An intervention mode construction method and system for a medical care plan prepared by an elderly patient

Abstract

The application discloses an intervention mode construction method and system for a medical care plan of an old patient, relates to the technical field of medical assistance, and comprises the following steps: acquiring ACP information of a target patient, extracting key information of the ACP information, and mapping each key information into a will preference vector according to a value preference dimension; acquiring a candidate medical scheme of the target patient, extracting a medical entity of the candidate medical scheme, and mapping each medical entity into a scheme preference vector according to the value preference dimension; generating a consistency comparison result of the will preference vector and the scheme preference vector; and generating a recommended value of the candidate medical scheme according to the consistency comparison result, so as to establish an association between the ACP information and the medical scheme, improve the efficiency and transparency of the doctor-patient joint decision, and effectively solve the problem of ACP execution difficulty in the care of the old patient.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of medical communication assistance, in particular to an intervention mode construction method and system for an advance care planning (ACP) of an elderly patient. BACKGROUND

[0002] Advance care planning (ACP) refers to that when a patient has autonomous decision-making ability, the patient expresses the will of medical care objectives and treatment measures in advance for possible serious diseases or disabled states in the future, and records and archives in written form. When the patient is disabled, medical staff can formulate a medical care plan according to the content of the ACP and the actual situation of the clinic, so as to reflect the autonomous will of the patient in the medical decision-making.

[0003] At present, ACP still faces great implementation difficulties in popularization and application. One of the important reasons is that the clinical scenarios covered by ACP are limited. When a clinical scenario not covered by ACP occurs, medical staff still need to rely on subjective judgment to make medical decisions, which reduces the implementation effect of ACP. Especially for the elderly population, their overall physical function is in a state of gradual decline, and the reason for disability is more complex and difficult to predict, which further increases the implementation difficulty of ACP in the elderly population.

[0004] However, the existing ACP management system mainly saves the ACP of the patient through the database, and the medical staff retrieves and refers to it before clinical decision-making, and lacks guidance suggestions for medical staff to provide medical scheme decisions in specific clinical scenarios according to the ACP information of the patient. SUMMARY

[0005] The purpose of the application is to solve the technical problem of how to establish the corresponding relationship between ACP and medical schemes not directly expressed, and provide an intervention mode construction method and system for an advance care planning of an elderly patient. By introducing a value preference dimension, the key information of the patient in the ACP and the medical entity in the medical scheme are quantified as corresponding preference vectors, whether the medical scheme can be used as the basis for expanding the ACP information is judged according to the consistency comparison result of the will preference vector and the scheme preference vector, so as to establish the association between the ACP information and the medical scheme, and effectively improve the problem of ACP implementation difficulty in the care of the elderly patients.

[0006] According to a first aspect of the application, the application claims an intervention mode construction method for an advance care planning of an elderly patient, comprising:

[0007] Obtaining ACP information of a target patient, extracting key information of the ACP information, and mapping each key information to a will preference vector according to a value preference dimension;

[0008] obtaining a candidate medical scheme for a target patient, extracting medical entities of the candidate medical scheme, mapping each medical entity into a scheme preference vector according to a value preference dimension;

[0009] generating a consistency comparison result between the will preference vector and the scheme preference vector;

[0010] generating a recommended value of the candidate medical scheme according to the consistency comparison result.

[0011] Preferably, after obtaining the ACP information of the target patient, it further comprises: if the ACP information contains at least two versions, generating the will preference vector based on the latest version of the ACP information.

[0012] Preferably, in the generating of the recommended value of the candidate medical scheme, it further comprises:

[0013] obtaining a fluctuation amplitude of each value preference dimension of the target patient according to the will preference vector of each ACP information;

[0014] generating a stability weight of the corresponding value preference dimension according to the fluctuation amplitude;

[0015] calculating a weighted result of the similarity between the will preference vector and the scheme preference vector according to the stability weight, to obtain the recommended value.

[0016] Preferably, in the mapping of the key information into the will preference vector, it further comprises: matching each key information with a medical entity library respectively, if the matching fails, matching a value preference interval according to the corresponding key information; in the generating of the consistency comparison result, generating a corresponding consistency comparison result according to whether the value of the scheme preference vector in the corresponding value preference dimension belongs to the corresponding value preference interval.

[0017] Preferably, in the matching of the value preference interval according to the corresponding key information, it further comprises:

[0018] judging whether there is key information matched successfully with the medical entity library, if there is, taking the value of each value preference dimension corresponding to the key information matched successfully with the medical entity library as a reference value, and obtaining the value preference interval according to the interval whose distance from the reference value does not exceed a value distance threshold.

[0019] Preferably, if there are at least two key information matched successfully with the medical entity library, after obtaining all the reference values, it further comprises: taking the interval whose distance from each reference value does not exceed the value distance threshold as a reference interval, and taking the intersection of all the reference intervals as the value preference interval.

[0020] Preferably, the method further comprises: comparing the scheme preference vector of each medical entity with all the will preference vectors respectively, taking the comparison result with the highest consistency as the consistency comparison result of the corresponding medical entity, and generating the recommended value of the candidate medical scheme according to the consistency comparison results of all the medical entities.

[0021] According to a second aspect of the present application, the present application claims a system for constructing an intervention mode of a pre-established medical care plan for an elderly patient, comprising:

[0022] An acquisition module is configured to acquire ACP information and a candidate medical scheme of a target patient.

[0023] An extraction module is configured to extract key information of the ACP information and medical entities of the candidate medical scheme.

[0024] A mapping module is configured to map each key information into a will preference vector and each medical entity into a scheme preference vector according to a value preference dimension.

[0025] A comparison module is configured to generate a consistency comparison result of the will preference vector and the scheme preference vector.

[0026] A generation module is configured to generate a recommended value of the candidate medical scheme according to the consistency comparison result.

[0027] Preferably, the acquisition module further comprises: if the ACP information contains at least two versions, generating the will preference vector based on the latest version of the ACP information.

[0028] Preferably, the generation module further comprises:

[0029] According to the will preference vector of each ACP information, obtaining a fluctuation amplitude of the target patient in each value preference dimension.

[0030] According to the fluctuation amplitude, generating a stable weight of the corresponding value preference dimension.

[0031] According to the stable weight, calculating a weighted result of the similarity between the will preference vector and the scheme preference vector to obtain the recommended value.

[0032] Preferably, the mapping module further comprises: matching each key information with a medical entity library respectively, if the matching fails, matching the corresponding key information with a value preference interval; and when generating the consistency comparison result, generating a corresponding consistency comparison result according to whether the value of the scheme preference vector in the corresponding value preference dimension belongs to the corresponding value preference interval.

[0033] Preferably, the mapping module further comprises: judging whether there is key information matched successfully with the medical entity library, if yes, taking the value of each value preference dimension corresponding to the key information matched successfully with the medical entity library as a reference value, and obtaining the value preference interval according to the interval in which the distance to the reference value is not more than the value distance threshold.

[0034] Preferably, if there are at least two key information matched successfully with the medical entity library, the mapping module further comprises: taking the interval in which the distance to each reference value is not more than the value distance threshold as a reference interval, and taking the intersection of all reference intervals as the value preference interval.

[0035] Preferably, the comparison module further comprises comparing each medical entity corresponding scheme preference vector with all will preference vectors respectively, and taking the comparison result with the highest consistency as the consistency comparison result corresponding to the medical entity; and the generating module further comprises generating the recommended value of the candidate medical scheme according to the consistency comparison results of all medical entities.

[0036] According to a third aspect of the present application, an intervention mode construction device for an elderly patient to make a medical care plan in advance is provided, comprising a processor and a memory, the memory stores computer readable instructions, when the computer readable instructions are executed by the processor, the steps in the method of the first aspect are executed.

[0037] The present application has the following beneficial effects:

[0038] 1. By introducing the value preference dimension, the key information of the patient in the ACP and the medical entity in the medical scheme are quantified into corresponding preference vectors, and the consistency comparison result of the will preference vector and the scheme preference vector is used to judge whether the medical scheme can be used as the basis for expanding the ACP information. The higher the consistency, the more the medical scheme conforms to the value preference expressed by the patient in the ACP information. Even if there is a part in the medical scheme that is not covered by the ACP information, it can also be recommended as the medical scheme of the patient. If the consistency is too low, the medical scheme does not conform to the value preference expressed by the patient in the ACP information, and the medical scheme needs to be adjusted or the patient and / or the patient's family members need to be communicated. The method establishes the association between the ACP information and the medical scheme. When there is content in the medical scheme that does not conform to the content directly expressed by the patient in the ACP information, the real will of the patient can also be captured by quantifying the value preference, which can make the medical staff quickly focus on the treatment path most consistent with the value preference of the patient in the clinical application, improve the efficiency and transparency of the medical staff-patient joint decision, improve the operability of the ACP in the actual clinical scene, and effectively solve the problem of difficulty in executing the ACP in the care of the elderly patients.

[0039] 2. The historical version of the willingness preference vector is used as the basis for dimension-level statistical analysis, and the stability weight is set for each value preference dimension according to the fluctuation amplitude. The higher the stability weight, the higher the value of the corresponding dimension confidence. The lower the stability weight, the lower the value of the corresponding dimension confidence. Thus, the historical stable value preference of the target patient and the latest value preference are taken into account, the influence of accidental factors in the latest value preference on the value preference of the target patient is reduced, and the accuracy is improved.

[0040] 3. The mapping mode of the interval value preference dimension not only allows the implicit expression to obtain a quantitative result, but also provides a more flexible compatibility judgment in the consistency comparison process, so that the matching result between the candidate scheme and the patient's willingness is more in line with the clinical practice.

[0041] 4. The value preference interval is generated by taking the value of the matching medical entity as the reference value. This generation method of the value preference interval can adapt to the individual preference characteristics of the target patient, that is, the mapping representation from the value preference willingness corpus to the value preference dimension and the mapping standard from the medical entity to the value preference dimension adapt to the individual differences of the patient, and improve the clinical interpretability. BRIEF DESCRIPTION OF DRAWINGS

[0042] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure and serve to explain the principles of the present disclosure together with the specification.

[0043] Figure 1 A flowchart of a method for constructing an intervention mode of a pre-established medical care plan for an elderly patient according to an embodiment of the present application;

[0044] Figure 2 A flowchart of a recommended value generation method according to an embodiment of the present application;

[0045] Figure 3 A flowchart of a willingness preference vector calculation method according to an embodiment of the present application;

[0046] Figure 4 A structural diagram of a system for constructing an intervention mode of a pre-established medical care plan for an elderly patient according to an embodiment of the present application;

[0047] Figure 5 A structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0048] The application provides a method and system for constructing an intervention mode of a medical care plan for an elderly patient.

[0049] It should be noted that similar reference numerals and letters refer to similar items throughout the accompanying drawings, and therefore, once an item is defined in one drawing, it is not necessary to further define and explain it in subsequent drawings. Meanwhile, in the description of the application, the terms "first", "second", and the like relational terms are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations.

[0050] According to a first aspect of the application, the application claims a method for constructing an intervention mode of a medical care plan for an elderly patient, referring to the accompanying drawings, Figure 1 as shown, comprising:

[0051] S1: obtaining ACP information of a target patient.

[0052] It should be noted that the ACP information refers to the record of the will and preference of the target patient for future medical care when the target patient has the ability to make autonomous decisions. It can be collected in the form of a designed questionnaire, or obtained through text recording, voice recording, and the like.

[0053] In this embodiment, the ACP information recorded in the form of text is taken as an example for detailed description, and ACP information in other forms can be converted into text through voice recognition or other feasible processing modes, or other feasible processing modes.

[0054] In the embodiment, if the ACP information of the target patient contains at least two versions, the latest version is selected as the basis for generating the will preference vector according to the collection time of each ACP information, so as to ensure that the will preference vector conforms to the latest value preference of the target patient in different disease courses.

[0055] S2: Extract the key information of the ACP information.

[0056] It should be noted that the key information refers to the sentences or phrases directly or indirectly expressing the medical care will extracted from the ACP information. For example, "willing to use a breathing machine" is extracted from the text or voice recording.

[0057] In the embodiment, step S2 further includes: S21, performing text cleaning and word segmentation processing on each ACP information. S22, extracting key sentences and / or phrases from the ACP information to remove text content with low relevance. Each ACP can be mapped to a word vector by using natural language processing technology, and then input into a pre-trained text extraction model to output the key information. The text extraction model can be a long short-term memory network model, a convolutional neural network model, a model based on a Transformer architecture, etc.

[0058] It should be noted that in step S22, it also includes: collecting text samples containing medical care will related and unrelated in advance. After preprocessing such as word segmentation, stop word removal and relevance annotation, a training data set is obtained. The training data set is divided into a training set and a test set. Each preprocessed sample data in the training set is input into the pre-built text extraction model, and the value of the loss function is obtained according to the difference between the corresponding relevance annotation and the output of the text extraction model. The text extraction model is trained by minimizing the loss function. The text extraction model is evaluated using the test set, and the model is optimized according to the evaluation result, and the hyperparameter values and structure of the model are adjusted to obtain the final text extraction model.

[0059] It should be noted that the extracted key information can be one or more. The embodiment does not limit the number of key information.

[0060] S3: Map each key information to a corresponding preference vector according to the value preference dimension to obtain a will preference vector.

[0061] It should be noted that the value preference dimension is used to describe a multi-dimensional set of aspects valued by the patient in medical care decision-making, such as life extension, economic burden, pain management, and invasiveness acceptance. The value preference dimension can be obtained in a pre-set manner. The value of each value preference dimension can also be obtained in a pre-set manner.

[0062] In the embodiment, the will preference vector refers to a vector representation of the key information after quantization on the value preference dimension, denoted as . Wherein d represents the dimension number representing the value preference dimension. When the value of the corresponding dimension is -1, it means no preference; when the value of the corresponding dimension is 1, it means preference; when the value of the corresponding dimension is 0, it means neutral or not mentioned.

[0063] In the embodiment, step S3 further comprises the following sub-steps:

[0064] S31, respectively judge the information type of each key information, that is, judge whether the key information corresponds to the acceptance will of the medical entity or the value preference. If the key information belongs to the acceptance will of the medical entity, continue to execute step S32; if the key information belongs to the value preference, continue to execute step S34.

[0065] It should be noted that the value mapping rule of "medical entity-value dimension" can be established in advance. According to the pre-set value preference dimension and scoring standard, the clinical experts score all medical entities of the medical institution. For example, in the "life extension" dimension, 0 represents neutral and / or not mentioned, and 1 represents significantly prolonging survival time; in the "pain management" dimension, negative value represents increasing pain or discomfort, and positive value represents relieving pain; in the "economic burden" dimension, negative value represents high medical cost, and positive value represents low medical cost; in the "invasiveness acceptance" dimension, negative value represents low acceptance, and only positive value represents high acceptance.

[0066] In the embodiment, the key information is matched with all medical entities in the pre-constructed value mapping rule through natural language processing and the like. If the matching is successful, the corresponding key information is regarded as the acceptance will of the medical entity, and if the matching fails, the corresponding key information is regarded as the value preference.

[0067] It should be noted that the information type of the key information can also be determined in other feasible embodiments, such as by pre-constructing a corpus including the acceptance will of the medical entity and the corpus of the value preference, pre-training the neural network model through the corpus, and distinguishing the type of the corresponding key information through the neural network model. The type determination method of the key information is not limited further in the application.

[0068] S32, taking the value of each of the value preference dimensions corresponding to the medical entity matched successfully with the key information as a reference value.

[0069] S33, extracting whether the patient accepts or refuses the corresponding medical entity in the key information. If the key information indicates that the target patient accepts the corresponding medical entity, the key information can directly use the reference value as the score of the corresponding dimension; if the key information indicates that the target patient refuses the corresponding medical entity, the key information can directly use the opposite of the reference value as the score of the corresponding dimension. The value of the key information in all value preference dimensions is obtained to obtain the willingness preference vector.

[0070] For example, if the score of the medical entity "mechanical ventilation" in "pain management" is -0.5, the key information "willing to accept mechanical ventilation" corresponds to -0.5 in "pain management", and the key information "refuse mechanical ventilation" corresponds to 0.5 in "pain management".

[0071] S34, the standard value of each sample corpus in the pre-constructed value preference willingness corpus can be set in each value preference dimension in an artificial marking manner. For example, the sample "hope not too painful", "don't want to be too painful", "want to be comfortable" corresponds to the value marked as 1 in the "pain management" dimension, and the values of the remaining dimensions are marked as 0. The neural network model is pre-trained through the value preference willingness corpus, and the key information is mapped to the willingness preference vector through the neural network model.

[0072] S4: obtaining a candidate medical scheme of a target patient, the candidate medical scheme indicating a selected medical scheme to be selected by medical staff.

[0073] S5: extracting medical entities of the candidate medical scheme. The medical entities include medical schemes and / or operation units that can be adopted by the patient.

[0074] It should be noted that the medical entities can be obtained in a pre-set manner, for example, "tracheal intubation", "short-term mechanical ventilation", "long-term mechanical ventilation", and "kidney dialysis treatment" can be selected as treatment schemes, and "ventilator" and the like can be used as medical equipment.

[0075] S6: generating a preference vector corresponding to each medical entity according to the matching result of each medical entity in the value mapping rule of "medical entity-value dimension" to obtain the scheme preference vector, which is denoted as V PALN .

[0076] S7: generating a consistency comparison result of the willingness preference vector and the scheme preference vector.

[0077] It should be noted that the vector similarity between the will preference vector and the scheme preference vector can be calculated based on cosine similarity, Euclidean distance similarity, etc. to obtain the consistency comparison result.

[0078] It should be noted that the consistency comparison result can be fed back to medical staff, and further used as a basis for adjusting the medical scheme and a key point for communication with the target patient's family. For example, if the target patient has a conflict between the value preference tendency of the ACP information and the value preference tendency of the medical scheme, the medical staff can adjust the medical scheme or communicate with the patient's family focusing on the conflicting value preference dimension.

[0079] In the embodiment, if the number of the extracted key information and the number of the medical entities are both 1, the recommended value is directly calculated according to the consistency comparison result of the will preference vector and the scheme preference vector.

[0080] In the embodiment, if the number of the extracted key information is 1, but the number of the extracted medical entities is greater than 1, the scheme preference vector corresponding to each medical entity is compared with the will preference vector, and the recommended value of the candidate medical scheme is generated according to the consistency comparison result of all medical entities. For example, the recommended value of the candidate medical scheme can be obtained according to the minimum value, maximum value or average value, etc. in the similarity corresponding to all medical entities.

[0081] In the embodiment, if the number of the extracted key information is greater than 1, and the number of the extracted medical entities is greater than or equal to 1, the scheme preference vector corresponding to each medical entity is compared with all will preference vectors, the comparison result with the highest consistency is taken as the consistency comparison result corresponding to the medical entity, and the recommended value of the candidate medical scheme is generated according to the consistency comparison result of all medical entities. For example, the extracted key information includes A1 and A2, the extracted medical entity is B, the similarity between A1 and B is a1, and the similarity between A2 and B is a2. The consistency comparison result of the medical entity B is max (a1, a2), thereby reducing the risk of deviation amplification caused by local inconsistency.

[0082] S8: generating a recommended value of a candidate medical scheme according to the consistency comparison result. The higher the consistency between the will preference vector and the scheme preference vector, the greater the recommended value of the corresponding candidate medical scheme; the lower the consistency between the will preference vector and the scheme preference vector, the smaller the recommended value of the corresponding candidate medical scheme.

[0083] It should be noted that the recommended value can be directly displayed to the medical staff and / or patient family members to provide decision-making assistance and improve communication efficiency. The medical treatment scheme can also be screened according to the ranking result of the recommended value to exclude the medical communication scheme with too low willingness of the target patient and improve the patient autonomy.

[0084] In a feasible implementation, referring to the accompanying drawings Figure 2 , step S8 further includes the following sub-steps:

[0085] S81, obtaining the fluctuation amplitude of the target patient in each value preference dimension according to the willingness preference vector of each ACP information.

[0086] It should be noted that the fluctuation amplitude of the value preference dimension is used to describe the discrete degree of the preference value of the target patient reflected in the value preference dimension in different versions of the ACP information relative to the mean value. It can be represented by statistical quantities such as variance and standard deviation. The greater the fluctuation amplitude, the more unstable and changeable the preference value of the target patient in the value preference dimension is; the smaller the fluctuation amplitude, the more stable and difficult to change the preference value of the target patient in the value preference dimension is.

[0087] In the embodiment, all the ACP information of the target patient is sequentially recorded as ACP t , where t = 1, …, i, …, T. ACP i represents the i-th version of the ACP information, ACP T represents the latest version of the ACP information. The value preference dimension is recorded as K, where K = 1, …, k, …, d. ACP i The value in the k-th dimension is recorded as v i，k . The values of all the ACP in the k-th dimension form a time sequence value V k , where Vj = {v 1，k , …, v i，k , …, v T，k}.

[0088] In the embodiment, the standard deviation is taken as an example to be described in detail as the calculation standard of the fluctuation amplitude. The calculation method of the fluctuation amplitude is as follows:

[0089] ;

[0090] ;

[0091] wherein, represents the mean value of the ACP of the target patient in the k-th dimension.

[0092] S82. Generate stable weights for the corresponding value preference dimension based on the fluctuation amplitude. The stable weights are denoted as S, where S = {s1, ..., s2}. k , ..., s d}. Among them, s k This represents the stable weight of the k-th dimension.

[0093] It should be noted that the value of the stabilizing weight is based on the mapping result of the fluctuation amplitude of the corresponding dimension within a preset weight value range. The value of the stabilizing weight is inversely proportional to the value of the fluctuation amplitude. The smaller the fluctuation amplitude, the larger the value of the stabilizing weight; the larger the fluctuation amplitude, the larger the value of the stabilizing weight. The fluctuation amplitude can be converted into the stabilizing weight by using it as input to mapping methods such as linear mapping, parametric mapping, or sigmoid mapping.

[0094] In this embodiment, the range of the stable weights is [0, 1], and the sum of all the stable weights is 1.

[0095] S83. Calculate the weighted result of the similarity between the intention preference vector and the alternative preference vector based on the stable weights to obtain the recommended value.

[0096] In this embodiment, the method for calculating the recommendation value Rec includes:

[0097] ;

[0098] Wherein, Dk represents the difference between the value of the intention preference vector in the k-th dimension and the value of the solution preference vector in the k-th dimension.

[0099] In this embodiment, when the target patient has only one version of the ACP information, s k The values ​​can all be set to the default value. .

[0100] In one feasible implementation, refer to the appendix. Figure 3 As shown, step S31 matches each key piece of information with the medical entity database. If the matching fails, proceed to step S34. Step S34 matches the value preference interval based on the corresponding key information. In step S7, a consistency comparison result is generated based on whether the value of the solution preference vector in the corresponding value preference dimension belongs to the corresponding value preference interval. If the value of the solution preference vector in the corresponding value preference dimension does not belong to the value preference interval, the comparison result of the corresponding value preference dimension is marked as dissimilar, i.e., the corresponding similarity value is 0; if the value of the solution preference vector in the corresponding value preference dimension belongs to the value preference interval, the comparison result of the corresponding value preference dimension is marked as similar, i.e., the corresponding similarity value is 1.

[0101] It should be noted that the medical entity library is obtained according to all the medical entities of the medical institution, and can also be obtained by integrating all the medical entities in the value mapping rule of the medical entity-value dimension.

[0102] It should be noted that the value interval corresponding to each sample corpus in the corpus of value preference intention can be pre-set in each dimension, the feature vector of the key information and the feature vector of each sample corpus are extracted by the neural network model, the cosine similarity between the feature vector of the key information and the feature vector of each sample corpus is calculated, and the value interval corresponding to the sample corpus with the maximum cosine similarity is taken as the value preference interval of the key information. Of course, other feasible implementations can also be used.

[0103] In the embodiment, step S34 further includes: S341, judging whether there is key information matched successfully with the medical entity library. If yes, step S342 is continued to be executed, and if not, step S344 is continued to be executed. S342, taking the value of each value preference dimension corresponding to the key information matched successfully with the medical entity library as a reference value, denoted as R k . S343, obtaining the value preference interval according to the interval with a distance from the reference value not exceeding a value distance threshold, that is, the value preference interval can be identified as [R k -δ, R k +δ]. Wherein, δ represents the value distance threshold, and the specific value thereof can be obtained by pre-setting. S344, pre-setting the corresponding value interval for each sample corpus, extracting the feature vector of the key information and the feature vector of each sample corpus in the pre-collected corpus of value preference intention by the neural network model respectively, calculating the cosine similarity between the feature vector of the key information and the feature vector of each sample corpus, and taking the value interval corresponding to the sample corpus with the maximum cosine similarity as the value preference interval of the key information.

[0104] It should be noted that the value interval corresponding to each sample corpus in the corpus of value preference intention can be obtained by artificial pre-setting.

[0105] It should be noted that the value corresponding to the matched medical entity is taken as the reference value for generating the corresponding value preference interval, and this generation manner of the value preference interval can adapt to the individual preference characteristics of the target patient, that is, the mapping representation of the corpus of value preference intention to the value preference dimension is adaptive to the individual differences of the patient according to the mapping standard of the medical entity to the value preference dimension, and the clinical interpretability is improved.

[0106] In the embodiment, if there are at least two key information matched with the medical entity library successfully, after obtaining all reference values, the interval with a distance not more than the value distance threshold from each reference value is taken as a reference interval, and the intersection of all reference intervals is taken as a value preference interval, so as to extract the preference interval supported by multiple parties from the mapping of multiple specific medical entities, and thus the same intention of the target patient in the dimension is more strictly "consensus screened".

[0107] According to a second aspect of the present application, the present application claims a kind of old patient pre- medical care plan intervention mode construction system, refer to the drawings Figure 4 As shown in the drawings, comprising:

[0108] The acquisition module is used to acquire the ACP information of the target patient and the candidate medical scheme;

[0109] The extraction module is used to extract the key information of the ACP information and the medical entity of the candidate medical scheme;

[0110] The mapping module is used to map each key information into an intention preference vector and each medical entity into a scheme preference vector according to the value preference dimension;

[0111] The comparison module is used to generate a consistency comparison result of the intention preference vector and the scheme preference vector;

[0112] The generation module is used to generate a recommended value of the candidate medical scheme according to the consistency comparison result.

[0113] In a feasible embodiment, the acquisition module further comprises: if the ACP information contains at least two versions, generating the intention preference vector based on the latest version of the ACP information.

[0114] In a feasible embodiment, the generation module further comprises:

[0115] According to the intention preference vector of each ACP information, obtaining the fluctuation amplitude of the target patient in each value preference dimension;

[0116] According to the fluctuation amplitude, generating a stable weight of the corresponding value preference dimension;

[0117] According to the stable weight, calculating a weighted result of the similarity between the intention preference vector and the scheme preference vector, and obtaining the recommended value.

[0118] In a feasible embodiment, the mapping module further comprises matching each key information with the medical entity library, if the matching fails, matching the value preference interval according to the corresponding key information; when generating the consistency comparison result, generating a corresponding consistency comparison result according to whether the value of the scheme preference vector in the corresponding value preference dimension belongs to the corresponding value preference interval.

[0119] In a feasible implementation, the mapping module further comprises: judging whether there is key information matched successfully with the medical entity library, if there is, taking the value of each value preference dimension corresponding to the key information matched successfully with the medical entity library as a reference value, and obtaining the value preference interval according to the interval in which the distance to the reference value is not more than the value distance threshold.

[0120] In a feasible implementation, if there are at least two key information matched successfully with the medical entity library, the mapping module further comprises: taking the interval in which the distance to each reference value is not more than the value distance threshold as a reference interval, and taking the intersection of all reference intervals as the value preference interval.

[0121] In a feasible implementation, the comparison module further comprises: comparing each medical entity corresponding scheme preference vector with all will preference vectors respectively, and taking the comparison result with the highest consistency as the consistency comparison result corresponding to the medical entity; and the generation module further comprises: generating the recommended value of the candidate medical scheme according to the consistency comparison result of all medical entities.

[0122] Referring to the accompanying Figure 5 The embodiments of the present application provide an electronic device, comprising: a processor and a memory, the processor and the memory are interconnected and communicate with each other through a communication bus and / or other forms of connection mechanism (not marked), the memory stores a computer program executable by the processor, when the computing device runs, the processor executes the computer program to execute the system in any optional implementation of the above-mentioned embodiments.

[0123] The embodiments of the present application provide a storage medium, when the computer program is executed by the processor, the system in any optional implementation of the above-mentioned embodiments is executed. Wherein, the storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0124] It should be understood that the disclosed system can be implemented in other ways. The above-described system embodiments are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, another division manner can be used. For example, a plurality of modules or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed systems or units can be indirect coupling or communication connection through some communication interfaces, and can be electrical, mechanical or other forms.

[0125] In addition, the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0126] In addition, the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0127] Flowcharts herein are used to illustrate the steps of the method through the embodiments of the disclosure. It should be understood that the preceding or subsequent steps do not necessarily proceed in sequence. On the contrary, various steps can be evaluated in reverse order or simultaneously. Other operations can also be added to these processes.

[0128] Unless otherwise defined, all terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It should also be understood that terms such as those defined in generally used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or extremely formalized sense, unless expressly defined herein.

[0129] The above provides a method and system for constructing an intervention mode of a pre-established medical care plan for an elderly patient. The principles and implementation of the present application are described using specific examples. The above description of the embodiments is merely illustrative of the present application and is used to help understand the method and system for constructing an intervention mode of a pre-established medical care plan for an elderly patient, and does not limit the scope of protection of the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for constructing an intervention model for advance medical care planning for elderly patients, characterized in that, include: Obtain ACP information of the target patient, extract key information from the ACP information, and map each key information into a willingness preference vector according to the value preference dimension; Each value preference dimension has a value range of [-1, 1]. When the value of the corresponding dimension is -1, it indicates no preference; when the value of the corresponding dimension is 1, it indicates preference; when the value of the corresponding dimension is 0, it indicates neutrality or no mention. Obtain candidate medical plans for the target patient, extract the medical entities of the candidate medical plans, and map each medical entity into a plan preference vector according to the value preference dimension; Generate consistency comparison results between the intention preference vector and the alternative preference vector; Recommended values ​​for candidate medical solutions are generated based on the consistency comparison results; The step of extracting key information from ACP information also includes: performing text cleaning and word segmentation on each ACP information, mapping each ACP information to a word vector, and then inputting it into a pre-trained text extraction model to output the key information.

2. The method for constructing an intervention model for advance medical care planning for elderly patients according to claim 1, characterized in that, After obtaining the ACP information of the target patient, the process also includes: if the ACP information contains at least two versions, generating a preference vector based on the latest version of the ACP information.

3. The method for constructing an intervention model for advance medical care planning for elderly patients according to claim 2, characterized in that, The recommended values ​​for generating candidate medical plans also include: The fluctuation range of the target patient in each value preference dimension is obtained based on the willingness and preference vector of each ACP information. Stable weights for the corresponding value preference dimension are generated based on the volatility. The recommended value is obtained by calculating the weighted similarity between the intention preference vector and the alternative preference vector based on the stable weights.

4. The method for constructing an intervention model for advance medical care planning for elderly patients according to claim 1, characterized in that, The process of mapping key information to a preference vector also includes matching each key information to a medical entity database. If a match fails, a value preference interval is matched based on the corresponding key information. When generating a consistency comparison result, the corresponding consistency comparison result is generated based on whether the value of the scheme preference vector in the corresponding value preference dimension belongs to the corresponding value preference interval.

5. The method for constructing an intervention model for advance medical care planning for elderly patients according to claim 4, characterized in that, Within the matching of value preference ranges based on corresponding key information, it also includes: Determine if there is key information that successfully matches the medical entity database. If so, use the value of the key information that successfully matches the medical entity database in each value preference dimension as a reference value. Obtain the value preference range based on the range where the distance from the reference value does not exceed the value distance threshold.

6. The method for constructing an intervention model for advance medical care planning for elderly patients according to claim 5, characterized in that, If there are at least two key pieces of information that successfully match the medical entity database, after obtaining all reference values, the following steps are also taken: take the interval whose distance from each reference value does not exceed the value distance threshold as the reference interval, and take the intersection of all reference intervals as the value preference interval.

7. A method for constructing an intervention model for advance medical care planning for elderly patients according to any one of claims 1-6, characterized in that, The method further includes: comparing the solution preference vector corresponding to each medical entity with all willingness preference vectors, taking the comparison result with the highest consistency as the consistency comparison result of the corresponding medical entity, and generating the recommended value of the candidate medical solution based on the consistency comparison results of all medical entities.

8. A system for constructing an intervention model for advance medical care planning for elderly patients, characterized in that, include: The acquisition module is used to acquire ACP information and candidate treatment plans for the target patient; The extraction module is used to extract key information from ACP information and medical entities from candidate medical solutions; The step of extracting key information from ACP information also includes: performing text cleaning and word segmentation on each ACP information, mapping each ACP information to word vectors, and then inputting them into a pre-trained text extraction model to output the key information. The mapping module is used to map each key piece of information to a willingness preference vector and each medical entity to a solution preference vector according to the value preference dimension. The value range of each value preference dimension is [-1, 1]. When the value of the corresponding dimension is -1, it means no preference; when the value of the corresponding dimension is 1, it means preference; when the value of the corresponding dimension is 0, it means neutral or not mentioned. The comparison module is used to generate consistency comparison results between the intention preference vector and the alternative preference vector; The generation module is used to generate recommended values ​​for candidate medical solutions based on the consistency comparison results.

Images