A method, device and system for uploading double front pages of medical records
By combining SBERT and TF-IDF algorithms to generate mapping tables and set verification rules, the problem of low encoding conversion efficiency is solved, and the automatic generation and verification of dual homepages are realized, improving the accuracy of encoding and verification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHONGGONGWANG MEDICAL TREATMENT IT CO LTD
- Filing Date
- 2023-01-10
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technology cannot simultaneously meet the different coding requirements of the National Health Commission and the National Healthcare Security Administration, which forces coding personnel to manually convert codes, resulting in low efficiency. Furthermore, the verification rules are too simplistic and cannot integrate the requirements of multiple departments.
A mapping table combining SBERT and TF-IDF algorithms is used to generate a dual-encoded interface. Through data cleaning, verification, and transcoding, medical insurance settlement lists and health records homepages are generated, and verification rules for the relationship between the primary diagnosis and the primary surgery are set.
This system enables the generation of two homepages with a single encoding, simplifying the encoding process, standardizing encoding criteria, and improving the accuracy of encoding and verification results.
Smart Images
Figure CN116230151B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of hospital information management technology, specifically to a method, apparatus, and system for uploading dual homepages of medical records. Background Technology
[0002] The inpatient medical record summary is a data summary of a patient's case, compiled by the health administration department based on relevant information during their hospitalization. It includes four aspects: basic patient information, hospitalization process information, treatment information, and cost information. The medical record summary is the essence of the entire medical record, and its data forms the foundation of hospital management.
[0003] Due to the 2020 reform of medical insurance DRG (Diagnosis Related Groups) payment, the concept of a medical insurance settlement list was introduced. The medical insurance fund settlement list (referred to as the "medical insurance settlement list") refers to the data list submitted by designated medical institutions to the medical insurance department when applying for expense settlement after providing inpatient, outpatient, and chronic / special disease medical services. Medical insurance DRG / DIP payments are based on the medical insurance settlement list. The treatment information in the list comes from the homepage, but it differs from the homepage. The main difference between the homepage and the settlement list lies in the fact that the homepage is used for performance evaluation, emphasizing the "severity of the disease," while the list is used for payment, emphasizing the "consumption of medical resources." With only one homepage, it is impossible to simultaneously meet the needs of both management functions.
[0004] Completing the medical record cover sheet and settlement statement is a crucial method for collecting medical information and serves as the basis for medical insurance payments. According to the management requirements of the Health Commission and Medical Insurance Bureau for medical record data, after completing the compilation of the medical record cover sheet data, hospitals must simultaneously upload the medical record cover sheet data to the Health Commission and the medical insurance settlement statement data to the Medical Insurance Bureau. If only one cover sheet is available, the coder must modify it according to the medical insurance settlement statement reporting requirements before uploading it to the medical insurance bureau, and then restore and archive it again.
[0005] Two issues arose during the process described above. First, the National Health Commission and the National Healthcare Security Administration use two different coding systems. Hospitals need to manually code the medical record cover pages using their coding personnel before uploading them to the different departments. Second, in addition to the coding versions, the National Health Commission and the National Healthcare Security Administration have different quality control requirements for the content of the medical record data. Hospitals need to make targeted adjustments and manage the medical record data before uploading.
[0006] Currently, most medical record verification rule products have relatively simple rule settings. Some only involve health and wellness rules, some only involve medical insurance enrollment rules, and some only involve list upload rules. They cannot comprehensively verify the requirements of multiple departments at once and meet the needs of different departments for uploading medical records. In addition, different coding versions of settlement lists and medical record front pages are currently mainly mapped manually. Therefore, the upload of medical insurance settlement lists and health and wellness medical record front pages requires hospitals to process them in different systems, which is inefficient. Furthermore, the verification of the relationship between the primary diagnosis and the primary surgery is a crucial part of the verification process. Summary of the Invention
[0007] To address the shortcomings of existing technologies, and to achieve the goal of generating two homepages (medical insurance and health commission) with a single coding, simplify the coding process, and standardize coding standards,
[0008] This invention proposes a method for uploading dual-page medical records, comprising the following steps:
[0009] Perform data cleaning and extraction on the input raw data;
[0010] Verify the cleaned and extracted data;
[0011] Based on the mapping table, the verified data is converted into medical insurance code and health code respectively.
[0012] Generate a dual-encoding interface based on the converted medical insurance code and health code;
[0013] Based on the generated dual-code interface, generate the medical insurance settlement list and the health and medical record homepage respectively;
[0014] Upload the generated medical insurance settlement list to the medical insurance system, and upload the generated health and wellness medical record homepage to the health and wellness system.
[0015] The creation of the mapping table specifically involves:
[0016] Regarding the SBERT algorithm, model training and optimization were performed using empirical data and the differences in specific encodings for different types of directories to obtain the SBERT model. The original encoded name and the standard encoded name were converted into numerical vectors using the SBERT model, and the first standard encoded name with the highest similarity was obtained using cosine similarity.
[0017] Regarding the TF-IDF algorithm, a word segmentation model is used. After personalized processing of stop words, medical terminology related dictionaries, and partial word weights, synonym and antonym logic is added to obtain word segmentation vectors. Cosine similarity is used to obtain the second standard encoded name with the highest similarity.
[0018] The first standard encoding name and the second standard encoding name are compared, and the optimal standard encoding name is obtained based on the reliability of the results of different algorithms in each similarity segment.
[0019] A mapping table is obtained based on the original encoding name and the optimal standard encoding name.
[0020] Before verifying the cleaned and extracted data, verification rules are set for the health and medical record homepage and the medical insurance settlement list. These verification rules include verification of the relationship between the primary diagnosis and the primary surgery.
[0021] The verification method for the relationship between the primary diagnosis and the primary surgery is as follows:
[0022] The data obtained through cleaning and extraction is denoted as A, wherein data A includes the relationship data between the primary diagnosis and the primary surgery A1, and the relationship data between the primary diagnosis and the primary surgery A1 further includes the primary surgery data A1.1 and the primary diagnosis data A1.2;
[0023] Read the main surgical data A1.1 and the main diagnostic data A1.2. Match the read main surgical data A1.1 with the surgical diagnosis matching table to obtain the matching diagnostic list set D. Compare the diagnostic list set D with the main diagnostic data A1.2 to obtain the intersection d.
[0024] If the intersection d = empty, invoke the DRG grouper or DIP grouper:
[0025] When the DRG grouper is invoked, if the invoked DRG grouper fails to return a normal result, the first prompt will be issued.
[0026] When a DIP grouper is invoked, if the invoked DIP grouper fails to return a normal result, a second prompt will be issued.
[0027] If the intersection d is not empty, start polling.
[0028] The polling includes:
[0029] Determine the diagnostic combinations to participate in the polling: diagnostic combination 1, diagnostic combination 2, and so on, up to diagnostic combination N;
[0030] The DRG / DIP grouper is invoked sequentially using the virtual medical records participating in the polling. The baseline score corresponding to the grouping result obtained by the DRG / DIP grouper is viewed. The main diagnosis of the virtual medical record with the highest baseline score constitutes the inverse main diagnosis set M. The virtual medical record is the diagnostic combination and the original surgical combination of the medical record.
[0031] If the main diagnostic data A1.2 belongs to the main diagnostic set M, then exit the verification;
[0032] If the primary diagnostic data A1.2 does not belong to the primary diagnostic set M, then a third prompt is issued.
[0033] The present invention also proposes a device for uploading dual front pages of medical records, comprising:
[0034] The data cleaning and extraction module is used to clean and extract the input raw data.
[0035] The verification module is used to verify the cleaned and extracted data;
[0036] The transcoding module is used to transcode the verified data based on the mapping table, converting it into medical insurance code and health code respectively;
[0037] The interface generation module is used to generate a dual-encoding interface based on the converted medical insurance code and health code;
[0038] The list and homepage generation module is used to generate medical insurance settlement lists and health and medical record homepages based on the generated dual-coded interface.
[0039] The upload module is used to upload the generated medical insurance settlement list to the medical insurance system and the generated health and health medical record homepage to the health and health system.
[0040] The creation of the mapping table specifically involves:
[0041] Regarding the SBERT algorithm, model training and optimization were performed using empirical data and the differences in specific encodings for different types of directories to obtain the SBERT model. The original encoded name and the standard encoded name were converted into numerical vectors using the SBERT model, and the first standard encoded name with the highest similarity was obtained using cosine similarity.
[0042] Regarding the TF-IDF algorithm, a word segmentation model is used. After personalized processing of stop words, medical terminology related dictionaries, and partial word weights, synonym and antonym logic is added to obtain word segmentation vectors. Cosine similarity is used to obtain the second standard encoded name with the highest similarity.
[0043] The first standard encoding name and the second standard encoding name are compared, and the optimal standard encoding name is obtained based on the reliability of the results of different algorithms in each similarity segment.
[0044] A mapping table is obtained based on the original encoding name and the optimal standard encoding name.
[0045] The system also includes a verification rules module, which is used to set verification rules for the health and medical record homepage and medical insurance settlement list. The verification rules include verification of the relationship between the primary diagnosis and the primary surgery.
[0046] The verification rule module is used for:
[0047] The data obtained through cleaning and extraction is denoted as A, wherein data A includes the relationship data between the primary diagnosis and the primary surgery A1, and the relationship data between the primary diagnosis and the primary surgery A1 further includes the primary surgery data A1.1 and the primary diagnosis data A1.2;
[0048] Read the main surgical data A1.1 and the main diagnostic data A1.2. Match the read main surgical data A1.1 with the surgical diagnosis matching table to obtain the matching diagnostic list set D. Compare the diagnostic list set D with the main diagnostic data A1.2 to obtain the intersection d.
[0049] If the intersection d = empty, invoke the DRG grouper or DIP grouper:
[0050] When the DRG grouper is invoked, if the invoked DRG grouper fails to return a normal result, the first prompt will be issued.
[0051] When a DIP grouper is invoked, if the invoked DIP grouper fails to return a normal result, a second prompt will be issued.
[0052] If the intersection d is not empty, start polling.
[0053] The polling includes:
[0054] Determine the diagnostic combinations to participate in the polling: diagnostic combination 1, diagnostic combination 2, and so on, up to diagnostic combination N;
[0055] The DRG / DIP grouper is invoked sequentially using the virtual medical records participating in the polling. The baseline score corresponding to the grouping result obtained by the DRG / DIP grouper is viewed. The main diagnosis of the virtual medical record with the highest baseline score constitutes the inverse main diagnosis set M. The virtual medical record is the diagnostic combination and the original surgical combination of the medical record.
[0056] If the main diagnostic data A1.2 belongs to the main diagnostic set M, then exit the verification;
[0057] If the primary diagnostic data A1.2 does not belong to the primary diagnostic set M, then a third prompt is issued.
[0058] The present invention also proposes a system for uploading dual homepages of medical records, including a processor and a memory storing a computer program, wherein the computer program is executed by the processor to perform the aforementioned method.
[0059] Compared with the prior art, the present invention has the following significant advantages:
[0060] First, coders can use the system to generate two homepages (medical insurance and health) with a single coding operation, thereby simplifying the coding process, standardizing coding criteria, and improving accuracy.
[0061] Second, an innovative encoding mapping algorithm is proposed, which can efficiently complete the encoding mapping task;
[0062] Third, it innovatively proposed a verification method for the relationship between the primary surgery and the primary diagnosis, which further improved the accuracy of the medical record verification results. Attached Figure Description
[0063] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0064] Figure 1 This is a flowchart illustrating the steps of the method for uploading dual homepages of medical records according to the present invention;
[0065] Figure 2 This is a schematic diagram of the device for uploading dual medical record homepages according to the present invention;
[0066] Figure 3 This is a schematic diagram of the system for uploading medical records on a dual homepage according to the present invention. Detailed Implementation
[0067] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.
[0068] The medical record cover sheet can reflect the medical outcomes, workflow, and efficiency to a certain extent. It is a summary of patient information during hospitalization, compiled by medical staff using text, symbols, codes, and numbers in a specific table. The International Classification of Diseases (ICD) is an internationally unified disease classification method developed by the WHO. Based on the etiology, pathology, clinical manifestations, and anatomical and locational characteristics of diseases, it categorizes diseases into ordered groups, represented by a coding system. It serves as a carrier of medical information and is the foundation for medical data mining, disease diagnosis grouping, performance evaluation, and DRG-based medical insurance billing. In practice within domestic medical institutions, different regions have expanded the coding system differently based on the characteristics of clinical diseases. Furthermore, descriptive variations exist even within the same disease. For example, in GB-2016 ICD-10, "A00.100 Cholera, caused by Vibrio cholerae group O1, Eltor biotype" differs from "A00.101 Eltor biotype cholera" in BJ-V6.01 in terms of coding and terminology description. This inconsistency between multiple versions seriously affects data interoperability and the mining and application of medical data within the industry. A single medical record coder needs to code it twice to meet the different coding requirements of health and medical insurance departments, resulting in low efficiency.
[0069] Verification is generally divided into three types: pre-verification, in-process verification, and post-verification. From the perspective of medical record verification, pre-verification provides reminders about which diagnoses and surgeries should be filled in before the doctor fills in the diagnosis. In-process verification provides certain prompts based on the information entered by the doctor or medical record department during the filling process. Post-verification checks the accuracy of the medical record filling and provides feedback after the medical record department has completed the filling and before uploading it to medical insurance.
[0070] This solution primarily serves as a post-event verification system. Before hospitals upload medical insurance information, the system checks for problems and prompts for corrections, preventing grouping errors caused by inaccurate medical record entries. After post-event verification is complete, it can be integrated with the Hospital Information System (HIS) to embed the verification function into the work processes of doctors and the medical records department.
[0071] like Figure 1 As shown, a method for uploading dual homepages of medical records according to the present invention includes the following steps:
[0072] Step 1: Clean and extract data from the input raw data;
[0073] Step 2: Verify the cleaned and extracted data;
[0074] Step 3: Based on the mapping table, the verified data is converted into medical insurance codes and health codes respectively;
[0075] Step 4: Generate a dual-encoding interface based on the converted medical insurance code and health code;
[0076] Step 5: Based on the generated dual-code interface, generate the medical insurance settlement list and the health and medical record homepage respectively;
[0077] Step 6: Upload the generated medical insurance settlement list to the medical insurance system, and upload the generated health and wellness medical record homepage to the health and wellness system.
[0078] In this invention, the raw data is the data uploaded by the hospital system.
[0079] In this invention, before verifying the cleaned and extracted data, verification rules are set for the health and medical record homepage and the medical insurance settlement list. These verification rules include 91 categories of integrity rules, 73 categories of normative rules, 37 categories of logical rules, 7 categories of key rules for health and medical record uploads, and 5 categories of medical insurance bureau inclusion rules. The medical insurance bureau inclusion rules include verification of the relationship between the primary diagnosis and the primary surgery.
[0080] In one embodiment of the present invention, the verification rules include verification of the relationship between the primary diagnosis and the primary surgery.
[0081] In one embodiment of the present invention, the verification method for the relationship between the primary diagnosis and the primary surgery is as follows:
[0082] The data obtained through cleaning and extraction is denoted as A, wherein data A includes the relationship data between the primary diagnosis and the primary surgery A1, and the relationship data between the primary diagnosis and the primary surgery A1 further includes the primary surgery data A1.1 and the primary diagnosis data A1.2.
[0083] Read the main surgical data A1.1 and the main diagnostic data A1.2. Match the main surgical data A1.1 with the surgical diagnosis matching table to obtain a set of matching diagnostic lists D. Compare the set of diagnostic lists D with the main diagnostic data A1.2 to obtain the intersection d. Since A1.1 and A1.2 are associated through the patient ID, some fields of A1.1 and A1.2 are the same, so the intersection d can be obtained. Specifically, verify that the main surgical case = 01 and other surgical cases = 0pr (excluding 01). Find the PACK_CODE corresponding to 01 in the tb_pack_operation table, take the diagnostic set D corresponding to the PACK_CODE in the tb_pack_disease table, and take the intersection of all the original diagnoses of the medical records in set D to obtain the intersection d.
[0084] (3.3) If the intersection d = empty (the diagnostic set D has no intersection with all the original diagnoses in the medical record), then retrieve the DRG and / or DIP (Big Data Diagnosis-Intervention Packet, payment by disease type) grouper:
[0085] When the DRG grouper is invoked, if the invoked DRG grouper fails to return a normal result, the first prompt will be issued, such as "The main diagnosis and the main surgery do not match. It is recommended to select the diagnosis related to the prompt word" (the prompt word is the prompt word of PACK_CODE corresponding to 01 in the table tb_pack_disease).
[0086] When the DIP grouper is invoked, if the invoked DIP grouper fails to return a normal result, a second prompt is issued; for example, the prompt is "According to the existing grouper, the diagnosis and surgery cannot be successfully enrolled, the primary diagnosis and primary surgery do not match, it is recommended to select the diagnosis related to the prompt word" (the prompt word is the prompt word of PACK_CODE corresponding to 01 in the table tb_pack_disease).
[0087] If the intersection d ≠ empty, start polling. Determine the diagnostic combinations to participate in the polling: Diagnostic combination 1 {Main diagnosis = d1, Other diagnoses = All original diagnoses in the medical record - d1}, Diagnostic combination 2 {Main diagnosis = d2, Other diagnoses = All original diagnoses in the medical record - d2}, and so on.
[0088] The DRG / DIP grouper is invoked sequentially using the virtual medical records participating in the polling. The baseline score corresponding to the grouping results obtained by the DRG / DIP grouper is then viewed. The primary diagnosis of the virtual medical record with the highest baseline score constitutes the inferred primary diagnosis set M. It should be noted that there may be multiple virtual medical records with the highest baseline score; a "virtual medical record" refers to both the diagnostic combination and the original surgical combination of the medical record.
[0089] If the main diagnostic data A1.2 belongs to the main diagnostic set M, then exit the verification;
[0090] If the primary diagnosis data A1.2 does not belong to the primary diagnosis set M, a third prompt will be issued; for example, the prompt will say "The primary diagnosis and the primary surgery do not match, it is recommended to select M as the primary diagnosis".
[0091] In addition, the completeness rules include mandatory fields and conditionally mandatory fields. Mandatory fields include: name, gender, age, etc.; conditionally mandatory fields include: primary diagnosis and external cause code for injury or poisoning, diagnosis related to traumatic brain injury and duration of coma, etc.
[0092] The normative rules include basic information and coding information. Basic information includes: the range of newborn birth weight, age, etc.; coding information includes: the diagnosis code adopts the disease classification code clinical version 2.0 code (IDC10), and the diagnosis name adopts the name corresponding to the medical insurance disease classification code, etc.
[0093] Logical rules include verification of primary diagnosis and discharge method, general diagnosis coding, and verification of the relationship between surgery and cost.
[0094] The rules for uploading medical records by the Health Commission include key points related to complications, key points related to the primary diagnosis and admission condition, and key points related to surgical incision healing.
[0095] The inclusion rules of the Medical Insurance Bureau include verification of the primary diagnosis, verification of the primary surgery, and verification of the relationship between the primary diagnosis and the primary surgery.
[0096] In one embodiment of the present invention, the steps for creating the mapping table are as follows: A method combining the SBERT (Siamese network) natural language processing algorithm with the TF-IDF (term frequency–inverse document frequency) algorithm is employed. This method adds synonyms based on medical expertise (e.g., cervix: uterine cervix, ca: malignant tumor), antonyms (e.g., catheter placement: catheter removal, incision: closure), stop words (e.g., of, Johnson & Johnson 1962, Youxinwei), and dictionary content (e.g., post-operative, LEEP knife, injection procedure), combining multiple logical elements for mapping. A medical insurance dictionary table and a health and wellness dictionary table are pre-set. The dictionary tables are standardized databases with multiple specified fields. By calling the medical insurance dictionary table and the health and wellness dictionary table, and using an intelligent mapping algorithm, medical insurance codes and health and wellness codes are generated.
[0097] Furthermore, regarding the SBERT algorithm, empirical data and the differences in specific encodings for different types of directories are used for model training and optimization to obtain the SBERT model. The original encoded name and the standard encoded name are converted into numeric vectors using the SBERT model, and the first standard encoded name with the highest similarity is obtained using cosine similarity. For example, if the original name "anemia" is unclear, the first digit of the original encoding can be increased, such as "O anemia", and the model will process it accordingly.
[0098] Regarding the TF-IDF algorithm, a word segmentation model is used. After personalized processing of stop words, medical terminology-related dictionaries, and partial word weights, synonym and antonym logic is added to obtain word segmentation vectors. Cosine similarity is used to obtain the second standard encoded name with the highest similarity. For example, when the words before and after conversion contain a pair of antonyms, the cosine similarity is forcibly set to 0.
[0099] The first standard encoding name and the second standard encoding name are compared. Based on the reliability of the results of different algorithms in each similarity segment, the optimal standard encoding name is obtained. A mapping table is obtained based on the encoding name before conversion and the optimal standard encoding name.
[0100] This invention also proposes an intelligent medical record dual-page upload system, such as... Figure 2 As shown, it includes a data cleaning and extraction module, a verification module, a transcoding module, an interface generation module, a list and homepage generation module, and an upload module.
[0101] Specifically, the data cleaning and extraction module is used to clean and extract the input raw data; the verification module is used to verify the cleaned and extracted data; the transcoding module is used to transcode the verified data based on the mapping table, converting it into medical insurance codes and health codes respectively; the interface generation module is used to generate a dual-coded interface based on the converted medical insurance codes and health codes; the list and homepage generation module is used to generate medical insurance settlement lists and health medical record homepages based on the generated dual-coded interfaces respectively; and the upload module is used to upload the generated medical insurance settlement lists to the medical insurance system and the generated health medical record homepages to the health system.
[0102] In one embodiment of the present invention, the system further includes a verification rule module for setting verification rules for the health and medical record homepage and the medical insurance settlement list, wherein the verification rules include verification of the relationship between the primary diagnosis and the primary surgery.
[0103] In one embodiment of the present invention, the verification rule module is further configured to: take the cleaned and extracted data as A, wherein the data A includes the main diagnosis and main surgery relationship data A1, and the main diagnosis and main surgery relationship data A1 further includes the main surgery data A1.1 and the main diagnosis data A1.2; read the main surgery data A1.1, read the main diagnosis data A1.2, match the read main surgery data A1.1 with the surgical diagnosis matching table to obtain a matching diagnosis list set D; compare the diagnosis list set D with the main diagnosis data A1.2 to obtain the intersection d.
[0104] If the intersection d = empty, invoke the DRG grouper or DIP grouper:
[0105] When the DRG grouper is invoked, if the invoked DRG grouper fails to return a normal result, a first prompt is issued; when the DIP grouper is invoked, if the invoked DIP grouper fails to return a normal result, a second prompt is issued; if the intersection d ≠ empty, polling begins.
[0106] In one embodiment of the present invention, the polling includes: determining the diagnostic combinations participating in the polling: diagnostic combination 1, diagnostic combination 2, and so on, up to diagnostic combination N.
[0107] The DRG / DIP grouper is invoked sequentially using the virtual medical records participating in the polling. The baseline score corresponding to the grouping result obtained by the DRG / DIP grouper is viewed. The main diagnosis of the virtual medical record with the highest baseline score constitutes the inverse main diagnosis set M. The virtual medical record is the diagnostic combination and the original surgical combination of the medical record.
[0108] If the main diagnostic data A1.2 belongs to the main diagnostic set M, the verification is terminated; if the main diagnostic data A1.2 does not belong to the main diagnostic set M, a third prompt is issued.
[0109] In one embodiment of the present invention, the mapping table is specifically constructed as follows: Regarding the SBERT algorithm, model training and optimization are performed using empirical data and differences in the specific encoding of different types of directories to obtain the SBERT model; the original encoded name and the standard encoded name are converted into numerical vectors using the SBERT model, and the first standard encoded name with the highest similarity is obtained using cosine similarity; Regarding the TF-IDF algorithm, a word segmentation model is used, and after personalized processing of stop words, medical terminology-related dictionaries, and partial word weights, synonym and antonym logic is added to obtain word segmentation vectors, and the second standard encoded name with the highest similarity is obtained using cosine similarity; the first standard encoded name and the second standard encoded name are compared, and the optimal standard encoded name is obtained based on the reliability of the results of different algorithms in each similarity segment; a mapping table is obtained based on the original encoded name and the optimal standard encoded name.
[0110] like Figure 3 As shown, the present invention also proposes a system for uploading dual medical record homepages, including a processor and a memory storing a computer program. When the computer program is run by the processor, it executes the aforementioned method for uploading dual medical record homepages.
[0111] The aforementioned intelligent medical record dual-page upload method, device, and system enable coders to generate two page covers with a single coding operation, simplifying the coding process, standardizing coding criteria, and improving the accuracy of medical record completion. Furthermore, the aforementioned coding mapping algorithm efficiently completes the coding mapping task, and the designed method for verifying the relationship between the primary surgery and the primary diagnosis further enhances the accuracy of medical record verification results.
[0112] Compared with existing technologies, the significant advantages of this solution are:
[0113] First, coders can use the system to generate two homepages (medical insurance and health) with a single coding operation, thereby simplifying the coding process, standardizing coding criteria, and improving accuracy.
[0114] Second, an innovative encoding mapping algorithm is proposed, which can efficiently complete the encoding mapping task;
[0115] Third, it innovatively proposed a verification method for the relationship between the primary surgery and the primary diagnosis, which further improved the accuracy of the medical record verification results.
[0116] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for uploading dual homepages of medical records, characterized in that, Includes the following steps: Perform data cleaning and extraction on the input raw data; Verify the cleaned and extracted data; Based on the mapping table, the verified data is converted into medical insurance code and health code respectively. Generate a dual-encoding interface based on the converted medical insurance code and health code; Based on the generated dual-code interface, generate the medical insurance settlement list and the health and medical record homepage respectively; Upload the generated medical insurance settlement list to the medical insurance system, and upload the generated health and wellness medical record homepage to the health and wellness system; The creation of the mapping table specifically involves: Regarding the SBERT algorithm, model training and optimization were performed using empirical data and the differences in specific encodings for different types of directories to obtain the SBERT model. The original encoded name and the standard encoded name were converted into numerical vectors using the SBERT model, and the first standard encoded name with the highest similarity was obtained using cosine similarity. Regarding the TF-IDF algorithm, a word segmentation model is used. After personalized processing of stop words, medical terminology related dictionaries, and partial word weights, synonym and antonym logic is added to obtain word segmentation vectors. Cosine similarity is used to obtain the second standard encoded name with the highest similarity. The first standard encoding name and the second standard encoding name are compared, and the optimal standard encoding name is obtained based on the reliability of the results of different algorithms in each similarity segment. Based on the original encoding name and the optimal standard encoding name, a mapping table is obtained; Among them, before verifying the cleaned and extracted data, verification rules are set for the health and medical record homepage and medical insurance settlement list. The verification rules include verification of the relationship between the main diagnosis and the main surgery. The verification method for the relationship between the primary diagnosis and the primary surgery is as follows: The data obtained through cleaning and extraction is denoted as A, wherein data A includes the relationship data between the primary diagnosis and the primary surgery A1, and the relationship data between the primary diagnosis and the primary surgery A1 further includes the primary surgery data A1.1 and the primary diagnosis data A1.2; Read the main surgical data A1.1 and the main diagnostic data A1.
2. Match the read main surgical data A1.1 with the surgical diagnosis matching table to obtain the matching diagnostic list set D. Compare the diagnostic list set D with the main diagnostic data A1.2 to obtain the intersection d. If the intersection d = empty, invoke the DRG grouper or DIP grouper: When the DRG grouper is invoked, if the invoked DRG grouper fails to return a normal result, the first prompt will be issued. When a DIP grouper is invoked, if the invoked DIP grouper fails to return a normal result, a second prompt will be issued. If the intersection d is not empty, start polling.
2. The method for uploading dual homepages of medical records according to claim 1, characterized in that, The polling includes: Determine the diagnostic combinations to participate in the polling: diagnostic combination 1, diagnostic combination 2, and so on, up to diagnostic combination N; The DRG / DIP grouper is invoked sequentially using the virtual medical records participating in the polling. The baseline score corresponding to the grouping result obtained by the DRG / DIP grouper is viewed. The main diagnosis of the virtual medical record with the highest baseline score constitutes the inverse main diagnosis set M. The virtual medical record is the diagnostic combination and the original surgical combination of the medical record. If the main diagnostic data A1.2 belongs to the main diagnostic set M, then exit the verification; If the primary diagnostic data A1.2 does not belong to the primary diagnostic set M, then a third prompt is issued.
3. A device for uploading dual homepages of medical records, characterized in that, include: The data cleaning and extraction module is used to clean and extract the input raw data. The verification module is used to verify the cleaned and extracted data; The transcoding module is used to transcode the verified data based on the mapping table, converting it into medical insurance code and health code respectively; The interface generation module is used to generate a dual-encoding interface based on the converted medical insurance code and health code; The list and homepage generation module is used to generate medical insurance settlement lists and health and medical record homepages based on the generated dual-coded interface. The upload module is used to upload the generated medical insurance settlement list to the medical insurance system and the generated health and wellness medical record homepage to the health and wellness system. The creation of the mapping table specifically involves: Regarding the SBERT algorithm, model training and optimization were performed using empirical data and the differences in specific encodings for different types of directories to obtain the SBERT model. The original encoded name and the standard encoded name were converted into numerical vectors using the SBERT model, and the first standard encoded name with the highest similarity was obtained using cosine similarity. Regarding the TF-IDF algorithm, a word segmentation model is used. After personalized processing of stop words, medical terminology related dictionaries, and partial word weights, synonym and antonym logic is added to obtain word segmentation vectors. Cosine similarity is used to obtain the second standard encoded name with the highest similarity. The first standard encoding name and the second standard encoding name are compared, and the optimal standard encoding name is obtained based on the reliability of the results of different algorithms in each similarity segment. Based on the original encoding name and the optimal standard encoding name, a mapping table is obtained; The system also includes a verification rules module, which is used to set verification rules for the health and medical record homepage and medical insurance settlement list. The verification rules include verification of the relationship between the primary diagnosis and the primary surgery. The verification rule module is used for: The data obtained through cleaning and extraction is denoted as A, wherein data A includes the relationship data between the primary diagnosis and the primary surgery A1, and the relationship data between the primary diagnosis and the primary surgery A1 further includes the primary surgery data A1.1 and the primary diagnosis data A1.2; Read the main surgical data A1.1 and the main diagnostic data A1.
2. Match the read main surgical data A1.1 with the surgical diagnosis matching table to obtain the matching diagnostic list set D. Compare the diagnostic list set D with the main diagnostic data A1.2 to obtain the intersection d. If the intersection d = empty, invoke the DRG grouper or DIP grouper: When the DRG grouper is invoked, if the invoked DRG grouper fails to return a normal result, the first prompt will be issued. When a DIP grouper is invoked, if the invoked DIP grouper fails to return a normal result, a second prompt will be issued. If the intersection d is not empty, start polling.
4. The device for uploading dual homepages of medical records according to claim 3, characterized in that, The polling includes: Determine the diagnostic combinations to participate in the polling: diagnostic combination 1, diagnostic combination 2, and so on, up to diagnostic combination N; The DRG / DIP grouper is invoked sequentially using the virtual medical records participating in the polling. The baseline score corresponding to the grouping result obtained by the DRG / DIP grouper is viewed. The main diagnosis of the virtual medical record with the highest baseline score constitutes the inverse main diagnosis set M. The virtual medical record is the diagnostic combination and the original surgical combination of the medical record. If the main diagnostic data A1.2 belongs to the main diagnostic set M, then exit the verification; If the primary diagnostic data A1.2 does not belong to the primary diagnostic set M, then a third prompt is issued.
5. A system for uploading dual-page medical records, characterized in that: It includes a processor and a memory storing a computer program, which, when executed by the processor, performs the method as described in any one of claims 1-2.