A primary medical treatment prescription intelligent generation method, system, device and medium

Abstract

This application relates to a method, system, device, and medium for intelligent prescription generation in primary healthcare, belonging to the technical field of medical auxiliary diagnosis. The method includes: real-time collection and analysis of doctor-patient dialogue content to extract diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history; extraction of key pathophysiological parameters based on the diagnostic information; generation of current treatment information based on the key pathophysiological parameters and filling it into a medical record template, the current treatment information including a set of diagnoses for the current visit and a set of treatment goals for each diagnosis; generation of a personalized prescription suggestion draft based on the current treatment information, and formation of a formal prescription after review and confirmation by a doctor; and a qualification review of the formal prescription, and generation of a personalized medication guidance sheet after approval. This application has the beneficial effect of alleviating the workload of primary healthcare doctors and improving prescription quality.

Description

Technical Field

[0001] This application relates to the technical field of medical auxiliary diagnosis, and in particular to a method, system, device and medium for intelligent generation of prescriptions in primary healthcare. Background Technology

[0002] As the "last mile" of my country's medical and health service system, primary healthcare bears the important responsibility of providing basic medical care, health management, and other services to the vast majority of people at the grassroots level. The quality of its services is directly related to the improvement of the overall health level of the population.

[0003] Currently, primary healthcare institutions generally face problems such as a shortage of doctors, a heavy workload in diagnosis and treatment, and a low degree of standardization in medical records. As a core medical document in the diagnosis and treatment process, the efficiency, accuracy, and standardization of prescription generation not only affect the smoothness of the diagnosis and treatment process, but also relate to the patient's medication safety and treatment effectiveness.

[0004] However, primary care physicians need to juggle multiple tasks during diagnosis and treatment, including doctor-patient communication, disease assessment, medical record keeping, and prescription writing. This dispersed attention leads to frequent problems such as omissions in medical record keeping and non-standard prescription writing. Therefore, there is an urgent need for an efficient and intelligent prescription generation method to alleviate the workload of primary care physicians and improve prescription quality. Summary of the Invention

[0005] To alleviate the workload of primary care physicians and improve the quality of prescriptions, this application provides a method, system, device, and medium for intelligent generation of prescriptions in primary healthcare.

[0006] Firstly, this application provides a method for intelligently generating prescriptions for primary healthcare, employing the following technical solution: A method for intelligently generating prescriptions for primary healthcare includes: The system collects and analyzes doctor-patient dialogue content in real time to extract diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history. Based on the diagnostic information, extract key pathophysiological parameters; Based on the key pathophysiological parameters, current diagnosis and treatment information is generated and populated into the medical record template. The current diagnosis and treatment information includes the set of diagnoses for this visit and the set of treatment goals for each diagnosis. Based on the current medical information, a draft personalized prescription suggestion is generated, and after the doctor reviews and confirms it, a formal prescription is formed. The formal prescription is reviewed for eligibility, and a personalized medication guidance sheet is generated upon successful review.

[0007] By adopting the above-mentioned technical solutions, fragmented doctor-patient dialogues are transformed into structured diagnostic and treatment goals. This not only reduces the time doctors spend manually compiling medical records, allowing primary care physicians to devote more energy to communicating with patients and assessing their conditions, but also reduces information bias caused by human error in record-keeping through standardized medical record filling processes, improving the standardization and accuracy of primary care medical documents. Prescription suggestion drafts based on personalized diagnosis and treatment information combine the individual characteristics of patients' conditions with professional references for primary care physicians with relatively less experience, effectively compensating for the shortcomings of uneven distribution of primary care medical resources and differences in doctors' professional levels, allowing patients to receive more targeted treatment plans. Subsequent prescription compliance review and personalized medication guidance form build a strong defense for medical safety from the end of the diagnosis and treatment process. The review mechanism can promptly intercept potential medication risks, while the medication guidance form helps patients clearly understand the medication guidelines in an easy-to-understand format, improving patients' medication adherence. Ultimately, a complete diagnosis and treatment closed loop of "accurate data collection - intelligent analysis - standardized generation - safety verification - considerate guidance" is formed, which improves the service efficiency, diagnosis and treatment quality and patient medical experience of primary healthcare, and alleviates the workload of primary healthcare doctors.

[0008] Optionally, the specific steps for generating a draft personalized prescription suggestion based on the current medical information include: Based on the built-in recommendation rules and drug guide knowledge base, a corresponding set of recommended drug candidates is generated for each diagnosis in the diagnostic set. The recommendation rules include efficacy, safety, cost-effectiveness and availability. Based on the recommended drug candidate set and the treatment target set for each diagnosis association, the positive synergistic optimization logic, the reverse conflict avoidance logic, and the drug interaction pre-screening logic are executed sequentially. The positive synergistic optimization logic indicates that the candidate drugs can simultaneously cover multiple treatment targets. The reverse conflict avoidance logic indicates whether the contraindications of all candidate drugs match the patient's condition. The drug interaction pre-screening logic indicates that at the recommended combination level, it assesses whether there are pharmacological interactions between candidate drugs. The integrated logic processing results generate a suggested prescription draft, which includes a set of suggested drugs and a suggested initial dose and rationale for each drug.

[0009] By adopting the above technical solutions, relying on built-in recommendation rules and an authoritative drug guideline knowledge base, candidate drugs are matched for each diagnosis from multiple dimensions such as efficacy, safety, economy, and availability. This ensures both the professionalism and standardization of drug recommendations and meets the actual needs of drug accessibility and patients' economic affordability in primary healthcare settings. The triple logical verification of positive synergistic selection, negative conflict avoidance, and drug interaction pre-screening constructs a rigorous prescription optimization mechanism. Positive synergy maximizes the therapeutic value of drugs, allowing a single drug to cover multiple treatment goals and reducing the types and frequency of medications taken by patients. Negative conflict avoidance starts from the individual patient's condition, excluding drugs with contraindications and avoiding medication risks from the source. Drug interaction pre-screening further predicts potential pharmacological contraindications at the combination level, avoiding adverse reactions that may occur when multiple drugs are used in combination. The final integrated draft prescription not only matches the patient's individual condition and treatment needs, but also undergoes multi-dimensional professional verification. It can not only provide primary care physicians with scientific prescription references and reduce irrational medication due to lack of experience, but also improve the efficiency of prescription generation, allowing primary care physicians to obtain more accurate medication plans in a short time.

[0010] Optionally, the steps following the generation of the suggested prescription draft also include: For each drug in the proposed prescription draft, an individualized dosage range is calculated based on the patient's individual parameters; The draft prescription is revised based on the individualized dosage range corresponding to each drug to obtain a personalized prescription recommendation draft.

[0011] By adopting the above technical solution, we break away from the traditional fixed model of prescribing according to universal dosage. We calculate the individualized dosage range based on the patient's own parameters, taking into account the influence of individual characteristics such as the patient's age, weight, liver and kidney function, and underlying diseases on drug metabolism. This makes the drug dosage no longer a "one-size-fits-all" standard value, but a precise range that fits the patient's physical condition, which can effectively avoid insufficient drug efficacy or adverse drug reactions caused by improper dosage.

[0012] Optionally, the specific steps for conducting a conformity review of the formal prescription include: Check that the formal prescription meets the format requirements; The following steps are performed sequentially: contraindication verification, drug interaction identification, and identification of duplicate and contradictory medication use. Compare the actual prescribed dose of each drug in the formal prescription with the maximum permissible dose of the individualized dose range; Based on the audit results, an audit conclusion and corresponding intervention measures are generated.

[0013] By adopting the above technical solutions, starting with basic checks on format requirements, the standardization of prescription documents is ensured. The progressive substantive review, including contraindication verification, pharmacokinetic and pharmacodynamic interaction identification, and identification of duplicate and contradictory medications, provides in-depth verification from the perspectives of individual patient suitability and drug combination safety. This not only identifies conflicts between drugs and individual patient conditions but also proactively identifies potential pharmacological risks associated with combined drug use, compensating for medication details that primary care physicians might overlook due to lack of experience or busy schedules. Comparing the actual prescribed dosage with the maximum permissible dosage within the individualized dosage range further ensures dosage accuracy, preventing safety risks caused by exceeding dosage limits and ensuring that dosage is more aligned with the patient's individual metabolic capacity. The conclusions and interventions generated based on the review results provide primary care physicians with clear directions for correction, helping them quickly locate problems and adjust prescriptions. This not only enhances the practicality of prescription review but also subtly improves the prescription-writing abilities of primary care physicians through repeated review feedback.

[0014] Optionally, the specific steps for generating a personalized medication instruction sheet include: After the review is approved, the standard usage and common side effects of each drug in the formal prescription will be extracted from the drug information database. Integrate the review results with the core treatment rationale in the formal prescription; Based on the diagnosis presented during this visit, additional dietary and exercise recommendations are provided. The standard usage and common side effects of each drug extracted from the official prescription, the integrated review results and core treatment rationale, as well as additional dietary and exercise recommendations, are filled into the medication guidance template to generate a personalized medication guidance sheet for the patient.

[0015] By employing the aforementioned technical solutions, standard usage and common side effects are extracted from the drug information in formal prescriptions. This allows patients to clearly understand the usage guidelines and potential physical reactions of each medication, avoiding adverse effects due to incorrect dosage. It also enables timely management of minor side effects, reducing unnecessary panic and frequent medical visits. Integrating review results with core treatment rationale reveals the diagnostic and treatment logic behind the prescription, helping patients understand "why this medication is used" and "what effect it has." This not only improves patients' understanding and acceptance of the treatment plan but also enhances their willingness to actively cooperate with treatment, fundamentally improving medication adherence. Additional dietary and exercise recommendations combine medication guidance with patients' daily health management, providing lifestyle adjustments tailored to specific diagnoses. This helps patients shift from "passive medication" to "proactive health management," especially for patients with chronic diseases such as hypertension and diabetes. This multi-dimensional guidance effectively supports drug treatment and improves overall treatment outcomes. By focusing on patients' actual needs and transforming professional medical information into easily understandable and practical guidelines, the overall patient experience is further enhanced.

[0016] Optionally, the specific steps for generating a corresponding set of recommended drug candidates for each diagnosis in the diagnostic set, based on the built-in recommendation rules and drug guide knowledge base, include: Based on the built-in recommendation rules and drug guide knowledge base, an initial set of recommended drug candidates is generated for each diagnosis in the diagnostic set; Retrieve current inventory information for all drugs and categorize them into three levels based on inventory adequacy: sufficient inventory, tight inventory, and shortage inventory; the inventory information includes real-time inventory data, expiration date information, and inventory turnover rate. For each candidate drug in the initial recommended drug candidate set, match the corresponding inventory information; For candidate drugs with sufficient inventory, they are retained in the initial recommended drug candidate set with priority, and the real-time inventory balance and expected supply duration are marked. For candidate drugs with tight inventory, the number of available treatment courses is calculated based on the patient's treatment needs for this visit and the inventory turnover rate of the candidate drug. If the number of available treatment courses meets the treatment needs for this visit, the candidate drug is retained and an inventory warning is issued. If the treatment needs for this visit cannot be met, the drug is moved to the end of the initial recommended drug candidate set and associated with alternative drugs. For candidate drugs with inventory shortages, they are removed from the initial recommended drug candidate set, and alternative drugs are added based on the drug guide knowledge base.

[0017] By adopting the above technical solutions, a deep linkage between the recommended drug candidate set and the drug inventory at the grassroots level has been achieved, breaking the traditional limitation of "emphasizing diagnosis and treatment while neglecting inventory". By integrating inventory factors into the pre-prescription generation process, the embarrassing situation of "no medicine available after prescription is issued" is avoided, reducing the number of times patients need to visit medical institutions and improving the convenience of medical treatment.

[0018] Optionally, the specific steps for extracting key pathophysiological parameters based on the diagnostic information include: Extract symptom descriptions, durations, frequency of attacks, and accompanying symptoms from the chief complaint and present medical history, and extract the patient's basic physiological parameters, which include at least age, weight, and gender. Analyze allergy history and past medical history information, and record in a structured manner the name of the allergic drug, the type of allergic reaction, and the name and stage of the diagnosis of previous chronic diseases; Based on the symptom description and the basic physiological parameters, derived physiological indicators are calculated. The derived physiological indicators include at least body surface area, estimated creatinine clearance rate, and potential liver and kidney function classification. Based on the duration, frequency of attacks, and accompanying symptoms, a clinical scoring scale was matched to calculate the severity index and the probability of complication risk. Based on the correlation analysis between past medical history and current symptoms, predictive values ​​of organ function impairment level and drug metabolism capacity are generated. By integrating the aforementioned basic physiological parameters, derived physiological indicators, disease severity index, complication risk probability, organ function impairment level, and drug metabolism capacity prediction values, key pathophysiological parameters are generated.

[0019] By employing the aforementioned technical solutions and extracting and cross-validating multi-dimensional information, a comprehensive portrait of the patient's pathophysiological state was achieved. Not only were basic pharmacokinetic parameters such as body surface area calculated using fundamental physical signs, but the severity of the condition was also quantified by combining dynamic symptom information such as disease duration and attack frequency, compensating for the limitations of single static parameters. The organ function impairment level and drug metabolism capacity prediction values ​​generated based on the correlation analysis of medical history and current condition effectively solved the problem of inaccurate disease assessment caused by a lack of diagnostic equipment in primary healthcare, ensuring the safety and effectiveness of treatment plans.

[0020] Secondly, this application provides an intelligent prescription generation system for primary healthcare, which adopts the following technical solution: A primary healthcare prescription intelligent generation system includes: The voice processing module is used to collect and analyze the doctor-patient dialogue content in real time and extract diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history. The parameter processing module is used to extract key pathophysiological parameters based on the diagnostic information. The medical record processing module is used to generate current diagnosis and treatment information based on the key pathophysiological parameters and fill it into the medical record template. The current diagnosis and treatment information includes the set of diagnoses for this visit and the set of treatment goals for each diagnosis. The prescription generation module is used to generate a personalized prescription suggestion draft based on the current diagnosis and treatment information, and to form a formal prescription after the doctor reviews and confirms it; The prescription review module is used to review the eligibility of the formal prescription and generate a personalized medication guidance sheet after the review is passed.

[0021] Thirdly, this application provides a computer device that adopts the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the intelligent prescription generation method for primary healthcare as described in the first aspect.

[0022] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program capable of being loaded by a processor and executed as described in the first aspect of the intelligent generation method for primary healthcare prescriptions. Attached Figure Description

[0023] Figure 1 This is a first flowchart of an embodiment of the method of this application; Figure 2 This is a second flowchart of an embodiment of the method of this application; Figure 3 This is a third flowchart of an embodiment of the method of this application; Figure 4 This is the fourth flowchart of an embodiment of the method of this application; Figure 5 This is the fifth flowchart of an embodiment of the method of this application; Figure 6 This is the sixth flowchart of an embodiment of the method of this application. Detailed Implementation

[0024] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1-6 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0025] The first embodiment of this application discloses a method for intelligent generation of prescriptions in primary healthcare. (Refer to...) Figure 1The method includes S110-S150: S110 collects and analyzes doctor-patient dialogue content in real time, extracts diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history; S120: Extract key pathophysiological parameters based on diagnostic information; S130: Based on key pathophysiological parameters, generate current diagnosis and treatment information and populate it into the medical record template. The current diagnosis and treatment information includes the set of diagnoses for this visit and the set of treatment goals for each diagnosis. S140: Based on the current medical information, a draft personalized prescription suggestion is generated, and after the doctor reviews and confirms it, a formal prescription is formed. S150 conducts a qualification review of formal prescriptions and generates a personalized medication guidance sheet after the review is passed.

[0026] Specifically, for step S110, an end-to-end speech recognition model based on deep learning (such as Whisper-large-v3) is used as the core acquisition tool, coupled with a dedicated microphone array for noise reduction in the clinic environment. This filters out interference signals such as background noise and equipment operation sounds, enabling real-time speech capture and transcription of doctor-patient dialogue. During transcription, an intelligent medical terminology correction module is embedded to immediately correct common errors such as misidentifying "cough" as "keshou" and "penicillin" as "penicillin fungus," ensuring the accuracy of the transcribed text. The dialogue text analysis employs a Named Entity Recognition (NER) model from Natural Language Processing (NLP). This model is pre-trained using a dataset of common primary care disease cases, constructing an entity dictionary containing categories such as symptoms, diseases, medications, and medical history. It can parse the transcribed text sentence by sentence and extract key diagnostic information. For example, when a doctor-patient dialogue is "I've been coughing for the past three days, with a fever. I've been allergic to penicillin before, and I was diagnosed with hypertension last year," the model can automatically identify the chief complaint as "coughing with fever for the past three days," the present illness as "cough and fever, lasting for three days," the allergy history as "penicillin allergy," and the past medical history as "diagnosed with hypertension last year." After extraction, the data is categorized and stored according to a preset structured JSON format, corresponding to the four fields: "chief complaint," "present illness," "allergy history," and "past medical history."

[0027] Reference Figure 2 S120, the specific steps for extracting key pathophysiological parameters based on diagnostic information include S210-S260: S210: Extract symptom descriptions, durations, frequency of attacks, and accompanying symptoms from the chief complaint and present medical history, and extract the patient's basic physiological parameters, which include at least age, weight, and gender. S220 analyzes allergy history and past medical history information, and records the names of allergic drugs, types of allergic reactions, and the names and stages of previous chronic disease diagnoses. S230, based on symptom description and basic physiological parameters, calculates derived physiological indicators, which include at least body surface area, estimated creatinine clearance rate, and potential liver and kidney function classification. S240, based on duration, frequency of attacks and accompanying symptoms, matches a clinical scoring scale to calculate the severity index and probability of complication risk; S250 generates predictive values ​​for organ function impairment level and drug metabolism capacity based on correlation analysis of past medical history and current symptoms. S260 integrates basic physiological parameters, derived physiological indicators, disease severity index and complication risk probability, as well as organ function impairment level and drug metabolism capacity prediction value to generate key pathophysiological parameters.

[0028] Specifically, for step S210, symptom-related information extraction employs a dual logic combining text classification and entity extraction. First, a text classification model distinguishes four information types: symptom description, duration, frequency of occurrence, and accompanying symptoms. Then, an entity extraction model extracts the specific content. During model training, common symptom expression habits at the grassroots level are incorporated (e.g., "coughing up phlegm" is often expressed as "having phlegm," and "fever" is often expressed as "fever") to ensure extraction accuracy. For example, for the text "I cough every morning after waking up, each cough lasts about 10 minutes, I've been coughing for 5 days, and I also have a sore throat and runny nose," the model can accurately extract the symptom description as "cough, sore throat, runny nose," the duration as "5 days," the frequency of occurrence as "every morning after waking up, each lasting about 10 minutes," and the accompanying symptoms as "sore throat, runny nose."

[0029] The extraction of basic physiological parameters adopts a three-dimensional approach: "dialogue extraction + electronic health record (EHR) association + doctor supplementation". First, relevant information is automatically identified from the doctor-patient dialogue. For example, from "I am 35 years old, weigh 60 kg, and am female", the system directly extracts age 35, weight 60 kg, and gender female. If the patient has already established an EHR at a primary healthcare institution, the system automatically associates the record with the patient's ID number or medical card number, retrieves the basic physiological parameters, and supplements any information not mentioned in the dialogue. If the information is not mentioned in the dialogue and there is no EHR record, the system displays a prompt box through the in-clinic interaction interface, allowing the doctor to manually enter the information. After entry, the data is automatically synchronized to the system database, and the reasonableness of the entered data is simultaneously verified.

[0030] For step S220, the allergy history analysis employs a combination of the NER model and rule matching. The NER model identifies the names of the allergenic drugs, while the rule matching module analyzes the allergy symptoms described in the dialogue based on a pre-defined dictionary of allergy reaction types (such as skin allergies, respiratory allergies, gastrointestinal allergies, etc.) to classify the reactions. For example, from the statement "I am allergic to amoxicillin and cephalosporins; after taking them, I develop rashes and difficulty breathing," the model identifies the allergenic drugs as "amoxicillin and cephalosporins." The rule matching module classifies "rash" as a skin allergy and "difficulty breathing" as a respiratory allergy, while simultaneously recording the specific manifestations of the reactions.

[0031] Past medical history analysis also employs a NER model to extract chronic disease diagnoses, combined with a text semantic analysis model to determine the disease stage. The text semantic analysis model identifies keywords such as disease duration and disease control status (e.g., "3 years" and "stable" in "diagnosed 3 years ago, consistently on medication, blood sugar basically stable"), and, in conjunction with primary care chronic disease treatment guidelines, classifies disease stages into types such as "chronic disease course, stable condition," "chronic disease course, unstable condition," and "acute exacerbation." Structured records use standardized tables. Allergy history records include four fields: "name of allergic drug," "type of allergic reaction," "specific manifestation of reaction," and "time of discovery." Past chronic disease records include five fields: "diagnosis name," "time of diagnosis," "disease stage," "control status," and "medication history." All records are associated with a unique patient identifier.

[0032] For step S230, the body surface area is calculated using the DuBois formula commonly used in primary healthcare institutions, namely, body surface area (BSA) = 0.007184 × body weight (kg). 0.425 × Height (cm) 0.725 The results should be rounded to two decimal places, and the basis for the calculation should be noted. For example, if a patient weighs 60 kg and is 165 cm tall, the calculated BSA is 0.007184 × 60. 0.425 ×165 0.725The area was approximately 1.68 square meters, and the calculation was labeled "Calculation basis: DuBois formula, weight 60 kg, height 165 cm". The estimated creatinine clearance rate (Ccr) was calculated using the Cockcroft-Gault formula, differentiated by gender: male Ccr = (140 - age) × weight (kg) / (72 × serum creatinine (mg / dl)), female Ccr = male Ccr × 0.85. Serum creatinine data was automatically obtained from the patient's recent (within 7 days) serum creatinine test results through the system's interface with the testing equipment of primary healthcare institutions. If no test was performed, the system prompted the doctor to determine whether supplementary testing was necessary. If no supplementary testing was required, the normal reference values ​​for adults (male 70-106 μmol / L, female 53-97 μmol / L) were used for conversion and calculation. Potential liver and kidney function grading is based on the simplified grading standards used in primary healthcare institutions. Kidney function grading is based on estimated creatinine clearance: Ccr ≥ 90 ml / min is normal, 60-89 ml / min is mildly impaired, 30-59 ml / min is moderately impaired, and < 30 ml / min is severely impaired. Liver function grading combines symptom description (e.g., presence of jaundice, ascites, fatigue) and baseline physiological parameters (e.g., age, alcohol consumption history), referencing the simplified Child-Pugh classification, and is divided into four levels: normal, mildly impaired, moderately impaired, and severely impaired. For example, a patient without jaundice or ascites and with normal albumin is considered to have normal liver function; if mild jaundice and a slight decrease in albumin are present, liver function is considered mildly impaired.

[0033] For step S240, a clinical scoring scale database for common diseases at the grassroots level is constructed, covering prevalent diseases such as the common cold, pneumonia, hypertensive emergencies, diabetic ketoacidosis, and acute gastroenteritis. Each scale corresponds to a clear scoring standard, with scoring items covering the duration of illness, frequency of attacks, accompanying symptoms, and basic physiological parameters. For example, in the common cold scoring scale, 1 point is awarded for illness duration <3 days, 2 points for 3-7 days, and 3 points for >7 days; 1 point is awarded for occasional attacks, and 2 points for frequent attacks; 1 point is awarded for no accompanying symptoms, 2 points for fever, and 3 points for dyspnea. Scale matching uses a keyword matching algorithm. The system automatically matches the corresponding scoring scale based on the extracted duration of illness, frequency of attacks, and accompanying symptoms. If the patient's symptoms involve multiple diseases, the scale for the most relevant primary disease is matched. The severity index is calculated by adding up the scores of each item and combining them with the scale's preset grading standard (0-3 points for mild, 4-6 points for moderate, and 7-9 points for severe). For example, a patient with a 5-day cold, frequent cough, and fever has a total score of 2+2+2=6 points, indicating a moderate condition with a severity index of 6 points. The probability of complication risk is calculated using a logistic regression machine learning model. This model is trained on a historical case dataset from primary healthcare institutions (containing labels such as patient symptoms, baseline parameters, and whether complications have occurred). It takes duration, frequency of attacks, accompanying symptoms, baseline physiological parameters, and derived physiological indicators as input features and outputs the probability of complication risk, while also labeling the risk level (low risk <20%, medium risk 20%-50%, high risk >50%). For example, if a patient has a 5-day cold with fever, the model outputs a 15% probability of pneumonia complications, labeling it as low risk, and suggests "closely monitoring body temperature changes to avoid complications."

[0034] For step S250, the association analysis employed the Apriori association rule mining algorithm, combined with a primary healthcare case database, to uncover the associations between past medical history and current symptoms, as well as organ dysfunction. For example, a strong association was found between "history of hypertension + current dizziness and chest tightness" and "impaired cardiac function," and between "history of diabetes + current worsening of polydipsia and polyuria" and "impaired renal function." The classification of organ dysfunction levels referenced the "Guidelines for Organ Function Assessment in Primary Healthcare Institutions," classifying major organs such as the heart, liver, kidneys, and lungs into grades I-IV based on patient symptoms and baseline parameters. For instance, a patient with a history of hypertension who currently experiences occasional chest tightness without lower extremity edema is classified as having grade I cardiac dysfunction; if chest tightness, shortness of breath, and mild lower extremity edema occur, the patient is classified as having grade II cardiac dysfunction. The predictive value of drug metabolism capacity is determined by a quantitative scoring method with a maximum score of 100 points. The scoring indicators include liver and kidney function grade (normal 30 points, mild impairment 20 points, moderate impairment 10 points, severe impairment 0 points), age (<60 years old 25 points, 60-70 years old 15 points, >70 years old 5 points), history of metabolic-related diseases (none 25 points, present 10 points), and body mass index (normal 20 points, abnormal 10 points). The predictive value is obtained by adding up the scores of each item. 80-100 points indicates strong metabolic capacity, 60-79 points indicates moderate metabolic capacity, and <60 points indicates weak metabolic capacity. For example, a 35-year-old woman with normal liver and kidney function (30 points), age < 60 years (25 points), no history of metabolic diseases (25 points), and a body mass index (BMI) of 24.5 kg / m² (overweight, 10 points) would have a total score of 30 + 25 + 25 + 10 = 90 points, indicating strong drug metabolism capacity. Conversely, a 72-year-old man with mild kidney impairment (20 points), normal liver function (30 points), a history of diabetes (10 points), and a BMI of 28 kg / m² (obese, 10 points) would have a total score of 20 + 30 + 10 + 10 = 70 points, indicating moderate drug metabolism capacity. The system stores the organ function impairment level and the predicted drug metabolism capacity in a structured label format, such as "Heart function grade I impairment, moderate drug metabolism capacity."

[0035] For step S260, a weighted calculation model based on multi-dimensional feature fusion is employed. First, weights are assigned to each input parameter. These weights are set based on primary healthcare clinical guidelines and expert experience, such as 15% for basic physiological parameters (age, weight, gender), 30% for derived physiological indicators (body surface area, estimated creatinine clearance, liver and kidney function classification), 25% for disease severity index and complication risk probability, and 30% for organ function impairment level and predicted drug metabolism capacity. Then, normalization is performed to unify parameters of different dimensions to the 0-100 range. For example, age (years), creatinine clearance (ml / min), disease severity index (points), and predicted drug metabolism capacity (points) are linearly normalized according to their corresponding reference ranges, and then weighted and summed to obtain a comprehensive score for key pathophysiological parameters, while simultaneously decomposing sub-parameter labels for each dimension. For example, for a 35-year-old female patient with hypertension, the normalized baseline physiological parameters scored 85 points (weight 15%, corresponding to 12.75 points), derived physiological indicators scored 92 points (weight 30%, corresponding to 27.6 points), the severity index and complication risk probability scored 78 points (weight 25%, corresponding to 19.5 points), and the predicted organ function and metabolic capacity scored 90 points (weight 30%, corresponding to 27 points). The comprehensive score was 12.75 + 27.6 + 19.5 + 27 = 86.85 points. The sub-parameter labels included "age 35 years, weight 60 kg, BSA 1.68 m², Ccr 80.28 ml / min, normal renal function, NYHA Class I heart function, strong drug metabolism, moderate hypertension, and low risk of pneumonia." The system outputs key pathophysiological parameters in a structured report format, including the comprehensive score, core sub-parameters, and clinical interpretation.

[0036] For step S130, a diagnostic matching library for common primary care diseases is constructed. Each disease entry in the library is associated with matching rules for key pathophysiological parameters. For example, the matching rule for hypertension is "systolic blood pressure ≥140 mmHg and / or diastolic blood pressure ≥90 mmHg (combined with symptoms) + age ≥35 years + creatinine clearance ≥60 ml / min", and the matching rule for diabetes is "fasting blood glucose ≥7.0 mmol / L or 2-hour postprandial blood glucose ≥11.1 mmol / L + history of diabetes + normal / mildly impaired renal function classification". The system performs multi-rule parallel matching between the aforementioned key pathophysiological parameters and the diagnostic matching library, while also performing double verification based on the doctor's preliminary diagnosis in the clinic. If multiple disease entries are successfully matched and all match the patient's current symptoms and baseline parameters, it is determined that the patient has multiple coexisting diseases, and the system automatically generates a diagnostic set containing multiple diagnoses; if the two are consistent, the diagnostic set is determined; if there are discrepancies, the system pops up a prompt box for the doctor to confirm the final diagnosis (including the priority order of multiple disease diagnoses). For example, if a patient's key pathophysiological parameters show "moderate hypertension, normal renal function, moderate drug metabolism, fasting blood glucose of 8.2 mmol / L, history of diabetes, and occasional chest tightness," and the doctor's preliminary diagnosis is "primary hypertension grade 2 (intermediate risk), type 2 diabetes, and coronary heart disease (stable)," the system will automatically generate a diagnostic set of ["primary hypertension grade 2 (intermediate risk)," "type 2 diabetes," and "coronary heart disease (stable)"], and sort them according to "primary diagnosis - secondary diagnosis" (primary hypertension is the primary diagnosis, and the rest are secondary diagnoses).

[0037] For each diagnosis, the system sets standardized treatment goals based on authoritative documents such as the "National Guidelines for the Prevention and Management of Hypertension in Primary Care," the "Guidelines for the Comprehensive Prevention and Treatment of Diabetes in Primary Care," and the "Guidelines for the Prevention and Treatment of Coronary Heart Disease in Primary Care." It also considers the need for coordinated treatment of multiple diseases and avoids conflicts in treatment goals. For example, the treatment goal for hypertension is "blood pressure controlled below 140 / 90 mmHg, while also protecting target organs in diabetes and coronary heart disease." The treatment goal for diabetes is "fasting blood glucose 4.4-7.0 mmol / L, 2-hour postprandial blood glucose <10.0 mmol / L, glycated hemoglobin <7.0%, avoiding interactions with antihypertensive and cardioprotective drugs." The treatment goal for coronary heart disease is "relieving chest tightness symptoms, preventing angina attacks, and protecting myocardial function." The system fills multiple diagnostic sets and corresponding treatment goal sets (with each diagnosis priority marked) into a preset primary healthcare electronic medical record template. The template includes standardized fields such as patient basic information, chief complaint, present medical history, diagnosis (divided into primary / secondary), and treatment goals (corresponding to each diagnosis), generating structured current diagnosis and treatment information. At the same time, it supports doctors to manually modify treatment goals and diagnostic priorities to ensure that they are tailored to the individual circumstances of patients with multiple coexisting diseases.

[0038] Reference Figure 3The specific steps in S140 for generating a draft personalized prescription suggestion based on current medical information include S310-S350: S310 generates a corresponding set of recommended drug candidates for each diagnosis in the diagnostic set based on built-in recommendation rules and drug guide knowledge base. The recommendation rules include efficacy, safety, cost-effectiveness and availability. S320, based on the recommended drug candidate set and treatment target set for each diagnosis association, sequentially executes the positive synergistic optimization logic, the reverse conflict avoidance logic, and the drug interaction pre-screening logic; the positive synergistic optimization logic characterizes whether the candidate drugs can simultaneously cover multiple treatment targets, the reverse conflict avoidance logic characterizes whether the contraindications of all candidate drugs match the patient's condition, and the drug interaction pre-screening logic characterizes the assessment of whether there are pharmacological interactions between candidate drugs at the recommended combination level. S330, integrate the results of the logic processing and generate a draft prescription. The draft prescription includes a set of recommended drugs and the recommended initial dose and rationale for each drug. S340, for each drug in the suggested prescription draft, calculates an individualized dosage range based on the patient's individual parameters; S350, based on the individualized dosage range corresponding to each drug, revise the proposed prescription draft to obtain a personalized prescription recommendation draft.

[0039] Reference Figure 4 S310, the specific steps for generating a corresponding set of recommended drug candidates for each diagnosis in the diagnostic set based on the built-in recommendation rules and drug guide knowledge base include S410-S460: S410 generates an initial set of recommended drug candidates for each diagnosis in the diagnostic set based on the built-in recommendation rules and drug guide knowledge base; S420 retrieves the current inventory information of all drugs and classifies them into three levels according to inventory adequacy: sufficient inventory, tight inventory, and shortage of inventory; the inventory information includes real-time inventory data, expiration date information, and inventory turnover rate. S430: For each candidate drug in the initial recommended drug candidate set, match the corresponding inventory information; S440 prioritizes candidate drugs with sufficient inventory in the initial recommended drug candidate set and marks the real-time inventory balance and expected supply duration. S450: For candidate drugs with tight inventory, calculate the number of treatment courses that can be supplied based on the patient's treatment needs for this visit and the inventory turnover rate of the candidate drug. If the number of treatment courses that can be supplied meets the treatment needs for this visit, the candidate drug is retained and an inventory warning is marked. If the treatment needs for this visit cannot be met, the drug is moved to the end of the initial recommended drug candidate set and associated with alternative drugs. S460 removes candidate drugs with inventory shortages from the initial recommended drug candidate set and supplements them with alternative drugs based on the drug guide knowledge base.

[0040] Specifically, for step S410, the drug guide knowledge base integrates authoritative content such as the National Essential Medicines List, the Drug Use List for Primary Healthcare Institutions, and the Guidelines for the Diagnosis and Treatment of Common Diseases in Primary Care. Drug information is stored by disease category, and each drug entry includes core attributes such as drug name, dosage form, specifications, indications, efficacy rating, safety rating, medical insurance type, primary care supply status, and synergy / conflict indicators with other disease drugs. The built-in recommendation rules adopt a multi-dimensional scoring system with a maximum score of 100 points, including 40 points for efficacy, 30 points for safety, 20 points for cost-effectiveness, and 10 points for availability. For scenarios involving multiple coexisting diseases, an additional score (0-10 points) is added for "synergistic drug compatibility." If a drug can simultaneously meet the treatment needs of multiple diagnoses without mutual conflict, 5-10 points are added; if there is a slight conflict with other diagnostic drugs but it can be adjusted, 0-3 points are added; if there is a serious conflict, 10 points are deducted.

[0041] The efficacy score is based on the actual efficacy data of the drug in primary care clinical practice. For example, amlodipine, a high-tension drug, is rated A (excellent) and receives 40 points, while nifedipine is rated B (good) and receives 35 points. The safety score refers to the incidence of adverse drug reactions. For example, metoprolol has a low incidence of adverse reactions and receives 30 points, while reserpine has a high incidence of adverse reactions and receives 20 points. The cost-effectiveness score is based on the unit price of the drug and the medical insurance reimbursement ratio. For example, amlodipine costs 1.2 yuan per tablet and is fully reimbursed by medical insurance, receiving 20 points, while some imported antihypertensive drugs cost 5 yuan per tablet and are partially reimbursed by medical insurance, receiving 12 points. The availability score is based on the drug supply situation in primary healthcare institutions. Sufficient supply receives 10 points, tight supply receives 5 points, and shortage of supply receives 0 points. The system matches a corresponding drug entry for each diagnosis in the diagnostic set. It calculates a comprehensive score (including bonus points for synergistic use) for each drug based on recommendation rules, sorts the drugs by score from highest to lowest, and selects drugs with scores ≥60 to form the initial recommended drug candidate set for each diagnosis. It also indicates the compatibility of each drug with other diagnoses. For example, for the diagnostic set ["Primary Hypertension Grade 2 (Intermediate Risk)", "Type 2 Diabetes", "Coronary Artery Disease (Stable)"], amlodipine (comprehensive score 95, including 5 bonus points for synergistic use, which can help improve myocardial blood supply in coronary artery disease) and valsartan (comprehensive score 90, including 5 bonus points for synergistic use, with no adverse effects on blood sugar) are matched for hypertension; metformin (comprehensive score 88, including 3 bonus points for synergistic use, with no conflict with antihypertensive and cardioprotective drugs) and gliclazide (comprehensive score 82) are matched for coronary artery disease; and aspirin (comprehensive score 89, including 5 bonus points for synergistic use, which can prevent thrombosis and is compatible with antihypertensive and hypoglycemic drugs) is matched. The initial recommended drug candidate set was generated by integrating the following drugs: amlodipine, valsartan, metformin, aspirin, metoprolol, and gliclazide. Each drug was rated with a score, core recommendation criteria, and synergistic compatibility. For example, "Amlodipine: efficacy 40 points, safety 30 points, cost-effectiveness 20 points, availability 10 points, synergistic bonus 5 points, overall score 95 points. It is a first-line antihypertensive drug with sufficient supply at the grassroots level. It can help improve myocardial blood supply in patients with coronary heart disease and has no adverse effects on blood sugar."

[0042] For step S420, the system establishes a real-time data interface with the drug management system (PMS) of primary healthcare institutions. This interface synchronizes drug inventory data periodically (e.g., every 5 minutes) to ensure the timeliness of inventory information. Inventory information collection includes the unique code of each drug (National Drug Approval Number + Generic Name + Dosage Form + Specification), real-time inventory quantity (current warehouse inventory), expiration date information (near-expiration drugs are marked with remaining expiration date, such as "3 months remaining"), and inventory turnover rate. The inventory turnover rate is calculated using monthly turnover times, with the formula "Monthly Turnover Times = Monthly Drug Outbound Volume / Monthly Average Inventory Quantity". The system automatically calculates the turnover rate based on the outbound volume and average inventory of the past three months.

[0043] The inventory adequacy grading standards are preset by primary healthcare institutions based on their own diagnosis and treatment needs. Specifically: sufficient inventory means that the real-time inventory quantity is ≥3 times the average monthly outbound volume, and there are no drugs nearing their expiration date (remaining shelf life >6 months); tight inventory means that the real-time inventory quantity is between 1 and 3 times the average monthly outbound volume, or there are drugs nearing their expiration date (remaining shelf life 3-6 months); and insufficient inventory means that the real-time inventory quantity is <1 times the average monthly outbound volume, or there are drugs with serious near-expiration dates (remaining shelf life <3 months). For example, amlodipine (5mg*7 tablets / box) has a real-time inventory of 150 boxes, an average monthly outflow of 30 boxes, a turnover rate of 3 times / month, and a remaining shelf life of 12 months, which is considered sufficient inventory; valsartan (80mg*7 tablets / box) has a real-time inventory of 80 boxes, an average monthly outflow of 50 boxes, a turnover rate of 1.6 times / month, and a remaining shelf life of 4 months, which is considered tight inventory; metoprolol (25mg*20 tablets / box) has a real-time inventory of 20 boxes, an average monthly outflow of 40 boxes, a turnover rate of 0.5 times / month, and a remaining shelf life of 2 months, which is considered insufficient inventory. The system categorizes and organizes all drug inventory information according to hierarchical standards, generating a drug inventory status list.

[0044] For step S430, the system uses the unique drug code in the initial recommended drug candidate set as the association key to match it with the generated drug inventory status list, automatically retrieving real-time inventory data, expiration date information, inventory turnover rate, and inventory adequacy level for each candidate drug. During the matching process, the system handles synonymous drug names, such as uniformly identifying "amlodipine besylate tablets" and "amlodipine" as the same drug to avoid matching omissions; at the same time, for the same drug with multiple specifications, it matches the corresponding inventory information for each specification separately. For example, if the initial recommended drug candidate set is ["Amlodipine", "Valsartan", "Metoprolol", "Nifedipine"], the system will match and obtain the inventory information for each drug: Amlodipine (5mg*7 tablets / box), real-time inventory 150 boxes, remaining shelf life 12 months, turnover rate 3 times / month, sufficient inventory; Valsartan (80mg*7 tablets / box), real-time inventory 80 boxes, remaining shelf life 4 months, turnover rate 1.6 times / month, tight inventory; Metoprolol (25mg*20 tablets / box), real-time inventory 20 boxes, remaining shelf life 2 months, turnover rate 0.5 times / month, shortage inventory; Nifedipine (10mg*100 tablets / bottle), real-time inventory 100 bottles, remaining shelf life 10 months, turnover rate 2 times / month, sufficient inventory. After matching, the system will integrate the inventory information with the initial recommended drug candidate set to form a related data group of "drug name + specification + inventory information".

[0045] For step S440, the system filters out candidate drugs with sufficient inventory, retaining their original priority in the initial recommended drug candidate set. Priority is determined by the comprehensive score of the recommendation rules; a higher score indicates higher priority. Subsequently, the estimated supply duration is calculated based on real-time inventory quantity and average monthly outflow, using the formula: "Estimated supply duration (days) = (Real-time inventory quantity × Single box / bottle dosage / Average daily patient dosage) / (Average monthly outflow × Single box / bottle dosage / 30)". If the average daily patient dosage is not specified, it is calculated based on the adult standard dosage recommended in the drug's instructions. For example, the adult standard dosage of amlodipine is 5mg / day, with a single box containing 7 tablets of 5mg each, meaning each box can supply for 1.4 days. With a real-time inventory of 150 boxes and an average monthly outflow of 30 boxes (approximately 4.3 boxes / day), the estimated supply duration = 150 / 4.3 ≈ 35 days. While retaining priority, the system labels drugs with sufficient inventory with information such as "Real-time inventory balance: 150 boxes, Estimated supply duration: 35 days".

[0046] For step S450, the system presets the treatment duration for each diagnosis based on the generated set of treatment goals and standard treatment courses for common primary care diseases. For example, the initial treatment course for hypertension is 1 month, the initial treatment course for diabetes is 3 months, and the treatment course for acute bronchitis is 7-14 days. The patient's treatment needs for this visit are confirmed or entered by the doctor in the consultation room. If not entered, the system defaults to the standard treatment course for the corresponding diagnosis. The formula for calculating the number of available treatment courses is "Number of available treatment courses = (Real-time inventory quantity × Single box / bottle specification dosage / Patient's average daily dosage) / (Total dosage for standard treatment course)", where the total dosage for standard treatment course = Patient's average daily dosage × Treatment duration. The system compares the calculated number of available treatment courses with the patient's treatment needs for this visit. If the number of available treatment courses is greater than or equal to the required number, it means that the drug in short supply can meet the treatment needs for this visit, and the system retains the drug but marks it with an inventory warning message: "Stock shortage, remaining shelf life 4 months, it is recommended to replenish the stock in time." If the number of available treatment courses is less than the required number, it means that the drug cannot meet the full course of treatment, and the system moves it to the end of the initial recommended drug candidate set. At the same time, based on the drug guideline knowledge base, it searches for and associates with alternative drugs of the same category, with equivalent efficacy and sufficient stock. For example, if the available treatment courses for valsartan are 18.7 days < 30 days, the system moves it to the end of the initial recommended candidate set and associates it with the alternative drug "telmisartan (40mg*7 tablets / box)," which is an ARB antihypertensive drug with equivalent efficacy to valsartan and sufficient stock.

[0047] For step S460, the system identifies candidate drugs with stock shortages and removes them directly from the initial recommended drug candidate set to avoid treatment disruptions due to drug unavailability. Subsequently, based on the drug recommendation rules for the same category and indication in the drug guideline knowledge base, alternative drugs are searched. During the search, priority is given to drugs with similar efficacy and safety ratings to the original drug and sufficient stock. Simultaneously, the multi-dimensional scoring criteria of the recommendation rules are followed to ensure that the comprehensive score of the alternative drug is not lower than 60 points. For example, metoprolol (25mg*20 tablets / box) is a drug with stock shortages. After removing it, the system searches for β-blocker alternatives in the drug guideline knowledge base, prioritizing bisoprolol (5mg*10 tablets / box). This drug is a first-line antihypertensive drug with an efficacy rating of A, a good safety rating, a high reimbursement rate under medical insurance, and sufficient stock, achieving a comprehensive score of 89 points, meeting the recommendation requirements. The system marks the removed drug with stock shortages as "Stock Shortage, Removed" and inserts the selected alternative drugs into the corresponding positions in the final recommended drug candidate set according to the priority of the recommendation rules. If no suitable alternative medication is found, the system will pop up a notification box to inform the doctor, who can then manually select another medication to supplement the prescription. Ultimately, the system integrates readily available medications that meet the treatment requirements but are in short supply, along with the supplementary alternatives, to form a complete final recommended medication candidate set. For example, the final candidate set might be ["Amlodipine (Priority 1, Adequate Stock)", "Telmisartan (Priority 2, Alternative to Valsartan, Adequate Stock)", "Nifedipine (Priority 4, Adequate Stock)"], ensuring that the doctor's prescription is feasible.

[0048] For step S320, the positive collaborative optimization logic employs a multi-objective optimization algorithm. Using the comprehensive coverage of multiple diagnostic and treatment goals as the core indicator, it calculates the matching degree of each candidate drug to multiple diagnostic and treatment goals, prioritizing drugs that can simultaneously meet multiple diagnostic and treatment needs and have good synergistic effects. For example, for the diagnostic set ["Primary hypertension stage 2 (intermediate risk)", "Type 2 diabetes", "Coronary artery disease (stable)"], the treatment goal set is "blood pressure control target + blood glucose control target + relief of chest tightness and prevention of angina." Amlodipine can simultaneously lower blood pressure and assist in improving myocardial blood supply in coronary artery disease, covering the treatment goals of two diagnoses, with a matching score of 100. Aspirin can prevent coronary artery disease thrombosis and has no conflict between lowering blood pressure and blood glucose, with a matching score of 90. Metformin only meets the blood glucose lowering goal, with a matching score of 80. The system prioritizes drugs with high matching degrees for inclusion in the prescription combination, while also considering the core treatment needs of each diagnosis, ensuring that no diagnosis is missed.

[0049] The reverse conflict avoidance logic uses a contraindication rule matching library, which contains explicit contraindications for each drug (e.g., beta-blockers are contraindicated in patients with bronchial asthma, ACE inhibitors are contraindicated in patients with bilateral renal artery stenosis, and sulfonylureas are contraindicated in patients with severe hepatic or renal insufficiency). The system matches the contraindications of candidate drugs with the patient's allergy history, past medical history, current symptoms, and all diagnoses one by one. If a contraindication match is found (e.g., in patients with bronchial asthma complicated by hypertension and coronary heart disease, metoprolol is among the candidate drugs), the system immediately marks the drug as a "contraindicated drug" and automatically removes it from the candidate set. At the same time, it prompts the doctor that "this drug has a contraindication and has been removed. It is recommended to choose an alternative drug (e.g., bisoprolol, which has less impact on the bronchi)."

[0050] The drug interaction pre-screening logic is based on a drug interaction knowledge base, which integrates the "Guidelines for Clinical Drug Interactions" and commonly used drug interaction data at the primary care level. This knowledge base includes three levels of interaction relationships: absolute contraindications, relative contraindications, and caution. For multi-disease medication scenarios, it performs multi-directional matching on all drugs in the recommended drug candidate set to predict whether there are any absolute pharmacological contraindications (e.g., the combination of cephalosporins and alcoholic drugs can cause a disulfiram-like reaction, which is an absolute contraindication; the combination of antihypertensive drugs and hypoglycemic drugs requires caution regarding hypoglycemia and hypotension). If an absolute contraindication drug combination exists, the system automatically marks it and prompts the doctor, while simultaneously recommending adjustments to the drug combination (e.g., adjusting hypoglycemic drugs that may cause hypoglycemia to a more compatible combination with antihypertensive drugs) to ensure the safety of multi-disease medication use.

[0051] For step S330, the system integrates the logical processing results and selects drug combinations with no contraindications, no absolute drug interactions, and high positive synergistic matching as the core drug group for the suggested prescription draft. At the same time, it ensures that each diagnosis has a corresponding core treatment drug to avoid missing the treatment needs of any diagnosis. It is recommended that the initial dose be determined based on the drug instructions, primary care guidelines, the patient's key pathophysiological parameters, and the requirements for synergistic use of multiple drugs. Dosage compatibility between drugs should be considered to avoid adverse reactions caused by dose superposition. For example, the routine initial dose of amlodipine for adults is 5 mg / day. If the patient also has coronary heart disease or diabetes and has moderate drug metabolism, the initial dose should be adjusted to 2.5 mg / day to avoid hypotension and hypoglycemia caused by superposition with aspirin or metformin. For children, the initial dose should be calculated based on age and weight. For example, if a child has a cold complicated by mild pneumonia, the initial dose of acetaminophen is 10-15 mg / kg / dose, every 4-6 hours, not exceeding 4 times a day. This can be combined with cefaclor, a commonly used drug for pediatric pneumonia, with the dosage adjusted according to weight to avoid liver and kidney damage caused by superposition with acetaminophen. The rationale for the recommendation should be written in detail, taking into account the patient's individual condition of multiple coexisting diseases, the characteristics of the drug, the synergistic effect of medication, and the logical processing results. It should clearly explain the basis for choosing the drug and its suitability for each diagnosis. For example, "Amlodipine tablets, initial dose 2.5mg / day, once a day, recommended rationale: The patient is diagnosed with primary hypertension grade 2 (intermediate risk), type 2 diabetes, and coronary heart disease (stable). The drug has moderate drug metabolism capacity. This drug has excellent efficacy and high safety. It can simultaneously lower blood pressure and help improve myocardial blood supply in coronary heart disease. It has no adverse effect on blood sugar. There is no significant interaction with metformin and aspirin. Moreover, the stock is sufficient and is expected to meet the needs of this one-month treatment course. Metformin tablets, initial dose 0.5g / day, twice a day, recommended rationale: For the patient's diagnosis of type 2 diabetes, this drug has a stable hypoglycemic effect, few adverse reactions, and is synergistically suitable with amlodipine and aspirin. It does not affect blood pressure control and coronary heart disease treatment. The dosage is in line with the patient's drug metabolism capacity."

[0052] It is recommended that the prescription draft adopt a standardized prescription format, including fields such as patient basic information, multiple diagnoses (divided into primary / secondary), drug name, dosage form, specifications, initial dose, usage and dosage, and reasons for recommendation (indicating the appropriate diagnosis). After generation, it should be automatically saved to the system and displayed in the doctor's consultation interface for the doctor to view and modify.

[0053] For step S340, an individualized dosage calculation model is constructed. The core input parameters are the patient's baseline physiological parameters (age, weight, sex), derived physiological indicators (body surface area, estimated creatinine clearance, liver and kidney function classification), predicted drug metabolism capacity, and disease severity indices for each diagnosis. This is combined with the drug's pharmacokinetic characteristics (such as half-life, bioavailability, and metabolic pathway) and the characteristics of multi-drug synergistic metabolism to calculate the individualized dosage range for each drug. For drugs primarily metabolized by the kidneys (such as cephalosporins and aminoglycosides), the dosage range is adjusted based on the estimated creatinine clearance, while also considering the dosages of other drugs metabolized by the kidneys to avoid excessive metabolic burden on the kidneys. For example, for ceftazidime, when the creatinine clearance is ≥50 ml / min, the dosage range is 1-2 g / dose, every 8-12 hours. If the patient is also using metformin (partially metabolized by the kidneys), the dosage range is reduced to 0.5-1.5 g / dose, every 12 hours. For drugs primarily metabolized by the liver (such as erythromycin and chloramphenicol), the dosage range is adjusted according to liver function classification. For mild liver impairment, the dosage is halved; for moderate and severe impairment, the dosage is reduced to one-third or discontinued. Simultaneously, avoid concurrent use with other hepatically metabolized drugs. For scenarios involving multiple medications used in combination, additional consideration is given to the dosage effects between drugs. For example, when antihypertensive drugs are used with hypoglycemic drugs, the antihypertensive drug dosage needs to be appropriately reduced to avoid hypoglycemia; when cardioprotective drugs are used with antihypertensive drugs, both dosages need to be reduced to avoid hypotension. Furthermore, the dosage is further refined based on the patient's age, weight, and predicted drug metabolism capacity. For example, after calculating the dosage for children based on weight, it is adjusted according to age (the dosage for infants and young children is slightly lower than the weight-based calculation). For patients with weak drug metabolism capacity, the dosage range is reduced by 30%-50%, and the reduction percentage can be appropriately increased (not exceeding 60%) when using multiple medications. For example, a 68-year-old male, weighing 70 kg, with a creatinine clearance of 65 ml / min (mildly impaired renal function), moderate drug metabolism, diagnosed with stage 2 essential hypertension (intermediate risk), type 2 diabetes, and stable coronary artery disease, would have an individualized dosage range of 2.5-4 mg / day for amlodipine (2.5-5 mg / day for a single diagnosis of hypertension, but the upper limit is lowered due to the presence of diabetes and coronary artery disease); the individualized dosage range for metformin would be 0.5-1 g / day, to avoid the risk of hypotension and hypoglycemia caused by superposition with amlodipine. The system will mark the individualized dosage range for each drug after the corresponding drug entry in the suggested prescription draft, clearly indicating "Individualized dosage range: XX-XX (units / times / day)", and explaining the basis for dosage adjustment (including reasons for synergistic compatibility with multiple diseases).

[0054] For step S350, the system automatically compares the suggested initial dose of each drug in the draft prescription with the individualized dose range. If the initial dose is within the individualized dose range, it remains unchanged. If the initial dose is below the lower limit of the individualized dose range, the system prompts the doctor, "The initial dose is too low; it is recommended to adjust it to the individualized dose range (XX-XX)," and provides adjustment suggestions. If the initial dose is above the upper limit of the individualized dose range, the system automatically adjusts the dose to the upper limit and marks it as "The initial dose is too high; it has been automatically adjusted to the upper limit of the individualized dose range XX." After correction, a personalized prescription suggestion draft is generated. This draft retains all fields from the original draft prescription and adds an "Individualized Dose Adjustment Explanation," detailing the reasons and basis for dose adjustments for each drug. Subsequently, the system pushes the personalized prescription suggestion draft to the doctor's consultation interface for manual review. The doctor can manually adjust the drug type, dosage, and usage based on the patient's specific condition and clinical experience. After adjustment, the system automatically records the adjustment content and reasons. After the doctor confirms that the prescription is correct, he clicks the "Confirm Prescription" button and the system automatically generates a formal prescription. The formal prescription adopts an electronic prescription format that conforms to the standards of primary healthcare institutions and includes required fields such as prescription number, patient information, diagnosis information, drug information, dosage and administration, doctor's signature, and date of issuance. At the same time, it is automatically synchronized to the drug management system (PMS) and electronic health record (EHR).

[0055] Reference Figure 5 The specific steps for verifying the eligibility of formal prescriptions in S150 include S510-S540: S510, Check whether the formal prescription meets the format requirements; S520 involves sequentially verifying contraindications, identifying drug interactions, and identifying duplicate and contradictory medications. S530 compares the actual prescribed dose of each drug in the formal prescription with the maximum permissible dose of the individualized dose range. S540 generates audit conclusions and corresponding intervention measures based on the audit results.

[0056] Specifically, for step S510, the system pre-defines the electronic prescription format specifications for primary healthcare institutions, including three core modules: mandatory field verification, format consistency verification, and content completeness verification. Mandatory field verification checks for required fields such as prescription number, patient name, gender, age, medical card number, diagnosis, drug name, dosage form, specifications, dosage and administration, doctor's signature, and date of issuance. If any field is missing, the system will display a prompt box clearly informing the doctor, "Field XX is missing; please complete it." If a field is entered incorrectly (e.g., the patient's age is entered as "abc"), the system will automatically identify and prompt, "Age entered incorrectly; please enter a valid number." Format consistency verification checks for the standardization of drug specifications and dosage and administration. For example, drug specifications must be labeled "XXmg*XX tablets / box," and dosage and administration must be labeled "XX (dosage) / time, XX times / day." If the description is not standardized (e.g., "Amlodipine 5mg" without dosage instructions), the system will prompt, "Dosage and administration description is not standardized; please complete it." Content integrity verification checks the matching between diagnosis and medication. If the indication for a medication in a prescription does not match the diagnosis (e.g., a prescription for hypoglycemic drugs for a patient with hypertension), the system will prompt "Drug indication does not match diagnosis; please verify and adjust." After all verifications are completed, the system generates a formatted verification report, marked "Verification passed" or "Verification failed with suggestions for improvement." Only after verification passes can the subsequent review stage begin; if verification fails, the doctor must make corrections and resubmit for verification.

[0057] For step S520, the contraindication check is further refined based on S320, adopting a two-way check model of "drug-patient status-multiple diagnoses". On the one hand, it checks whether the contraindications of each drug in the prescription match the patient's allergy history, past medical history, current symptoms and all diagnoses. On the other hand, it checks whether the patient's special status (such as pregnant women, breastfeeding women, children, elderly people) and multiple coexisting diseases have corresponding drug use contraindications. For example, ribavirin is contraindicated in pregnant women, erythromycin is contraindicated in breastfeeding women, sulfonylurea hypoglycemic drugs are contraindicated in patients with diabetes mellitus and renal insufficiency, and beta-blockers are contraindicated in patients with hypertension and bronchial asthma. If the system finds a contraindication match, it immediately marks it and prompts the doctor "This drug has a contraindication (when it is suitable for XX diagnosis), it is recommended to adjust it", and at the same time provides alternative drug suggestions (suitable for all diagnoses).

[0058] Drug interaction identification (including pharmacokinetic and pharmacodynamic interactions) employs a hierarchical identification model. For multi-disease, multi-drug scenarios, it involves multi-directional identification of the entire drug combination. Pharmacokinetic interactions primarily examine the effects of drugs on absorption, distribution, metabolism, and excretion, as well as additive effects (e.g., omeprazole inhibits gastric acid secretion, which reduces itraconazole absorption; if a patient is using omeprazole, itraconazole, and metformin simultaneously, the cumulative metabolic effects of these three drugs on the liver need to be assessed). Pharmacodynamic interactions primarily examine the synergistic and antagonistic effects between drugs and their impact on multiple diagnoses (e.g., β-blockade). When used in combination with digitalis drugs, the myocardial depressant effect may be enhanced. If the patient also has hypertension or coronary heart disease, the dosage needs to be adjusted. When antihypertensive drugs are used in combination with hypoglycemic drugs, the risk of hypoglycemia and hypotension may be increased, requiring close monitoring. The system comprehensively identifies all drug combinations in the prescription based on the drug interaction knowledge base, and marks the interaction level (absolute contraindication, relative contraindication, use with caution) and risk warnings, such as "Amlodipine and metoprolol may enhance the antihypertensive effect and lead to hypotension. If the patient also has diabetes, the risk of hypoglycemia should be noted. Use with caution and it is recommended to monitor blood pressure and blood sugar."

[0059] The identification of duplicate and contradictory medication use mainly checks whether there are duplicate prescriptions containing drugs with the same or similar pharmacological effects (e.g., prescribing amlodipine and nifedipine simultaneously, both calcium channel blockers, constitutes duplicate medication use; prescribing metformin and gliclazide simultaneously, although both are hypoglycemic drugs, have different mechanisms of action and do not constitute duplicate medication use, but should be marked as synergistic hypoglycemic). It also checks for contradictions in drug usage and dosage, and contradictions in drug compatibility with multiple diagnoses (e.g., the prescription does not include the core treatment drug for a certain diagnosis, or the drug dosage does not match the synergistic needs of multiple diagnoses). If duplicate medication use, contradictory usage and dosage, or incompatibility exist, the system will mark it and prompt the doctor with "Duplicate medication use / contradictory usage and dosage / insufficient diagnostic compatibility exists, please adjust the prescription."

[0060] For step S530, the system automatically extracts the actual dosage of each drug in the formal prescription and retrieves the individualized dosage range of the drug calculated in S340, clarifying the maximum allowable dosage within the individualized dosage range. Subsequently, the actual dosage is compared with the maximum allowable dosage one by one, and handled in three ways: First, if the actual dosage is less than or equal to the maximum allowable dosage and within the individualized dosage range, the system marks it as "dosage compliant"; second, if the actual dosage is greater than the maximum allowable dosage, the system marks it as "dosage exceeds the limit" and prompts the doctor, "The actual dosage XX exceeds the maximum allowable individualized dosage XX, please adjust it to a reasonable range"; third, if the actual dosage is less than the lower limit of the individualized dosage range, the system marks it as "dosage too low" and prompts the doctor, "The actual dosage XX is lower than the lower limit of the individualized dosage XX, it is recommended to assess whether dosage adjustment is necessary." The comparison results are simultaneously included in the review report.

[0061] For step S540, the system integrates format verification results, check results, and dosage comparison results, and generates a review conclusion using a tiered review logic. The criteria for passing the review are: format verification passed, no contraindications, no absolute drug interactions, no medication duplication or contradictions, and the actual prescribed dosage is within the individualized dosage range. In this case, a "Review Passed" conclusion is generated, with the intervention measure being "It is recommended to follow the prescription and monitor medication use." The criteria for issuing a review warning are: format verification passed, but minor issues exist (such as dosage being too low but not affecting treatment, the existence of drug interactions requiring caution, no contraindications but attention should be paid to medication use in special populations). In this case, a "Review Warning" conclusion is generated, with the intervention measure being "Warning: XX issue (such as dosage being too low, it is recommended to assess and adjust; blood pressure monitoring is required when using XX medication), it can be followed as prescribed, and the patient's medication response should be closely observed." The criteria for issuing a warning / interception during the review process are: format validation failure, presence of contraindications, presence of absolute drug interactions, serious medication duplication or contradictions, or actual dosage significantly exceeding the maximum allowable dose (20% or more above the maximum permissible dose). In these cases, a "Warning / Interception" conclusion is generated, with the intervention measure being: "Warning: XX issue (e.g., contraindications, significantly excessive dosage), do not proceed with the current prescription. Please adjust the prescription immediately and resubmit for review." After the review conclusion and intervention measure are generated, they are automatically pushed to the doctor's consultation interface and medication dispensing interface. The doctor must adjust the prescription according to the warning / interception prompt and resubmit for review until it is approved. For review prompts, the doctor determines whether to adjust based on the clinical situation; resubmission is not required, but the prompt content and handling opinion must be recorded in the medical record.

[0062] Reference Figure 6 The specific steps for generating a personalized medication instruction sheet in S150 include S510-S540: S610, after approval, extracts the standard usage and common side effects of each drug in the official prescription from the drug information database; S620 integrates the core treatment rationale from the review results and the formal prescription; S630, based on the diagnosis set of this visit, additional dietary and exercise recommendations; S640 extracts the standard usage and common side effects of each drug from the official prescription, integrates the review results and core treatment rationale, and adds additional dietary and exercise recommendations to the medication instruction template to generate a personalized medication instruction for the patient.

[0063] Specifically, for step S610, after the formal prescription is approved, the system automatically retrieves the standard usage and common side effects of the corresponding drug from the drug information database using the drug's unique code as the association key. Standard usage includes the route of administration (oral, subcutaneous injection, intravenous drip, etc.), time of administration (before meals, after meals, on an empty stomach, etc.), handling for missed doses, and adjustments for special populations (pregnant women, children, the elderly). Common side effects include common adverse reactions (such as dizziness, nausea, diarrhea), rare adverse reactions (such as rash, dyspnea), and methods for managing side effects. For example, the standard usage for amlodipine retrieved is "oral, once daily, 5mg each time, before or after meals. If a dose is missed, and less than 12 hours have passed since the next scheduled dose, no additional dose is needed; simply take the medication as scheduled next time. Do not double the dose; the initial dose for the elderly should be halved." Common side effects include "common dizziness, headache, and facial flushing, usually occurring at the beginning of treatment and resolving spontaneously; rare rash and ankle edema, if these occur, discontinue use immediately and seek medical attention." The extracted information is automatically organized into structured text to ensure that the expression is concise, accurate, and in line with the comprehension ability of patients at the grassroots level.

[0064] For step S620, the system automatically extracts the aforementioned review conclusions and intervention measures, and at the same time extracts the core treatment rationale for each drug in the formal prescription, marks the diagnostic and synergistic advantages of each drug, and integrates the two to form a unified "Review and Treatment Instructions". During the integration process, the content is organized according to the logic of "review conclusion + intervention measures + core treatment rationale". For example, "Review conclusion: Review passed; Intervention measures: It is recommended to follow the prescription and monitor medication use; Core treatment rationale: 1. Primary hypertension grade 2 (intermediate risk): Amlodipine tablets, dose 2.5mg / day, within the individualized dose range, can simultaneously help improve myocardial blood supply in coronary heart disease, has no adverse effects on blood sugar, and has sufficient stock; 2. Type 2 diabetes: Metformin tablets, dose 0.5g / day, has a stable hypoglycemic effect, has no significant interaction with amlodipine and aspirin, and is consistent with the patient's drug metabolism capacity; 3. Coronary heart disease (stable): Aspirin enteric-coated tablets, dose 100mg / day, can prevent thrombosis, relieve chest tightness symptoms, and is synergistically compatible with antihypertensive and hypoglycemic drugs; Synergistic medication instructions: There is no absolute interaction between all drug combinations, the dose is suitable for the coexistence of multiple diseases, and can simultaneously cover the treatment goals of three diagnoses. Attention should be paid to monitoring the risk of hypotension and hypoglycemia."

[0065] For step S630, a database of dietary and exercise recommendations for common diseases at the grassroots level is constructed. The recommendations are stored according to diagnosis, with the database content referencing the *Chinese Dietary Guidelines* and the *Guidelines for the Health Management of Chronic Diseases at the Grassroots Level*, and designed in conjunction with the lifestyle habits of patients at the grassroots level. For example, for the diagnostic set ["Grade 2 Primary Hypertension (Intermediate Risk)", "Type 2 Diabetes", "Coronary Artery Disease (Stable)"], the dietary recommendations are: "A low-salt, low-sugar, and low-fat diet; daily salt intake not exceeding 5g; daily added sugar intake not exceeding 25g; reduce intake of fatty meat, animal organs, fried foods, sweets, and pickled products; eat more fresh vegetables (more than 500g daily), low-sugar fruits (200-350g daily, such as apples and grapefruits), whole grains, and high-quality protein (such as lean meat, eggs, and soy products); quit smoking and limit alcohol consumption; daily alcohol intake not exceeding 25g for men and 15g for women; avoid drinking on an empty stomach (to prevent hypoglycemia and hypotension)." The recommended exercise regimen is "moderate, gentle aerobic exercise, such as brisk walking, jogging, Tai Chi, or square dancing, 3-5 times a week, about 30 minutes each time, avoiding strenuous exercise; monitor blood pressure and blood sugar during exercise, and stop exercising immediately and rest if symptoms such as dizziness, chest tightness, palpitations, or sweating occur; choose to exercise 1-2 hours after meals to avoid hypoglycemia caused by exercising on an empty stomach, and avoid exercising during the morning peak blood pressure period (to prevent coronary heart disease attacks)." If the patient is diagnosed with "hypertension + diabetes," the dietary recommendations focus on low salt and low sugar, and the exercise recommendations are to avoid strenuous exercise and monitor blood pressure and blood sugar. If the diagnosis is "hypertension + coronary heart disease," the dietary recommendations focus on low salt and low fat, and the exercise recommendations are mainly gentle exercise, avoiding fatigue. The system automatically matches corresponding synergistic dietary and exercise recommendations based on multiple diagnostic sets, prioritizing synergistic recommendations for multiple disease combinations, while supplementing specific recommendations for each individual diagnosis.

[0066] For step S640, a pre-set template for personalized medication guidance in primary care is provided. This template includes fields such as patient basic information (name, gender, age, medical card number), diagnosis information, prescription drug details (drug name, specifications, dosage), standard drug usage, common side effects and their management, review and treatment instructions, dietary recommendations, exercise recommendations, and precautions (e.g., regular follow-up appointments, follow-up visits for discomfort). The system automatically fills in the standard drug usage and common side effects extracted in S610, the review and treatment instructions integrated in S620, the dietary and exercise recommendations added in S630, and the drug details, patient basic information, and diagnosis information from the formal prescription, one by one, according to the template fields. For example, the "Drug Details" field is filled with "Amlodipine Tablets, 5mg*7 tablets / box, Dosage: 0.5 tablet (2.5mg) once daily, orally, Contraindicated Diagnoses: Primary Hypertension Grade 2 (Intermediate Risk), Coronary Artery Disease (Stable)", the "Standard Dosage" field is filled with the corresponding extracted content, the "Precautions for Multiple Drug Use" field is filled with targeted tips, and the "Diet and Exercise Recommendations" field is filled with suggestions for coordinating multiple diseases. After completion, the system automatically optimizes the layout, using clear paragraphs and bolding of key content (such as drug name, dosage, contraindicated diagnoses, precautions for multiple drug use, and other precautions), while also indicating the date the medication instruction sheet was generated and the prescribing doctor's name.

[0067] After generating a personalized medication instruction sheet, the system automatically pushes the content to the clinic's high-definition display screen via the clinic's display interface. The sheet is presented in a scrolling or static display format, facilitating doctors' explanation of medication-related knowledge to patients. Patients can also directly view the content on the display screen to understand drug usage, side effects, and dietary and exercise recommendations. Simultaneously, the system automatically generates a corresponding QR code containing the complete medication instruction sheet content. Using a high-definition QR code generation algorithm, the QR code is ensured to be clear and scannable, and includes tamper-proof markings to prevent content modification. The generated QR code is displayed on the clinic's display screen and also printed in the upper right corner of the paper medication instruction sheet. Patients can scan the code with their mobile phones to save the medication instruction sheet to their phones for easy access later. Furthermore, the system supports pushing the medication instruction sheet to patients' mobile phones via SMS, WeChat official accounts, etc., allowing patients to choose their preferred viewing method to ensure effective delivery of medication guidance information, helping patients use medication correctly and improve treatment outcomes.

[0068] Based on the above method embodiments, the second embodiment of this application discloses a primary healthcare prescription intelligent generation system. The primary healthcare prescription intelligent generation system of this application can implement any of the above-described primary healthcare prescription intelligent generation methods, and the specific working process of each module in the primary healthcare prescription intelligent generation system can refer to the corresponding process in the above method embodiments.

[0069] For ease of understanding, an example is as follows: A primary healthcare prescription intelligent generation system includes: The voice processing module is used to collect and analyze doctor-patient dialogue content in real time and extract diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history. The parameter processing module is used to extract key pathophysiological parameters based on diagnostic information. The medical record processing module is used to generate current diagnosis and treatment information based on key pathophysiological parameters and populate it into the medical record template. The current diagnosis and treatment information includes the set of diagnoses for this visit and the set of treatment goals for each diagnosis. The prescription generation module is used to generate a draft of a personalized prescription suggestion based on the current medical information, and to form a formal prescription after the doctor reviews and confirms it; The prescription review module is used to review the eligibility of formal prescriptions and generate personalized medication guidance sheets after the review is approved.

[0070] The third embodiment of this application provides a computer device, which may include a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement a method for intelligent generation of primary healthcare prescriptions.

[0071] The memory can communicate with the processor via a communication bus, which can be an address bus, a data bus, a control bus, etc.

[0072] Additionally, the memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device.

[0073] Furthermore, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0074] The fourth embodiment of this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as a method for intelligent generation of primary healthcare prescriptions.

[0075] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0076] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for intelligently generating prescriptions for primary healthcare, characterized in that, include: The system collects and analyzes doctor-patient dialogue content in real time to extract diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history. Based on the diagnostic information, extract key pathophysiological parameters; Based on the key pathophysiological parameters, current diagnosis and treatment information is generated and populated into the medical record template. The current diagnosis and treatment information includes the set of diagnoses for this visit and the set of treatment goals for each diagnosis. Based on the current medical information, a draft personalized prescription suggestion is generated, and after the doctor reviews and confirms it, a formal prescription is formed. The formal prescription is reviewed for eligibility, and a personalized medication guidance sheet is generated upon successful review.

2. The method for intelligent generation of prescriptions for primary healthcare as described in claim 1, characterized in that, The specific steps for generating a draft personalized prescription suggestion based on the current medical information include: Based on the built-in recommendation rules and drug guide knowledge base, a corresponding set of recommended drug candidates is generated for each diagnosis in the diagnostic set. The recommendation rules include efficacy, safety, cost-effectiveness and availability. Based on the recommended drug candidate set and the treatment target set for each diagnosis association, the positive synergistic optimization logic, the reverse conflict avoidance logic, and the drug interaction pre-screening logic are executed sequentially. The positive synergistic optimization logic indicates that the candidate drugs can simultaneously cover multiple treatment targets. The reverse conflict avoidance logic indicates whether the contraindications of all candidate drugs match the patient's condition. The drug interaction pre-screening logic indicates that at the recommended combination level, it assesses whether there are pharmacological interactions between candidate drugs. The integrated logic processing results generate a suggested prescription draft, which includes a set of suggested drugs and a suggested initial dose and rationale for each drug.

3. The method for intelligent generation of prescriptions for primary healthcare according to claim 2, characterized in that, The steps following generating the suggested prescription draft also include: For each drug in the proposed prescription draft, an individualized dosage range is calculated based on the patient's individual parameters; The draft prescription is revised based on the individualized dosage range corresponding to each drug to obtain a personalized prescription recommendation draft.

4. The method for intelligent generation of prescriptions for primary healthcare according to claim 3, characterized in that, The specific steps for conducting a conformity review of the formal prescription include: Check that the formal prescription meets the format requirements; The following steps are performed sequentially: contraindication verification, drug interaction identification, and identification of duplicate and contradictory medication use. Compare the actual prescribed dose of each drug in the formal prescription with the maximum permissible dose of the individualized dose range; Based on the audit results, an audit conclusion and corresponding intervention measures are generated.

5. The method for intelligent generation of prescriptions for primary healthcare as described in claim 1, characterized in that, The specific steps for generating a personalized medication instruction sheet include: After the review is approved, the standard usage and common side effects of each drug in the formal prescription will be extracted from the drug information database. Integrate the review results with the core treatment rationale in the formal prescription; Based on the diagnosis presented during this visit, additional dietary and exercise recommendations are provided. The standard usage and common side effects of each drug extracted from the official prescription, the integrated review results and core treatment rationale, as well as additional dietary and exercise recommendations, are filled into the medication guidance template to generate a personalized medication guidance sheet for the patient.

6. The method for intelligent generation of prescriptions for primary healthcare according to claim 2, characterized in that, The specific steps for generating a corresponding set of recommended drug candidates for each diagnosis in the diagnostic set, based on the built-in recommendation rules and drug guide knowledge base, include: Based on the built-in recommendation rules and drug guide knowledge base, an initial set of recommended drug candidates is generated for each diagnosis in the diagnostic set; Retrieve current inventory information for all drugs and categorize them into three levels based on inventory adequacy: sufficient inventory, tight inventory, and shortage inventory; the inventory information includes real-time inventory data, expiration date information, and inventory turnover rate. For each candidate drug in the initial recommended drug candidate set, match the corresponding inventory information; For candidate drugs with sufficient inventory, they are retained in the initial recommended drug candidate set with priority, and the real-time inventory balance and expected supply duration are marked. For candidate drugs with tight inventory, the number of available treatment courses is calculated based on the patient's treatment needs for this visit and the inventory turnover rate of the candidate drug. If the number of available treatment courses meets the treatment needs for this visit, the candidate drug is retained and an inventory warning is issued. If the treatment needs for this visit cannot be met, the drug is moved to the end of the initial recommended drug candidate set and associated with alternative drugs. For candidate drugs with inventory shortages, they are removed from the initial recommended drug candidate set, and alternative drugs are added based on the drug guide knowledge base.

7. The method for intelligent generation of prescriptions for primary healthcare as described in claim 1, characterized in that, Based on the diagnostic information, the specific steps for extracting key pathophysiological parameters include: Extract symptom descriptions, durations, frequency of attacks, and accompanying symptoms from the chief complaint and present medical history, and extract the patient's basic physiological parameters, which include at least age, weight, and gender. Analyze allergy history and past medical history information, and record in a structured manner the name of the allergic drug, the type of allergic reaction, and the name and stage of the diagnosis of previous chronic diseases; Based on the symptom description and the basic physiological parameters, derived physiological indicators are calculated. The derived physiological indicators include at least body surface area, estimated creatinine clearance rate, and potential liver and kidney function classification. Based on the duration, frequency of attacks, and accompanying symptoms, a clinical scoring scale was matched to calculate the severity index and the probability of complication risk. Based on the correlation analysis between past medical history and current symptoms, predictive values ​​of organ function impairment level and drug metabolism capacity are generated. By integrating the aforementioned basic physiological parameters, derived physiological indicators, disease severity index, complication risk probability, organ function impairment level, and drug metabolism capacity prediction values, key pathophysiological parameters are generated.

8. A primary healthcare prescription intelligent generation system, characterized in that, The method for intelligent generation of primary healthcare prescriptions as described in any one of claims 1 to 7 includes: The voice processing module is used to collect and analyze the doctor-patient dialogue content in real time and extract diagnostic information, including chief complaint information, present illness history, allergy history, and past medical history. The parameter processing module is used to extract key pathophysiological parameters based on the diagnostic information. The medical record processing module is used to generate current diagnosis and treatment information based on the key pathophysiological parameters and fill it into the medical record template. The current diagnosis and treatment information includes the set of diagnoses for this visit and the set of treatment goals for each diagnosis. The prescription generation module is used to generate a personalized prescription suggestion draft based on the current diagnosis and treatment information, and to form a formal prescription after the doctor reviews and confirms it; The prescription review module is used to review the eligibility of the formal prescription and generate a personalized medication guidance sheet after the review is passed.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent prescription generation method for primary healthcare as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The system contains a computer program that can be loaded by a processor and executed as described in any one of claims 1 to 7 for intelligent generation of primary healthcare prescriptions.

Images