A prescription intelligent auditing method and system based on multi-dimensional information

Abstract

The present application belongs to the technical field of medical information technology and clinical rational drug use management, and specifically relates to a prescription intelligent auditing method and system based on multi-dimensional information, comprising the following steps: obtaining prescription related data and constructing a structured prescription judgment context; converting multi-source medical evidence into structured auditing rules with evidence level identification and risk action mode, and triggering applicable structured auditing rules based on the prescription judgment context to generate risk markers; based on the risk markers, combining disease diagnosis and treatment stage information and medication purposes, performing two-axis comprehensive evaluation on the prescription on the safety evaluation axis and the effectiveness evaluation axis to generate a comprehensive rationality evaluation result; the problems of the existing prescription auditing mode, such as dependence on artificial experience, non-uniform standards, single auditing dimension, neglect of patient individual and disease stage dynamic information, rule update lag, evidence management confusion, emphasis on safety interception and lack of effectiveness comprehensive evaluation, are solved.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of medical information technology and clinical rational drug use management, specifically involving a prescription intelligent review method and system based on multidimensional information. Background Technology

[0002] The rationality review of prescriptions and medical orders is a core element in ensuring patient medication safety, improving treatment outcomes, and optimizing the utilization of medical resources. With the surge in medical data and the increasing complexity of clinical treatment protocols, traditional prescription review models face severe challenges, and existing technologies have the following limitations: Firstly, the manual experience-based review model relies heavily on the pharmacist's personal professional knowledge and experience. While it offers some flexibility, it suffers from issues such as strong subjectivity, inconsistent standards, and low efficiency. This makes it difficult to meet the needs of large-scale and highly complex prescription reviews, and the review quality cannot be standardized and consistent.

[0003] Secondly, electronic prescription review systems based on static rule bases significantly improve efficiency by automatically reviewing prescriptions through preset rules. However, their rule bases are primarily built upon drug instructions, leading to outdated updates and difficulty in integrating the latest evidence-based medicine. More importantly, their review logic is typically drug-centric, failing to adequately consider or rigidly process dynamic contextual information such as individual patient characteristics (e.g., age, liver and kidney function, comorbidities) and specific stages of disease treatment (e.g., perioperative period, critical illness period). This results in poor adaptability of review results in complex clinical scenarios, often leading to the dilemma of either overloaded alerts or missed detections.

[0004] Third, research on prescription risk prediction based on machine learning has recently attempted to use machine learning models to predict the risk of irrational drug use from historical data. However, these models are mostly black-box predictions, and the output results (such as the probability of irrationality) are difficult to directly translate into clinically actionable review actions or graded intervention strategies. They have low integration with existing prescription review workflows and lack interpretability of the decision-making basis, which limits their reliable application in key medical decision support scenarios.

[0005] Fourth, the fragmented prescription quality and safety indicator system has multiple indicators for specific drugs (such as opioids) or specific risks. However, these indicators are often isolated and have a narrow scope of application, lacking a unified framework that can integrate across drugs, patients, and disease stages, thus failing to form a systematic closed loop for review and risk management.

[0006] In summary, existing technologies share the following common problems: a single review dimension, failing to systematically integrate drug-patient-disease three-dimensional information; multiple sources of review evidence but chaotic management, lacking a unified evidence grading and conflict resolution mechanism; an emphasis on safety interception, with insufficient comprehensive evaluation of treatment effectiveness and clinical suitability; and a disconnect between risk identification results and clinical review and handling processes, hindering accurate risk stratification and resource allocation. Therefore, clinicians urgently need an intelligent prescription review technology solution that can deeply integrate multi-dimensional information, manage multi-source evidence, and achieve comprehensive evaluation and tiered decision-making to improve the level of rational drug use management. Summary of the Invention

[0007] To address the problems of existing prescription review models, such as reliance on human experience and inconsistent standards; single review dimensions that ignore dynamic information about individual patients and disease stages; lagging rule updates and chaotic evidence management; emphasis on safety interception and lack of comprehensive effectiveness evaluation; and the disconnect between risk identification results and clinical treatment procedures, making them difficult to implement, this invention provides a prescription intelligent review method and system based on multi-dimensional information.

[0008] This invention is achieved using the following technical solution: A prescription intelligent review method based on multidimensional information includes the following steps: Acquire prescription-related data and construct a structured prescription judgment context, which includes at least drug dimension information, patient dimension information, and disease and treatment dimension information; Multi-source medical evidence is transformed into structured review rules with evidence level identifiers and risk action modes, and based on the prescription judgment context, the applicable structured review rules are triggered to generate risk markers; Based on the aforementioned risk markers, and combined with information on the disease diagnosis and treatment stage and the purpose of medication, a dual-axis comprehensive evaluation of the prescription is conducted on the safety evaluation axis and the efficacy evaluation axis to generate a comprehensive rationality evaluation result. Based on the comprehensive rationality evaluation results, the prescriptions are ranked and stratified by risk to determine their risk levels; Based on the risk level, prescriptions are assigned to the corresponding review and processing channels, and pharmacists make the final review decisions based on the risk warning information provided by the system.

[0009] Furthermore, the drug dimension information includes at least the drug identifier, generic name, specifications, manufacturer, route of administration, single dose, frequency of administration, course of treatment, and drug risk attribute information; the patient dimension information includes at least age, gender, height, weight, body surface area, comorbidity information, allergy history, history of adverse drug reactions, and vital sign data; the disease and treatment dimension information includes at least diagnostic information, type of visit, hospitalization schedule, disease treatment stage information, whether the patient is in the perioperative period or intensive care unit, medication purpose label, and medication-related laboratory tests and examination data.

[0010] Furthermore, the multi-source medical evidence includes drug instructions, clinical guidelines, evidence-based medicine literature, expert consensus, and internal regulations of medical institutions; the evidence level is predefined into nine levels from level 0 to level 8, wherein: Level 8 corresponds to evidence sources such as explicit contraindications in the drug instructions, prohibited combinations and interactions, mandatory prohibitions under national laws and regulations, or national safety warnings. Its risk effect is strongly binding, and the corresponding review and handling action is direct system interception. Level 7 corresponds to evidence sources such as serious warnings in drug instructions, situations explicitly not recommended by authoritative guidelines, key deduction items in national rational drug use assessments, or clear serious interactions. Its risk mechanism is strongly binding, and the corresponding review and handling action is to forcibly transfer it to high-priority manual review. Level 6 corresponds to evidence sources such as non-recommended protocols in clinical guidelines, significant risks indicated by evidence-based literature, use outside the scope of policy restrictions, or violation of explicit restrictions in the instructions. The risk effect is either suggestive or strongly binding, and the corresponding review and handling action is to transfer to manual review. Level 5 corresponds to evidence sources such as guidelines recommending caution in use, risk warnings of insufficient evidence consistency, risk monitoring indicators, or off-label drug use with evidence-based support. Its risk mechanism is warning-type, and the corresponding review and handling action is to transfer to manual review. The sources of evidence for Level 4, Level 3, Level 2, Level 1, and Level 0 are respectively normative recommendations, efficacy optimization tips, management cost-effectiveness tips, general precautions, and informational rules. Their risk action is either suggestive or risk-free judgment, and the corresponding review and handling actions are system release and recording or only for information recording.

[0011] Furthermore, the process of transforming multi-source medical evidence into structured review rules with evidence level identifiers and risk action modes also includes: The candidate rules were evaluated for inclusion / exclusion, clinical rationality, and operability through a two-round Delphi consultation method conducted by multidisciplinary experts. Correlation analysis and dimensionality reduction of the rule features are performed using principal component analysis and cluster analysis to eliminate rule redundancy.

[0012] Furthermore, the biaxial comprehensive evaluation includes: On the safety evaluation axis, based on the evidence level corresponding to the risk marker, prescriptions that trigger Level 8 or Level 7 rules are marked as having a clear medication safety risk, prescriptions that trigger Level 6 or Level 5 rules are marked as having a potential medication safety risk, and prescriptions that trigger Level 4 or lower rules are marked as having a low safety risk. On the efficacy evaluation axis, prescriptions are marked as treatment efficacy fit, efficacy fit is insufficient, or efficacy evidence is insufficient by judging the degree of match between the drug indication and the current diagnosis and disease stage, as well as the consistency between the purpose of medication and the recommended treatment goals.

[0013] Furthermore, the risk ranking and stratification of prescriptions based on the comprehensive rationality evaluation results specifically includes: classifying prescriptions marked as having clear medication safety risks and matching treatment effectiveness as high-risk; classifying prescriptions marked as having potential medication safety risks or insufficient matching effectiveness as medium-risk; and classifying prescriptions marked as having low safety risks and matching treatment effectiveness as low-risk.

[0014] Furthermore, the review and processing channels include an automatic pass channel, a pharmacist review channel, and an interception channel; wherein, prescriptions with low risk levels are assigned to the automatic pass channel, prescriptions with medium risk levels are assigned to the pharmacist review channel, prescriptions with high risk levels that trigger level 8 rules are assigned to the system interception channel for direct interception, and prescriptions that trigger level 7 rules are assigned to the pharmacist review channel and given the highest priority prompt.

[0015] Furthermore, the method also includes a feedback learning step: Record the pharmacist's review and decision-making results and the physician's adoption of intervention recommendations; Collect clinical outcome information related to prescriptions, including adverse reaction records, length of hospital stay, drug costs, medical expenses, drug ratio, and disease outcome; Based on the review decision results, adoption status, and clinical outcome information, optimize the weights or thresholds in the risk ranking and stratification strategies.

[0016] A prescription intelligent review system based on multidimensional information includes: The information construction module is used to acquire prescription-related data and construct a structured prescription judgment context, which includes drug dimension information, patient dimension information and disease and treatment dimension information. The rules engine module is used to store structured review rules that transform multi-source medical evidence, and to trigger applicable rules based on the prescription judgment context to generate risk markers; The comprehensive evaluation module is used to perform a dual-axis comprehensive evaluation of the prescription based on the risk markers, on the safety evaluation axis and the efficacy evaluation axis, and generate a comprehensive rationality evaluation result. The risk stratification module is used to rank and stratify prescriptions based on the comprehensive rationality evaluation results, and determine their risk levels. The decision allocation module is used to allocate prescriptions to the corresponding review and processing channels according to the risk level, and output risk warning information for pharmacists to conduct final review.

[0017] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method described in this invention.

[0018] This invention provides a prescription intelligent review method and system based on multi-dimensional information, which has the following advantages compared with the prior art: First, by constructing a structured prescription judgment context that integrates drugs, patients, and diseases, it has achieved a leap from single drug review to a panoramic clinical review, fundamentally solving the problem of prompt overload or missed risk detection caused by the fragmentation of information dimensions in traditional technologies.

[0019] Second, by establishing a hierarchical governance and structured rule transformation mechanism for multi-source medical evidence, and predefining evidence level labels from level 0 to level 8 and their corresponding risk effects, a dynamic knowledge system with clear responsibilities and maintainability has been constructed, solving the problems of chaotic sources of review rules, delayed updates, and difficulty in adjudicating conflicts.

[0020] Third, an innovative dual-axis comprehensive evaluation mechanism for safety and efficacy is introduced, which enables the review conclusions to take into account both medication safety and treatment benefits, thus solving the problem that traditional review focuses too much on safety interception and lacks judgment on treatment suitability.

[0021] Fourth, based on the comprehensive evaluation results, the risk quantification and hierarchical handling of prescriptions are realized, and different risk levels (high, medium and low) are seamlessly connected to differentiated review and processing channels (automatic approval, pharmacist review, and system interception), which solves the problems of mismatch of review resources and difficulty in implementing risk identification results.

[0022] Fifth, it establishes a credible human-machine collaborative boundary for system early warning and pharmacist decision-making. The system provides interpretable prompts based on clear evidence levels and risk markers, while pharmacists retain the final decision-making power, ensuring the safety and controllability of the medical decision-making process and the implementation of professional responsibilities. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0024] A prescription intelligent review system based on multidimensional information includes: The information construction module is used to acquire prescription-related data and construct a structured prescription judgment context, which includes drug dimension information, patient dimension information and disease and treatment dimension information. The rules engine module is used to store structured review rules that transform multi-source medical evidence, and to trigger applicable rules based on the prescription judgment context to generate risk markers; The comprehensive evaluation module is used to perform a dual-axis comprehensive evaluation of the prescription based on the risk markers, on the safety evaluation axis and the efficacy evaluation axis, and generate a comprehensive rationality evaluation result. The risk stratification module is used to rank and stratify prescriptions based on the comprehensive rationality evaluation results, and determine their risk levels. The decision allocation module is used to allocate prescriptions to the corresponding review and processing channels according to the risk level, and output risk warning information for pharmacists to conduct final review.

[0025] A prescription intelligent review method based on multidimensional information is proposed. This method is implemented based on the above system, and its core process is shown in the appendix. Figure 1 As shown, the main steps include: Step S1: Obtain prescription-related data and construct a structured prescription judgment context.

[0026] This step corresponds to the information construction module.

[0027] The system comprehensively collects basic data related to prescription rationality evaluation from the Hospital Information System (HIS), Electronic Medical Record System (EMR), Laboratory Information System (LIS), Picture Archiving and Communication System (PACS), and Nursing Information System (NIS). After cleaning and standardization, the data is uniformly structured into a prescription judgment context, which serves as the unified input for all subsequent review and evaluation modules.

[0028] The prescription judgment context includes at least the following three dimensions of structured information: a. Drug-related information: This includes at least the drug identifier (ID), generic name / ingredients, specifications, manufacturer, route of administration, single dose, frequency of administration, course of treatment, drug category attributes, and high-risk attribute labels (such as whether it is a key monitored drug).

[0029] b. Patient-specific information: including at least age, gender, height, weight, body surface area (BSA), comorbidity set (such as hypertension, diabetes), allergy history, history of adverse drug reactions, vital sign status (derived from data such as body temperature, blood pressure, and heart rate, such as fever / no fever), and blood glucose / blood pressure control status.

[0030] c. Disease and treatment information: This includes at least the primary / secondary diagnosis, past medical history / family history labels, type of visit (outpatient / inpatient), hospitalization schedule, disease treatment stage information (e.g., acute phase, remission phase), treatment scenario label (whether in the perioperative period or intensive care), medication purpose label (treatment / prevention / symptomatic / supportive), and medication-related laboratory tests and examination data.

[0031] Among them, medication-related laboratory test indicators (such as liver function, kidney function, blood routine, coagulation function, electrolytes) and vital signs data are further quantified or discretized into status labels that can be directly called by the rule engine, thereby dynamically and objectively reflecting the patient's real-time physiological and pathological state.

[0032] Output: A structured prescription decision context object containing a set of all key fields across three dimensions: drug, patient, disease, and treatment.

[0033] Step S2: Transform multi-source medical evidence into structured review rules and generate risk markers based on context-triggered rules.

[0034] This step corresponds to the rules engine module.

[0035] The system establishes a governable and maintainable evidence grading rule engine. Its core is to transform diverse medical evidence of varying strengths into structured review rules with clear impact and that can be automatically executed by the system.

[0036] a. Source of the rules: The multi-source medical evidence on which the review rules are based includes: drug instructions approved by the National Medical Products Administration (which have legal effect), clinical diagnosis and treatment guidelines and standards issued by national or industry authoritative institutions, high-quality evidence-based medical literature (such as systematic reviews and randomized controlled trials), expert consensus and internal medication guidelines of medical institutions.

[0037] b. Evidence Level Identification and Structured Rule Definition: The system predefines evidence into nine consecutive levels from level 0 to level 8 (as defined in claim 3) based on its authority, standardization, and risk certainty. Each review rule is defined as a rule unit containing the following elements: Triggering conditions: Describes the prescription context in which the rule is activated, which is a logical combination of conditions from three dimensions: drug, patient, and disease.

[0038] Scope of application: Defines the effective boundaries of the rules (such as specific populations or medical scenarios).

[0039] Evidence level label: Reflects the specific level of the evidence on which the rule is based in the 0-8 level system.

[0040] Risk Action Mode: The strength of the rule's effect is determined based on the evidence level. Levels 8 and 7 correspond to a strong binding effect, used to identify clear and definite medication risks; Level 6 can correspond to either a strong binding or advisory effect, depending on the specific rules; Levels 5, 4, and below correspond to an advisory effect or a no-risk assessment, used to identify potential risks, rationale issues, or simply provide information.

[0041] c. Rule Governance and Consolidation: To ensure the clinical rationality and engineering usability of the rule system, a governance mechanism is introduced: Candidate rules are evaluated and confirmed by multidisciplinary experts from pharmacy, clinical practice, and nursing using a multi-round consensus method (such as the Delphi method); Principal component analysis (PCA), cluster analysis, and other methods are used to perform correlation analysis and dimensionality reduction on rule features to eliminate redundancy, retain core rule combinations, and improve engine operating efficiency.

[0042] d. Engine Operation and Risk Tag Generation: During actual review, the rule engine receives the prescription judgment context as input. The engine filters out all applicable rules and determines whether they are activated based on the triggering conditions. When a rule is activated, the engine does not directly adjudicate whether the prescription is reasonable, but generates a corresponding risk tag based on the rule's evidence level and risk action mode. For example, it generates a risk message tagged "Level 8 - Strong Constraint: Contraindicated in patients with hepatic impairment" or "Level 5 - Warning: Use with caution in elderly patients." All generated risk tags are passed as input to the next module.

[0043] Step S3: Based on risk labeling, conduct a comprehensive evaluation on both the safety and effectiveness axes.

[0044] This step corresponds to the comprehensive evaluation module.

[0045] This module merges and processes risk markers from the rules engine, and incorporates information on the disease diagnosis and treatment stage and the purpose of medication, to comprehensively evaluate prescriptions on the following two evaluation axes: a. Safety evaluation axis: The risk is classified according to the level of evidence corresponding to the risk marker.

[0046] If a prescription triggers a Level 8 or Level 7 rule (strong binding type), it is marked as a clear medication safety risk.

[0047] If a prescription triggers a Level 6 or Level 5 rule (indicative or strongly binding), it is marked as a potential medication safety risk.

[0048] If a prescription only triggers a level 4 or lower rule or has no related risk, it is marked as having low safety risk. b. Efficacy Evaluation Axis: By combining information on the stage of disease diagnosis and treatment (such as perioperative preoperative prevention) and the drug's intended use label (such as treatment / prevention), the therapeutic suitability of the drug in the current clinical context can be determined.

[0049] The therapeutic effectiveness of a prescription is evaluated by assessing the degree to which the drug's indications match the current diagnosis and disease stage, and by the consistency between the purpose of medication and the treatment goals recommended in guidelines or evidence-based medicine.

[0050] The output results are: treatment effectiveness fit, inadequate effectiveness fit, or insufficient evidence of effectiveness.

[0051] c. Dual-axis mapping and integrated output: The safety evaluation results and the effectiveness evaluation results are combined and mapped to generate a comprehensive rationality evaluation output.

[0052] This output is not limited to a simple binary conclusion of reasonable / unreasonable, but rather comprehensively reflects the overall reasonableness of the prescription across both the safety and efficacy dimensions, in the form of a comprehensive score, reasonableness level, or a combination of labels. This output serves as input for subsequent risk ranking.

[0053] Step S4: Based on the comprehensive evaluation results, perform risk ranking and stratification.

[0054] This step corresponds to the risk stratification module.

[0055] This module receives the comprehensive rationality evaluation output from the comprehensive evaluation module. Its core objective is to transform prescriptions from a line-by-line equal-weight review to a hierarchical review based on risk priority.

[0056] a. Risk ranking: Based on the degree of safety risk (clear, potential, low) and efficacy suitability (suitable, insufficient, insufficient evidence) of the prescription and their combination, the relative risk priority is ranked.

[0057] Prescriptions marked as having clear medication safety risks and matching treatment effectiveness will be assigned the highest risk priority.

[0058] Prescriptions marked as having potential medication safety risks or inadequate efficacy suitability will be assigned a medium-risk priority.

[0059] Prescriptions marked as having low safety risk and appropriate treatment effectiveness will be assigned a lower risk priority.

[0060] b. Risk level classification: Based on the above ranking results, prescriptions are classified into high-risk, medium-risk, and low-risk levels.

[0061] The classification criteria can comprehensively consider the combination of dual-axis evaluation results, the number and type of risk markers, and can be dynamically optimized by combining historical audit feedback data (the optimization objects are limited to the ranking and stratification strategies themselves, without changing the underlying rule evidence). Step S5: Allocate review channels according to risk level and implement human-machine collaborative decision-making.

[0062] This step corresponds to the decision allocation module.

[0063] Based on the risk level determined by the prescription, this module assigns it to a preset differentiated review and processing channel to achieve precise allocation of review resources.

[0064] a. Allocation of review channels: Automatic passage: Prescriptions with low-risk levels are assigned. The system automatically releases the prescription.

[0065] Pharmacist review channel: Distribute prescriptions with medium risk levels.

[0066] System interception channel and high-priority pharmacist review: Prescriptions with high risk levels are assigned. Prescriptions that trigger level 8 rules are directly intercepted through the system interception channel; prescriptions that trigger level 7 rules are assigned to the pharmacist review channel and given the highest priority.

[0067] b. Human-machine collaborative decision-making: In the pharmacist review and key intervention channels, the system provides pharmacists with a clear risk warning information panel. This panel includes at least: the specific triggering rule, the corresponding level of evidence, the mode of risk action, and a summary of the comprehensive evaluation results of the prescription across the safety and efficacy axes.

[0068] Based on system prompts and in conjunction with the patient's actual clinical condition, treatment goals, and the doctor's intentions, the pharmacist makes the final review decision (approve, suggest adjustments, or reject).

[0069] The system does not replace pharmacists in making final decisions, and always retains the pharmacist's leading role and professional discretion in judging the rationality of prescriptions, forming a closed-loop collaborative workflow of system early warning, pharmacist intervention, and physician feedback.

[0070] The following embodiments further illustrate the prescription intelligent review method and system based on multi-dimensional information provided by the present invention. Obviously, the described embodiments are only a part of the embodiments, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Example 1

[0071] Example of off-label drug use risk identification in children based on evidence grading rules and biaxial evaluation This embodiment uses the prescription review process of a child patient as an example to illustrate the specific implementation process of the prescription rationality review method of the present invention in the context of pediatric medication.

[0072] The patient is a 6-year-old female, weighing 24 kg, clinically diagnosed with mycoplasma pneumonia. During outpatient (or inpatient) treatment, the physician prescribed the following medication: Doxycycline hydrochloride capsules, 0.1 g / capsule, prescribed dose is 80 mg orally twice a day; Simultaneously administer budesonide inhalation suspension, 2 mL: 1 mg, 1 mg each time, twice a day, via nebulization.

[0073] S1: Constructing the prescription decision context The system uses a prescription 3D information construction module to collect, clean, standardize, and structure information related to the prescription to form a prescription judgment context.

[0074] The drug-related information includes: The drug ID is 101273, the generic name is doxycycline hydrochloride capsules, the specification is 0.1 g / capsule, the manufacturer is pharmaceutical company B, the route of administration is oral, the single dose is 80 mg, the frequency of administration is twice a day, the drug category is tetracycline antibiotics; the drug risk attribute information is high risk related to pediatric drug restrictions / age contraindications (avoid use in children under 8 years old), the risk of adverse reactions related to bone and teeth development requires special attention, and the prescription prompt is potential off-label use.

[0075] The drug ID is 101293, the generic name is budesonide inhalation suspension, the specification is 2 mL: 1 mg / vial, the manufacturer is pharmaceutical company A, the route of administration is nebulized inhalation, the single dose is 1 mg, the frequency of administration is twice daily, and the drug category is inhaled corticosteroids. The drug risk information includes: routine risks / adverse reactions requiring monitoring (such as oropharyngeal candidiasis, hoarseness), long-term or high-dose use requiring attention to systemic hormone-like effects (such as impact on growth and development), and instructions for rinsing the mouth after use and proper device usage.

[0076] Patient-related information includes: The patient is a 6-year-old female, 120 cm tall, weighing 24 kg, with a body surface area of ​​0.89 m². 2 The patient has no known comorbidities; no history of drug allergies; and no history of adverse drug reactions.

[0077] Laboratory test results showed that the patient's liver and kidney function indicators were within the normal range, thus the liver and kidney function status was labeled as normal; during hospitalization, vital sign monitoring data showed that the patient's general condition was stable, and there were no obvious abnormalities in body temperature, heart rate and respiration, thus the vital sign data status was labeled as stable.

[0078] Meanwhile, routine nursing monitoring data showed that the patient's blood glucose monitoring results were within the normal range, and the blood pressure monitoring results were within the normal blood pressure reference range for children. After standardization, the above values ​​were respectively derived as blood glucose status label as normal and blood pressure status label as normal, which were used to support the triggering of prescription review rules and subsequent comprehensive evaluation.

[0079] Disease and treatment information includes: The patient's clinical diagnosis is mycoplasma pneumonia, which is the primary diagnosis; the type of visit is hospitalization; the current hospitalization period is day 5; the disease treatment stage information is marked as acute infection treatment stage; the label for being in the perioperative period or intensive care is no; the label for the purpose of this prescription is therapeutic medication, used to control infection and relieve related respiratory symptoms.

[0080] In addition, the disease and treatment dimension information also includes examination and test results related to prescription judgment, including: blood routine test results indicating infection-related changes; liver and kidney function test results within the normal range; and no other abnormal test results that clearly affect the safety judgment of this prescription. The above examination and test results are structured into status tags to support the subsequent application of prescription review rules and comprehensive evaluation.

[0081] S2: Rule Triggering and Risk Marker Generation The system invokes the evidence grading rule engine module to match the above prescription judgment context with the rule base and triggers the following multiple prescription review rules: The first rule is the age-contraindication rule for the avoidance of tetracycline antibiotics in children under 8 years of age. The applicable conditions are children under 8 years of age, the level of evidence is the drug instructions and related evidence-based data, and the risk action mode is strongly binding. The second rule is the dosage guideline for doxycycline use in children, which requires calculation based on body weight. The level of evidence is derived from the "Guidelines for the Diagnosis and Treatment of Mycoplasma pneumoniae in Children (2023 Edition)," and the mode of risk action is suggestive. The third rule is the Dosage Form Appropriateness Rule for Oral Preparations in Children, which is used to indicate the feasibility and risks of administering capsules to young children. The risk is indicated in a suggestive manner.

[0082] Based on the above rule triggering results, the system generates one strong constraint risk marker (involving contraindications for children's age) and multiple warning risk markers (involving the rationality of dosage calculation and the appropriateness of dosage form selection).

[0083] S3: Comprehensive evaluation of effectiveness and safety (dual-axis) The system inputs the risk markers into the effectiveness-safety comprehensive evaluation module for dual-axis evaluation.

[0084] On the safety evaluation axis, because the prescription triggered a strong binding rule based on a high level of evidence—namely, tetracycline antibiotics have a clear age contraindication in children under 8 years old—the system marked the prescription as having a clear medication safety risk. Simultaneously, a suggestive risk marker further indicated that the medication may have adverse effects on children's dental and skeletal development.

[0085] On the efficacy evaluation axis, the system considers both disease diagnosis and medication purpose. It concludes that doxycycline, as a novel tetracycline antibiotic, has some therapeutic value in specific situations such as drug-resistant mycoplasma pneumonia. However, according to current guidelines, its primary target population is children aged 8 years and older. In this case, the patient was 6 years old, and the prescribed dose was 80 mg per dose, which did not strictly adhere to the guideline-recommended dose of 2 mg / kg every 12 hours (the theoretical recommended dose for this patient is approximately 48 mg per dose). Furthermore, the dosage form was inappropriate. Therefore, this prescription was evaluated as having insufficient efficacy suitability.

[0086] Based on the above dual-axis evaluation results of safety and efficacy, the system generates a comprehensive rationality evaluation output for the prescription, which is characterized as a comprehensive rationality status with high-risk characteristics.

[0087] S4: Risk Ranking and Stratification The risk ranking and stratification module receives the comprehensive rationality evaluation results and, based on the combined characteristics of clear safety risks and insufficient effectiveness suitability, classifies the prescription as a high-risk level and assigns it a higher risk priority, so that it can be given priority to enter the manual review and intervention process.

[0088] S5: Review Channel Allocation and Human-Machine Collaborative Decision-Making The review channel allocation module assigns the prescription to the key intervention channel based on the risk level. The system highlights risk markers related to the child's age contraindications, dosage calculation errors, and dosage form suitability in the pharmacist's interface, along with the corresponding evidence sources.

[0089] After reviewing the system prompts and considering the relevant recommendations in the "Guidelines for the Diagnosis and Treatment of Mycoplasma Pneumonia in Children (2023 Edition)," the pharmacist determined that the use of doxycycline in this child was an off-label use and that there were clear age-related safety risks. Therefore, the pharmacist communicated with the physician and suggested prioritizing anti-mycoplasma treatment options that are more suitable for younger children, or fully assessing the risks and obtaining informed consent from the parents if doxycycline use is necessary.

[0090] Description of technical effects of the embodiments As can be seen from this embodiment, the prescription review method of the present invention can accurately identify multiple risk factors such as off-label use, unreasonable dosage, and inappropriate dosage form in the context of children's medication, based on evidence grading rules and a dual-axis comprehensive evaluation mechanism of efficacy and safety. Through risk stratification and human-machine collaborative decision-making mechanism, it guides pharmacists to intervene in high-risk prescriptions in a timely manner, thereby improving the safety and standardization of children's medication. Example 2

[0091] Example of Helicobacter pylori eradication regimen rationality review based on evidence grading rules and biaxial evaluation This embodiment uses the prescription review process of a middle-aged patient as an example to illustrate the specific implementation process of the prescription rationality review method of the present invention in the scenario of multi-drug combination therapy and interaction.

[0092] The patient is a 45-year-old female, clinically diagnosed with gastric ulcer (Helicobacter pylori Hp positive) and hyperlipidemia. The physician prescribed the following for Helicobacter pylori eradication therapy and chronic disease management: Omeprazole enteric-coated capsules, 20 mg, orally, once a day (labeled as being taken after meals); Amoxicillin capsules, 0.25 g / capsule, single dose 0.5 g, orally, twice a day; Clarithromycin extended-release tablets, 0.5 g / tablet, single dose 0.5 g, orally, twice a day; Colloidal bismuth pectin capsules, 0.1 g / capsule, single dose 0.2 g, orally, three times a day (taken 30 minutes before meals). Atorvastatin calcium tablets, 20 mg, orally, once nightly (before bedtime).

[0093] S1: Constructing the prescription decision context The system uses a prescription 3D information construction module to collect, clean, standardize, and structure drug information, patient information, and disease and treatment information related to the prescription to form a prescription judgment context.

[0094] Drug-related information includes the following: The drug ID is 101201; the generic name is omeprazole enteric-coated capsules; the specification is 20 mg / capsule; the manufacturer is pharmaceutical company A; the route of administration is oral; the single dose is 20 mg; the frequency of administration is once a day; the course of treatment is according to the Helicobacter pylori eradication regimen; the drug category is proton pump inhibitor; the drug risk attribute information is a regulatory risk that requires attention to the impact of the timing of administration on efficacy.

[0095] The drug ID is 101315; the generic name is amoxicillin capsules; the specification is 0.25 g / capsule; the manufacturer is pharmaceutical company A; the route of administration is oral; the single dose is 0.5 g; the frequency of administration is twice a day; the course of treatment is according to the Helicobacter pylori eradication regimen; the drug category is β-lactam antibiotic; the drug risk attribute information is that attention should be paid to the adequacy of the dosage and the risk of drug resistance during eradication therapy.

[0096] The drug ID is 101422; the generic name is clarithromycin extended-release tablets; the specification is 0.5 g / tablet; the manufacturer is pharmaceutical company A; the route of administration is oral; the single dose is 0.5 g; the frequency of administration is twice a day; the course of treatment is according to the Helicobacter pylori eradication regimen; the drug category is macrolide antibiotic; the drug risk information is that it is a strong CYP3A4 inhibitor, and the risk of drug interactions needs to be closely monitored.

[0097] The drug ID is 101509; the generic name is colloidal bismuth pectin capsules; the specification is 0.1 g / capsule; the manufacturer is pharmaceutical company A; the route of administration is oral; the single dose is 0.2 g; the frequency of administration is three times a day; the timing of administration is 30 minutes before meals; the course of treatment is according to the Helicobacter pylori eradication regimen; the drug category is bismuth preparation; the drug risk attribute information is routine risk, and attention should be paid to medication adherence and gastrointestinal tolerability.

[0098] The drug ID is 101608; the generic name is atorvastatin calcium tablets; the strength is 20 mg / tablet; the manufacturer is pharmaceutical company A; the route of administration is oral; the single dose is 20 mg; the frequency of administration is once daily (taken at bedtime); the course of treatment is long-term; the drug category is HMG-CoA reductase inhibitor (statin); the drug risk information indicates that the risk of myopathy and hepatotoxicity when used in combination with CYP3A4 inhibitors requires special attention.

[0099] Patient-related information includes: The patient is a 45-year-old female, 160 cm tall, weighing 65 kg, with a body surface area of ​​1.69 m². 2 Comorbidities include hyperlipidemia; no history of drug allergies or adverse drug reactions.

[0100] Laboratory test results showed that the patient's liver and kidney function indicators were within the normal range, and the liver and kidney function status was labeled as normal. During the consultation, vital sign monitoring data showed that the patient's general condition was stable, and there were no obvious abnormalities in body temperature, heart rate and respiration, and the vital sign data status was labeled as stable.

[0101] Meanwhile, routine nursing monitoring data showed that the patient's fasting blood glucose monitoring results were within the normal range, and the blood pressure monitoring results were within the normal blood pressure range. After standardization, the above values ​​were respectively derived as blood glucose status label as normal and blood pressure status label as normal, which were used to support the triggering of prescription review rules and comprehensive evaluation.

[0102] Disease and treatment information includes: The patient's primary diagnosis was gastric ulcer (Hp positive), and the secondary diagnosis was hyperlipidemia; the visit type was outpatient; no hospitalization schedule was involved; the disease treatment stage information was marked as Hp eradication therapy; the label for being in the perioperative period or intensive care was no; the label for the purpose of this prescription was therapeutic medication (Hp eradication) and long-term management medication for chronic diseases.

[0103] In addition, the disease and treatment information also includes examination and test results related to this prescription assessment, including: a positive Helicobacter pylori test result; no significant abnormalities in liver and kidney function tests; and no other abnormal test results that would affect the safety assessment of this prescription. These examination and test results are structured into status tags to support the subsequent application of prescription review rules and comprehensive evaluation.

[0104] S2: Rule Triggering and Risk Marker Generation The system invokes the evidence grading rule engine module to match the above prescription judgment context with the rule base and triggers the following multiple prescription review rules: The first rule is the standard rule for the timing of proton pump inhibitor administration, which states that omeprazole should be taken on an empty stomach 30 minutes before meals. The level of evidence is based on the "2022 Chinese Guidelines for the Treatment of Helicobacter pylori Infection", and the mode of risk action is suggestive. The second rule is the amoxicillin dosage guideline in the Helicobacter pylori eradication protocol. The recommended standard dose is 1g / dose, twice a day. The level of evidence is based on relevant clinical guidelines, and the risk action mode is suggestive. The third rule concerns the drug interaction between clarithromycin and atorvastatin. It states that clarithromycin is a strong inhibitor of CYP3A4 and can significantly increase the blood concentration of atorvastatin, thereby increasing the risk of myopathy and hepatotoxicity. The level of evidence is based on the drug instructions and evidence-based data, and the mode of risk is strongly constrained.

[0105] Based on the above rule triggering results, the system generates multiple risk marker information, including suggestive risk markers involving improper dosing timing and insufficient antibiotic dosage, as well as strong constraint risk markers involving serious drug interactions.

[0106] S3: Comprehensive evaluation of effectiveness and safety (dual-axis) The system inputs the risk markers into the effectiveness-safety comprehensive evaluation module for dual-axis evaluation.

[0107] On the safety evaluation axis, due to the strong interaction risk between clarithromycin and atorvastatin in the prescription, the system marked the prescription as having a clear medication safety risk, indicating that adverse reactions such as myopathy and liver function impairment should be closely monitored.

[0108] On the efficacy evaluation axis, the system combined disease diagnosis and medication purpose to make a judgment. It was found that the dosage of amoxicillin in the current Helicobacter pylori eradication regimen was lower than the guideline recommendation standard, and the timing of omeprazole administration was not in accordance with the norm, which may reduce the acid suppression effect and affect the Helicobacter pylori eradication rate. Therefore, the prescription was evaluated as having insufficient efficacy suitability.

[0109] Based on the above dual-axis evaluation results of safety and efficacy, the system generates a comprehensive rationality evaluation output for the prescription, which is characterized as a comprehensive rationality status with high-risk characteristics.

[0110] S4: Risk Ranking and Stratification The risk ranking and stratification module receives the comprehensive rationality evaluation results and, based on the combined characteristics of clear safety risks and insufficient effectiveness, classifies the prescription as a high-risk level and assigns it a higher risk priority, so that it can be given priority to enter the manual review process.

[0111] S5: Review Channel Allocation and Human-Machine Collaborative Decision-Making The review channel allocation module assigns the prescription to the key intervention channel based on the risk level. The system highlights drug interaction risks and relevant risk markers related to eradication regimens in the pharmacist's interface, along with the corresponding evidence sources and modes of risk action.

[0112] After reviewing the system prompts, the pharmacist, considering the patient's long-term medication history for hyperlipidemia, communicated with the physician and suggested adjustments to the Helicobacter pylori eradication regimen. These adjustments included standardizing the timing of omeprazole administration, increasing the amoxicillin dose to the guideline-recommended level, and avoiding the combined use of clarithromycin and atorvastatin. If necessary, alternative antibiotics or adjustments to the statin dose could be considered, along with enhanced monitoring of relevant laboratory indicators.

[0113] Description of technical effects of the embodiments As can be seen from this embodiment, the prescription review method of the present invention can, in the case of multi-drug combination therapy, simultaneously identify dosing standardization issues, risks of insufficient efficacy and serious drug interaction risks based on evidence grading rules and a dual-axis comprehensive evaluation mechanism of efficacy and safety. Through risk ranking and human-machine collaborative decision-making mechanism, it guides pharmacists to systematically intervene in high-risk prescriptions, thereby improving the safety and treatment success rate of complex combination drug regimens. Example 3

[0114] Example of medication review for cardiovascular risk in elderly patients based on evidence grading and biaxial assessment This example uses a 72-year-old male patient diagnosed with coronary heart disease and stable angina. He is 175cm tall and weighs 80kg. During his hospitalization, he was prescribed 10mg of domperidone tablets orally three times a day due to abdominal distension.

[0115] S1: Constructing the prescription decision context: The system collects data through the information construction module to build a structured prescription judgment context: Drug information: Drug ID is 105395, generic name is domperidone tablets, specification is 10mg / tablet, manufacturer is pharmaceutical company A, route of administration is oral, single dose is 10mg, dosing frequency is three times a day, course of treatment is as needed, drug category is prokinetic drug, and drug risk attribute information is key monitoring / high risk (heart-related).

[0116] Patient information: Age 72, Male, Height 175cm, Weight 80kg, Body surface area 1.95 m² 2 Comorbidity information includes coronary heart disease, no history of drug allergy, no history of adverse drug reactions, liver and kidney function status is normal, and vital signs data status is stable.

[0117] Disease and Treatment Dimension Information: Diagnosis information is coronary heart disease, stable angina pectoris, visit type is hospitalization, hospitalization period is day 3, disease treatment stage information is chronic disease treatment period, the label for perioperative period or intensive care is no, and the label for medication purpose is symptomatic treatment (relieving abdominal distension).

[0118] S2: Rule Triggering and Risk Marker Generation: The rules engine module pre-stores relevant review rules. The system matches the above context with the rule base and triggers two rules simultaneously: one for elderly people over 60 years old who need physician guidance for medication (evidence level: Level 5; mode of action: suggestive); and the other for patients with cardiac conduction abnormalities or heart disease who are contraindicated (evidence level: Level 8; mode of action: strong binding). Based on this, two risk markers are generated: the first is a suggestive risk marker, which relates to medication guidance for the elderly; the second is a strong binding risk marker, which relates to contraindications for patients with heart disease.

[0119] S3: Dual-axis comprehensive evaluation: The comprehensive evaluation module processes the risk markers.

[0120] Safety evaluation axis: Marked as a clear medication safety risk due to triggering Level 8 strong constraint safety rule.

[0121] Efficacy evaluation axis: Combining the diagnostic and medication purposes, it was determined that the indications of the drug domperidone (indigestion, abdominal distension, etc.) have a weak direct correlation with the current primary diagnosis (coronary heart disease), and there are better and safer alternatives. Therefore, the efficacy was evaluated as insufficiently suitable.

[0122] Overall Output: A comprehensive evaluation result on the rationality of the high-risk features generated by the mapping.

[0123] S4: Risk Ranking and Stratification The risk stratification module receives the above results. Combinations of clearly identified security risks and insufficient effectiveness are classified as the highest risk priority and placed in the high-risk category.

[0124] S5: Review Channel Allocation and Human-Machine Collaborative Decision-Making The decision-making and allocation module assigns the prescription to the pharmacist review channel (due to triggering level 8 rules, the system should actually intercept the request; this is just to demonstrate the hierarchical logic, assuming the system performs the review action). The pharmacist's interface highlights two risk markers along with their detailed basis and level of evidence. After assessing the risk based on the patient's cardiac history, the pharmacist decides to communicate with the physician and recommend switching to a gastrointestinal motility drug or other treatment option with less impact on the heart.

[0125] Description of technical effects of the embodiments As can be seen from this embodiment, the prescription review method of the present invention can effectively identify drug cardiac safety contraindications and treatment suitability issues in the medication scenarios of elderly patients and patients with cardiovascular diseases, based on a clear level of evidence (Level 8 contraindication) and a dual-axis evaluation mechanism. It can also accurately classify prescriptions into high-risk levels through a risk stratification mechanism, triggering corresponding review and intervention processes, thereby preventing serious medication risks.

[0126] In summary, the prescription intelligent review method and system based on multi-dimensional information provided by this invention can achieve the following significant beneficial effects: Firstly, by constructing a structured prescription judgment context that integrates drug, patient, and disease diagnosis and treatment information, the review decision is based on the perception of the complete clinical context, fundamentally solving the problem of prompt overload or risk omission caused by the single information dimension of traditional review technology.

[0127] Secondly, by establishing a hierarchical, governance, and structured rule transformation mechanism for multi-source medical evidence, a dynamic knowledge system with clear responsibilities, maintainability, and traceability has been constructed, ensuring the scientific nature, timeliness, and consistency of the review basis.

[0128] Third, it innovatively introduces a dual-axis comprehensive assessment mechanism that combines safety evaluation and treatment efficacy evaluation, so that the review conclusions not only focus on medication safety, but also support rational medication decisions guided by clinical treatment goals, and avoid ineffective or inappropriate treatment.

[0129] Fourth, it establishes a credible human-machine collaborative boundary for system early warning and pharmacist decision-making. The system provides risk alerts based on clear evidence, while retaining the final judgment on the rationality of prescriptions to professional pharmacists, ensuring the safety and controllability of the medical decision-making process and the implementation of professional responsibility.

[0130] Fifth, the system provides risk warnings based on clear evidence, while retaining the final decision-making power to the pharmacist, establishing a credible and transparent boundary for human-machine collaboration, and ensuring the safety and controllability of the medical process and the implementation of professional responsibilities.

[0131] Sixth, by establishing a continuous learning and optimization mechanism based on review feedback and clinical outcomes, the system's risk identification strategy can be dynamically adapted to local clinical practices while maintaining the stability of core rules, thus possessing the ability to evolve over the long term.

[0132] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A prescription intelligent review method based on multidimensional information, characterized in that: Includes the following steps: Acquire prescription-related data and construct a structured prescription judgment context, which includes at least drug dimension information, patient dimension information, and disease and treatment dimension information; Multi-source medical evidence is transformed into structured review rules with evidence level identifiers and risk action modes, and based on the prescription judgment context, the applicable structured review rules are triggered to generate risk markers; Based on the aforementioned risk markers, and combined with information on the disease diagnosis and treatment stage and the purpose of medication, a dual-axis comprehensive evaluation of the prescription is conducted on the safety evaluation axis and the efficacy evaluation axis to generate a comprehensive rationality evaluation result. Based on the comprehensive rationality evaluation results, the prescriptions are ranked and stratified by risk to determine their risk levels; Based on the risk level, prescriptions are assigned to the corresponding review and processing channels, and pharmacists make the final review decisions based on the risk warning information provided by the system.

2. The prescription intelligent review method based on multidimensional information according to claim 1, characterized in that: The drug-related information includes at least the drug identifier, generic name, specifications, manufacturer, route of administration, single dose, frequency of administration, course of treatment, and drug risk attributes. The patient-related information includes at least age, gender, height, weight, body surface area, comorbidities, allergy history, history of adverse drug reactions, and vital signs. The disease and treatment-related information includes at least diagnostic information, type of visit, hospitalization schedule, stage of disease treatment, whether the patient is in the perioperative period or intensive care unit, label indicating the purpose of medication, and relevant laboratory tests and examinations.

3. The prescription intelligent review method based on multidimensional information according to claim 1, characterized in that: The multi-source medical evidence includes drug instructions, clinical guidelines, evidence-based medicine literature, expert consensus, and internal regulations of medical institutions; the evidence level is predefined into nine levels from 0 to 8, where: Level 8 corresponds to evidence sources such as explicit contraindications in the drug instructions, prohibited combinations and interactions, mandatory prohibitions under national laws and regulations, or national safety warnings. Its risk effect is strongly binding, and the corresponding review and handling action is direct system interception. Level 7 corresponds to evidence sources such as serious warnings in drug instructions, situations explicitly not recommended by authoritative guidelines, key deduction items in national rational drug use assessments, or clear serious interactions. Its risk mechanism is strongly binding, and the corresponding review and handling action is to forcibly transfer it to high-priority manual review. Level 6 corresponds to evidence sources such as non-recommended protocols in clinical guidelines, significant risks indicated by evidence-based literature, use outside the scope of policy restrictions, or violation of explicit restrictions in the instructions. The risk effect is either suggestive or strongly binding, and the corresponding review and handling action is to transfer to manual review. Level 5 corresponds to evidence sources such as guidelines recommending caution in use, risk warnings of insufficient evidence consistency, risk monitoring indicators, or off-label drug use with evidence-based support. Its risk mechanism is warning-type, and the corresponding review and handling action is to transfer to manual review. The sources of evidence for Level 4, Level 3, Level 2, Level 1, and Level 0 are respectively normative recommendations, efficacy optimization tips, management cost-effectiveness tips, general precautions, and informational rules. Their risk action is either suggestive or risk-free judgment, and the corresponding review and handling actions are system release and recording or only for information recording.

4. The prescription intelligent review method based on multidimensional information according to claim 3, characterized in that: The process of transforming multi-source medical evidence into structured review rules with evidence level identifiers and risk action modes also includes: The candidate rules were evaluated for inclusion / exclusion, clinical rationality, and operability through a two-round Delphi consultation method conducted by multidisciplinary experts. Correlation analysis and dimensionality reduction of the rule features are performed using principal component analysis and cluster analysis to eliminate rule redundancy.

5. The prescription intelligent review method based on multidimensional information according to claim 3, characterized in that: The biaxial comprehensive evaluation includes: On the safety evaluation axis, based on the evidence level corresponding to the risk marker, prescriptions that trigger Level 8 or Level 7 rules are marked as having a clear medication safety risk, prescriptions that trigger Level 6 or Level 5 rules are marked as having a potential medication safety risk, and prescriptions that trigger Level 4 or lower rules are marked as having a low safety risk. On the efficacy evaluation axis, prescriptions are marked as treatment efficacy fit, efficacy fit is insufficient, or efficacy evidence is insufficient by judging the degree of match between the drug indication and the current diagnosis and disease stage, as well as the consistency between the purpose of medication and the recommended treatment goals.

6. The prescription intelligent review method based on multidimensional information according to claim 5, characterized in that: The risk ranking and stratification of prescriptions based on the comprehensive rationality evaluation results specifically includes: classifying prescriptions marked as having clear medication safety risks and matching treatment effectiveness as high-risk; classifying prescriptions marked as having potential medication safety risks or insufficient matching effectiveness as medium-risk; and classifying prescriptions marked as having low safety risks and matching treatment effectiveness as low-risk.

7. The prescription intelligent review method based on multidimensional information according to claim 6, characterized in that: The review and processing channels include an automatic approval channel, a pharmacist review channel, and an interception channel. Prescriptions with low risk are assigned to the automatic approval channel, prescriptions with medium risk are assigned to the pharmacist review channel, prescriptions with high risk that trigger level 8 rules are assigned to the system interception channel for direct interception, and prescriptions that trigger level 7 rules are assigned to the pharmacist review channel and given the highest priority.

8. The prescription intelligent review method based on multidimensional information according to claim 1, characterized in that: The method also includes a feedback learning step: Record the pharmacist's review and decision-making results and the physician's adoption of intervention recommendations; Collect clinical outcome information related to prescriptions, including adverse reaction records, length of hospital stay, drug costs, medical expenses, drug ratio, and disease outcome; Based on the review decision results, adoption status, and clinical outcome information, optimize the weights or thresholds in the risk ranking and stratification strategies.

9. A prescription intelligent review system based on multidimensional information, characterized in that, include: The information construction module is used to acquire prescription-related data and construct a structured prescription judgment context, which includes drug dimension information, patient dimension information and disease and treatment dimension information. The rules engine module is used to store structured review rules that transform multi-source medical evidence, and to trigger applicable rules based on the prescription judgment context to generate risk markers; The comprehensive evaluation module is used to perform a dual-axis comprehensive evaluation of the prescription based on the risk markers, on the safety evaluation axis and the efficacy evaluation axis, and generate a comprehensive rationality evaluation result. The risk stratification module is used to rank and stratify prescriptions based on the comprehensive rationality evaluation results, and determine their risk levels. The decision allocation module is used to allocate prescriptions to the corresponding review and processing channels according to the risk level, and output risk warning information for pharmacists to conduct final review.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 8.

Images