Patient medication risk early warning method

Abstract

The invention discloses a patient medication risk early warning method, and relates to the technical field of medication risk evaluation. The method comprises the following steps: acquiring multi-dimensional data such as age, past medical history, allergy history, clinical examination data and prescription data of a patient, extracting chronic disease diagnosis information from the multi-dimensional data, and carrying out standardized coding to generate a common disease combination tag; calling a preset disease weight system, and calculating a common disease severity score in combination with an age factor; extracting metabolism-related indexes to construct a drug metabolism capability feature vector of the patient; mapping the drug combination, the co-disease tag and the metabolic feature vector to a clinical drug knowledge base to generate a comprehensive conflict detection result; and generating corresponding early warning information based on the conflict detection result and a preset early warning rule base, wherein the rule base comprises a mapping relationship between the conflict type and the early warning level. According to the method, drug use risk early warning can be realized according to the multi-disease coexistence state of the patient, and the drug use safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medication risk assessment, and particularly to a method for warning of patients' medication risks. Background Art

[0002] With the acceleration of the global aging process, the diagnosis and treatment needs of elderly patients with multiple co-existing diseases have become increasingly prominent. Such patients face significant medication safety challenges due to the decline of multiple system functions, the intertwining of disease pathologies, and the complexity of the drug metabolism network.

[0003] In the prior art, by integrating patient information and drug knowledge bases for knowledge reasoning and using risk prediction models to generate medication risk degrees, certain achievements have been made in the detection of medication risks for single diseases. However, for the common scenario of multiple co-existing diseases in elderly patients, the existing solutions mainly match drug taboos based on single disease diagnoses and do not consider the physiological decline characteristics of the elderly population, making it difficult to comprehensively evaluate the comprehensive impact of multi-dimensional risk factors. Summary of the Invention

[0004] To solve the technical problems existing in the prior art, the present invention provides a method for warning of patients' medication risks.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A method for warning of patients' medication risks, comprising the following steps:

[0007] Step 1, obtaining multi-dimensional data in the patient's diagnosis and treatment process, where the multi-dimensional data includes patient age, past medical history, allergy history, clinical examination data, and prescription data issued by a doctor.

[0008] Step 2, extracting chronic disease diagnosis information from the clinical examination data and past medical history, performing standardized processing on disease coding, and generating a co-morbidity combination label.

[0009] Step 3, calling a preset disease weight system, and combining with the patient age factor, calculating a co-morbidity severity score, where the disease weight system sets weight coefficients according to the influence degree of the disease on the drug metabolism pathway, the organ function damage level, and the complication risk.

[0010] Step 4, extracting indicators related to drug metabolism from the clinical examination data, and constructing a patient drug metabolism ability feature vector.

[0011] Step 5, mapping the drug combination in the prescription data, the patient's co-morbidity combination label, and the drug metabolism ability feature vector to a clinical medication knowledge base, and generating a comprehensive conflict detection result.

[0012] Step 6, generate corresponding warning information based on the comprehensive conflict detection result and a preset warning rule library, where the warning rule library contains the mapping relationship between conflict types and warning levels.

[0013] Preferably, the step of calling a preset disease weight system and combining with the patient's age factor to calculate the comorbidity severity score includes:

[0014] When the patient's age reaches a preset age threshold, up-regulate and correct the weight coefficient of chronic diseases affecting drug metabolism, and calculate the comorbidity severity score based on the corrected weight coefficient.

[0015] Preferably, the step of extracting indicators related to drug metabolism from clinical examination data and constructing a patient drug metabolism ability feature vector includes:

[0016] Extract indicators reflecting liver and kidney functions from clinical examination data, normalize the indicators, and construct a patient drug metabolism ability feature vector based on the normalized indicators.

[0017] Preferably, the step of mapping the drug combination in the prescription data, the comorbidity combination label of the patient, and the drug metabolism ability feature vector to the clinical medication knowledge base to generate a comprehensive conflict detection result includes:

[0018] Map the drug combination in the prescription data and the comorbidity combination label of the patient to the disease-drug contraindication rule library in the clinical medication knowledge base to generate a disease-drug conflict detection result;

[0019] Map the drug combination in the prescription data and the patient's drug metabolism ability feature vector to the drug metabolism rule library in the clinical medication knowledge base to generate a drug metabolism conflict detection result;

[0020] Generate a comprehensive conflict detection result based on the disease-drug conflict detection result and the drug metabolism conflict detection result.

[0021] Preferably, the step of directly generating corresponding warning information based on the comprehensive conflict detection result and a preset warning rule library includes:

[0022] Based on the comprehensive conflict detection result, match the corresponding warning rule from the preset warning rule library;

[0023] Generate a warning information containing a description of the conflict mechanism and a disposal suggestion according to the matched warning rule.

[0024] Preferably, the step of extracting chronic disease diagnosis information from clinical examination data and past medical history includes:

[0025] Perform natural language processing on clinical examination data and past medical history to identify chronic disease diagnosis information therein; the natural language processing uses a rule-based named entity recognition method.

[0026] Preferably, the step of constructing a patient drug metabolism ability feature vector based on the normalized indicators includes:

[0027] Perform feature combination on the drug metabolism-related indicators after normalization processing to construct a drug metabolism ability feature vector; the feature combination uses a weighted summation method.

[0028] Preferably, the step of mapping the drug combination in the prescription data, the comorbidity combination label of the patient, and the drug metabolism ability feature vector to the clinical medication knowledge base includes:

[0029] Convert the drug combination, the comorbidity combination label, and the drug metabolism ability feature vector into a unified data representation form;

[0030] Based on the unified data representation form, perform matching retrieval in the clinical medication knowledge base.

[0031] The beneficial effects of the present invention are at least one of the following:

[0032] By obtaining multi-dimensional information such as the patient's age, past medical history, clinical examination data, and prescription data, it avoids the risk of missed detection caused by single-dimensional data, provides a complete data basis for the medication risk assessment of patients with multiple co-existing diseases, and can especially capture the impact of age-related physiological decline characteristics of elderly patients on medication safety.

[0033] Perform encoding standardization processing on the chronic disease diagnosis information, unify the diagnosis terms of different medical institutions into standardized disease codes, solve the problem of co-morbidity analysis errors caused by inconsistent diagnosis terms, improve the consistency of co-morbidity combination labels, and provide a reliable disease dimension analysis basis for subsequent conflict detection.

[0034] Set weight coefficients based on the degree of influence of diseases on drug metabolism, the level of organ function damage, etc., and dynamically adjust the weights in combination with the patient's age factor, solve the problem of underestimated risks caused by the traditional scheme not considering the decline of metabolic function in the elderly population, and make the co-morbidity severity score more in line with the actual medication risks of elderly patients.

[0035] Map the drug combination, the co-morbidity label, and the metabolic feature vector to the clinical medication knowledge base, and parallelly detect multi-dimensional conflicts such as disease-drug taboos and abnormal drug metabolism, solve the problem of missed detection caused by traditional single-dimensional detection. Description of the Drawings

[0036] Figure 1 It is a schematic diagram of the method flow of the present invention. Detailed implementation manners

[0037] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] This embodiment provides a method for warning of patients' medication risks, as Figure 1 shown, including the following steps:

[0039] Step 1, obtaining multi-dimensional data in the patient's diagnosis and treatment process, where the multi-dimensional data includes patient age, past medical history, allergy history, clinical examination data, and prescription data issued by a doctor.

[0040] To solve the problem of missed detection of medication risks caused by traditional solutions relying on single-dimensional data, exemplarily, basic information such as patient age, past medical history, and allergy history is obtained through the interface of the hospital information system (HIS); liver and kidney function indicators and clinical examination data of blood routine are collected in real time through the laboratory information management system (LIS); prescription data issued by a doctor, including drug name, dosage form, dose, administration route, and frequency, is obtained through the electronic prescription system (EPRS).

[0041] This embodiment provides a comprehensive data basis for medication risk assessment, enabling the assessment dimensions to cover the patient's basic characteristics, disease status, metabolic capacity, and medication plan, and helping to capture risk scenarios intertwined with multiple factors.

[0042] Step 2, extracting chronic disease diagnosis information from clinical examination data and past medical history, performing standardized processing of disease coding, and generating a comorbidity combination label.

[0043] To solve the problem of comorbidity analysis errors caused by a large number of synonyms, aliases, and abbreviations in medical terms, in a possible implementation manner, the step of extracting chronic disease diagnosis information from clinical examination data and past medical history includes:

[0044] Performing natural language processing on clinical examination data and past medical history to identify the chronic disease diagnosis information therein; the natural language processing uses a rule-based named entity recognition method.

[0045] Using a rule-based named entity recognition method, constructing a medical term dictionary (including chronic disease names, synonyms, and abbreviations), and matching diagnostic terms such as "hypertension" and "diabetes" in the medical record text through regular expressions. For example, for the text "The patient has a history of type 2 diabetes for 5 years", "type 2 diabetes" is extracted as the diagnosis information.

[0046] Map the recognized diagnostic terms to a unified disease classification and code standard. For example, "type 2 diabetes" corresponds to the code E11.9. Generate comorbidity combination labels based on the standardized codes, such as "hypertension (I10) + type 2 diabetes (E11.9) + chronic kidney disease (N18.9)".

[0047] In this embodiment, the consistency of the comorbidity combination labels is improved through standardized coding, making the basic data for subsequent drug-disease conflict detection more reliable and reducing misjudgments caused by term ambiguity.

[0048] Step 3: Invoke a preset disease weight system and combine the patient's age factor to calculate the comorbidity severity score. The disease weight system sets weight coefficients according to the degree of influence of the disease on the drug metabolism pathway, the level of organ function damage, and the risk of complications.

[0049] To address the problem of underestimating the medication risk caused by the traditional solution's failure to consider the decline of organ function in elderly patients (such as reduced liver and kidney function), in a possible embodiment, the step of invoking a preset disease weight system and combining the patient's age factor to calculate the comorbidity severity score includes:

[0050] When the patient's age reaches the preset age threshold, the weight coefficient of the chronic disease affecting drug metabolism is adjusted upward, and the comorbidity severity score is calculated based on the adjusted weight coefficient.

[0051] Exemplarily, medical experts preset weight coefficients for each disease according to the degree of influence of the disease on drug metabolism (such as chronic kidney disease affecting drug excretion), the level of organ function damage (such as the Child-Pugh classification of liver function), and the risk of complications (such as diabetes complicated with cardiovascular disease). For example, the weight coefficient of chronic kidney disease (N18.9) is 0.8, and the weight coefficient of hypertension (I10) is 0.5.

[0052] When the patient's age ≥ 65 years old, the elderly correction mechanism is automatically triggered, and the weight coefficients of the diseases affecting drug metabolism (such as chronic kidney disease, liver disease) are increased by 20%. For example, the weight coefficient of chronic kidney disease for patients over 65 years old is adjusted to 0.8 × 1.2 = 0.96.

[0053] Sum up the weight coefficients of each disease to obtain the comorbidity severity score. For example, the score of "chronic kidney disease (0.96) + hypertension (0.5)" = 1.46.

[0054] In this embodiment, through the dynamic correction of age-related weights, the comorbidity severity score is made more consistent with the actual metabolic capacity and medication risk of elderly patients, providing a quantitative basis for personalized risk assessment.

[0055] Step 4: Extract indicators related to drug metabolism from clinical examination data and construct a characteristic vector of the patient's drug metabolism ability.

[0056] To solve the problem of incomparable evaluation dimensions caused by the dimensional differences of different metabolic indicators, in a possible implementation manner, the step of extracting indicators related to drug metabolism from clinical examination data and constructing a characteristic vector of the patient's drug metabolism ability includes:

[0057] Extract indicators reflecting liver and kidney functions from clinical examination data, normalize the indicators, and construct a characteristic vector of the patient's drug metabolism ability based on the normalized indicators.

[0058] In a possible implementation manner, the step of constructing a characteristic vector of the patient's drug metabolism ability based on the normalized indicators includes:

[0059] Perform feature combination on the normalized indicators related to drug metabolism to construct a characteristic vector of drug metabolism ability; the feature combination uses the weighted summation method.

[0060] Exemplarily, extract indicators such as serum creatinine (evaluating kidney function), alanine aminotransferase (ALT, evaluating liver function), aspartate aminotransferase (AST), etc. from clinical examination data; if there is gene detection data, extract CYP450 enzyme genotypes.

[0061] Use the Z-score standardization method to convert the indicator values into standard normal distribution values. For example, if a patient's serum creatinine value is 150 μmol / L, the mean value of the same age group is 80 μmol / L, and the standard deviation is 20 μmol / L, then the standardized value = (150 - 80) / 20 = 3.5.

[0062] Perform weighted summation on the standardized liver and kidney function indicators and gene typing results (such as CYP2D6*4 is 1 and wild type is 0) to generate a multi-dimensional characteristic vector. It should be noted that CYP2D6 is a specific gene in the CYP2 subfamily of the CYP450 family and is mainly involved in drug metabolism (such as antidepressants, analgesics, etc.). Therefore, "CYP450 enzyme genotype" is a general term for the category, and "CYP2D6 genotype" is the typing of a specific gene in this family.

[0063] For example, serum creatinine (weight 0.4), ALT (weight 0.3), and CYP2D6 genotype (weight 0.3) are combined into a three-dimensional vector [3.5×0.4, 2.1×0.3, 1×0.3].

[0064] In this implementation manner, the individual metabolic ability of the patient is transformed into computable quantitative characteristics, making the detection of abnormal drug metabolism (such as dosage adjustment requirements) more targeted and improving the recognition accuracy of metabolism-related risks.

[0065] Step 5: Map the drug combination in the prescription data, the comorbidity combination label of the patient, and the drug metabolism ability feature vector to the clinical medication knowledge base to generate a comprehensive conflict detection result.

[0066] In a possible implementation manner, the step of mapping the drug combination in the prescription data, the comorbidity combination label of the patient, and the drug metabolism ability feature vector to the clinical medication knowledge base includes:

[0067] Convert the drug combination, the comorbidity combination label, and the drug metabolism ability feature vector into a unified data representation form;

[0068] Based on the unified data representation form, perform a matching search in the clinical medication knowledge base.

[0069] Exemplarily, convert the prescription drugs into a standardized coding (such as ATC coding), the comorbidity combination label uses ICD-10 coding, and the metabolic feature vector is converted into a numerical array to form a unified data representation form. For example, aspirin corresponds to the ATC coding B01AC06, and the comorbidity label of "hypertension + diabetes" is "I10 + E11.9".

[0070] In a possible implementation manner, the step of mapping the drug combination in the prescription data, the comorbidity combination label of the patient, and the drug metabolism ability feature vector to the clinical medication knowledge base to generate a comprehensive conflict detection result includes:

[0071] Map the drug combination in the prescription data and the comorbidity combination label of the patient to the disease-drug contraindication rule base in the clinical medication knowledge base to generate a disease-drug conflict detection result;

[0072] Exemplarily, map the comorbidity coding and the drug ATC coding to the disease-drug contraindication rule base (constructed based on the National Formulary). For example, "diabetes (E11.9)" and "thiazide diuretics (N03BA)" trigger a contraindication warning.

[0073] Map the drug combination in the prescription data and the drug metabolism ability feature vector of the patient to the drug metabolism rule base in the clinical medication knowledge base to generate a drug metabolism conflict detection result.

[0074] Exemplarily, match the metabolic feature vector with the drug metabolism rule base. For example, when the standardized value of serum creatinine > 3, it is prompted that "drugs excreted through the kidneys need to be reduced by 50%".

[0075] Generate a comprehensive conflict detection result based on the disease-drug conflict detection result and the drug metabolism conflict detection result.

[0076] Exemplarily, map the drug combination to an interaction database (such as DrugBank), and analyze the competition relationship of metabolic enzymes. For example, "Statins (C10AA) + Antibiotics (J01MA)" triggers a CYP3A4 enzyme inhibition warning.

[0077] Step 6, based on the comprehensive conflict detection result and a preset warning rule library, generate corresponding warning information. The warning rule library contains the mapping relationship between conflict types and warning levels.

[0078] Exemplarily, input the comprehensive conflict detection result into a warning rule library (constructed based on "Instructions for Clinical Drug Use"), for example:

[0079] Disease-drug contraindications correspond to high-risk warnings (red markings); metabolic abnormalities requiring dose adjustment correspond to medium-risk warnings (yellow markings); potential drug interactions correspond to low-risk warnings (blue markings).

[0080] In a possible implementation manner, the step of directly generating corresponding warning information based on the comprehensive conflict detection result and a preset warning rule library includes:

[0081] Based on the comprehensive conflict detection result, match the corresponding warning rule from the preset warning rule library;

[0082] According to the matched warning rule, generate warning information including a description of the conflict mechanism and disposal suggestions.

[0083] Exemplarily, the warning rule library is stored in a structured data format. Each rule contains the following core elements: rule unique identifier (such as "Drug-Disease Contraindication-001"); trigger conditions: including disease coding (such as "Type 2 diabetes" corresponding to the International Classification of Diseases coding E11.9), drug classification coding (such as "Thiazide diuretics" corresponding to the ATC coding N03BA), and conflict type (such as "Absolute contraindication"); warning level (high risk / medium risk / low risk); disposal suggestions (such as "It is recommended to discontinue thiazide diuretics and replace them with potassium-sparing diuretics").

[0084] In a specific implementation manner, extract the disease coding (such as E11.9) and drug coding (such as N03BA) from the comprehensive conflict detection result; quickly locate all relevant candidate rules through a composite index (such as rules that simultaneously contain "E11.9" and "N03BA"); verify one by one whether the conflict type of the candidate rule is consistent with the detection result (such as confirming whether it is "Absolute contraindication").

[0085] Exemplarily, high-risk warning generation scenarios:

[0086] Conflict detection result: The patient is diagnosed with type 2 diabetes (International Classification of Diseases code E11.9). The current prescription includes hydrochlorothiazide (a thiazide diuretic, ATC code N03BA01), and the serum creatinine index indicates reduced renal function.

[0087] Rule matching process: Extract the disease code E11.9 and the drug code N03BA, and match the "Drug-Disease Taboo-001" rule; confirm that the conflict type is "absolute taboo", triggering a high-risk warning.

[0088] Generated warning information: High-risk warning; Conflict details: There is an absolute taboo between hydrochlorothiazide and type 2 diabetes; Conflict mechanism: Thiazide diuretics can increase blood sugar and exacerbate the condition of diabetes (based on the "Chinese Guidelines for the Prevention and Treatment of Type 2 Diabetes (2023 Edition)"); Disposal suggestion: Immediately discontinue hydrochlorothiazide and replace it with indapamide;

[0089] Scenario for generating medium-risk warning:

[0090] Conflict detection result: A 68-year-old patient is diagnosed with stage 3 chronic kidney disease (International Classification of Diseases code N18.3). The prescription includes drug A excreted through the kidneys, and the serum creatinine index indicates moderate reduction in renal function. Through the semantic analysis process: Identify "stage 3 chronic kidney disease" and "abnormal serum creatinine" as characteristics of reduced renal function; Extract the keyword "excreted through the kidneys" and match the severity weight of 0.7. Generated warning information: Medium-risk warning; Conflict details: The excretion of drug A is reduced in patients with stage 3 chronic kidney disease, with a risk of accumulation; Conflict mechanism: Reduced renal function leads to abnormal drug metabolism (based on the "Guidelines for Medication Use in Patients with Chronic Kidney Disease"); Disposal suggestion: Reduce the dose of drug A by 50% and monitor the blood drug concentration every 3 days.

[0091] In summary, this solution forms a complete closed-loop of medication risk warning through multi-dimensional data integration, disease coding standardization, dynamic weight adjustment for elderly patients, quantitative analysis of metabolic capacity, multi-dimensional conflict detection, and a clinical guideline-driven warning mechanism. It covers multi-dimensional data such as patient basic information, disease status, metabolic capacity, and medication regimen, avoiding evaluation biases caused by single factors. Through standardized processing, dynamic weights, and quantitative feature construction, it realizes personalized risk assessment for elderly patients and patients with multiple co-existing diseases. The conflict detection results are directly mapped to the clinical guideline rule base, and the warning information has clear evidence-based support, enhancing the credibility of clinical applications; providing structured warnings including disposal suggestions for doctors, directly supporting clinical medication decisions, helping to reduce medication risks, and improving the medication safety of patients.

[0092] The above-described embodiments merely represent specific implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention.

Claims

1. A method for warning of the risk of patient medication, characterized in that It includes the following steps: Step 1: Obtain multi-dimensional data during the patient's diagnosis and treatment process. The multi-dimensional data includes patient age, past medical history, allergy history, clinical examination data, and prescription data issued by doctors; Step 2: Extract chronic disease diagnosis information from the clinical examination data and past medical history, perform disease coding standardization processing, and generate comorbidity combination labels; Step 3: Call a preset disease weight system, and combine with the patient age factor to calculate the comorbidity severity score. The disease weight system sets weight coefficients according to the degree of influence of the disease on the drug metabolism pathway, the level of organ function damage, and the complication risk; Step 4: Extract indicators related to drug metabolism from the clinical examination data, and construct a patient drug metabolism ability feature vector; Step 5: Map the drug combination in the prescription data, the patient's comorbidity combination label, and the drug metabolism ability feature vector to the clinical medication knowledge base to generate a comprehensive conflict detection result; Step 6: Based on the comprehensive conflict detection result and a preset warning rule library, generate corresponding warning information. The warning rule library contains the mapping relationship between conflict types and warning levels.

2. The patient medication risk warning method according to claim 1, wherein The step of calling a preset disease weight system and combining with the patient age factor to calculate the comorbidity severity score includes: When the patient's age reaches a preset age threshold, up-regulate and correct the weight coefficient of chronic diseases that affect drug metabolism, and calculate the comorbidity severity score based on the corrected weight coefficient.

3. The patient medication risk warning method according to claim 1, wherein The step of extracting indicators related to drug metabolism from the clinical examination data and constructing a patient drug metabolism ability feature vector includes: Extract indicators reflecting liver and kidney functions from the clinical examination data, perform normalization processing on the indicators, and construct a patient drug metabolism ability feature vector based on the normalized indicators.

4. The patient medication risk warning method according to claim 1, characterized in that The step of mapping the drug combination in the prescription data, the patient's comorbidity combination label, and the drug metabolism ability feature vector to the clinical medication knowledge base to generate a comprehensive conflict detection result includes: Map the drug combination in the prescription data and the patient's comorbidity combination label to the disease-drug contraindication rule library in the clinical medication knowledge base to generate a disease-drug conflict detection result; Map the drug combination in the prescription data and the patient's drug metabolism ability feature vector to the drug metabolism rule library in the clinical medication knowledge base to generate a drug metabolism conflict detection result; Based on the disease-drug conflict detection result and the drug metabolism conflict detection result, generate a comprehensive conflict detection result.

5. The patient medication risk warning method according to claim 1, characterized in that, Generating corresponding warning information based on the comprehensive conflict detection result and a preset warning rule library includes: Based on the comprehensive conflict detection result, match the corresponding warning rule from the preset warning rule library; According to the matched warning rule, generate warning information including a description of the conflict mechanism and disposal suggestions.

6. The patient medication risk warning method according to claim 1, wherein The step of extracting chronic disease diagnosis information from the clinical examination data and past medical history includes: Perform natural language processing on the clinical examination data and past medical history to identify the chronic disease diagnosis information therein; the natural language processing uses a rule-based named entity recognition method.

7. The patient medication risk warning method according to claim 3, wherein, The step of constructing a patient drug metabolism ability feature vector based on the normalized indicators includes: Perform feature combination on the drug metabolism-related indicators after normalization to construct a feature vector of drug metabolism ability; the feature combination uses the weighted summation method.

8. The patient medication risk warning method according to claim 1, characterized in that The step of mapping the drug combination, comorbidity combination label, and drug metabolism ability feature vector in the prescription data to the clinical medication knowledge base includes: Convert the drug combination, comorbidity combination label, and drug metabolism ability feature vector into a unified data representation form; Based on the unified data representation form, perform matching retrieval in the clinical medication knowledge base.

Images