty association relationships is shown. Figure 3 The method shown is a feature selection model for medical high-dimensional feature space, which can evaluate the predictive ability of each FRIs for outcome labels through a variety of single-factor Filter-type feature selection methods.

[0090] Figure 3 First, remove variables with too many missing values. Variables with too many missing values ​​refer to situations where there are vacancies in the terms. For example, the term medication status is identified as a frailty risk term, but the specific data recorded in the medication status is blank or very little, which means that there are too many missing values ​​in the medication status and should be removed. Alternatively, when there are not many sub-concepts under a variable, for example, in the knowledge in this field, the sub-concepts of abnormal liver function include: elevated alanine aminotransferase level, elevated aspartate aminotransferase level, and elevated glutamine transpeptidase level. Assume that the term abnormal liver function is extracted from the medical data, but there is no sub-concept, then abnormal liver function has too many missing values. The definition of vacancies in this application can be set freely, and this application does not limit this.

[0091] After obtaining the frailty feature subset through the single-factor Filter-type feature selection method, weighted voting is performed to exclude irrelevant variables.

[0092] After obtaining the frailty feature subset through the single-factor Filter-type feature selection method, the Filter-type multi-factor feature selection method can be applied from the candidate FRIs subset to gradually delete irrelevant, redundant, and collinear FRIs, and finally form the FRIs subset for eFI. The above searches for potential, independent frailty risk predictors in EMR, tests multiple FRIs feature combinations, and selects different FRIs concept granularities.

[0093] In the above embodiment, the frailty risk terms obtained from the medical records are screened to obtain frailty risk factors with better prediction effects, thereby optimizing the frailty assessment strategy and improving the prediction accuracy of the frailty index.

[0094] In one embodiment of the present application, the method further includes: obtaining parent concepts and child concepts at all levels of each frailty risk term in the multiple frailty risk terms to obtain an expanded set; based on prediction targets and / or frailty association relationships, screening out multiple target frailty risk terms from the expanded set, the frailty association relationships being associations between frailty risk terms and frail status, and the prediction targets being coping strategies to be taken when the user is determined to be in a frail status.

[0095] The FRIs terminified with SNOMED CT can further trace back the parent concepts and child concepts of the target term at all levels by cascading the hierarchical relationships between terms (i.e., the is-a relationship in SNOMED CT), and incorporate these concepts into the feature selection process.

[0096] In the past, some frailty assessment systems selected fewer but more precise indicators for the convenience of clinical medical staff. However, with the popularity of EMR today, it is not a bad idea to expand the range of indicators, especially those that can be easily and objectively collected in EMR, which may be conducive to the construction of eFI. In the era of big data, the use of big data and semantic technology to conduct multi-dimensional searches for various risk factors related to aging in EMR may bring "dimensionality gospel", that is, high-dimensional data may contain rich information, bringing new possibilities for problem solving. For example, the combination of several FRIs may have better significance for frailty classification, and then using outcome event labels for machine learning can discover potential FRIs that are highly correlated with the degree of frailty.

[0097] By cascading and expanding the “IS-A” hierarchical relationship of positive terms, we further explore the “most appropriate granularity” of the target concept and use it as FRIs to construct eFI.

[0098] Step 120 , constructing a semantic network of the plurality of target frailty risk terms according to the position of each target frailty risk term in the systematic clinical terminology system.

[0099] Step 130, calculating the weight of the target frailty risk term according to the position of the target frailty risk term in the semantic network.

[0100] The inventors found that the weighted score of FRIs in the previous eFI was the sum of the scores of each item divided by the actual number of items, which was the FI score. The more defects accumulated, the greater the probability of frailty, and the weight of each FRI was equal. However, not all FRIs have the same predictive ability for frailty, so giving different FRIs different weights can achieve higher prediction accuracy.

[0101] In the present application, the positions of the target frailty risk terms in the systematic clinical terminology system are used to construct a semantic network of the multiple target frailty risk terms, and the weights of the target frailty risk terms are calculated through the positions of the target frailty risk terms in the semantic network.

[0102] Semantic networks are often used as a form of knowledge representation. They are actually directed graphs where vertices represent concepts and edges represent semantic relationships between these concepts. A semantic network is a graphical structure used to represent knowledge in a pattern of interconnected nodes and arcs, i.e., a directed graph used for knowledge representation. What all semantic networks have in common is a declarative graphical representation that can be used to represent knowledge and support automated systems for reasoning about knowledge. Some versions are very informal, but others are formally defined logical systems. Common semantic networks include: Definitional networks, Assertional networks, Implicational networks, and Executable networks.

[0103] In this application, the position of frailty risk factors in the systematic clinical terminology system refers to the hierarchical relationship of frailty risk factors in the systematic clinical terminology system. For example, "left heart failure -> heart failure -> heart disease" are three levels in the systematic clinical terminology system. In this application, a term is regarded as a node, and the node is connected with its superior concept and subordinate concept to form an implicated network.

[0104] The weight is calculated according to the position of the frailty risk term in the semantic network. For example, a graph network similarity calculation algorithm is used to calculate the weight of each frailty risk term.

[0105] Using semantic networks to calculate weights can avoid the situation where two phenotypes have a similarity score of 0 due to different granularity. For example, a FRI of eFI is the parent concept (or sibling concept) of a frailty phenotype of the target patient. In this case, the traditional calculation scores it as 0, but the semantic similarity will be scored appropriately depending on the length of the "semantic path".

[0106] In one embodiment of the present application, the semantic network of the multiple frailty risk terms is constructed based on the position of each target frailty risk term in the systematic clinical terminology, including: obtaining the standard term corresponding to each target frailty risk term, the standard term being the term corresponding to the frailty risk term in the systematic clinical terminology system; and constructing the semantic network, the semantic network being the path for all standard terms to reach the general term, the general term being the common parent term of all standard terms.

[0107] Obtain the corresponding terms of frailty risk terms in the systematic clinical terminology system. For example, there are frailty risk factors: height 150, abnormal upper and lower body proportions, inconspicuous beard, decreased libido, and vesicoureteral reflux. According to the mapping relationship between FRI and standard terms, obtain the corresponding standard terms of these frailty risk factors in the systematic clinical terminology system, and obtain: short stature, abnormal upper and lower body proportions, loss of secondary sexual characteristics, hypogonadism, and vesicoureteral reflux.

[0108] After obtaining standard terminology, such as Figure 4 As shown, Figure 4 It is a schematic diagram of a semantic network provided by this application. Figure 4 For each standard term, the terms of the previous level are obtained step by step until the intersection of all standard terms is reached. This intersection is the common parent term of all standard terms. Figure 4 In the semantic network, a term is taken as a node and the hierarchical relationships between terms are taken as edges to construct a semantic network, which is actually a directed acyclic graph.

[0109] A certain FRI of eFI is the parent concept (or sibling concept) of a certain frailty phenotype of the target patient. In this case, the traditional calculation equals 0, but the semantic similarity will be scored according to the length of the "semantic path". For example, taking the term "abnormal phenotype" (HP:0000118) in the HPO terminology set as an example, its semantic network is actually a directed acyclic graph. Considering the semantic depth and distance between concepts, the semantic similarity of the concepts of "hypogonadism" and "absence of secondary sexual characteristics" is close, while the semantic similarity of the concepts of "short stature" and "vesicoureteral reflux" is far. Moreover, the scoring based on semantic similarity can avoid the situation where the similarity score of two phenotypes is 0 due to different granularity. For example, a certain FRI of eFI is the parent concept (or sibling concept) of a certain frailty phenotype of the target patient. In this case, the traditional calculation equals 0, but the semantic similarity will be scored according to the length of the "semantic path".

[0110] By utilizing the hierarchical structure of the concepts in the term set, semantic parameters such as depth and distance (i.e., topological structure features) between concepts are extracted, and then a semantic similarity function is established to improve the accuracy of semantic similarity estimation of medical terms.

[0111] The semantic similarity of each standard term is calculated using the semantic network, and the semantic similarity of the standard term is used as the weight of the frailty risk term corresponding to the standard term.

[0112] In the above embodiment, a semantic network is constructed through the semantic paths of frailty risk factors in a systematic clinical terminology system, which can accurately calculate the semantic similarities of different frailty risk factors, obtain the weights of frailty risk factors, and improve the prediction accuracy of the frailty index.

[0113] Step 140, calculating the average value of the weights of all target frailty risk terms to obtain the frailty index of the user, wherein the frailty index is used to predict whether the user is in a frail state.

[0114] After obtaining the weight of the target frailty risk factor, the user's frailty index can be calculated. For example, the weights of height 150, abnormal upper and lower body proportions, inconspicuous beard, decreased libido, and vesicoureteral reflux are 0.3, 0.3, 0.2, 0.2, and 0.5 respectively. The average of these is 0.3, and the user's frailty index is 0.3.

[0115] The above-mentioned frailty index calculation method obtains frailty risk terms in the user's medical data, assigns weights to the frailty risk terms through the semantic network of the frailty risk terms, evaluates the predictive significance of different frailty risk terms, and incorporates frailty risk terms of different granularities into the calculation through the semantic network, thereby improving the prediction accuracy of the frailty index.

[0116] In one embodiment of the present application, if the frailty index characterizes that the user is in a frail state, the method further includes: extracting terms from the user's medical data to obtain new terms; and adding the new terms to the frailty risk term pool. After the user is deemed to be in a frail disease state based on the frailty index, the user's medical data can be extracted again to obtain multiple terms. For example, a CRF (conditional random field)-based algorithm is used to extract terms from medical data. The algorithm is pre-trained by manually annotating the corpus and extracting multiple lexical and part-of-speech features for CRF training, which can complete the recognition of SNOMED CT terms in clinical texts.

[0117] These terms can be added to the term combination of outcome events according to the user's outcome events, and can also be used to test the effectiveness of the previous frailty risk factor pool.

[0118] More importantly, when these terms exist that are not in the frailty risk term pool, these terms are added to the frailty risk factor pool to discover potential, uncollected phenotypic terms, which can further enrich the SNOMED CT Chinese term dictionary. The hybrid model based on the two models has better recognition efficiency.

[0119] In the above embodiment, when a user is determined to be in a frail state, term mining is performed in the user's medical data to discover potential phenotypic terms and improve the term coverage of the frailty risk term pool, thereby obtaining frailty risk factors more comprehensively.

[0120] In one embodiment of the present application, a method for discovering new words from medical data of users diagnosed with frailty will achieve better results when an artificial intelligence algorithm is used. For example, models such as Conditional Random Fields (CRF), Bidirectional Long Short-Term Memory (Bi-LSTM), and Bidirectional Encoder Representations from Transformers BERT are trained to extract frailty risk factors from medical data through the trained models.

[0121] Localized terminology resources such as HPO, SNOMED CT and UMLS can help identify terms in EMR. At the same time, artificial intelligence algorithms can be used to extract daily idiomatic descriptions in EMR, which in turn can further enrich the synonymous vocabulary of terminology sets such as SNOMED CT.

[0122] In the Chinese environment, technicians still lack out-of-the-box terminology recognition tools and public medical annotation corpora. This application scheme breaks through this bottleneck, extracts and analyzes the phenotypic vocabulary commonly used by Chinese medicine practitioners in electronic medical records, applies a variety of similarity algorithms combined with manual review, completes the semantic mapping task of some Chinese clinical descriptions of frailty to standard terminology concepts, and then uses rule-based and deep learning methods to identify frailty. On this basis, the existing frailty knowledge system is integrated with the domestic hospital case database, and an electronic frailty index that is more in line with the evaluation of Chinese people is formed through feature selection and other methods.

[0123] Through the method provided in the present application, the long-term, multi-dimensional diagnosis and treatment data in EMR, and the variables reflecting a variety of complex health information, can be centrally converted into FRIs (or standardized clinical phenotypic terms) and the corresponding eFI can be constructed. The eFI construction method of the present application can be applied to the Chinese environment, filling the gap in the technical field.

[0124] The frailty index calculated by the method of this application can help understand the health characteristics of the elderly related to frailty. At the same time, the early identification of high-risk factors for frailty mined from medical records is conducive to early detection and diagnosis of patients, and even early intervention of relevant high-risk factors to delay the aging process, reduce the burden of disease, and overall reflect the diversity of the human aging process.

[0125] During the implementation of this application, a method for standardizing the terminology of aging characteristics is also provided. The term standardization can enable multi-source heterogeneous EMR systems to uniformly identify complex aging characteristics with different codes, and is also conducive to the sinking of the level of aging-related diagnosis to basic medical service institutions.

[0126] The method for constructing eFI based on semantic web and big data technology provided in this application can, after being embedded in EMR, instantly assess the aging fragility of hospitalized patients to provide early warning, thereby enabling the management of high-risk aging patients to reduce the occurrence of aging-related adverse events and alleviate the disease burden.

[0127] like Figure 5 As shown, Figure 5 A frailty index calculation device 500 provided in an embodiment of the present application is shown. The frailty index calculation device 500 includes: an acquisition module 510, a construction module 520, a weight calculation module 530 and an index calculation module 540.

[0128] An acquisition module 510, for acquiring a plurality of target frailty risk terms from the user's medical data; A construction module 520, for constructing a semantic network of the plurality of target frailty risk terms according to a position of each target frailty risk term in the systematic clinical terminology system; A weight calculation module 530, for calculating the weight of each target frailty risk term according to the position of the target frailty risk term in the semantic network; The index calculation module 540 is used to calculate the average value of the weights of all target frailty risk terms to obtain the frailty index of the user, and the frailty index is used to predict whether the user is in a frail state.

[0129] In an optional embodiment of the present application, the construction module 520 is specifically used to obtain the standard term corresponding to each target frailty risk term, and the standard term is the term corresponding to the frailty risk term in the systematic clinical terminology system; construct the semantic network, and the semantic network is the path for all standard terms to reach the general term, and the general term is the common parent term of all standard terms.

[0130] In an optional embodiment of the present application, the acquisition module 510 is specifically used to obtain multiple frailty risk terms from the user's medical records; based on the prediction target and / or frailty association relationship, multiple target frailty risk terms are screened out from the multiple frailty risk terms, the frailty association relationship is the association relationship between the frailty risk terms and the frail status, and the prediction target is the coping strategy adopted when the user is determined to be in a frail status.

[0131] In an optional embodiment of the present application, the acquisition module 510 is specifically used to match a preset frailty risk term pool with the medical data to obtain multiple target frailty risk terms, wherein the frailty risk term pool includes multiple preset terms. In an optional embodiment of the present application, the device also includes an enrichment module, which is used to map each term in the preset term set to a systematic clinical term system to obtain a standard term set; map each term in the standard term set to a first set to obtain a first term set, the first set includes a private term set, a Chinese tangram term set, and a Chinese systematic clinical term system; map each term in the standard term set to a second set to obtain a second term set, the second set includes a unified medical language system; map each term in the second term set to a third set to obtain a third term set, the third set includes a Chinese medical term system; map each term in the second term set to a fourth set to obtain a fourth term set, the fourth set includes a human phenotype term set; map each term in the fourth term set to a fifth set to obtain a fifth term set, the fifth set includes a Chinese human phenotype term set; the preset term set, the standard term set, the first term set, the second term set, the third term set, the fourth term set and the fifth term set are used as the frailty risk term pool.

[0132] In an optional embodiment of the present application, if the frailty index is used to characterize the user as being in a frail state, the device further includes a new module, which is used to extract terms from the user's medical data to obtain new terms; and add the new terms to the frailty risk term pool.

[0133] In an optional embodiment of the present application, the construction module 520 is also used to obtain a term combination corresponding to a prediction target, and the term combination is obtained by extracting terms from the medical data of all patients under the prediction target; and obtaining frailty risk terms in the term combination from the multiple frailty risk terms to obtain the multiple target frailty risk factors.

[0134] The frailty index calculation device provided in the embodiment of the present application has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0135] like Figure 6 As shown, Figure 6 A schematic block diagram of an electronic device 600 provided in an embodiment of the present application is shown. The electronic device 600 includes: a processor 610 and a memory 620; The processor 610, the memory 620 and each component are directly or indirectly electrically connected to each other to achieve data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The memory 620 is used to store computer programs, such as storing Figure 5The software function module shown in the figure is the frailty index calculation device 500. The frailty index calculation device 500 includes at least one software function module that can be stored in the memory 620 in the form of software or firmware or fixed in the operating system (OS) of the electronic device 600. The processor 610 is used to execute the executable module stored in the memory 620, such as the software function module or computer program included in the frailty index calculation device 500.

[0136] Among them, the memory 620 can be, but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.

[0137] The processor 610 may be an integrated circuit chip with signal processing capabilities. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a microprocessor, etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The various methods, steps and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. Alternatively, the processor 610 may also be any conventional processor, etc.

[0138] The embodiment of the present application further provides a non-volatile computer-readable storage medium (hereinafter referred to as storage medium), on which a computer program is stored. When the computer program is run by a computer such as the above-mentioned electronic device 600, the above-mentioned frailty index calculation method is executed.

[0139] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0140] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0141] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0142] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a computer-readable storage medium, including several instructions for enabling a computer device (which can be a personal computer, a laptop, a server, or an electronic device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0143] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for calculating a frailty index, characterized in that: The method comprises: obtaining a plurality of target frailty risk terms from the user's medical data; constructing a semantic network of the plurality of target frailty risk terms according to a position of each target frailty risk term in the systematic clinical terminology system; Calculating the weight of each target frailty risk term according to the position of the target frailty risk term in the semantic network; The average value of the weights of all target frailty risk terms is calculated to obtain the frailty index of the user, and the frailty index is used to predict whether the user is in a frail state.

2. The method according to claim 1, characterized in that The step of constructing a semantic network of the plurality of frailty risk terms according to the position of each target frailty risk term in the systematic clinical terminology includes: Obtaining a standard term corresponding to each target frailty risk term, wherein the standard term is a term corresponding to the frailty risk term in a systematic clinical terminology system; The semantic network is constructed, wherein the semantic network is a path for all standard terms to reach the general term, and the general term is a common parent term of all standard terms.

3. The method according to claim 1, characterized in that The method obtains a plurality of target frailty risk terms from the user's medical records, including: Obtain multiple frailty risk terms from the user's medical records; Based on the prediction target and / or frailty association relationship, multiple target frailty risk terms are screened out from the multiple frailty risk terms, the frailty association relationship is the association relationship between the frailty risk terms and the frailty status, and the prediction target is the coping strategy adopted when the user is determined to be in a frail status.

4. The method according to claim 1, characterized in that: The method further comprises: A preset frailty risk term pool is matched with the medical data to obtain a plurality of target frailty risk terms, wherein the frailty risk term pool includes a plurality of preset terms.

5. The method according to claim 4, characterized in that The method further comprises: Map each term in the preset terminology set to the systematic clinical terminology system to obtain a standard terminology set; Map each term in the standard terminology set to a first set to obtain a first terminology set, wherein the first set includes a private terminology set, a Chinese tangram terminology set, and a Chinese systematized clinical terminology system: Mapping each term in the standard term set to a second set to obtain a second term set, wherein the second set includes a unified medical language system; Mapping each term in the second term set to a third set to obtain a third term set, wherein the third set includes a Chinese medical term system; Mapping each term in the second term set to a fourth set to obtain a fourth term set, wherein the fourth set includes a human phenotype term set; Mapping each term in the fourth term set to a fifth set to obtain a fifth term set, wherein the fifth set includes a Chinese human phenotype term set; The preset term set, the standard term set, the first term set, the second term set, the third term set, the fourth term set and the fifth term set are used as the frailty risk term pool.

6. The method according to claim 4, characterized in that If the frailty index indicates that the user is in a frail state, the method further includes: Performing term extraction on the medical data of the user to obtain newly added terms; The newly added terms are added to the frailty risk term pool.

7. The method according to claim 2, characterized in that The method of screening out a plurality of target frailty risk terms from the plurality of frailty risk terms based on the prediction target includes: Acquire a term combination corresponding to a prediction target, wherein the term combination is obtained by performing term extraction on medical data of all patients under the prediction target; The frailty risk terms in the term combination are acquired from the multiple frailty risk terms to obtain the multiple target frailty risk terms.

8. A frailty index calculation device, characterized in that: include: an acquisition module, for acquiring a plurality of target frailty risk terms from medical data of a user; A construction module, configured to construct a semantic network of the plurality of target frailty risk terms according to a position of each target frailty risk term in the systematic clinical terminology system; A weight calculation module, used to calculate the weight of each target frailty risk term according to the position of the target frailty risk term in the semantic network; The index calculation module is used to calculate the average value of the weights of all target frailty risk terms to obtain the frailty index of the user, and the frailty index is used to predict whether the user is in a frail state.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the processor is connected to the memory; The memory is used to store programs; The processor is used to call the program stored in the memory to execute the method according to any one of claims 1 to 9.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is executed.

Images