Outpatient Main Diagnosis Error Medical Record Screening Method and System

Abstract

The present invention provides a method and system for screening medical records with incorrect primary diagnoses, including: obtaining all entities in each medical record in a set of medical records to be screened; the entities include: symptoms, diseases, and signs; calculating a weight value of each entity in each medical record based on the primary diagnosis of each medical record; screening out a first set of medical records with incorrect primary diagnoses from the set of medical records to be screened based on the weight values; the first set of medical records is a set of medical records in which the weight value of each entity is less than a preset threshold. The present invention alleviates the technical problem of low screening accuracy of medical records with incorrect primary diagnoses existing in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of medical record screening, and in particular, to a method and system for screening outpatient primary diagnosis error medical records. Background Art

[0002] With the rapid popularization of electronic medical record systems in medical institutions, a large amount of important medical-related information is stored in medical information systems in electronic form. After continuous accumulation, various forms of electronic medical systems have generated a huge amount of medical big data. These data record important information in clinical medicine. For example, the patient's chief complaint, current medical history, examinations and diagnoses, etc. In recent years, with the development of artificial intelligence, we can mine effective information from electronic medical records to realize related applications of intelligent medicine. The most common applications include recommending diagnosis and treatment plans based on the content of medical records, or recommending the patient's primary diagnosis and registrable departments, etc.

[0003] However, the number of outpatient visits in the hospital is large every day, and the workload of doctors writing electronic medical records is also very large. In the early stage, due to the lack of quality control of medical record content and other reasons, many outpatient medical record diagnoses were written incorrectly or irregularly, and the outpatient diagnosis was different from the first page diagnosis of the inpatient medical record and was not cataloged by the medical record department, so these errors were not corrected in time. In the later stage, when using big data to manage outpatient historical medical records, only the diagnosis serial number can be marked according to the diagnosis order filled in by the doctor at that time, and the one written in the first place is the primary diagnosis of this visit. This will lead to many medical records with incorrect primary diagnoses that do not match the description information of the outpatient medical records. Currently, in many applications of recommending diagnosis and treatment plans based on patients, many applications recommend the primary diagnosis and the treatment plan of the primary diagnosis. Therefore, if these medical records with incorrect primary diagnoses are used as the learning set, it will affect the accuracy of the recommendation.

[0004] In the prior art, there is a method for screening incorrect labels based on a clustering method. However, because the number of historical electronic medical records is huge and the patient's condition information is relatively complex, the primary diagnoses of similar patients are very likely to be different. Therefore, the accuracy of clustering is also relatively low, which in turn leads to the technical problem of low screening accuracy for medical records with incorrect primary diagnoses. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method and system for screening outpatient primary diagnosis error medical records to alleviate the technical problem of low screening accuracy for medical records with incorrect primary diagnoses existing in the prior art.

[0006] In a first aspect, an embodiment of the present invention provides a method for screening medical records with incorrect primary diagnoses in outpatient clinics, including: obtaining all entities in each medical record in a set of medical records to be screened; the entities include: symptoms, diseases, and signs; calculating a weight value of each entity under each medical record based on the primary diagnosis of each medical record; screening out a first set of medical records with incorrect primary diagnoses from the set of medical records to be screened based on the weight value; the first set of medical records is a set of medical records in which the weight value of each entity is less than a preset threshold.

[0007] Further, obtaining all entities in each medical record in the set of medical records to be screened includes: obtaining the entities from the chief complaint information and the current medical history information of each medical record in the set of medical records to be screened based on a natural language processing method.

[0008] Further, calculating a weight value of each entity under each medical record based on the primary diagnosis of each medical record includes: calculating the weight value through the following formula: IMPT NB = log(p(x i = 1|y j = 1)) - log(p(x i = 1|y j = 0)); IMPT NB is the weight value, log(p(x i = 1|y j = 1)) is the logarithmic probability value of having entity x under the condition that the primary diagnosis y j appears, and log(p(x i = 1|y i = 0)) is the logarithmic probability value of having entity x under the condition that the primary diagnosis y j does not appear. j appears. i appears.

[0009] Further, after screening out the first set of medical records with incorrect primary diagnoses from the set of medical records to be screened, the method further includes: determining, based on a preset similarity algorithm, a set of medical records in a preset medical record library that have a similarity exceeding a preset similarity threshold with a target medical record, to obtain a second set of medical records; the target medical record is a medical record in the first set of medical records; determining whether the primary diagnosis of the target medical record is consistent with the primary diagnosis that appears most frequently in the second set of medical records; if so, determining that the target medical record is a normal medical record; if not, determining that the target medical record is a medical record with an incorrect primary diagnosis in outpatient clinics.

[0010] In a second aspect, an outpatient primary diagnosis error medical record screening system according to an embodiment of the present invention includes: an acquisition module, a calculation module, and a screening module; wherein, the acquisition module is configured to acquire all entities in each medical record in a set of medical records to be screened; the entities include: symptoms, diseases, and signs; the calculation module is configured to calculate a weight value of each entity under each medical record based on the primary diagnosis of each medical record; the screening module is configured to screen out a first set of medical records with incorrect primary diagnoses from the set of medical records to be screened based on the weight value; the first set of medical records is a set of medical records in which the weight value of each entity is less than a preset threshold.

[0011] Further, the acquisition module is further configured to: acquire the entities from the chief complaint information and the current medical history information of each medical record in the set of medical records to be screened based on a natural language processing method.

[0012] Further, the calculation module is further configured to: calculate the weight value through the following formula: IMPT NB = log(p(x i = 1|y j = 1)) - log(p(x i = 1|y j = 0)); IMPT NB is the weight value, log(p(x i = 1|y j = 1)) is the logarithmic probability value of having entity x j under the condition that the primary diagnosis y i appears, log(p(x i = 1|y j = 0)) is the logarithmic probability value of having entity x j under the condition that the primary diagnosis y i does not appear.

[0013] Further, the system further includes a determination module, configured to: determine, based on a preset similarity algorithm, a set of medical records in a preset medical record library whose similarity to a target medical record exceeds a preset similarity threshold, to obtain a second set of medical records; the target medical record is a medical record in the first set of medical records; determine whether the primary diagnosis of the target medical record is consistent with the primary diagnosis that appears most frequently in the second set of medical records; if so, determine that the target medical record is a normal medical record; if not, determine that the target medical record is an outpatient primary diagnosis error medical record.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the steps of the method described in the first aspect above are implemented.

[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable medium having non-volatile program code executable by a processor, and the program code causes the processor to execute the method described in the first aspect above.

[0016] The present invention provides a method and a system for screening medical records with incorrect primary diagnoses. First, all entities in each medical record in the set of medical records to be screened are obtained. Then, based on the primary diagnosis of each medical record, the weight value of each entity under each medical record is calculated. Finally, based on the weight value, a first set of medical records with incorrect primary diagnoses is screened out from the set of medical records to be screened; the first set of medical records is a set of medical records in which the weight value of each entity is less than a preset threshold. By using the weight values of each entity under the primary diagnosis for judgment, the present invention makes the screening of incorrect medical records more accurate, and alleviates the technical problem of low screening accuracy of medical records with incorrect primary diagnoses existing in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a flowchart of a method for screening medical records with incorrect primary diagnoses provided by an embodiment of the present invention;

[0019] Figure 2 It is a flowchart of another method for screening medical records with incorrect primary diagnoses provided by an embodiment of the present invention;

[0020] Figure 3 It is a schematic diagram of a system for screening medical records with incorrect primary diagnoses provided by an embodiment of the present invention;

[0021] Figure 4 It is a schematic diagram of another system for screening medical records with incorrect primary diagnoses provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The following will clearly and completely describe the technical solutions of the present invention with reference to the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the protection scope of the present invention.

[0023] Embodiment 1:

[0024] Figure 1It is a flowchart of a method for screening outpatient primary diagnosis error medical records provided according to an embodiment of the present invention. As Figure 1 shown, the method specifically includes the following steps:

[0025] Step S102, obtain all entities in each medical record in the set of medical records to be screened; the entities include: symptoms, diseases, and signs. For example, disease: coronary heart disease, symptoms: chest pain, chest tightness, sign: systolic blood pressure 100.

[0026] In the embodiments of the present invention, the medical records involved are all electronic medical records. Specifically, an electronic medical record (EMR) is also called a computerized medical record system or a computer-based patient record. It is a digital medical record of a patient saved, managed, transmitted, and reproduced by electronic devices (computers, health cards, etc.), replacing the handwritten paper medical record.

[0027] The content of the electronic medical record includes all the information of the paper medical record. Specifically, it includes chief complaint information, current medical history information, past medical history information, etc., and each electronic medical record corresponds to a label of the primary diagnosis. Among them, the outpatient diagnosis doctor and other professionals make a set of diseases that the patient suffers from based on the patient's symptoms, medical history, and medical examination results, etc. The primary diagnosis is the disease that causes the main reason for the patient's current hospitalization in this set of diseases.

[0028] Optionally, based on the natural language processing (NLP, Natural Language Processing) method, obtain entities from the chief complaint information and current medical history information of each medical record in the set of medical records to be screened.

[0029] Step S104, calculate the weight value of each entity under each medical record based on the primary diagnosis of each medical record.

[0030] Specifically, calculate the weight value through the following formula:

[0031] IMPT NB =log(p(x i =1|y j =1))-log(p(x i =1|y j =0))

[0032] IMPT NB is the weight value, log(p(x i =1|y j =1)) is the logarithmic probability value of having entity x under the condition that the primary diagnosis y j appears, log(p(x i =1|y i =0)) is the logarithmic probability value of having entity x under the condition that the primary diagnosis y j does not appear j under the condition of having entity xi Logarithmic probability value.

[0033] In the embodiments of the present invention, for the logarithmic probability value of having a certain entity (such as symptoms, diseases, and signs) under a certain main diagnosis condition, it can be calculated through a Bayesian network or directly obtained from the statistical data of a preset medical record library.

[0034] Step S106, based on the weight values, screen out the first medical record set with incorrect main diagnosis from the medical record set to be screened; the first medical record set is the set of medical records in which the weight value of each entity is less than a preset threshold.

[0035] Specifically, for each medical record in the medical record set to be screened, determine whether the weight values of all entities are less than the preset threshold. If the weight values of all entities are less than the preset threshold, it is considered that the medical record has an incorrect main diagnosis, and the medical record is placed in the first medical record set.

[0036] The embodiments of the present invention provide a method for screening outpatient medical records with incorrect main diagnosis. By using the weight values of each entity under the main diagnosis for judgment, the screening of incorrect medical records is made more accurate, alleviating the technical problem of low screening accuracy of medical records with incorrect main diagnosis in the prior art.

[0037] Optionally, after step S106, the method provided by the embodiments of the present invention further includes the following steps:

[0038] Step S108, based on a preset similarity algorithm, determine the set of medical records with a similarity exceeding the preset similarity threshold to the target medical record from the preset medical record library, and obtain the second medical record set; the target medical record is the medical record in the first medical record set.

[0039] Optionally, the method for setting the preset similarity threshold is to first set a base value, and then verify it on the test set after manual annotation, and continuously adjust according to the screening accuracy until a relatively high screening accuracy is achieved.

[0040] Step S110, determine whether the main diagnosis of the target medical record is consistent with the main diagnosis that appears most frequently in the second medical record set; if so, execute step S112; if not, execute step S114.

[0041] Step S112, determine that the target medical record is a normal medical record.

[0042] Step S114, determine that the target medical record is an outpatient medical record with incorrect main diagnosis.

[0043] In the embodiments of the present invention, due to many problems such as non-standard or colloquial descriptions in outpatient electronic medical records, when using NLP for entity recognition, there may be cases where entities are not correctly recognized. As a result, there may be correct medical records in the first medical record set screened in step S106. Therefore, in the embodiments of the present invention, the medical records in the first case set are further screened and judged through the above steps S108 - S114. Specifically, through a similarity algorithm, for each medical record in the first medical record set, a second medical record set with a similarity exceeding a preset similarity threshold (e.g., 90%) in the preset medical record library is calculated. Then, it is judged whether the main diagnosis of this medical record is consistent with the majority of the main diagnoses in the second case set. If they are consistent, it is considered that this case is a medical record with a correct main diagnosis; if they are not consistent, it is finally determined that this medical record is a medical record with a wrong main diagnosis.

[0044] As can be seen from the above description, the embodiments of the present invention provide a method for screening outpatient medical records with wrong main diagnoses. By combining the Bayesian network with similarity, the method provided by the embodiments of the present invention can achieve a higher screening accuracy rate, and can provide better data cleaning for clinical auxiliary main diagnosis and department disease recommendation. Further, the method provided by the embodiments of the present invention can screen outpatient medical records with wrong main diagnoses without restricting the types of main diagnoses, and has achieved a high accuracy rate of more than 99% in clinical verification.

[0045] Embodiment Two:

[0046] Figure 2 It is a flowchart of another method for screening outpatient medical records with wrong main diagnoses provided according to the embodiments of the present invention. As Figure 2 shown, this method specifically includes the following steps:

[0047] Step (1) First, according to the natural language (NLP) processing method, each entity (such as symptoms, diseases, and signs) in the chief complaint and the current history of the electronic medical record is identified.

[0048] Step (2) Calculate the weight value of each symptom under each main diagnosis through the following formula to obtain a weight value dictionary of all relevant entities under each diagnosis.

[0049] IMPT NB =log(p(x i =1|y j =1)) - log(p(x i =1|y j =0))

[0050] where log(p(x i =1|y j =1)) represents the probability of having a certain symptom x under the condition of a certain diagnosis y j being presenti The logarithmic probability value, log(p(x i = 1|y j = 0)) represents the logarithmic probability value of having a certain symptom x under the condition that a certain diagnosis y does not occur. j i is present.

[0051] Step (III): Assign the weights calculated in Step (II) to all entities of each medical record respectively.

[0052] Step (IV): Set a threshold, and compare in turn whether the weight value of each entity in a medical record is less than the threshold.

[0053] Step (V): Calculate all medical records according to Step (IV) to obtain a set of medical records where all primary diagnoses may be incorrect.

[0054] Step (VI): For the set of medical records where the primary diagnosis may be incorrect screened in Step (V), select one medical record from it, and then calculate, according to the similarity algorithm, the set of medical records in the historical database whose similarity to this medical record exceeds 90%. If the primary diagnosis of this medical record is also inconsistent with the majority of the primary diagnoses of its highly similar medical records, then this disease is considered a medical record with an incorrect primary diagnosis.

[0055] Step (VII): Loop Step (VI) to obtain the finally screened set of medical records with incorrect primary diagnoses.

[0056] Example III:

[0057] Figure 3 is a schematic diagram of a screening system for outpatient medical records with incorrect primary diagnoses provided according to an embodiment of the present invention. As Figure 3 shown, the system includes: an acquisition module 10, a calculation module 20, and a screening module 30.

[0058] Specifically, the acquisition module 10 is used to acquire all entities in each medical record in the set of medical records to be screened; the entities include: symptoms, diseases, and signs.

[0059] The calculation module 20 is used to calculate the weight value of each entity under each medical record based on the primary diagnosis of each medical record.

[0060] The screening module 30 is used to screen out a first set of medical records with incorrect primary diagnoses from the set of medical records to be screened based on the weight values; the first set of medical records is a set of medical records where the weight value of each entity is less than a preset threshold.

[0061] The embodiment of the present invention provides a screening system for outpatient medical records with incorrect primary diagnoses. By using the weight values of each entity under the primary diagnosis for judgment, the screening of incorrect medical records is made more accurate, alleviating the technical problem of low screening accuracy for medical records with incorrect primary diagnoses existing in the prior art.​

[0062] Optionally, the obtaining module 10 is further configured to: obtain entities from the chief complaint information and the current medical history information of each medical record in the medical record set to be screened based on a natural language processing method.

[0063] Optionally, the calculating module 20 is further configured to:

[0064] Calculate the weight value through the following formula:

[0065] IMPT NB = log(p(x i = 1|y j = 1)) - log(p(x i = 1|y j = 0))

[0066] IMPT NB is the weight value, log(p(x i = 1|y j = 1)) is the logarithmic probability value of having entity x under the condition of the occurrence of the main diagnosis y j ; log(p(x i = 1|y i = 0)) is the logarithmic probability value of having entity x under the condition of the non-occurrence of the main diagnosis y j ; j i i .

[0067] Figure 4 is a schematic diagram of another outpatient main diagnosis error case screening system provided by an embodiment of the present invention. As Figure 4 shown, the system further includes a determining module 40, configured to:

[0068] Based on a preset similarity algorithm, determine a set of medical records whose similarity to the target medical record in a preset medical record library exceeds a preset similarity threshold, and obtain a second medical record set; the target medical record is a medical record in the first medical record set;

[0069] Determine whether the main diagnosis of the target medical record is consistent with the main diagnosis that appears most frequently in the second medical record set;

[0070] If so, determine that the target medical record is a normal medical record;

[0071] If not, determine that the target medical record is an outpatient main diagnosis error medical record.

[0072] An embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the methods in the first and second embodiments above are implemented.

[0073] An embodiment of the present invention further provides a computer-readable medium having non-volatile program code executable by a processor, and the program code causes the processor to execute the methods in Embodiment 1 and Embodiment 2 above.

[0074] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for screening medical records with incorrect main outpatient diagnoses, characterized in that, including: obtaining all entities in each medical record in the medical record set to be screened; the entities include: symptoms, diseases, and signs; calculating a weight value for each entity under each medical record based on the primary diagnosis of each medical record; calculating the weight value through the following formula: IMPT NB = log(p(x i = 1|y j = 1)) - log(p(x i = 1|y j = 0)); IMPT NB is the said weight value, log(p(x i = 1|y j = 1)) is the log probability value of having entity x j under the condition of the occurrence of the main diagnosis y i and log(p(x i = 1|y j = 0)) is the log probability value of having entity x j under the condition of the non - occurrence of the main diagnosis y i ; screening out a first medical record set with incorrect primary diagnoses from the medical record set to be screened based on the weight value; the first medical record set is a set of medical records in which the weight value of each entity is less than a preset threshold.

2. The method according to claim 1, wherein obtaining all entities in each medical record in the medical record set to be screened, including: obtaining the entities from the chief complaint information and the current medical history information of each medical record in the medical record set to be screened based on a natural language processing method.

3. The method according to claim 1, characterized in that, after screening out a first medical record set with incorrect primary diagnoses from the medical record set to be screened, the method further includes: determining, based on a preset similarity algorithm, a set of medical records in a preset medical record library whose similarity to a target medical record exceeds a preset similarity threshold, to obtain a second medical record set; the target medical record is a medical record in the first medical record set; judging whether the primary diagnosis of the target medical record is consistent with the primary diagnosis that appears most frequently in the second medical record set; if so, determining that the target medical record is a normal medical record; if not, determining that the target medical record is a medical record with an incorrect outpatient primary diagnosis.

4. A screening system for outpatient main diagnosis error medical records, characterized in that, including: an obtaining module, a calculating module, and a screening module; wherein, the obtaining module is configured to obtain all entities in each medical record in the medical record set to be screened; the entities include: symptoms, diseases, and signs; the calculating module is configured to calculate a weight value for each entity under each medical record based on the primary diagnosis of each medical record; the calculating module is further configured to: calculate the weight value through the following formula: IMPT NB = log(p(x i = 1|y j = 1)) - log(p(x i = 1|y j = 0)) IMPT NB For the weight value, log(p(x i = 1|y j = 1)) is the log probability value of having entity x j under the condition of the occurrence of the primary diagnosis y i ; log(p(x i = 1|y j = 0)) is the log probability of having entity x j under the condition of the non-occurrence of the primary diagnosis y i ; the screening module is configured to screen out a first medical record set with incorrect primary diagnoses from the medical record set to be screened based on the weight value; the first medical record set is a set of medical records in which the weight value of each entity is less than a preset threshold.

5. The system according to claim 4, characterized in that, the obtaining module is further configured to: obtain the entities from the chief complaint information and the current medical history information of each medical record in the medical record set to be screened based on a natural language processing method.

6. The system according to claim 4, wherein the system further includes a determining module, configured to: determine, based on a preset similarity algorithm, a set of medical records in a preset medical record library whose similarity to a target medical record exceeds a preset similarity threshold, to obtain a second medical record set; the target medical record is a medical record in the first medical record set; judge whether the primary diagnosis of the target medical record is consistent with the primary diagnosis that appears most frequently in the second medical record set; if so, determine that the target medical record is a normal medical record; if not, determine that the target medical record is a medical record with an incorrect outpatient primary diagnosis.

7. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, the steps of the method described in any one of claims 1 to 3 above are implemented.

8. A computer-readable medium having non-volatile program code executable by a processor, characterized in that, The program code causes the processor to execute the method described in any one of claims 1 - 3.

Images