Novel hospitalization electronic medical record generation method

Abstract

The invention provides a novel hospitalization electronic medical record generation method, which comprises the following steps of: taking a patient from hospitalization to discharge as a static timeline, regarding different medical advices, signs, examinations and the like as receipts, distributing and updating the collected receipts according to a time sequence time window, and automatically generating an initial hospitalization electronic medical record through different time sequences and different medical record types. And taking the initial hospitalization electronic medical record as input data, giving a summarized patient disease progress and giving a current corresponding treatment scheme by combining a patient chief complaint updated by a doctor, a small knot generated by physical examination content and a large model in the medical droop field, and generating a hospitalization electronic medical record of a corresponding type by combining subjective medical examination manually input by the doctor.

Description

Technical Field

[0001] The present invention relates to the technical field of clinical medicine, and in particular to a novel method for generating electronic medical records of hospitalization. Background Art

[0002] At present, doctors need to conduct daily ward rounds for hospitalized patients to obtain the latest changes in their conditions and analyze tests and examinations to form hospitalization records. Complete hospitalization records have corresponding rules. Doctors need to first comprehensively analyze the condition of patients based on daily complaints, physical examination reports, nurses' registration of vital signs (blood pressure, heart rate, respiration, pulse, etc.), reports issued by the laboratory, reports issued by the examination department, and hospitalization records, and then complete further tests and examinations and analyze them again. Each step needs to be recorded in detail, and the test reports and the purpose of medication must be analyzed and explained. Finally, the medication method and dosage must be started, stopped, or adjusted, thereby relying on manual typing or template copying and pasting to write a single traditional hospitalization electronic medical record.

[0003] The scoring rules for writing inpatient records vary in specific details among different provinces, municipalities and autonomous regions. Doctors can only complete the writing of medical records within the prescribed time limit according to local rules, through a series of operations such as templates, which is inefficient and has serious content homogeneity.

[0004] Since the development and treatment of a patient's disease is a one-way process in time sequence, each record written during the hospitalization process must refer to and be analyzed from all previous records. However, in busy or urgent situations, doctors may not have time to review past materials, or the details of the condition may be lost in the medical record due to copying and pasting templates, causing doctors to only use drugs suitable for the current condition based on the current condition. This may conflict with the correct drugs that should be used according to the time sequence of the disease progression, affecting the optimal time for treatment. Summary of the invention

[0005] The purpose of the present invention is to generate an inpatient electronic medical record in a dynamic chronological order, which generates a summary of the contents processed by the doctor with objective and explainable records, and summarizes the current patient's condition progress on the inpatient electronic medical record based on the large model summary of the medical vertical field and gives a corresponding treatment plan.

[0006] In order to achieve the above-mentioned purpose of the invention, the technical solution of the present invention provides a novel method for generating electronic medical records of hospitalization, comprising the following steps:

[0007] Obtain the admission time of the patient according to the preset time interval and / or the user's active request, and poll the corresponding patient information according to the admission time, the patient information includes name, age, gender and nursing level, etc.;

[0008] Collect medical orders, physical signs, laboratory tests, examination documents, etc. according to the patient information, and classify them according to the document type conditions in the preset conditions to obtain event-based documents and non-event-based documents;

[0009] Split them into multiple modules according to the medical record type requirements, and preset corresponding limited time ranges according to different medical record types;

[0010] For event-based documents, dynamically adjust the time variables in the preset conditions and the operation variables in the preset operations according to the generation time sequence corresponding to the event-based documents, execute the action instructions of the corresponding operation set according to the conditions in the preset conditions, and combine the operation variables to obtain the final operation, and generate the first initial inpatient electronic medical record of the corresponding type according to the final operation;

[0011] For non-event-based documents, dynamically adjust the time variables in the preset conditions according to the generation time sequence corresponding to the non-event-based documents, and dynamically sort the non-event-based documents according to the modules according to the time variables and fill them into the modules to generate the second initial inpatient electronic medical record;

[0012] Use the content of the first initial inpatient electronic medical record and the second initial inpatient electronic medical record as the input of the medical vertical large model, output medication analysis, diagnosis basis and differential diagnosis, and fill them into the modules to obtain the final inpatient electronic medical record.

[0013] Preferably, when the single consumption duration of the timer for obtaining patient information is greater than the set time interval, an alarm prompt is given to guide the reset of the time interval.

[0014] Preferably, the event-based documents include documents corresponding to key medical behavior events such as surgery, consultation, transfer of departments, blood transfusion, etc.

[0015] Preferably, the medication analysis is obtained by analyzing the ward round chief complaint, medication orders, general information, current medical history, vital signs, physical examination, and laboratory reports in the first initial inpatient electronic medical record and the second initial inpatient electronic medical record according to the medical knowledge graph and the drug library.

[0016] Preferably, extract the characteristic information of the first initial inpatient electronic medical record and the second initial inpatient electronic medical record, and combine the preliminary diagnosis to retrieve the medical knowledge base using the medical knowledge graph for disease retrospective diagnosis to obtain the diagnosis basis.

[0017] Preferably, retrieve the medical knowledge base based on the characteristic information combined with the knowledge graph to obtain the first-strategy candidate disease group, perform screening for diseases of the same type according to the characteristic information to obtain the second-strategy candidate disease group, perform screening by identifying the text semantic similarity according to the characteristic information to obtain the third-strategy candidate disease group, and pass the first-strategy candidate disease group, the second-strategy candidate disease group, and the third-strategy candidate disease group through the reasoning process of simulating a medical expert, considering the individual characteristics and disease characteristics of the patient, and finally output the differential diagnosis diseases.

[0018] The technical solution of the present invention proposes a new method for generating an inpatient electronic medical record. Taking the time from the patient's hospitalization to discharge as a static timeline, regarding different medical orders, physical signs, tests, examinations, etc. as documents, allocating and updating the collected documents according to the chronological time window, automatically generating an initial inpatient electronic medical record through different time sequences and different medical record types, and using the initial inpatient electronic medical record as input data, combining the summary generated from the patient's chief complaint and physical examination content updated by the doctor, the large model in the medical vertical field summarizes the patient's condition progress and gives the current corresponding treatment plan, and combines the subjective medical examinations manually input by the doctor to generate the corresponding type of inpatient electronic medical record. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a logical diagram for the medication analysis module provided by the embodiment of the present invention to analyze the role and find the chief complaint, physical examination, test, and examination required for this diagnosis or to form this diagnosis;

[0020] Figure 2 It is a logical diagram for tracing the drug effect of the recommended medication obtained by combining the chief complaint, physical examination, test, and examination for diagnosis provided by the embodiment of the present invention;

[0021] Figure 3 It is a flowchart of the medication analysis module provided by the embodiment of the present invention;

[0022] Figure 4 It is a flowchart of the diagnostic basis module provided by the embodiment of the present invention;

[0023] Figure 5 It is a flowchart of the differential diagnosis module provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

[0025] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and gives detailed implementation manners and specific operation processes, but the protection scope of the present invention is not limited to the following embodiments.

[0026] An embodiment of the present invention provides a novel method for generating an inpatient electronic medical record, which includes two major steps: medical record generation and content analysis:

[0027] Specifically, the medical record generation step is as follows:

[0028] After the patient is admitted to the hospital, the timer obtains the admission time when the patient goes through the admission procedures according to a preset time interval and / or a user's active request, and polls the corresponding patient information according to the admission time. The patient information includes name, age, gender, nursing level, etc. The preset time interval is determined according to the patient data volume, that is, the preset time interval should be greater than or equal to the duration consumed by the timer to execute the acquisition of patient information. The preset time interval can be set to 24h, 1h, 30min, etc. When the duration consumed by the timer to execute the acquisition of patient information once is greater than the set time interval, an alarm prompt is given to guide the resetting of the time interval.

[0029] The task executor is bound behind the timer and is automatically started after the time polling operation is completed. The task executor includes a type editor, a rule engine, a document collector, a document sorter, and a sequencer.

[0030] The document collector will poll the patient data through the timer function, collect documents such as doctor's orders, signs, tests, and examinations for the patient, and classify them into event-based documents and non-event-based documents according to the document type conditions in the preset conditions. Event-based means that once such a document is collected, a separate medical record needs to be generated for recording, and non-event-based means otherwise.

[0031] Event-based includes key medical behavior events such as surgery, consultation, transfer of departments, blood transfusion, etc., and non-event-based includes objective medical data such as body temperature, heart rate, and drugs. The collected documents will have a unique ID and are default labeled as "unused".

[0032] For event-based documents, according to the conditions in the preset conditions, the action instructions of the corresponding operation set are executed, and the type editor executes the action instructions to achieve independent recording. For example, the condition in the preset conditions is that in the quality control scoring item, a blood transfusion record needs to be completed within 24 hours after each blood transfusion. Then, the "blood transfusion" document is set as an event-based document, the preset condition set is [document = "blood transfusion"] and [blood transfusion end time < blood transfusion record time < (blood transfusion end time + 24h)] and [type number] ≥ 0, and the corresponding operation set is [generate blood transfusion record]. That is, a blood transfusion record will be automatically generated within 24 hours after the blood transfusion, and at the same time, the status of this document will change from "unused" to "used".

[0033] The rule engine classifies the scoring items in the medical record quality control scoring rules into medical record types, number of types, time range, and generation time, and presets the condition set and operation set. The generation time of each medical record type cannot exceed its preset time range. The number of condition variables / operation variables in the set is ≥1, and the condition variables / operation variables in the set can be connected using and / or and other relationships. For example, the quality control scoring item requires that the first medical record be completed within 8 hours after admission, and this record type must be once during a single hospitalization, that is, the generation condition must meet [admission time < recording time < (admission time + 8h)] and [number of types = 1], where admission time to admission time + 8h is the preset time range, and the generation time is any time variable within the time range, corresponding to the operation set [generate first medical record]. After the patient is admitted to the hospital, the generation condition in the condition set is met, and the operation set will be executed according to the structured layout preset by the type editor, that is, a record type titled "First Medical Record" is generated. At this time, since the original preset module content contains empty values, the content displayed in the first medical record may be empty, and needs to wait for the document collector to collect relevant documents before filling in.

[0034] For non-event documents, when the timer is triggered, the document sorter will assign the documents to different modules according to the structured layout preset in the type editor. For example, in a non-event scenario, the temperature, pulse, blood pressure and other data collected within 8 hours of admission will be assigned to the "physical examination" module in the first medical record, and the doctor's prescriptions such as medicines collected during the period will be assigned to the treatment plan module. At the same time, the label of the assigned document will be changed from "unused" to "used".

[0035] First, distinguish the types of medical records involved in hospitalization medical records. According to national policy documents, complete electronic medical records for hospitalization are divided into: admission records, first medical records, daily medical records, ward rounds by senior doctors, stage summaries, transfer records, consultation records, invasive operation records, preoperative summaries, preoperative discussions, surgical records, first postoperative medical records, admission and discharge records within 24 hours, discharge records, etc. The one-way time from admission to discharge is used as a static timeline to record the complete treatment process according to the different medical methods taken during hospitalization.

[0036] The type editor is used to edit the quantity of different record types that can be generated and the content layout framework. For example, the number of generated records (type number) of the initial course of disease record is set to 1, indicating that during a patient's single complete hospitalization, there is exactly 1 copy of this record type; or the number of generated records (type number) of the daily course of disease record is set to ≥0, indicating that this record type can be not generated or repeatedly generated multiple times during a patient's single complete hospitalization. The type editor can, according to actual needs, edit the structured layout of the content of a certain record type and freely adjust each content module. The modules include but are not limited to the patient's general information, admission chief complaint, ward round chief complaint, current medical history, past medical history, vital signs, physical examination, laboratory report, examination report, preliminary diagnosis, diagnosis basis, differential diagnosis, treatment plan, new progress in treatment, medication order, medication analysis, course of treatment, precautions, etc. For example, the preset structured layout of the initial course of disease record may include several modules such as the patient's general information, admission chief complaint, current medical history, past medical history, physical examination, preliminary diagnosis, diagnosis basis, differential diagnosis, treatment plan, etc.

[0037] The sequencer will participate in the operation synchronously when the documents are sorted. Its function is to send instructions to the rule engine for the non-event document time collected by the document collector within the limited time range set for different medical record types under the linear time sequence from the patient's hospitalization to discharge, dynamically adjust the time variables of its condition set, and sort and update the generation time of the documents by the document sorter according to the dynamic time sequence, ensuring that all documents can be correctly assigned to different types of medical records and ensuring that there is no time crossing of documents under the linear time, that is, there is no situation where a document at 3 pm appears in a medical record at 10 am. For example: The generation time of the initial course of disease record is within the time range from the admission time to the admission time + 8h. Since there are null values in the originally preset module content, it is necessary to wait for the document collector to collect the data required for each module. When the document a1 with time n1 (admission time < n1 < admission time + 8h) is collected, the sequencer dynamically adjusts the time variable in the condition set to n1, and then the document sorter assigns the data to the corresponding module of the initial course of disease record. Therefore, the time of the initial course of disease record triggered by a1 is n1; when the document a2 with time n2 (n1 < n2 < admission time + 8h) is collected, the time variable is dynamically adjusted to n2 and the document sorter continues to assign the data. Therefore, the time of the initial course of disease record is updated to n2 by a2; when the document a3 with time n3 (admission time < n3 < n1) is collected, the sequencer does not affect the time variable in the condition set. Therefore, the medical record content updated by the data a3 does not change the generation time of the initial course of disease record.

[0038] For event documents, while the sequencer dynamically adjusts the time variables of the condition set in the rule engine, it also sends instructions to the operation set to affect the operation variables therein. The instructions include the creation, deletion, and update of different medical record types.

[0039] Affect the time variable in the condition set. For example, if there is a surgical operation within 8 hours after the patient is admitted to the hospital, that is, the document collector collects the surgical document c, which is an event-based document, and the document time is m1 (admission time < m1 < admission time + 8h). According to the quality control requirements, there should be a preoperative attending surgeon ward round record within 24 hours before the operation. Therefore, under the action of the sequencer, first, the time variable m2 (m2 < m1) will be added to the condition set. Secondly, under the action of the operation set, a preoperative attending surgeon ward round record with the time m2 will be created. For non-event-based documents collected between m2 and m1 subsequently, the document sorter will assign them to the preoperative attending surgeon ward round record, and at the same time, dynamically adjust the time variable of the record to ensure that m2 is not less than the latest document time before m1 and no time travel occurs. Therefore, the time of the preoperative attending surgeon ward round record triggered by c is m2, and it will be continuously updated backward in the time period from m2 to m1.

[0040] While affecting the time variable in the condition set, it also affects the operation variable in the operation set. For example, on the 3rd day after the patient is admitted to the hospital, there is a surgical operation, and the operation time is m3. And there is a daily ward round record before the time m3 on the 3rd day, and the record time is m4. Because a preoperative attending surgeon ward round record needs to be generated before the operation and there is a daily ward round record in the same time period at this time, the sequencer will first add the time variable m5 (admission time + 48h < m5 < m3) to the condition set. Secondly, change the operation variable in the operation set, delete the daily ward round record on the same day, convert the "used" documents in the daily ward round record back to "unused", and then create a preoperative attending surgeon ward round record with the time m5. At this time, if m5 > m4, all "unused" documents before the time m5 will be sorted into this preoperative attending surgeon ward round record, and at the same time, the label will be converted to "used"; if m5 < m4, all "unused" documents before the time m5 will be sorted into this preoperative attending surgeon ward round record, and at the same time, the label will be converted to "used". For the "unused" documents in the time period from m5 to m4, under the action of the sequencer's previous effect on the condition variables in the condition set, dynamic adjustment is realized until all documents are reallocated, so as to ensure that there is no time travel situation.

[0041] Regarding content analysis, the document content included in different medical record types is used as input and put into the medical vertical large model. Through the operation of the medical vertical large model and the recall path composed of the medical knowledge spectrum, the final required analysis content is output. The medical vertical large model includes a medication analysis module, a diagnosis basis module, and a differential diagnosis module. The modules participating in the calculation as input include but are not limited to: patient general information, admission chief complaint, ward round chief complaint, current medical history, past medical history, vital signs, physical examination, auxiliary examination, medication order, preliminary diagnosis. The modules that need to be calculated as output are: medication analysis, diagnosis basis, differential diagnosis.

[0042] First, there is a medication analysis module. Its main logic starts from the medication, analyzes the possible effects of the drug, and then searches for the corresponding diagnosis or the chief complaints, physical examinations, tests, and examinations required to form this diagnosis (such as Figure 1 shown). At the same time, it combines the capabilities of the medical vertical large model to propose suggestions for the diagnosis and treatment plan of the disease (such as Figure 2 shown). During the process, a knowledge graph and a drug library are required as dependencies.

[0043] The knowledge graph library is a database composed of triples as the smallest storage unit, that is, "entity - relationship - entity" is a triple component. For example, "proteinuria - symptom - kidney disease" means that "proteinuria" is a symptom of "kidney disease".

[0044] The drug library is a database of drug information, mainly including drug names, drug classifications, drug effects, etc. Among them, the drug classification is a content classification with three - level directories, including site - of - action classification, action - type classification, and drug names. For example, "Urapidil Hydrochloride Injection" (drug name) belongs to "Antihypertensive Drugs" (action - type classification) under "Cardiovascular System" (site - of - action classification), and "Nifedipine Capsule" (drug name) belongs to "Calcium Channel Blockers" (action - type classification) under "Cardiovascular System" (site - of - action classification). When performing the calculation steps of the medication analysis module, first, the medication orders issued by the doctor are classified according to the action type, and the classification basis comes from the drug library. Secondly, the effect of the drug is found in the library, and then according to Figure 1 shown, the corresponding diagnosis is searched through the knowledge graph library. Finally, the corresponding patient physical sign information is searched from the four modules of ward chief complaints, physical examination, test report, and examination report, and these contents are used as inputs to be passed to the medical vertical large model. Finally, the medication analysis content is given through the operation of the large model. As Figure 3 shown. The specific process of the medication analysis module is as follows:

[0045] After the doctor prescribes drugs according to the patient's condition, the patient's medical record characteristics, the doctor's diagnosis results, and the specific drug information are received.

[0046] Retrieve detailed information related to the drugs prescribed by the doctor from the preset knowledge base, including drug indications, usage and dosage, adverse reactions, etc.

[0047] Fuse the patient's medical record characteristics, the doctor's diagnosis results, and the drug information, and combine specific prompting words to generate professional medication analysis content through the medical domain large model. This analysis content includes the applicability of the drug, potential risks, and optimization suggestions for the patient's individual situation.

[0048] The diagnosis basis module extracts the patient case characteristic information and combines the preliminary diagnosis and the disease retrospective diagnosis basis in the medical knowledge graph.

[0049] For example: There is now a patient with the following case characteristics: "The patient is a 1-year-and-1-month-old female child; the onset is acute, and the course of the disease is 1 day; her mother reported that the child began to pass yellow loose watery stools without obvious inducement 1 day ago, about 10 times a day, with a small amount, no purulent bloody stools, vomited gastric contents after eating, no coffee-like substances, non-projectile, no fever, no chills or convulsions, no cough, expectoration, nasal congestion or runny nose, no shortness of breath or cyanosis. Now, for further diagnosis and treatment, the patient came to our department and was admitted to our department for inpatient treatment with the diagnosis of 'acute gastroenteritis'."

[0050] The case characteristic information is as follows: The patient's spirit and sleep were okay after the onset of the disease, the appetite was slightly poor, the urine was normal, and the weight gain or loss was unknown. The patient had a normal past medical history. Physical examination: T 37.7°C, HR 120 beats / min, R 32 beats / min, weight 10.5 kg. Conscious, slightly poor spirit, no signs of dehydration, normal skin elasticity, normal development, medium nutrition, no deformities in the head and facial features, no sunken orbits, red lips, congested pharynx, soft neck. On auscultation, the breath sounds of both lungs were clear, no rales were heard, the heart auscultation was normal, the abdomen was flat and soft, bowel sounds were active, no abnormalities were found in the spine and extremities, and no abnormalities were found in the neurological examination; Auxiliary examinations: Not available for the time being."

[0051] The preliminary diagnosis is: "Acute gastroenteritis".

[0052] The content retrieved from the medical knowledge graph in the medical knowledge base is: "The following is the information related to acute gastroenteritis: Symptoms of acute gastroenteritis: Vomiting, diarrhea, nausea. Clinical manifestations of acute gastroenteritis: Restlessness, high fever, mucoid stools, poor appetite, vomiting coffee-like substances. Signs of acute gastroenteritis: Hypokalemia." The model analyzes the three parts of the content received and finally gives the basis for the doctor to make the diagnosis of "acute gastroenteritis", so as to explain the logical derivation process behind the doctor's diagnosis of acute gastroenteritis. The specific process of the diagnosis basis Figure 4 is shown as follows:

[0053] First, collect the patient's medical record characteristics and preliminary diagnosis. The preliminary diagnosis is, for example, input as "pneumonia (right side)".

[0054] Standardize the preliminary diagnosis and convert it into a standardized diagnosis name. For example, convert "pneumonia (right side)" to "pneumonia". This process is achieved through preset standardization rules to ensure the standardization and consistency of the diagnostic information. Based on the standardized diagnosis name, retrieve medical information such as symptoms, signs, and clinical manifestations related to the disease in the medical knowledge base through the medical knowledge graph. The medical knowledge base stores verified medical knowledge, providing basic data support for subsequent analysis.

[0055] Fuse the patient's medical record features, preliminary diagnosis, and disease-related information retrieved from the knowledge base. On this basis, combined with specific prompting words, generate the diagnostic basis for this patient through a pre-trained model. This diagnostic basis synthesizes the patient's individual characteristics and disease-related medical information, ensuring the professionalism and pertinence of the output content. Finally, output the diagnostic basis content that meets professional requirements, providing comprehensive and accurate diagnostic references for doctors and assisting them in making more scientific medical decisions.

[0056] The differential diagnosis module mainly outputs two aspects of content: 1. Give the diseases to be differentiated; 2. Explain the diseases to be differentiated in combination with the patient's own situation. The differential diagnosis process is as follows:

[0057] Receive the patient's primary diagnosed disease information and medical record features as input, including but not limited to the patient's symptoms, signs, examination results, past medical history, etc.

[0058] Multi-strategy disease recall generates a candidate group of diseases to be differentiated through the following three strategies:

[0059] Strategy 1: Disease recall based on the knowledge base of the large model. The medical vertical large model uses the medical knowledge learned during its pre-training stage and combines the patient's own situation to automatically output a group of diseases to be differentiated. This process relies on the model's understanding and reasoning ability of medical knowledge.

[0060] Strategy 2: Recall of the same type of diseases based on ICD-10 coding. Through the ICD-10 coding matching rules in the medical record, automatically screen out the diseases of the same type as the primary diagnosed disease as candidate diseases and input them into the medical vertical large model. Let the large model select the disease that is most likely to be the one to be differentiated from these candidate diseases.

[0061] Strategy 3: Disease recall based on text semantic similarity. The system uses text semantic similarity calculation technology to perform semantic analysis on disease names, symptom descriptions, or medical record texts, and recall diseases highly relevant to the patient's condition. Finally, generate several diseases to be differentiated as a candidate group and input them into the large model, and the model selects the most likely one according to its own logical reasoning.

[0062] After receiving the candidate disease groups generated by the above three strategies, combined with the medical knowledge learned during its own pre-training stage, comprehensively analyze and screen the candidate diseases, and finally output the differential diagnosis disease that best suits the current patient's situation by simulating the reasoning process of medical experts, considering the patient's individual characteristics and disease characteristics.

[0063] The generated differential diagnosis results are output in a structured form, providing accurate diagnostic references for doctors and assisting them in making more scientific medical decisions.

[0064] Such as Figure 5As shown, taking the main diagnosed disease and patient information as input, the medical vertical large model outputs differential diagnosis diseases through its three disease recall strategies for differentiation. Strategy 1: Automatically output the diseases to be differentiated by combining the knowledge base of the medical vertical large model with patient information. Strategy 2: Use ICD-10 coding. Based on matching rules, the medical record system automatically inputs diseases of the same type as candidate diseases to the medical vertical large model. Strategy 3: Use text semantic similarity to recall relevant diseases, so as to give 5 diseases to be differentiated. These 5 diseases to be differentiated will be used as a candidate group and input to the large model. The large model combines the knowledge learned during its pre-training stage to screen the diseases. Thus, the differential diagnosis diseases that best match the current patient's own situation are given.

Claims

1. A novel method for generating electronic medical records of hospitalization, characterized in that: The following steps are involved: Obtain the admission time of the patient according to the preset time interval and / or the user's active request, and poll the corresponding patient information according to the admission time, the patient information includes name, age, gender and nursing level, etc.; Collect medical advice, physical signs, tests, examinations and other documents based on patient information, and classify them according to the document type conditions in the preset condition set to obtain event-type documents and non-event-type documents; It is divided into multiple modules according to the type of medical records, and the corresponding limited time range is preset according to different medical record types; For event-type documents, dynamically adjust the time variables in the preset condition set and the operation variables in the preset operation set according to the generation time sequence corresponding to the event-type documents, execute the action instructions of the corresponding operation set according to the conditions in the preset condition set, and obtain the final operation in combination with the operation variables, and generate the first initial hospitalization electronic medical record of the corresponding type according to the final operation; For non-event documents, dynamically adjust the time variables in the preset condition set according to the generation time sequence corresponding to the non-event documents, dynamically sort the non-event documents according to the module according to the time variables and fill them into the module to generate the second initial hospitalization electronic medical record; The contents of the first initial hospitalization electronic medical record and the second initial hospitalization electronic medical record are used as the input of the medical vertical model, and the medication analysis, diagnostic basis and differential diagnosis are output and filled into the module to obtain the final hospitalization electronic medical record.

2. A novel method for generating electronic medical records of hospitalization as claimed in claim 1, characterized in that: When the timer consumes more time than the set time interval to obtain patient information, an alarm is given to guide the user to reset the time interval.

3. A novel method for generating electronic medical records of hospitalization as claimed in claim 1, characterized in that: The event-type documents include documents corresponding to key medical behavior events such as surgery, consultation, transfer, blood transfusion, etc.

4. A novel method for generating electronic medical records of hospitalization as claimed in claim 1, characterized in that: The medication analysis is obtained by analyzing the chief complaint of ward rounds, medication orders, general information, current medical history, vital signs, physical examinations, and test reports in the first initial hospitalization electronic medical record and the second initial hospitalization electronic medical record according to the medical knowledge graph and the drug library.

5. A novel method for generating electronic inpatient medical records as claimed in claim 4, characterized in that: The characteristic information of the first initial hospitalization electronic medical record and the second initial hospitalization electronic medical record is extracted, and combined with the preliminary diagnosis, the medical knowledge graph is used to search the medical knowledge base for disease retrospective diagnosis to obtain the diagnostic basis.

6. A novel method for generating electronic medical records of hospitalization as claimed in claim 5, characterized in that: According to the characteristic information combined with the knowledge graph, the medical knowledge base is retrieved to obtain the first strategy disease candidate group. According to the characteristic information, the same type of diseases are screened to obtain the second strategy disease candidate group. According to the characteristic information, text semantic similarity recognition and screening are used to obtain the third strategy disease candidate group. The first strategy disease candidate group, the second strategy disease candidate group and the third strategy disease candidate group are combined through the reasoning process of simulating medical experts, considering the individual characteristics and disease characteristics of the patients, and finally the differential diagnosis of the disease is output.

Images