WO2020119383A1

WO2020119383A1 - Medical insurance supervision method, device, apparatus and computer readable storage medium

Info

Publication number: WO2020119383A1
Application number: PCT/CN2019/118810
Authority: WO
Inventors: 陈明东; 黄越; 胥畅
Original assignee: 平安医疗健康管理股份有限公司
Priority date: 2018-12-13
Filing date: 2019-11-15
Publication date: 2020-06-18
Also published as: CN109636645A

Abstract

A medical insurance supervision method, a device, an apparatus and a computer readable storage medium, the medical insurance supervision method being implemented on the basis of intelligent decision making. The medical insurance supervision method comprises the following steps: acquiring treatment data of an insured individual (S100); extracting a preset field in the treatment data according to a preset data model (S200); and analyzing the preset field on the basis of the preset model and extracting a suspected abnormal case according to the analysis result (S300). The described method helps to reduce the difficulty of medical insurance supervision, thereby improving the accuracy and strictness of medical insurance supervision.

Description

Medical insurance supervision method, equipment, device and computer readable storage medium The

This application requires the priority of the Chinese patent application submitted to the China Patent Office on December 13, 2018, with the application number 201811530813.6 and the invention titled "Medical Insurance Supervision Methods, Equipment, Devices, and Computer-readable Storage Media." The reference is incorporated in the application.

Technical field

This application relates to the technical field of medical insurance, in particular to a medical insurance supervision method, device, device and computer-readable storage medium.

Background technique

With the development of society and economy, the national medical insurance policy is gradually optimized, and more and more expenses incurred in the medical process can be reimbursed. However, in order to obtain more reimbursement, some insured persons have violated regulations such as falsifying the medical condition, splitting expenses, and excessive treatment, resulting in the loss of medical insurance funds. At present, medical insurance supervision mainly relies on the method of manual review. However, due to factors such as a large number of insured persons, huge medical data and strong specialization, manual supervision is very difficult. It not only requires a large amount of professional human resources, but also The accuracy is also low, and it is easy to produce regulatory loopholes.

Summary of the invention

The main purpose of this application is to provide a method of medical insurance supervision, which aims to solve the above technical problems of medical insurance supervision that are difficult, low in accuracy, and prone to regulatory loopholes, so as to reduce the difficulty of medical insurance supervision and improve the accuracy and rigor of medical insurance supervision .

To achieve the above purpose, this application provides a medical insurance supervision method, including the following steps:

Obtain medical data of the insured;

According to a preset model, extracting preset fields in the medical consultation data, wherein the medical consultation data includes information of the insured, medical records and diagnosis and treatment paths;

The preset field is analyzed based on the preset model, and a suspected abnormal case is extracted according to the analysis result.

To achieve the above purpose, the present application further proposes a medical insurance supervision device, the medical insurance supervision device includes: a memory, a processor, and computer-readable instructions stored on the memory and operable on the processor, the When the computer-readable instructions are executed by the processor, the steps of the medical insurance supervision method are implemented. The medical insurance supervision method includes the following steps: obtaining medical data of the insured, wherein the medical data includes information of the insured, medical records, and Diagnosis and treatment path; according to a preset model, extract the preset fields in the visit data; analyze the preset fields based on the preset model, and extract suspected abnormal cases according to the analysis results.

In order to achieve the above purpose, the present application also proposes a medical insurance supervision device, the medical insurance supervision device includes a data acquisition module, a field extraction module and an abnormality analysis module, wherein the data acquisition module is used to obtain medical data of the insured, Wherein, the medical consultation data includes information of the insured, medical records and diagnosis and treatment paths; the field extraction module is used to extract preset fields in the medical consultation data according to a preset model; and the abnormality analysis module is used to The preset model analyzes the preset field and extracts a suspected abnormal case according to the analysis result.

In order to achieve the above object, the present application further proposes a computer-readable storage medium having medical insurance supervision readable instructions stored on the computer-readable storage medium, the medical insurance supervision readable instructions being executed by a processor to implement a medical insurance supervision method Steps, the medical insurance supervision method includes the following steps: obtaining medical data of the insured, wherein the medical data includes information of the insured, medical records and diagnosis and treatment paths; according to a preset model, extracting Set a field; analyze the preset field based on the preset model, and extract a suspected abnormal case according to the analysis result.

In the technical solution of the present application, the medical insurance supervision method includes the following steps: obtaining medical data of the insured; extracting preset fields in the medical data according to a preset model, wherein the medical data includes information of the insured and medical records And diagnosis and treatment path; analyze the preset field based on the preset model, and extract the suspected abnormal case according to the analysis result. By extracting the preset fields in the insured's visit data, the redundant information in the visit data is removed, the key information is retained, and the analysis difficulty of the subsequent preset model is reduced. Further, the preset fields are analyzed based on the preset model. Generally, the suspected abnormal case has a larger difference from the normal case. According to the analysis result, the suspected abnormal case can be automatically extracted from many cases for further processing. By setting different preset models, the corresponding regulatory requirements can be achieved, thereby achieving automatic supervision of medical insurance without manual supervision, reducing the difficulty of supervision, greatly saving human resources, and improving the accuracy and rigor of medical insurance supervision , Effectively avoid the occurrence of regulatory loopholes.

BRIEF DESCRIPTION

FIG. 1 is a schematic flowchart of a first embodiment of a medical insurance supervision method for application;

2 is a schematic flowchart of a second embodiment of a medical insurance supervision application method;

FIG. 3 is a schematic flowchart of a third embodiment of a medical insurance supervision application method;

FIG. 4 is a detailed flowchart of step S200 in the fourth embodiment of the medical insurance supervision method of the application; FIG.

FIG. 5 is a detailed flowchart of step S240 in the fifth embodiment of the medical insurance supervision method of the application; FIG.

FIG. 6 is a detailed flowchart of step S300 in the sixth embodiment of the medical insurance supervision method of the application; FIG.

FIG. 7 is a schematic structural diagram of a medical insurance supervision device in a hardware operating environment involved in an embodiment of the present application.

The implementation, functional characteristics and advantages of the present application will be further described in conjunction with the embodiments and with reference to the drawings.

detailed description

It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

The main solution of the embodiment of the present application is to analyze the medical data of the insured based on a preset model and extract suspected abnormal cases.

In the prior art, medical insurance supervision is usually implemented manually, which is difficult to supervise, requires a large amount of professional human resources, and has low accuracy of manual supervision. It is also easy to produce supervision loopholes.

This application provides a solution to analyze the insured's medical data based on a preset model and extract suspected abnormal cases to achieve automatic supervision of medical insurance costs, reduce the difficulty of supervision, and improve the accuracy of supervision And rigor.

The first embodiment of the present application provides a medical insurance supervision method. As shown in FIG. 1, the medical insurance supervision method includes the following steps:

Step S100: Obtain medical data of the insured, wherein the medical data includes information of the insured, medical records and diagnosis and treatment paths;

Generally, medical insurance costs are settled directly when the insured swipes a card to pay for medical treatment, so as to facilitate the medical treatment of the insured and improve the circulation efficiency of hospitals and other medical institutions. Medical insurance costs cover all or part of the insured's medical expenses. In order to supervise the medical insurance costs, it is necessary to obtain the insured's medical data, and further determine whether there are any abnormal reimbursements based on the medical data. Among them, the consultation data specifically include the information of the insured, the condition record and the diagnosis and treatment path, etc. The condition record helps to monitor the development of the insured's condition and determine whether the diagnosis and treatment path adopted is reasonable and effective according to the development of the condition. Among them, diagnosis and treatment The path can specifically include the insured's inspection items and drug programs. The insured's visit data can be uploaded to the user's system (such as the Human Resources and Social Security Bureau) after the treatment (including outpatient treatment and inpatient treatment), or can be uploaded to the user's system regularly during the treatment In order to discover the possible loss of reimbursement or medical insurance costs in time to avoid the eventual loss of medical insurance costs.

Step S200: According to the preset model, extract the preset fields in the consultation data;

The medical data usually includes basic information such as the insured's name and age, the corresponding inspection information for various examinations performed by the insured, and medical record information such as medical symptoms and treatment plans recorded by the doctor. Among them, the preset fields included in the basic information can often be extracted directly from the form in a relatively simple manner, and the form of the inspection information and medical record information is varied, and the presets often need to be extracted based on natural language processing and the like Fields, for further analysis, the extraction of preset fields will be further elaborated later. The preset field is determined according to the preset model, and the preset model is determined according to the specific item to be supervised. For example, when it is necessary to supervise whether there are illegal hospitalization items such as decomposition hospitalization and false hospitalization, it is necessary to extract preset fields related to the insured's hospitalization. According to the preset model, the preset fields in the consultation data are extracted for further analysis.

Step S300: Analyze the preset field based on the preset model, and extract the suspected abnormal case according to the analysis result.

After the preset field is extracted, the preset field is analyzed based on the preset model, and the suspected abnormal case is extracted according to the analysis result. In general, the analysis of preset fields is mainly deviation analysis. Deviation refers to abnormal or irregular cases in the case to be analyzed, or cases where the consistency between the observation results and the results predicted by the preset model is poor. The goal of deviation analysis is to find out abnormal cases with meaningful differences from the reference object. Specific deviation analysis methods include clustering method, sequence anomaly method, nearest neighbor method, multidimensional data analysis method, etc., which will be detailed later Elaborate. According to the preset fields, analyze the deviation of a particular insured relative to other insureds. If the deviation of the insured is relatively large compared to other insureds, the case is extracted as a suspected abnormal case, and further abnormalities are to be confirmed Cases, recalculation of medical insurance costs, etc.

In this embodiment, the medical insurance supervision method includes the following steps: obtaining medical data of the insured, wherein the medical data includes information of the insured, medical records, and diagnosis and treatment paths; according to a preset model, extracting preset fields in the medical data ; Analyze the preset fields based on the preset model, and extract suspected abnormal cases based on the analysis results. By extracting the preset fields in the insured's visit data, the redundant information in the visit data is removed, the key information is retained, and the analysis difficulty of the subsequent preset model is reduced. Further, the preset fields are analyzed based on the preset model. Usually, the suspected abnormal case has a larger deviation from the normal case. According to the analysis result, the suspected abnormal case can be automatically extracted from many cases for further processing. By setting different preset models, the corresponding regulatory requirements can be achieved, thereby achieving automatic supervision of medical insurance without manual supervision, reducing the difficulty of supervision, greatly saving human resources, and improving the accuracy and rigor of medical insurance supervision , Effectively avoid the occurrence of regulatory loopholes.

Based on the above first embodiment, as shown in FIG. 2, in the second embodiment of the present application, after analyzing the preset fields based on the preset model and extracting the suspected abnormal case according to the analysis result, the medical insurance supervision method further includes The following steps:

Step S400, output a suspected abnormal case;

Step S500: Receive confirmation information corresponding to the suspected abnormal case;

Step S610: When the confirmation information confirms that the suspected abnormal case is abnormal, the medical insurance fee is amended.

In this embodiment, when a suspected abnormal case is extracted, it is further confirmed whether the case is an abnormal case by outputting the suspected abnormal case to improve the accuracy of supervision. The further confirmation may be automatic confirmation or manual confirmation, or a combination of automatic confirmation and manual confirmation. For example, for a suspected abnormal case of obvious abnormality, the range of the preset field corresponding to the obvious abnormality can be set, and the case can be directly confirmed by the supervisory readable command as an abnormal case, and corresponding confirmation information can be generated. For other cases, the relevant personnel will manually confirm and give the corresponding confirmation information. When receiving the confirmation information corresponding to the suspected abnormal case, if the confirmation information confirms that the suspected abnormal case is abnormal, the medical insurance fee is amended. Specifically, you can issue a deduction basis and a deduction notice to the relevant responsible unit or responsible person, and directly deduct medical insurance expenses to improve processing efficiency. In order to strengthen supervision, the statistical results of abnormal cases within a certain period of time can also be issued regularly, including the relevant responsible units or persons responsible for the abnormal cases, so as to further deal with the serious violations of the responsible units or persons responsible.

Based on the second embodiment described above, as shown in FIG. 3, in the third embodiment of the present application, after the step of receiving confirmation information corresponding to the suspected abnormal case, the following steps are further included:

Step S620: When the confirmation information denies that the suspected abnormal case is abnormal, optimize the preset model according to the suspected abnormal case, where the optimization of the preset model is implemented based on machine learning.

For relatively complicated suspected abnormal cases, after further confirmation by automatic or manual, if the suspected abnormal case is not abnormal, it indicates that the preset model has certain defects. According to the suspected abnormal case, or after accumulating a certain number of suspected abnormal cases with preliminary judgment errors, the preset model is optimized based on the method of machine learning to improve the accuracy and reliability of medical insurance supervision. Of course, it is also possible to make appropriate reforms to the medical insurance scheme based on preset models to meet the actual insurance needs of the insured.

Based on the above embodiments, as shown in FIG. 4, in the fourth embodiment of the present application, the step of extracting the preset fields in the consultation data according to the preset model includes:

Step S210, cleaning the medical data;

Step S220: Obtain the canonical text and non-normative text in the cleaned medical data, respectively;

Step S230: Extract the preset fields in the specification text according to the preset model;

Step S240: Analyze the non-standard text based on the recurrent neural network, and extract the preset fields in the non-standard text according to the analysis results of the preset model and the non-standard text.

Among them, data cleaning refers to checking and correcting the identifiable errors in the data, including checking the consistency of the data, removing duplicate data, correcting invalid and erroneous data, etc., in order to improve the efficiency of the execution of subsequent steps. Obtain the standardized text and non-standard text of the medical data after cleaning. Specifically, the standard text refers to the relatively fixed information such as the name and age of the insured person recorded in specific items such as tables and forms. Non-standard text refers to relatively complex information handwritten by doctors and other personnel such as examination information and medical record information. For the canonical text, the preset fields can be directly extracted according to the requirements of the preset model, while for the non-canonical text, the preset fields are often extracted based on the natural language processing method and based on recurrent neural network analysis .

Based on the fourth embodiment described above, as shown in FIG. 5, in the fifth embodiment of the present application, non-standard text is analyzed based on a recurrent neural network, and the non-standard text is extracted from the analysis results of the preset model and the non-standard text The steps of the preset fields include:

Step S241: Represent non-standard text as a sequence of vectors;

Step S242: Based on the semantic content and semantic distance of the vector sequence, encode the vector sequence into a sentence vector matrix based on the bidirectional recurrent neural network;

Step S243: According to the preset model, the attention mechanism is used to compress the sentence vector matrix into a sentence vector, and the preset field in the sentence vector is extracted.

Specifically, the non-normative text to be extracted is represented by a vector sequence, combined with the semantic content and semantic distance of the vector sequence, the vector sequence is encoded into a sentence vector matrix based on the bidirectional recurrent neural network model. The semantic content reflects the meaning of the text itself, while the semantic distance reflects the correlation between the texts, which can be expressed by correlation functions or correlation coefficients. For the sentence vector matrix, each row can be understood as a word vector, and the word vector is sensitive to the context of the sentence. Further, according to the preset model, the attention vector mechanism is used to compress the sentence vector matrix into sentence vectors, and the required preset fields are extracted therefrom, so that the consultation data is matched to the corresponding standardized preset fields for subsequent analysis . Among them, the attention mechanism makes the neural network have the ability to focus on its input (or feature) subset: select a specific input. Attention can be applied to any type of input regardless of its shape. In the case of limited computing power, the attention mechanism is a resource allocation scheme that solves the problem of information overload, so as to allocate computing resources to more important tasks.

Based on the above embodiments, as shown in FIG. 6, in the sixth embodiment of the present application, the step of analyzing the preset field based on the preset model and extracting the suspected abnormal case according to the analysis result includes:

Step S310: Calculate the distance metric of the preset field in the case corresponding to each insured;

Step S320: Calculate the outlier degree of the distance metric of the insured's preset field relative to the distance metric of other insured's preset fields;

Step S330, comparing the outlier degree and the preset outlier degree;

Step S340: When the outlier degree is greater than the preset outlier degree, mark the case corresponding to the insured as a suspected abnormal case.

Statistical analysis is used to analyze the insured's visit data. Among them, each insured's visit data can be regarded as a point in the statistical distribution graph, and whether there is a suspected abnormal case is determined according to the distribution position of the point.

Further, corresponding to different regulatory requirements, the preset fields may include at least one of symptoms of illness, examination items, medication plan, length of hospitalization, and length of hospitalization interval. In the following text, several specific regulatory requirements will be used as examples, and different deviation analysis or outlier analysis methods will be used to elaborate in detail.

In a specific example, the possible decomposition of hospitalization is monitored. Among them, decomposed hospitalization means that the insured person deliberately splits one hospitalization into multiple hospitalizations in order to obtain more reimbursement expenses, so as to reimburse them separately, resulting in the loss of medical insurance funds. When supervising the decomposition of hospitalization, an abnormal situation, you can perform deviation analysis or outlier analysis on the corresponding preset fields according to the characteristics of decomposition hospitalization. The distance metric can be obtained based on the absolute distance (Manhattan distance), Euclidean distance, and Mahalanobis distance of the preset fields. The preset field can be selected as the range, interquartile range, interquartile range, mean difference, standard deviation, etc. of the number of days between two consecutive hospitalizations of the insured, so as to preliminarily screen out the cases of suspected decomposition hospitalization. Among them, the larger the value of the data variability index, the greater the data variability and the wider the spread. The threshold for the number of days between intervals can be set to 10 to 15 days. For insured persons whose number of days between consecutive hospitalizations is less than the threshold, it can be classified as a suspected abnormal case. Further, by analyzing the medical condition in the medical data of the suspected abnormal case and the diagnosis and treatment path including the inspection items and medication plan, etc., the suspected abnormal case is further confirmed. If the same insured person is hospitalized for the same disease within a short period of time, or the disease path of the two hospitalizations is the same or belongs to a continuation relationship, it is determined to be a case of suspected decomposition hospitalization.

In another specific example, the possible false hospitalizations are monitored. Among them, false hospitalization refers to the insured person making hospital records without hospitalization in order to obtain more reimbursement expenses in order to defraud the reimbursement expenses. For example, when there is a medical hospitalization for more than 15 days or a surgical hospitalization for more than 30 days, but there is no treatment plan, it corresponds to a false hospitalization. The anomaly detection algorithm adopted by the preset model in identifying suspected abnormal cases is an index-based algorithm, that is, given a data set, the index-based algorithm uses a multi-dimensional index structure to find the phase of each case to be analyzed within a certain radius O case. Suppose M is the maximum number of objects within a certain radius of the case to be analyzed. If the M+l adjacent case of the case to be analyzed is found, the case to be analyzed is not a suspected abnormal case. Considering the complexity of the index-based algorithm, it has good scalability.

In yet another specific example, the situation in which the insured person is admitted to the hospital for treatment when the patient's condition does not reach the admission treatment indication is supervised. Specifically, the analysis is performed through a nested loop algorithm, that is, a nested loop algorithm and an index-based algorithm. Because the loop algorithm and the nested algorithm have the same complexity, and at the same time avoid the construction of the index structure, in order to minimize the number of data input and output, the memory buffer space is divided into two halves, and the data set is divided into several logical blocks. By carefully selecting the order in which the logic blocks are loaded into each buffer area, the efficiency of input and output is effectively improved.

In yet another specific example, the situation of splitting the charges of a project is monitored. Specifically, the unit-based algorithm analyzes the insured's medical data. By dividing the data space into units with side length equal to d/(2*k1/2). Each unit has two layers surrounding it. The thickness of the first layer is one unit, and the thickness of the second layer is (2*k1/2-1). The algorithm counts outliers on a unit-by-unit basis rather than on an object-by-object basis. For a given unit, three counts are accumulated: the number of objects in the unit, the number of objects in the unit and the first layer, and the number of objects in the unit and two layers. The algorithm changes the detection of outlier data for each element of the data set to the detection of outlier data for each unit, thereby improving the efficiency of the algorithm and reducing the complexity of the algorithm. It performs anomaly detection in this way: if the number of objects in the unit and the first layer is greater than a preset number, all objects in the unit are not abnormal; if the number of objects in the unit and the two layers is less than or equal to the above-mentioned preset number , All objects in the unit are abnormal; otherwise, some data in the unit may be abnormal. In order to detect these abnormal points, you need to join the processing object by object.

As shown in FIG. 7, FIG. 7 is a schematic structural diagram of a terminal of a hardware operating environment involved in the embodiment of the present application, that is, a medical insurance supervision device.

In this embodiment of the present application, the terminal may be a server, a PC, or a smart phone, tablet computer, e-book reader, MP3 (Moving Picture Experts Group Audio Layer III, motion picture expert compression standard audio layer 3 player, MP4 (Moving Picture Experts Group Audio Layer IV, the standard audio layer for motion picture experts compression 3) Players, portable computers and other mobile terminal devices with display functions.

As shown in FIG. 7, the terminal may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Among them, the communication bus 1002 is used to implement connection communication between these components. The user interface 1003 may include a display (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as disk storage. The memory 1005 may optionally be a storage device independent of the foregoing processor 1001.

Optionally, the terminal may also include a camera, RF (Radio Frequency (radio frequency) circuits, sensors, audio circuits, WiFi modules, etc. Among them, sensors such as light sensors, motion sensors and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display screen according to the brightness of the ambient light, and the proximity sensor may turn off the display screen and/or when the mobile terminal moves to the ear Backlight. As a type of motion sensor, the gravity acceleration sensor can detect the magnitude of acceleration in various directions (generally three axes), and can detect the magnitude and direction of gravity when at rest. It can be used for applications that recognize the posture of mobile terminals (such as horizontal and vertical screen switching, Related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tap), etc. Of course, the mobile terminal can also be configured with other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc. No longer.

A person skilled in the art may understand that the terminal structure shown in FIG. 7 does not constitute a limitation on the terminal, and may include more or less components than those shown in the figure, or a combination of certain components, or a different component arrangement.

As shown in FIG. 7, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and medical insurance supervision readable instructions.

In the terminal shown in FIG. 7, the network interface 1004 is mainly used to connect to the back-end server and perform data communication with the back-end server; the user interface 1003 is mainly used to connect to the client (user end) and perform data communication with the client; and the processor 1001 can be used to call the medical insurance supervision readable instructions stored in the memory 1005, and perform the following operations:

Obtain the insured's visit data, where the visit data includes the insured's information, medical record and diagnosis and treatment path;

According to the preset model, extract the preset fields in the consultation data;

Analyze the preset fields based on the preset model, and extract suspected abnormal cases based on the analysis results.

Further, the processor 1001 may be used to call the medical insurance supervision readable instructions stored in the memory 1005, analyze the preset field based on the preset model, and extract the suspected abnormal case according to the analysis result, and then perform the following operations:

Output suspected abnormal cases;

Receive confirmation information corresponding to suspected abnormal cases;

When the confirmation information confirms that the suspected abnormal case is abnormal, the medical insurance fee is amended.

Further, the processor 1001 may be used to call the medical insurance supervision readable instructions stored in the memory 1005, and after receiving the confirmation information corresponding to the suspected abnormal case, perform the following operations:

When the confirmation information denies that the suspected abnormal case is abnormal, the preset model is optimized according to the suspected abnormal case, wherein the optimization of the preset model is implemented based on machine learning.

Further, the processor 1001 may be used to call the medical insurance supervision readable instructions stored in the memory 1005, and the operation of extracting the preset fields in the medical data according to the preset model includes:

Clean the visit data;

Obtain the canonical text and non-normative text in the cleaned medical data, respectively;

According to the preset model, extract the preset fields in the specification text;

The non-standard text is analyzed based on the recurrent neural network, and the preset fields in the non-standard text are extracted according to the analysis results of the preset model and the non-standard text.

Further, the processor 1001 may be used to call the medical insurance supervision readable instructions stored in the memory 1005, analyze the non-standard text based on the recurrent neural network, and extract the pre-code in the non-standard text according to the analysis result of the preset model and the non-standard text The operations for setting fields include:

Represent non-standard text as a sequence of vectors;

According to the semantic content and semantic distance of the vector sequence, the vector sequence is encoded into a sentence vector matrix based on the bidirectional recurrent neural network;

According to the preset model, the attention mechanism is used to compress the sentence vector matrix into sentence vectors, and the preset fields in the sentence vectors are extracted.

Further, the processor 1001 may be used to call the medical insurance supervision readable instructions stored in the memory 1005, analyze the preset field based on the preset model, and extract the suspected abnormal case according to the analysis result includes:

Calculate the distance metric of the preset field in the case corresponding to each insured;

Calculate the outlier degree of the distance measure of the insured's preset field relative to the distance measure of other insured's preset fields;

Compare the outlier degree with the preset outlier degree;

When the outlier degree is greater than the preset outlier degree, the case corresponding to the insured is marked as a suspected abnormal case.

Further, the processor 1001 may be used to call the medical insurance supervision readable instructions stored in the memory 1005, and the preset fields include at least one of the symptom of the condition, the examination item, the medication plan, the length of hospitalization, and the length of the interval between hospitalizations.

In addition, an embodiment of the present application also proposes a medical insurance supervision device. The medical insurance supervision device includes:

The data acquisition module is used to obtain the medical data of the insured, wherein the medical data includes the information of the insured, the medical record and the path of diagnosis and treatment;

The field extraction module is used to extract the preset fields in the consultation data according to the preset model;

The abnormality analysis module is used to analyze the preset fields based on the preset model and extract suspected abnormal cases according to the analysis results.

Further, the medical insurance supervision device also includes:

Case output module to output suspected abnormal cases;

Information receiving module, used to receive confirmation information corresponding to suspected abnormal cases;

The expense revision module is used to revise the medical insurance expenses when the confirmation information confirms that the suspected abnormal case is abnormal.

Further, the medical insurance supervision device also includes:

The model optimization module is used to optimize the preset model according to the suspected abnormal case when the confirmation information denies that the suspected abnormal case is abnormal, wherein the optimization of the preset model is implemented based on machine learning.

Further, the field extraction module includes:

The data cleaning unit is used to clean up the medical data;

The text classification unit is used to obtain the canonical text and non-normative text in the cleaned visit data, respectively;

The field extraction unit is used to extract the preset fields in the specification text according to the preset model;

The field extraction unit is also used to analyze the non-standard text based on the recurrent neural network, and extract the preset fields in the non-standard text according to the analysis results of the preset model and the non-standard text.

Further, the field extraction unit includes:

Vector sequence subunit to represent non-standard text as a sequence of vectors;

Vector matrix subunit, used to encode the vector sequence into a sentence vector matrix based on the bidirectional recurrent neural network based on the semantic content and semantic distance of the vector sequence;

The vector compression subunit is used to compress the sentence vector matrix into a sentence vector according to a preset model, and extract a preset field in the sentence vector.

Further, the abnormality analysis module includes:

The distance measurement unit is used to calculate the distance measurement of the preset field in the case corresponding to each insured;

The outlier degree unit is used to calculate the outlier degree of the distance measure of the insured's preset field relative to the distance measure of the other insured's preset field;

The comparison unit is used to compare the outlier degree and the preset outlier degree;

The marking unit is used to mark the case corresponding to the insured as a suspected abnormal case when the outlier degree is greater than the preset outlier degree.

Further, the preset fields include at least one of illness symptom, examination item, medication plan, length of hospitalization and length of hospitalization interval.

In addition, the embodiments of the present application also propose a computer-readable storage medium, which may be a non-volatile readable storage medium; the computer-readable storage medium stores medical insurance regulatory readable instructions, and the medical insurance regulatory readable The instruction performs the following operations when executed by the processor:

Obtain the medical data of the insured, where the medical data includes the information of the insured, the medical record and the path of diagnosis and treatment;

Further, when the medical insurance supervision readable instruction is executed by the processor, after analyzing the preset field based on the preset model and extracting the suspected abnormal case according to the analysis result, the following operations are also performed:

Output suspected abnormal cases;

Receive confirmation information corresponding to suspected abnormal cases;

Further, when the medical insurance supervision readable instruction is executed by the processor, after receiving the confirmation information corresponding to the suspected abnormal case, it also performs the following operations:

Further, when the medical insurance supervision readable instruction is executed by the processor, the operation of extracting the preset field in the consultation data according to the preset model includes:

Clean the visit data;

Further, when the medical insurance supervision readable instructions are executed by the processor, the non-standard text is analyzed based on the recurrent neural network, and according to the analysis results of the preset model and the non-standard text, the operation of extracting the preset fields in the non-standard text includes:

Represent non-standard text as a sequence of vectors;

Further, when the medical insurance supervision readable instruction is executed by the processor, the operation of analyzing the preset field based on the preset model and extracting the suspected abnormal case according to the analysis result includes:

Compare the outlier degree with the preset outlier degree;

Further, when the medical insurance supervision readable instruction is executed by the processor, the preset field includes at least one of the condition symptom, the examination item, the medication plan, the length of hospitalization and the length of the interval between hospitalizations.

It should be noted that in this article, the terms "include", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system that includes a series of elements includes not only those elements, It also includes other elements that are not explicitly listed, or include elements inherent to this process, method, article, or system. Without more restrictions, the element defined by the sentence "include one..." does not exclude that there are other identical elements in the process, method, article or system that includes the element.

The sequence numbers of the above embodiments of the present application are for description only, and do not represent the advantages and disadvantages of the embodiments.

Through the description of the above embodiments, those skilled in the art can clearly understand that the methods in the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware, but in many cases the former is better Implementation. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence or part that contributes to the existing technology, and the computer software product is stored in a storage medium (such as ROM/RAM as described above) , Magnetic disk, optical disk), including several instructions to make a terminal device (which can be a mobile phone, computer, server, air conditioner, or network equipment, etc.) to perform the method described in each embodiment of the present application.

The above are only the preferred embodiments of the present application, and do not limit the scope of the patent of the present application. Any equivalent structure or equivalent process transformation made by the description and drawings of this application, or directly or indirectly used in other related technical fields , The same reason is included in the scope of patent protection in this application.

Claims

A medical insurance supervision method, characterized in that the medical insurance supervision method includes the following steps:

Obtain the medical data of the insured, wherein the medical data includes the information of the insured, the medical record and the path of diagnosis and treatment;

According to a preset model, extract preset fields in the medical data;

The preset field is analyzed based on the preset model, and a suspected abnormal case is extracted according to the analysis result.
The medical insurance supervision method according to claim 1, wherein after the step of analyzing the predetermined field based on the predetermined model and extracting a suspected abnormal case according to the analysis result, the medical insurance supervision method further comprises the following steps :

Output the suspected abnormal case;

Receive confirmation information corresponding to the suspected abnormal case;

When the confirmation information confirms that the suspected abnormal case is abnormal, the medical insurance fee is revised.
The medical insurance supervision method according to claim 2, wherein after the step of receiving confirmation information corresponding to the suspected abnormal case, the method further includes the following steps:

When the confirmation information denies that the suspected abnormal case is abnormal, the preset model is optimized according to the suspected abnormal case, wherein the optimization of the preset model is implemented based on machine learning.
The medical insurance supervision method according to claim 1, wherein the step of extracting a preset field in the medical data according to a preset model includes:

Clean the visit data;

Obtain the canonical text and non-normative text in the cleaned medical data, respectively;

Extract the preset fields in the specification text according to the preset model;

The non-normative text is analyzed based on the recurrent neural network, and the preset fields in the non-normative text are extracted according to the analysis results of the preset model and the non-normative text.
The medical insurance supervision method according to claim 4, wherein the non-standard text is analyzed based on a recurrent neural network, and the non-standard text is extracted according to the analysis result of the preset model and the non-standard text The steps of the preset fields include:

Express the non-standard text as a sequence of vectors;

According to the semantic content and semantic distance of the vector sequence, the vector sequence is encoded into a sentence vector matrix based on a bidirectional recurrent neural network;

According to the preset model, an attention mechanism is used to compress the sentence vector matrix into a sentence vector, and a preset field in the sentence vector is extracted.
The medical insurance supervision method according to claim 1, wherein the step of analyzing the preset field based on the preset model and extracting the suspected abnormal case according to the analysis result includes:

Calculate the distance metric of the preset field in the case corresponding to each insured;

Calculate the outlier degree of the distance measure of the insured's preset field relative to the distance measure of other insured's preset fields;

Compare the outlier degree with the preset outlier degree;

When the outlier degree is greater than the preset outlier degree, the case corresponding to the insured is marked as a suspected abnormal case.
The medical insurance supervision method according to claim 1, wherein the preset field includes at least one of a disease symptom, an examination item, a drug plan, a length of hospitalization, and a length of interval between hospitalizations.
A medical insurance supervision device, characterized in that the medical insurance supervision device includes: a memory, a processor, and computer-readable instructions stored on the memory and operable on the processor, the computer-readable instructions are When the processor executes, the following steps are realized:

Obtain the medical data of the insured, wherein the medical data includes the information of the insured, the medical record and the path of diagnosis and treatment;

According to a preset model, extract preset fields in the medical data;

The preset field is analyzed based on the preset model, and a suspected abnormal case is extracted according to the analysis result.
The medical insurance supervision device according to claim 8, wherein after the step of analyzing the preset field based on the preset model and extracting a suspected abnormal case according to the analysis result, the medical insurance supervision method further includes The following steps:

Output the suspected abnormal case;

Receive confirmation information corresponding to the suspected abnormal case;

When the confirmation information confirms that the suspected abnormal case is abnormal, the medical insurance fee is revised.
The medical insurance supervision device of claim 9, wherein after the step of receiving confirmation information corresponding to the suspected abnormal case, the method further includes the following steps:

When the confirmation information denies that the suspected abnormal case is abnormal, the preset model is optimized according to the suspected abnormal case, wherein the optimization of the preset model is implemented based on machine learning.
The medical insurance supervision device according to claim 8, wherein the step of extracting a preset field in the medical data according to a preset model includes:

Clean the visit data;

Obtain the canonical text and non-normative text in the cleaned medical data, respectively;

Extract the preset fields in the specification text according to the preset model;

The non-normative text is analyzed based on the recurrent neural network, and the preset fields in the non-normative text are extracted according to the analysis results of the preset model and the non-normative text.
The medical insurance supervision device according to claim 11, wherein the non-standard text is analyzed based on a recurrent neural network, and the non-standard text is extracted from the analysis results of the preset model and the non-standard text The steps of the preset fields include:

Express the non-standard text as a sequence of vectors;

According to the semantic content and semantic distance of the vector sequence, the vector sequence is encoded into a sentence vector matrix based on a bidirectional recurrent neural network;

According to the preset model, an attention mechanism is used to compress the sentence vector matrix into a sentence vector, and a preset field in the sentence vector is extracted.
The medical insurance supervision device according to claim 11, wherein the step of analyzing the preset field based on the preset model and extracting the suspected abnormal case according to the analysis result includes:

Calculate the distance metric of the preset field in the case corresponding to each insured;

Calculate the outlier degree of the distance measure of the insured's preset field relative to the distance measure of other insured's preset fields;

Compare the outlier degree with the preset outlier degree;

When the outlier degree is greater than the preset outlier degree, the case corresponding to the insured is marked as a suspected abnormal case.
A medical insurance supervision device, characterized in that the medical insurance supervision device includes:

The data acquisition module is used to obtain the medical data of the insured;

A field extraction module, used to extract a preset field in the medical data according to a preset model;

The abnormality analysis module is configured to analyze the preset field based on the preset model and extract a suspected abnormal case according to the analysis result.
The medical insurance supervision device of claim 14, wherein the medical insurance supervision device further comprises:

Case output module to output suspected abnormal cases;

Information receiving module, used to receive confirmation information corresponding to suspected abnormal cases;

The expense revision module is used to revise the medical insurance expenses when the confirmation information confirms that the suspected abnormal case is abnormal.
The medical insurance supervision device of claim 15, wherein the medical insurance supervision device further comprises:

The model optimization module is used to optimize the preset model according to the suspected abnormal case when the confirmation information denies that the suspected abnormal case is abnormal, wherein the optimization of the preset model is implemented based on machine learning.
The medical insurance supervision device of claim 14, wherein the field extraction module includes:

The data cleaning unit is used to clean up the medical data;

The text classification unit is used to obtain the canonical text and non-normative text in the cleaned visit data, respectively;

The field extraction unit is used to extract the preset fields in the specification text according to the preset model;

The field extraction unit is also used to analyze the non-standard text based on the recurrent neural network, and extract the preset fields in the non-standard text according to the analysis results of the preset model and the non-standard text.
The medical insurance supervision device according to claim 17, wherein the field extraction unit includes:

Vector sequence subunit to represent non-standard text as a sequence of vectors;

Vector matrix subunit, used to encode the vector sequence into a sentence vector matrix based on the bidirectional recurrent neural network based on the semantic content and semantic distance of the vector sequence;

The vector compression subunit is used to compress the sentence vector matrix into a sentence vector according to a preset model, and extract a preset field in the sentence vector.
The medical insurance supervision device of claim 14, wherein the abnormality analysis module includes:

The distance measurement unit is used to calculate the distance measurement of the preset field in the case corresponding to each insured;

The outlier degree unit is used to calculate the outlier degree of the distance measure of the insured's preset field relative to the distance measure of the other insured's preset field;

The comparison unit is used to compare the outlier degree and the preset outlier degree;

The marking unit is used to mark the case corresponding to the insured as a suspected abnormal case when the outlier degree is greater than the preset outlier degree.
A computer-readable storage medium, characterized in that medical insurance supervision readable instructions are stored on the computer readable storage medium, and the medical insurance supervision readable instructions are executed by a processor to implement the following steps:

Obtain the medical data of the insured, wherein the medical data includes the information of the insured, the medical record and the diagnosis and treatment path

According to the preset model, extract the preset fields in the medical data;

The preset field is analyzed based on the preset model, and a suspected abnormal case is extracted according to the analysis result.