Medication Adherence Prediction and Assessment System for Patients with Chronic Diseases

Abstract

This invention discloses a medication adherence prediction and assessment system for patients with chronic diseases, relating to the field of chronic disease management technology. The system includes: a data acquisition module that collects patients' medication behavior and treatment effect data; a behavior adherence calculation module and an effect adherence calculation module that calculate behavior adherence scores reflecting the degree to which medication behavior conforms to the plan and effect adherence scores reflecting the degree to which treatment effects are achieved, respectively; a contribution confidence calculation module that couples the above adherence scores to obtain a contribution confidence score characterizing the credibility of the behavior's contribution to the effect; an overall adherence assessment module that makes dynamic decisions based on this: high-confidence drugs are directly included in the calculation, low-confidence drugs are excluded, and medium-confidence drugs are included after weight adjustment; and a result output module that generates a medication adherence assessment result based on weighted calculation. This invention can effectively improve the accuracy of the assessment.

Description

Technical Field

[0001] This invention relates to the field of chronic disease management technology, specifically to a medication adherence prediction and assessment system for patients with chronic diseases. Background Technology

[0002] In the field of chronic disease management, accurate assessment of patient medication adherence plays a crucial role in treatment effectiveness and disease control, which requires comprehensive analysis of patient medication behavior data and treatment effect data.

[0003] In practical applications, due to the complex influences of drug interactions, individual physiological differences, environmental factors, and concomitant treatments, behavioral adherence and treatment efficacy often show significant inconsistencies. For example, a patient may strictly adhere to medication on time, but the efficacy may be poor due to drug antagonism; or a patient may not fully comply with the medication plan, but achieve ideal results due to other interventions. Ignoring such inconsistencies and directly calculating adherence scores based on behavioral data for overall evaluation will lead to serious biases: when behavioral adherence is high but treatment efficacy is poor, the contribution of the relevant drugs may be overestimated, distorting the overall evaluation results; conversely, when behavioral adherence is low but efficacy is good, its actual role may be underestimated, hindering the optimization of treatment plans.

[0004] Such assessment distortion not only impairs the reliability of results but may also mislead clinical decisions, leading to decreased treatment efficiency and waste of medical resources. Existing technologies lack effective mechanisms to identify inconsistencies between behavior and effects, cannot dynamically quantify the credibility of behavioral data's contribution to effects, and are even less able to adaptively adjust data weights in the overall assessment to avoid distortion caused by conflicts, thus failing to meet the needs of clinical practice for accuracy and applicability.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a medication adherence prediction and assessment system for patients with chronic diseases.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] In a first aspect, this invention discloses a medication adherence prediction and assessment system for patients with chronic diseases, comprising:

[0009] The data acquisition module is used to acquire patients' medication behavior data and treatment effect data;

[0010] The behavioral compliance calculation module is used to calculate the behavioral compliance score for each drug based on medication behavior data. The behavioral compliance score represents the degree to which the actual medication behavior conforms to the preset plan.

[0011] The efficacy adherence calculation module is used to calculate the efficacy adherence score for each drug based on treatment efficacy data. The efficacy adherence score represents the degree to which the treatment effect achieves the expected improvement goal.

[0012] The contribution confidence calculation module is used to obtain the contribution confidence of each drug through coupled processing based on behavioral compliance score and effect compliance score. The contribution confidence represents the degree of credibility of the contribution of behavioral compliance to the treatment effect.

[0013] The overall compliance assessment module is used to determine whether the contribution confidence level meets the preset high confidence conditions. If so, the overall compliance is calculated using the behavioral compliance score with preset clinical importance weights.

[0014] Otherwise, further determine whether the contribution confidence level meets the preset low confidence condition. If yes, exclude the corresponding drug from the overall compliance calculation; otherwise, adjust the clinical importance weight according to the contribution confidence level before calculating the overall compliance.

[0015] The results output module is used to output the medication adherence assessment results based on the overall adherence calculation, which is based on the behavioral adherence score of each drug and its corresponding weight.

[0016] Secondly, this invention discloses a method for predicting and assessing medication adherence in patients with chronic diseases, comprising the following steps:

[0017] Obtain patient medication behavior data and treatment effect data;

[0018] Behavioral compliance scores for each drug are calculated based on medication behavior data. These scores represent the degree to which actual medication behavior conforms to the pre-set plan.

[0019] The efficacy adherence score for each drug is calculated based on the treatment efficacy data. The efficacy adherence score represents the degree to which the treatment effect achieves the expected improvement goal.

[0020] Based on behavioral compliance scores and effect compliance scores, the contribution confidence of each drug is obtained through coupling processing. The contribution confidence represents the degree of confidence in the contribution of behavioral compliance to the treatment effect.

[0021] To determine whether the contribution confidence level meets the preset high confidence condition, if so, the behavioral compliance score is used to calculate the overall compliance with the preset clinical importance weight.

[0022] Otherwise, further determine whether the contribution confidence level meets the preset low confidence condition. If yes, exclude the corresponding drug from the overall compliance calculation; otherwise, adjust the clinical importance weight according to the contribution confidence level before calculating the overall compliance.

[0023] The output is the medication adherence assessment result based on the overall adherence calculation, which is based on the behavioral adherence score of each drug and its corresponding weight.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0025] 1. By introducing a contribution confidence calculation module, the consistency between behavioral compliance score and effect compliance score is dynamically quantified, effectively identifying data inconsistencies caused by drug interactions, individual physiological differences or environmental factors, avoiding assessment distortion caused by the disconnect between behavior and effect, and ensuring that the overall compliance results more accurately reflect the patient's medication status.

[0026] 2. By constructing a drug relationship map and detecting conflict patterns (such as antagonistic drug loops and repeated treatment combinations) through a treatment pathway consistency verification mechanism, the system calculates the confidence level of the individualized consistency impact factor correction contribution. This enables the system to comprehensively consider the overall rationality of the treatment plan, provide clinicians with more accurate compliance assessments and intervention suggestions, and optimize the treatment pathway.

[0027] 3. In the overall compliance assessment module, the clinical importance weight of drugs is dynamically adjusted according to the contribution confidence level, and low-confidence drugs are excluded and their weights are redistributed. This mechanism ensures the integrity and balance of the weight allocation, avoids the assessment bias caused by fixed weights in traditional methods, and improves the clinical rationality of the overall calculation. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is an overall block diagram of the system according to Embodiment 1 of the present invention;

[0030] Figure 2 This is an overall block diagram of the method in Embodiment 2 of the present invention;

[0031] Figure 3 This is a flowchart illustrating the overall execution process of the method in Embodiment 2 of the present invention. Detailed Implementation

[0032] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Application Overview: In the field of chronic disease management, medication adherence assessment relies on the integration of medication behavior data and treatment effect data. However, existing technologies have significant shortcomings in data processing. Specifically, the dynamic relationship between behavioral adherence scores and effect adherence scores is not accurately modeled, causing the system to fail to identify data inconsistencies caused by drug interactions, individual physiological differences, or environmental factors. When behavioral adherence scores are high but effect adherence scores are low—for example, when patients strictly follow their medication plan but fail to achieve the expected therapeutic effect due to drug antagonism—traditional methods still use fixed weights for data fusion, resulting in an incorrect quantification of the credibility of behavioral data's contribution to treatment effectiveness. Consequently, the overall adherence calculation results deviate from the true state, directly affecting the reliability of the assessment system and the effectiveness of clinical decision-making.

[0034] For example, in the long-term management of hypertension, a patient continuously uses antihypertensive medication. Their medication adherence data, recorded by a smart pillbox, shows a behavioral adherence score of 0.95, indicating a high degree of consistency between actual medication use and the pre-set plan. However, continuously monitored blood pressure values ​​did not reach the expected improvement target, resulting in an efficacy adherence score of only 0.40. Further clinical analysis confirmed that this phenomenon stemmed from a drug interaction caused by the patient concurrently taking nonsteroidal anti-inflammatory drugs (NSAIDs), which suppressed the antihypertensive effect. In this scenario, existing assessment systems, lacking a mechanism to identify conflicts between behavioral and efficacy data, directly use the behavioral adherence score for weighted calculation, leading to an abnormally high overestimation of the medication's contribution to adherence. Consequently, the overall assessment result incorrectly indicates good patient adherence, thus masking the potential risks of drug interactions and interfering with the optimization of the treatment plan.

[0035] If the problem of assessment distortion caused by inconsistencies between behavioral adherence and treatment effectiveness data is not addressed, the system will continue to output unreliable adherence assessment results. Consequently, clinical decisions may be made based on distorted data to adjust medication regimens, such as maintaining ineffective medications or incorrectly increasing dosages, thereby exacerbating patient health risks. Furthermore, biased assessment results will reduce the efficiency of healthcare resource allocation, making treatment interventions unable to accurately match patients' actual needs, ultimately impacting the overall effectiveness of chronic disease management.

[0036] Example 1:

[0037] like Figure 1As shown, the medication adherence prediction and assessment system for patients with chronic diseases includes:

[0038] The data acquisition module is used to acquire patients' medication behavior data and treatment effect data;

[0039] Specifically, the data acquisition and preprocessing module first obtains the patient's raw medical data from multiple data sources, including electronic health record systems, patient-reported medication records, and automatic monitoring data from medical devices. The acquired data falls into two main categories: medication behavior data and treatment effect data. Medication behavior data includes drug identifiers, precise medication timestamps, and the actual dosage for each dose. Treatment effect data includes various clinical indicator measurements and their corresponding timestamps. These clinical indicators vary depending on the specific type of chronic disease, and may include quantitative indicators such as blood glucose, blood pressure, and blood lipids.

[0040] After data acquisition, the system enters the data preprocessing stage to perform time alignment on data from different source systems. Since medication behavior data and treatment effect data may come from different acquisition devices and recording systems, their timestamp formats and recording frequencies often differ. Therefore, a unified time window partitioning mechanism is needed to divide continuous time-series data into equal-length analysis windows, typically 24 hours or one week, depending on the needs of clinical assessment. When partitioning time windows, the system uses a time nearest neighbor matching algorithm to associate each medication record with the nearest treatment effect measurement record within the same analysis window, ensuring consistency between behavioral and effect data in the time dimension.

[0041] Next, the system types and defines the directionality of clinical indicator data, which forms the basis for subsequent efficacy evaluation. Based on predefined rules in the medical knowledge base, each clinical indicator is classified as either a positive or negative indicator. Positive indicators, such as hemoglobin levels, represent better treatment outcomes with higher values; negative indicators, such as blood glucose and blood pressure, represent better treatment outcomes with lower values. This classification process is automatically completed by querying the medical ontology database. The system maps each indicator to standard medical terminology and then determines its directionality based on the predefined attributes of the terminology in the knowledge base.

[0042] After classifying the indicators, the system calculates a baseline and current value for each clinical indicator. The baseline value is calculated using a sliding window averaging method, taking the average of the indicator measurements over several consecutive periods (usually four periods) prior to the current time window. This eliminates the influence of random fluctuations from a single measurement. The current value is the most recent measurement within the current time window; if multiple measurements exist within the window, the most recent one is used as the representative value. Through this baseline-current value comparison mechanism, the system can effectively capture the dynamic trends in treatment efficacy.

[0043] For medication behavior data, the system also performs cleaning and standardization. First, drug identifiers are normalized, mapping potentially different drug codes from different sources to a standard drug dictionary to ensure consistent representation of the same drug across different data sources. Next, based on the preset dosing plan for each drug, including daily frequency and dosage, the system constructs the expected dosing pattern for each drug within each time window, serving as the benchmark for subsequent adherence calculations.

[0044] Finally, all the preprocessed data is organized into a standardized analysis dataset, which is stored in a hierarchical structure of patient identifier, time window, and drug identifier. Each record contains complete medication behavior characteristics and treatment effect indicators, providing high-quality data input for the subsequent adherence calculation module.

[0045] The formula for calculating the baseline values ​​of clinical indicators is as follows:

[0046]

[0047] in, This represents the baseline value of the clinical indicator, where N represents the number of historical time windows used to calculate the baseline value. This represents the measured value of the indicator within the t-th historical time window.

[0048] Nearest neighbor matching formula with time window alignment:

[0049]

[0050] in, Indicates the time window for matching. Indicates the timestamp of medication administration. Indicates the timestamp for measuring the treatment effect.

[0051] Taking a type 2 diabetes patient as an example, the system retrieves the following raw data from their electronic health record: Medication records show the patient took "metformin 500mg" with timestamps of "2024-03-20 08:30:00" and "2024-03-20 19:15:00"; blood glucose monitoring data shows "fasting blood glucose 6.8 mmol / L" with a measurement time of "2024-03-20 07:45:00". In the preprocessing stage, the system first maps the drug identifier "metformin 500mg" to the standard drug code "A10BA02", then divides the time window into days, placing all the above data into the "2024-03-20" analysis window. Since blood glucose is a negative indicator, the system calculates the patient's average fasting blood glucose value of 7.5 mmol / L over the past 4 days as the baseline value, and the current measurement value of 6.8 mmol / L as the current value. After this preprocessing, a standardized record is generated: {Patient ID, Time Window, Medication A10BA02, Dosage Frequency 2, Planned Dosage Frequency 2, Baseline Blood Glucose 7.5, Current Blood Glucose 6.8}, providing complete and consistent input data for subsequent adherence analysis.

[0052] The behavioral compliance calculation module is used to calculate the behavioral compliance score for each drug based on medication behavior data. The behavioral compliance score represents the degree to which the actual medication behavior conforms to the preset plan.

[0053] The efficacy adherence calculation module is used to calculate the efficacy adherence score for each drug based on treatment efficacy data. The efficacy adherence score represents the degree to which the treatment effect achieves the expected improvement goal.

[0054] Specifically, the behavioral compliance calculation module and the effect compliance calculation module receive standardized analysis datasets from preprocessing and execute two independent calculation processes in parallel: behavioral compliance score calculation and effect compliance score calculation. These two calculation processes share the same time window and drug identifiers, but employ completely different calculation logic and evaluation criteria, thereby forming a comprehensive and multi-dimensional assessment perspective of patient medication use.

[0055] In the behavioral compliance calculation process, the system first generates a desired medication pattern based on the preset medication plan for each drug. This pattern defines in detail the expected number of doses, dosage, and timing of doses within each time window. Next, the system compares actual recorded medication behaviors with this desired pattern, quantifying behavioral deviations by calculating time matching and dosage compliance. Specifically, for each planned dose, the system searches for a matching actual medication record within a reasonable time tolerance window. If a match is found, it is considered valid medication; otherwise, it is marked as a missed dose. Simultaneously, for each actual dose, the system verifies whether the dosage falls within the acceptable deviation range of the preset dosage. Finally, the behavioral compliance score is calculated as the ratio of actual valid doses to planned doses, taking into account the influence of dosage compliance.

[0056] The formula for calculating the behavioral compliance score is as follows:

[0057]

[0058] in, Indicates behavioral compliance score, This indicates the actual number of effective medication doses taken within the time window. This indicates the planned number of medication doses within the same time window.

[0059] The efficacy adherence calculation process focuses on the actual clinical effect of treatment rather than the medication behavior itself. The system first selects an appropriate improvement rate calculation formula for each clinical indicator based on the directionality determined in the preprocessing phase. For negative indicators, the improvement rate is calculated as the ratio of the difference between the baseline value and the current value to the baseline value; for positive indicators, it is calculated as the ratio of the difference between the current value and the baseline value to the baseline value. This calculation method ensures that the improvement rate accurately reflects the positive change in treatment effect regardless of the indicator's direction. Next, the system compares the calculated improvement rate with a preset clinical improvement threshold. If the threshold is reached or exceeded, the efficacy adherence score for that drug is marked as full; otherwise, it is marked as minimum.

[0060] The formula for calculating the improvement rate is as follows:

[0061] For negative indicators:

[0062]

[0063] For positive indicators:

[0064]

[0065] in, Indicates the improvement rate. This indicates the current indicator value. This represents the baseline index value.

[0066] Furthermore, the mapping formula for effect compliance scores is:

[0067]

[0068] in, Indicates compliance score. This indicates the preset clinical improvement threshold.

[0069] Key parameters involved in the above process, such as the time tolerance window width, the allowable range of dose deviation, and the clinical improvement threshold, are all centrally managed through the parameter configuration module, allowing for flexible adjustments based on different drug types, disease stages, and individual patient characteristics. This parameterized design enables the system to adapt to diverse clinical scenarios, providing accurate adherence assessments from stable chronic disease management to complex multi-disease coexistence situations.

[0070] For example, consider the management of a hypertensive patient currently taking two antihypertensive medications: amlodipine (once daily) and valsartan (once daily). Within a 7-day time window, the actual number of amlodipine doses recorded was 6, and the actual number of valsartan doses recorded was 7, both corresponding to the planned dosing frequency of 7. Simultaneously, blood pressure monitoring data shows that the patient's baseline blood pressure was 150 / 95 mmHg, and the average blood pressure within the current time window was 138 / 88 mmHg.

[0071] In the behavioral compliance calculation, the score for amlodipine was calculated as 6 / 7 × 100% = 85.7%, and the score for valsartan was 7 / 7 × 100% = 100%. In the effect compliance calculation, since blood pressure is a negative indicator, the improvement rate was calculated as (150-138) / 150 × 100% = 8% (systolic blood pressure) and (95-88) / 95 × 100% = 7.4% (diastolic blood pressure). Assuming the preset improvement threshold is 5%, both indicators met the threshold requirement, so the effect compliance scores for both drugs were marked as 1. Thus, the system generated complete two-dimensional evaluation results for each drug: amlodipine {behavioral score 85.7%, effect score 1}, valsartan {behavioral score 100%, effect score 1}, providing complete input data for subsequent confidence assessment.

[0072] The contribution confidence calculation module is used to obtain the contribution confidence of each drug through coupled processing based on behavioral compliance score and effect compliance score. The contribution confidence represents the degree of credibility of the contribution of behavioral compliance to the treatment effect.

[0073] Specifically, the contribution confidence calculation module receives behavioral compliance scores and effect compliance scores as inputs. It quantifies the actual contribution value of each drug in the treatment regimen by establishing a correlation model between the two dimensions. The core task of this module is to address the disconnect between behavior and effect in traditional assessment methods. By calculating contribution confidence, it identifies special cases where medication adherence is followed but treatment outcomes are poor, or where medication adherence is flawed but treatment outcomes are good.

[0074] The module first performs standardization preprocessing on the scores of the two input dimensions, normalizing behavioral compliance scores and effect compliance scores to the same numerical range, typically between zero and one. This standardization process ensures that scores of different dimensions can be compared and calculated within the same mathematical framework, laying the foundation for subsequent confidence level calculations. The standardization method employs a min-max normalization algorithm, performing a linear transformation based on the theoretical maximum and minimum values ​​of each score, preserving the distribution characteristics of the original data while eliminating dimensional differences.

[0075] After standardizing the scores, the module calculates the absolute difference between the two scores. This difference directly reflects the degree of inconsistency between behavioral compliance and treatment effect. A large difference indicates a significant deviation between the actual and expected therapeutic effect of the drug, which may be due to individual drug sensitivity, drug interactions, or other clinical factors. At this point, the module introduces a coupling coefficient to moderate the strength of this inconsistency's impact on the final confidence level. The magnitude of the coupling coefficient reflects the system's sensitivity to behavioral-effect divergence.

[0076] Based on the difference value and coupling coefficient, the module calculates the initial drug contribution confidence level through a linear transformation. The calculation process uses a baseline value minus the product of the difference value and the coupling coefficient to ensure that the confidence level decreases as the behavior-effect inconsistency increases. This calculation method effectively identifies drugs with a high degree of consistency between behavior and effect; these drugs typically have high confidence levels, indicating that the medication behavior indeed brings about the expected therapeutic effect.

[0077] The fraction normalization formula is as follows:

[0078]

[0079] in, This represents the normalized fraction. Represents the raw fraction. This represents the minimum theoretical value of this type of score. This represents the maximum theoretical value.

[0080] The formula for calculating the confidence level of a drug's contribution is:

[0081]

[0082] in, Indicates the confidence level of drug contribution. This represents the normalized behavioral compliance score. This represents the normalized effect compliance score. The coupling coefficient representing behavior and effect.

[0083] For example, a diabetic patient uses two hypoglycemic drugs, metformin and glimepiride. During the current assessment period, the behavioral compliance score for metformin was 90%, and the effect compliance score was 1 (based on achieving glycemic target improvement), which, after normalization, were 0.9 and 1.0, respectively; the behavioral compliance score for glimepiride was 80%, and the effect compliance score was 0 (glycemic target improvement not achieved), which, after normalization, were 0.8 and 0, respectively.

[0084] Assuming the current coupling coefficient is set to 1.0, the confidence level of metformin's contribution is calculated as follows: This indicates that the drug's behavior and effect are highly consistent, with a high confidence level. The contribution confidence level of glimepiride is calculated as follows: This indicates that although the medication was taken reasonably well, the therapeutic effect was poor, hence the low confidence level.

[0085] The overall compliance assessment module is used to determine whether the contribution confidence level meets the preset high confidence conditions. If so, the overall compliance is calculated using the behavioral compliance score with preset clinical importance weights.

[0086] Otherwise, further determine whether the contribution confidence level meets the preset low confidence condition. If yes, exclude the corresponding drug from the overall compliance calculation; otherwise, adjust the clinical importance weight according to the contribution confidence level before calculating the overall compliance.

[0087] Specifically, for drugs whose contribution confidence level reaches or exceeds the high confidence threshold, the system classifies them as high-confidence drugs. These drugs exhibit good behavior-effect consistency and are therefore assigned their full clinical importance weight, which is fully retained and participates in subsequent overall adherence calculations. These drugs typically represent core components of a treatment regimen with clear effects and good patient responses, and their weight in the overall assessment is not affected by discounting.

[0088] When a drug's contribution confidence level falls between high and low thresholds, the system identifies it as a medium-confidence drug, which then undergoes weight adjustment. The weight adjustment process first calculates a confidence-based adjustment factor, which is linearly determined by the relative position of the drug's actual confidence level within the threshold range, ensuring the continuity and smoothness of the adjustment. This adjustment factor is then multiplied by the drug's clinical importance weight to obtain the final participation weight. This calculation method results in a smaller weight discount for drugs with confidence levels closer to the high threshold, and a correspondingly larger weight discount for drugs closer to the low threshold.

[0089] For drugs whose contribution confidence level is lower than or equal to the low confidence threshold, the system includes them in the category of low confidence drugs. These drugs usually exhibit significant behavior-effect inconsistency and are therefore temporarily excluded from the overall compliance calculation for the current period, with their weight set to zero.

[0090] The formula for calculating the weighting adjustment factor is as follows:

[0091]

[0092] in, Indicates the weighting adjustment factor. Indicates the confidence level of drug contribution. Indicates a low confidence threshold. This indicates a high confidence threshold.

[0093] The formula for calculating the drug weights of medium confidence level is:

[0094]

[0095] in, This represents the final weight of the i-th medium-confidence drug. This indicates the clinical importance weight of the drug. This represents the corresponding weight adjustment factor.

[0096] Among them, clinical importance weight It is a static weight value that reflects the relative importance of drug i in the patient's current treatment plan. Specifically, drugs can be classified and assigned a baseline weight according to clinical guidelines and drug instructions (for example, the weight of core treatment drugs is 1.0, the weight of adjuvant treatment drugs is 0.7, and the weight of symptomatic treatment drugs is 0.5).

[0097] The results output module is used to output the medication adherence assessment results based on the overall adherence calculation, which is based on the behavioral adherence score of each drug and its corresponding weight.

[0098] The results output module, as the final stage of the entire evaluation process, is responsible for integrating all intermediate results generated by the preceding modules to produce a comprehensive and actionable adherence assessment report. This module receives the final weighting results, behavioral adherence scores for each drug, and processing records. Through an intelligent synthesis algorithm, it integrates these scattered assessment elements into a unified overall adherence index, while retaining necessary details for clinical decision-making reference.

[0099] The module first verifies the completeness of all input data, ensuring that all necessary evaluation elements are in place and their values ​​are within reasonable ranges. After successful verification, the system proceeds to the overall compliance calculation stage. This stage uses a weighted average algorithm to combine the behavioral compliance scores of each drug with their corresponding final weights. Since the system has already adjusted the weights during the weight calculation process, the weighted average here actually reflects the optimized drug contribution assessment.

[0100] After calculating the basic overall adherence score, the module further analyzes the characteristics of the weight distribution to identify key influencing factors in the assessment results. For example, the system checks whether certain high-weighted drugs have a decisive impact on the overall score, or whether a large group of drugs with lower weights has a cumulative effect on the results. This distribution characteristic analysis helps to understand the internal structure of overall adherence and provides directional guidance for subsequent personalized interventions. In addition to numerical calculations, the module also generates a structured assessment report. This report not only includes the final overall adherence score but also details the specifics of each component. The report clearly marks the drugs that were temporarily excluded and the reasons for their exclusion, records the specific process and basis for weight adjustment, and highlights any anomalies and special patterns found during the assessment process.

[0101] The formula for calculating the overall compliance score is as follows:

[0102]

[0103] in, This represents the overall compliance score. This represents the behavioral compliance score for the i-th drug. Let S represent the final weight of the i-th drug, and let S represent the total number of drugs participating in the evaluation.

[0104] For example, in the case of a hypertensive patient taking four antihypertensive drugs, the final weighting of the four drugs is as follows: amlodipine weight 1.0, behavior score 90%; valsartan weight 0.8, behavior score 85%; hydrochlorothiazide weight 0, behavior score 70% (excluded due to low confidence); atenolol weight 0.6, behavior score 80%.

[0105] Overall compliance score is calculated as follows:

[0106] ;

[0107] The system-generated assessment report details the following: hydrochlorothiazide was temporarily excluded because its contribution confidence level was below the threshold for two consecutive assessment periods; valsartan and atenolol had their weights adjusted due to their moderate confidence levels; amlodipine, as a high-confidence drug, retained its full weight. The report also indicates that the data completeness and calculation consistency of this assessment met excellent standards, and recommends that clinicians focus on the therapeutic efficacy of hydrochlorothiazide and consider whether adjustments to the treatment regimen for this drug are necessary.

[0108] Therefore, this technical solution effectively solves the problem of assessment distortion caused by inconsistencies between behavioral compliance and treatment effect data. Through a dynamic quantification mechanism that contributes confidence scores, the system can identify deviations between behavioral execution and effect feedback, and adaptively adjust the weight allocation strategy based on the confidence level. When behavioral and effect data are highly consistent, behavioral compliance scores dominate the assessment; when significant inconsistencies exist, the system automatically reduces or eliminates the influence of unreliable data, thus avoiding assessment bias caused by the simple assumption of a positive correlation in traditional methods. Ultimately, the overall compliance assessment results more accurately reflect the patient's true medication status, significantly improving the reliability and clinical applicability of the assessment, and providing more effective decision support for chronic disease management.

[0109] In some of the embodiments described above in this application, a method is proposed to exclude low-confidence drugs from the overall compliance calculation. However, in its implementation, the clinical importance weight of the excluded drugs is directly discarded, resulting in an imbalance in the weight ratio of the remaining high-confidence drugs. This fails to accurately reflect the relative importance of each drug in the treatment plan, thereby causing the overall compliance assessment results to deviate from the true situation and affecting the reliability of clinical decision-making.

[0110] In response, this application further proposes that when there are drugs excluded from the overall adherence calculation, a weight redistribution should also be included:

[0111] Calculate the sum of the clinical importance weights of all excluded drugs;

[0112] The sum of the weights is proportionally redistributed to the drugs that meet the high confidence condition; the formula for calculating the weight redistribution is as follows:

[0113]

[0114] in, This represents the weight of the i-th drug after redistribution. This represents the weight of the i-th drug before redistribution. This represents the sum of the clinical importance weights of all excluded drugs. This represents the clinical importance weight of the i-th drug. This represents the sum of the clinical importance weights of all high-confidence drugs.

[0115] The calculation of the sum of the clinical importance weights of all excluded drugs involves numerically accumulating the clinical importance weights of the excluded drugs to accurately quantify the total weight value that needs to be redistributed, thus avoiding arbitrariness in weight processing. Redistributing the sum of weights proportionally to drugs meeting the high-confidence criteria can be understood as allocating additional weights based on the original weight ratios of the high-confidence drugs. This aims to maintain a dynamic balance in the relative importance of drugs and prevent distortion of the treatment plan structure due to weight loss. The weight redistribution calculation formula refers to dynamically adjusting the weight values ​​through mathematical expressions. This aims to ensure that the weight redistribution process is closely related to the relative contribution of the drugs in the treatment plan, guaranteeing the rigor of the assessment.

[0116] Specifically, when the system determines that there are excluded drugs, the weighting unit first obtains the clinical importance weight data of the excluded drugs and performs an accumulation operation to obtain the total weight. Then, based on the original clinical importance weight ratio of the high-confidence drugs, the total weight is allocated to each high-confidence drug proportionally. In this process, the weight redistribution formula generates an adjustment term by multiplying the total weight of the excluded drugs by the weight ratio of the high-confidence drugs, and adds it to the original weight to form the redistributed weight value. This mechanism ensures the integrity of the total weight in the overall compliance calculation, while maintaining the proportional relationship of relative importance among high-confidence drugs, avoiding the assessment distortion caused by simply discarding weights, and effectively improving the clinical rationality of weight allocation.

[0117] For example, consider a hypertensive patient taking three medications: amlodipine, irbesartan, and hydrochlorothiazide. Hydrochlorothiazide is excluded from the overall adherence calculation because its contribution confidence level does not meet the criteria. The system first calculates the clinical importance weight of hydrochlorothiazide and sums it with the weights of the other excluded medications. Then, based on the original clinical importance weight ratios of the remaining high-confidence medications, amlodipine and irbesartan, this sum of weights is redistributed to these two medications. For example, if the original weight ratio of amlodipine to irbesartan is 3:2, the sum of weights is distributed according to this ratio, thereby adjusting the weight values ​​of each medication to ensure that the overall adherence assessment reflects the true treatment regimen structure.

[0118] The above technical solutions effectively solve the problem of weight imbalance caused by the loss of weight for excluded drugs, ensuring the integrity and clinical rationality of weight allocation in the overall adherence assessment, avoiding deviation of assessment results from the actual situation, and providing a more reliable basis for medication adherence assessment for clinical decision-making.

[0119] In the field of chronic disease management, accurate assessment of patient medication adherence is crucial for treatment effectiveness and disease control, which typically involves a comprehensive analysis of patient medication behavior and treatment outcome data. Traditionally, medication adherence assessment has relied primarily on medication behavior data, such as pillbox records or electronic monitoring devices, to quantify the consistency between actual medication use and the treatment plan. However, with the development of medical datafication, some methods have begun to incorporate treatment outcome indicators, such as improvements in physiological parameters or the degree of symptom relief, to more comprehensively reflect adherence. However, current technological solutions often employ simple weighting or direct aggregation when integrating these two types of data, assuming a stable positive correlation between behavioral adherence and treatment outcome—that is, high behavioral adherence necessarily leads to good treatment results. In reality, however, inconsistencies often arise between behavioral adherence and treatment outcome due to drug interactions, individual physiological differences, or environmental factors. For example, a patient may take medication on time but experience poor efficacy due to drug antagonism, or may not adhere strictly to medication but experience good results due to other treatment methods. If this inconsistency is ignored, directly calculating adherence scores based on behavioral data and conducting an overall assessment will lead to results that deviate from reality. For example, when behavioral adherence is high but treatment efficacy is poor, the contribution of the drug to adherence may be overestimated, thus distorting the overall assessment; conversely, when behavioral adherence is low but efficacy is good, its role may be underestimated, affecting the optimization of treatment plans. This bias not only reduces the reliability of the assessment but may also mislead clinical decisions, leading to decreased patient treatment efficiency or wasted resources.

[0120] In some of the schemes mentioned above in this application, contribution confidence is proposed to quantify the credibility of the contribution of behavioral compliance to treatment effect. However, this process does not consider the conflict patterns formed by drug interactions, which makes the contribution confidence unable to accurately reflect the impact of the overall rationality of the treatment plan on individual drug evaluation, thus leading to the overall compliance calculation results deviating from the actual situation.

[0121] To address this, this application further proposes that after obtaining the contribution confidence of each drug through coupling processing, it also includes treatment pathway consistency verification: constructing a drug relationship graph based on the patient's complete medication regimen, where nodes represent drugs and edges represent interactions between drugs, with edge weights determined based on conflict intensity values ​​in a drug interaction database; detecting conflict patterns in the drug relationship graph, including antagonistic drug loops and repeated treatment combinations; calculating an individualized consistency impact factor for each drug based on the conflict pattern detection results, used to quantify the negative impact of drug conflicts on the individual contribution confidence of that drug; and multiplying the individualized consistency impact factor of each drug with its contribution confidence to obtain an individualized corrected confidence based on the overall rationality of the treatment regimen.

[0122] Among them, treatment pathway consistency verification refers to the technical process of systematically evaluating the rationality of drug treatment regimen structure. This can be achieved using graph theory analysis or network topology methods, aiming to improve the accuracy of adherence assessment by identifying potential drug conflicts through structured detection. Drug relationship graphs can be understood as network models representing drug interactions, where nodes can be specific drug entities. Edge construction can be based on data from public drug interaction databases such as DrugBank or the FDA Adverse Drug Reaction Reporting System, and edge weights can be quantified according to the conflict intensity value of the interaction. The aim is to provide a precise structured data foundation to support conflict pattern detection. Conflict patterns refer to specific topological structures in drug combinations that lead to reduced treatment efficacy, such as antagonistic drug loops (forming closed loops). Antagonistic pathways and repeated treatment combinations (multiple drugs targeting the same treatment goal) can be identified using depth-first search algorithms or community detection techniques, aiming to specifically capture key conflict types affecting treatment efficacy; the individualized consistency impact factor is an indicator that quantifies the degree of negative impact of drugs in conflict, which can be achieved based on conflict penalty scores through linear decay or piecewise function mapping, aiming to reflect the actual role of drugs in specific conflicts and avoid individual assessment distortion caused by overall averaging; the individualized corrected confidence can be understood as the contribution confidence after conflict correction, which is obtained by multiplying the individualized consistency impact factor by the original contribution confidence, aiming to naturally decay the confidence value to dynamically integrate the conflict impact, so that subsequent assessments can exclude drug conflict interference.

[0123] Specifically, the proposed solution introduces a treatment pathway consistency verification mechanism after calculating the contribution confidence score. First, a drug relationship graph is constructed based on the patient's complete medication regimen, representing drugs as nodes and drug interactions as edges. Edge weights are assigned based on conflict intensity values ​​in the drug interaction database, forming a structured model reflecting the drug interaction network. Then, conflict patterns in the graph are detected, including identifying closed-loop antagonistic drug loops (e.g., a loop where drug A antagonizes B, B antagonizes C, and C antagonizes A) and repeated treatment combinations targeting the same therapeutic goal (e.g., multidrug therapy targeting the same target). These patterns directly relate to potential offsetting or overdose risks of treatment effects. Based on the detection results, an individualized consistency impact factor is calculated for each drug. This factor quantifies the depth and breadth of drug involvement in the conflict to characterize its negative impact. Finally, the individualized consistency impact factor is multiplied by the contribution confidence score to obtain the corrected confidence score. This process ensures that the confidence score not only reflects the consistency between behavior and effect but also incorporates dynamic considerations of the overall rationality of the treatment regimen, thus adaptively adjusting weights in the overall adherence assessment and avoiding assessment distortion caused by drug conflicts.

[0124] As a preferred embodiment, the specific implementation of this application is as follows: Consider a patient suffering from both hypertension and hyperlipidemia, whose medication regimen includes antihypertensive drug A, antihypertensive drug B, and lipid-lowering drug C. The system first constructs a drug relationship graph based on a drug interaction database, with nodes corresponding to the three drugs A, B, and C respectively; edge AB represents the potential antagonistic effect between two antihypertensive drugs, with its weight set to a higher value to reflect the intensity of significant conflict; edges AC and BC represent no significant interaction between the antihypertensive drug and the lipid-lowering drug, with their weights set to lower values. When detecting conflict patterns, the system identifies that edge AB forms an antagonistic drug loop (because both A and B are antihypertensive drugs but there is a cyclic antagonistic path), and A and B belong to a repeated treatment combination for hypertension. For drug A, the system calculates its conflict penalty score: based on the antagonistic loop and repeated combination involved, the mapped individualized consistency impact factor is a lower value; similarly, the impact factor of drug B is also a lower value, while the impact factor of drug C remains at 1.0 because it does not participate in the conflict. Multiplying the impact factor by the original contribution confidence level: the corrected confidence levels for drugs A and B decrease accordingly, while the confidence level for drug C remains unchanged. Finally, the overall compliance assessment module uses the corrected confidence levels for calculation, effectively avoiding assessment bias caused by drug conflicts between A and B.

[0125] Through the above scheme, this application can effectively identify the conflict patterns formed by drug interactions, so that the contribution confidence accurately reflects the impact of the overall rationality of the treatment plan on individual drug assessment, thereby avoiding the deviation of results caused by drug conflicts in the overall adherence calculation, and improving the accuracy and clinical reliability of medication adherence assessment.

[0126] In some of the embodiments described above in this application, an individualized consistency impact factor is proposed to quantify the negative impact of drug conflict. However, in its implementation, there is a lack of a specific method for calculating the conflict penalty score, which may lead to the calculation of the impact factor being too general or subjective. It is unable to effectively distinguish the severity differences of different conflict modes such as antagonistic drug loops and repeated treatment combinations, thereby affecting the accuracy of subsequent correction confidence and reducing the reliability of treatment pathway consistency verification.

[0127] In this regard, this application further proposes the calculation process for the individualized consistency impact factor, including:

[0128] Identify all conflict patterns involved by each drug, including its antagonistic drug loop and repeated treatment combinations;

[0129] Calculate the conflict penalty score for each conflict mode it participates in, and obtain the individualized consistency impact factor based on the conflict penalty score mapping;

[0130] The formula for calculating conflict penalty points is:

[0131]

[0132] in, Let i be the conflict penalty score for drug i. This represents the set of antagonistic rings containing drug i. This represents the conflict score of the m-th antagonistic ring. Indicates antagonistic weights, This indicates a combination of repeated treatments including drug i. This represents the overlap coefficient of the nth combination. This indicates the weight of repetition.

[0133] Among them, the antagonistic drug loop refers to a closed-loop path in the drug relationship graph consisting of drug nodes and edges representing antagonistic effects. It can be identified using depth-first search algorithms or Floyd's loop detection algorithms, and the antagonistic properties of the edges can be verified through a drug interaction database. Its purpose is to systematically capture drug-drug interaction conflicts, avoid missing key conflicts, and provide a complete data foundation for conflict quantification. Repeated treatment combinations refer to drug sets formed based on treatment goals according to a medical knowledge base. They can be implemented using a combination of treatment goal category classification and threshold detection, such as judging redundancy by setting a maximum allowed number of drugs. Its purpose is to identify treatment... The plan addresses the issue of duplicate medication, avoiding simplistic assessments based solely on the quantity of drugs. The conflict penalty score is a quantitative indicator comprehensively reflecting the severity of the conflict modes in which a drug participates. It can be obtained by multiplying the contribution scores of different conflict modes by their corresponding weights using a weighted summation framework. Its purpose is to provide an adaptive metric that dynamically distinguishes between high-risk and low-risk conflicts. The individualized consistency impact factor is a decay factor mapped from the conflict penalty score. This mapping can be achieved using exponential decay functions, linear decay functions, or piecewise functions. Its purpose is to transform the conflict penalty into a basis for correcting contribution confidence, ensuring that negative impacts are accurately attenuated.

[0134] Specifically, the scheme first performs a conflict pattern identification step, scanning the drug relationship graph to determine the antagonistic drug loop and repeated treatment combination to which each drug belongs. The output of this step serves as the input data source for subsequent calculations. Next, a conflict penalty score calculation step is performed. Based on the identified conflict patterns, the contribution values ​​of the antagonistic loop and repeated combination components are calculated separately. The antagonistic loop component obtains a conflict score by aggregating the drug interaction strength along the loop, while the repeated combination component obtains an overlap coefficient by integrating the similarity between the number of drugs and treatment intensity. The scores of each component are then multiplied by preset weights and summed to form a single conflict penalty score. Finally, a mapping step is performed, inputting the conflict penalty score into a preset function to generate an individualized consistency impact factor. This information flow from conflict pattern data through penalty score calculation to impact factor generation ensures the coherence and accuracy of the quantification process, effectively distinguishing the severity differences between different conflict patterns, thus providing a reliable basis for verifying treatment pathway consistency.

[0135] Through the above scheme, this application achieves refined calculation of conflict penalty scores, which can effectively distinguish the severity differences of different conflict patterns such as antagonistic drug loops and repeated treatment combinations, making the quantification of individualized consistency influencing factors more accurate, thereby improving the reliability of treatment pathway consistency verification and ensuring that medication adherence assessment results are more in line with clinical practice.

[0136] In some of the embodiments described above in this application, an individualized consistency impact factor is proposed to quantify the negative impact of drug conflict on contribution confidence. However, in its implementation, the mapping method of conflict penalty score lacks reasonable constraints, which may cause the impact factor to exceed the effective range or change discontinuously, resulting in distortion of the corrected confidence and thus affecting the accuracy of overall medication adherence assessment.

[0137] In response, this application further proposes that the process of obtaining the individualized consistency influence factor based on the conflict penalty score mapping is implemented using an exponential decay function:

[0138] The exponential decay function ensures that the higher the conflict penalty score, the lower the individualized consistency impact factor, and that the factor ranges from zero to one; the mapping calculation formula is:

[0139]

[0140] in, This represents the factor influencing the individualized consistency of drug i. This represents the conflict penalty score for drug i. This indicates the preset attenuation coefficient.

[0141] Among them, the exponential decay function refers to a nonlinear mapping mechanism, which can be implemented in the form of a natural exponential function. Its purpose is to ensure that the output value decreases continuously as the input increases and is limited to between zero and one. The decay coefficient k refers to a preset adjustment parameter, which can be implemented as a fixed constant or dynamically adjusted according to clinical needs. Its purpose is to control the sensitivity of the conflict penalty score to the influence of the individualized consistency factor.

[0142] Specifically, the proposed solution generates an individualized consistency impact factor by inputting the conflict penalty score into an exponential decay function. This factor is then corrected by multiplying it by the contribution confidence score. Due to the natural decay characteristics of the exponential function, the increase in the conflict penalty score leads to a smooth decrease in the impact factor, avoiding abrupt corrections caused by step changes. At the same time, the output range of the function is automatically constrained between zero and one, ensuring the mathematical validity and continuity of the correction process. This ensures that the negative impact of drug conflict on the contribution confidence score can be progressively quantified.

[0143] Through the above scheme, this application ensures that the individualized consistency influencing factors change continuously within the effective range, avoiding mutations and distortions in the correction process, thereby improving the accuracy of treatment pathway consistency verification and providing a more reliable data foundation for overall medication adherence assessment.

[0144] In some of the embodiments described above in this application, a conflict score is proposed to quantify the degree of conflict in the antagonistic loop. However, in its implementation, due to the lack of a clear definition of the specific calculation method for the conflict score, it is impossible to accurately capture the overall intensity distribution characteristics of the antagonistic effect between drugs. This results in a lack of objective basis for the calculation of the individualized consistency impact factor, which in turn causes deviations in the evaluation results of the treatment pathway consistency verification link and affects the reliability of the overall medication adherence assessment.

[0145] In this regard, this application further proposes a calculation process for the conflict score, including:

[0146] Identify all closed-loop paths in the drug relationship graph, where each edge represents an antagonistic effect between drugs;

[0147] For each identified antagonistic loop, the interaction strength of all drug pairs in the loop is extracted;

[0148] The conflict score of this loop is calculated using the geometric mean method; the formula for calculating the conflict score is:

[0149]

[0150] in, Let L represent the conflict score of the m-th antagonistic loop, and L represent the number of drugs in that loop. This represents the strength of the interaction between drugs i and j; a negative value indicates antagonism. This represents the m-th identified antagonistic drug ring.

[0151] Specifically, identifying closed-loop paths refers to identifying all closed-loop structures in the drug relationship graph. This can be achieved using loop detection algorithms in graph theory (such as depth-first search or breadth-first search). The aim is to focus on cyclic conflict scenarios composed of antagonistic interactions, avoiding interference from non-closed-loop paths in conflict assessment. Extracting interaction strength refers to obtaining the absolute value of the antagonistic strength of drug pairs. This can be retrieved from standardized drug interaction databases (such as DrugBank or Micromedex). The purpose is to ensure the objectivity and clinical relevance of the strength data, avoiding errors caused by subjective estimation. In practical applications, the geometric mean method refers to calculating the geometric mean of a set of values. This can be achieved using numerical calculation techniques (such as calculating the arithmetic mean after logarithmic transformation and then exponentially restoring). The purpose is to balance the comprehensive influence of the strength values ​​of multiple drug pairs in the loop, especially reflecting the synergistic characteristics of antagonistic effects in multi-drug loops, and avoiding the excessive dominance of individual high-intensity values ​​in the results.

[0152] Specifically, the proposed solution first identifies all closed-loop pathways consisting of antagonistic interactions, ensuring that conflict assessment focuses solely on cyclic conflict characteristics. Then, it extracts the absolute values ​​of the interaction strengths of all drug pairs on each antagonistic loop, providing an objective data foundation based on a drug interaction database. Finally, it applies a geometric mean method to calculate the conflict score. This method eliminates the influence of the sign of the antagonistic interaction by taking the absolute value and dynamically correlates the score with the number of drugs in the loop through a square root of L design. This provides comparable conflict strength quantification across loops of different sizes, effectively balancing the combined influence of multiple strength values, avoiding excessive dominance of individual high-intensity values, and ensuring that the conflict score accurately reflects the distribution characteristics of the overall antagonistic degree.

[0153] Through the above technical solution, this application can accurately quantify the overall intensity distribution characteristics of drug antagonism, providing a standardized and objective basis for the calculation of individualized consistency influencing factors, thereby reducing the assessment bias in the treatment pathway consistency verification process and improving the reliability of medication adherence assessment.

[0154] In some of the embodiments described above in this application, an overlap coefficient is proposed to quantify the degree of conflict in repeated treatment combinations. However, in its implementation, the calculation of the overlap coefficient lacks a specific method. Relying solely on the number of drugs may lead to inaccurate assessment of the degree of repeated treatment, failing to reflect the similarity difference in treatment intensity between drugs, thereby affecting the reliability of the conflict penalty score and ultimately weakening the correction effect of the individualized consistency influencing factor.

[0155] In this regard, this application further proposes a calculation process for the overlap coefficient, including:

[0156] Drugs are classified according to their treatment goals based on a medical knowledge base, and drugs under the same treatment goal category are included in the repeat treatment test.

[0157] The quantity overlap component is calculated based on the number of drugs in the category, and the intensity overlap component is calculated based on the similarity of the therapeutic intensity of drugs in the category.

[0158] The final repeated treatment overlap coefficient is obtained by combining the quantity overlap component and the intensity overlap component;

[0159] The formula for calculating the overlap factor of repeated treatments is:

[0160]

[0161] in, This represents the overlap coefficient of the nth treatment target category. This represents the set of drugs included in the nth treatment target category. This indicates the number of drugs in that category. This indicates the maximum number of drugs allowed in this category (preset). This indicates the similarity in therapeutic intensity between drug i and drug j. Indicates from The number of combinations of choosing 2 out of 10 drugs.

[0162] In practical applications, a medical knowledge base refers to an authoritative data source storing standardized drug treatment target information. This can be implemented using DrugBank, clinical practice guideline databases, or professional medical knowledge graphs. Its purpose is to ensure the medical accuracy and clinical relevance of drug classification, avoiding errors introduced by subjective grouping. Duplicate treatment detection refers to identifying whether there are functionally overlapping drug combinations within the same treatment target category. This can be achieved using rule-based matching algorithms or semantic similarity analysis models. Its purpose is to identify drug sets with potential duplicate treatment risks, providing a basis for quantitative analysis. Specifically, the quantity overlap component refers to a normalized assessment index based on the quantity of drugs, which can be implemented using... The linear proportional calculation method is used to transform absolute quantities into relative proportions, reflecting both the extent to which drug quantities exceed clinical safety thresholds and eliminating the influence of differences in the scale of different treatment target categories. The intensity overlap component refers to an averaged assessment index based on the similarity of drug treatment intensity, which can be achieved by summing the similarities of all drug pairs and dividing by the number of combinations. The standardized calculation method is used to capture the subtle differences in functional overlap between drugs and avoid the distortion of similarity calculation caused by changes in the number of drugs. In practical applications, the comprehensive quantity overlap component and intensity overlap component refer to the fusion of two dimensions of indicators through multiplication. This can be achieved by a non-linear combination of direct multiplication. The purpose is to simultaneously retain the macroscopic influence of quantity scale and the microscopic characteristics of intensity similarity, ensuring that the overlap coefficient can continuously reflect the breadth and depth of repeated treatment.

[0163] Specifically, the proposed solution first authoritatively categorizes drugs according to their treatment goals based on a medical knowledge base, ensuring that subsequent testing focuses on real clinical scenarios. Then, it independently performs a dual-dimensional quantitative calculation for each category: on the one hand, it calculates the quantity overlap component based on the number of drugs, using normalization to ensure the evaluation results are not affected by category size; on the other hand, it calculates the intensity overlap component based on the therapeutic intensity similarity of drug pairs, using averaging to eliminate the influence of drug quantity fluctuations. Finally, the two components are multiplied to obtain the overlap coefficient, dynamically coupling quantity and intensity similarity within the 0-1 range. This avoids the one-sidedness of a single dimension while ensuring the continuity and comparability of the output results, thus providing reliable input for the conflict penalty score.

[0164] Through the above technical solution, this application enables the overlap coefficient to simultaneously reflect the quantity and intensity of repeated treatments and the similarity characteristics of treatment intensity, avoiding the assessment bias caused by relying solely on quantity, thereby improving the accuracy of the conflict penalty score, providing a reliable basis for the correction of individualized consistency influencing factors, and ultimately optimizing the authenticity and clinical applicability of medication adherence assessment results.

[0165] Example 2:

[0166] like Figures 2-3 As shown, the method for predicting and assessing medication adherence in patients with chronic diseases includes the following steps:

[0167] Obtain patient medication behavior data and treatment effect data;

[0168] Behavioral compliance scores for each drug are calculated based on medication behavior data. These scores represent the degree to which actual medication behavior conforms to the pre-set plan.

[0169] The efficacy adherence score for each drug is calculated based on the treatment efficacy data. The efficacy adherence score represents the degree to which the treatment effect achieves the expected improvement goal.

[0170] Based on behavioral compliance scores and effect compliance scores, the contribution confidence of each drug is obtained through coupling processing. The contribution confidence represents the degree of confidence in the contribution of behavioral compliance to the treatment effect.

[0171] To determine whether the contribution confidence level meets the preset high confidence condition, if so, the behavioral compliance score is used to calculate the overall compliance with the preset clinical importance weight.

[0172] Otherwise, further determine whether the contribution confidence level meets the preset low confidence condition. If yes, exclude the corresponding drug from the overall compliance calculation; otherwise, adjust the clinical importance weight according to the contribution confidence level before calculating the overall compliance.

[0173] The output is the medication adherence assessment result based on the overall adherence calculation, which is based on the behavioral adherence score of each drug and its corresponding weight.

[0174] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0175] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0176] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A medication adherence prediction and evaluation system for patients with chronic diseases, characterized in that, The method comprises the following steps: a data acquisition module is configured to acquire medication behavior data and treatment effect data of a patient; a behavior adherence calculation module is configured to calculate a behavior adherence score of each drug based on the medication behavior data, the behavior adherence score representing the degree of compliance of actual medication behavior with a planned schedule; an effect adherence calculation module is configured to calculate an effect adherence score of each drug based on the treatment effect data, the effect adherence score representing the degree of treatment effect reaching an expected improvement target; a contribution confidence calculation module is configured to obtain a contribution confidence of each drug by coupling processing based on the behavior adherence score and the effect adherence score, the contribution confidence representing the confidence of the contribution of behavior adherence to treatment effect, the greater the gap between the behavior adherence score and the effect adherence score, the smaller the contribution confidence of the drug; a global adherence evaluation module is configured to determine whether the contribution confidence meets a preset high-confidence condition, if yes, the behavior adherence score is used to calculate global adherence with a preset clinical importance weight; if not, it is further determined whether the contribution confidence meets a preset low-confidence condition, if yes, the corresponding drug is excluded from global adherence calculation, otherwise, the clinical importance weight is adjusted based on the contribution confidence and then global adherence is calculated; the process of adjusting the clinical importance weight based on the contribution confidence comprises: calculating a weight adjustment factor based on the contribution confidence, the weight adjustment factor being determined according to the relative position of the contribution confidence within a preset high-low confidence threshold interval, multiplying the clinical importance weight by the weight adjustment factor to obtain an adjusted final weight; the calculation formula of the weight adjustment factor is: wherein, represents a weight adjustment factor, represents a contribution confidence, represents a low confidence threshold, represents a high confidence threshold; a result output module is configured to output a medication adherence evaluation result obtained based on global adherence calculation, the global adherence calculation being based on the behavior adherence score of each drug and the weight thereof.

2. The medication adherence prediction and evaluation system for chronic patients according to claim 1, characterized in that: The calculation process of the contribution confidence comprises: normalizing the behavior adherence score and the effect adherence score to map them to a unified numerical interval; calculating the absolute difference between the normalized behavior adherence score and the effect adherence score; subtracting the product of the absolute difference and a preset coupling coefficient from a reference value to obtain the contribution confidence; the calculation formula of the contribution confidence is: wherein, denotes a contribution confidence, denotes a normalized behavior adherence score, denotes a normalized effect adherence score, denotes a coupling coefficient. 3.The system for predicting and evaluating medication adherence of a chronic disease patient according to claim 1, wherein: when there are drugs excluded from global adherence calculation, weight redistribution is further included: calculating the sum of the clinical importance weights of all excluded drugs; redistributing the sum of the weights to drugs meeting the high-confidence condition in proportion; the calculation formula of the weight redistribution is: wherein, represents the weight of the ith drug after redistribution, represents the weight of the ith drug before redistribution, represents the sum of the clinical importance weights of all excluded drugs, represents the clinical importance weight of the ith drug, represents the sum of the clinical importance weights of all high-confidence drugs. 4.The system for predicting and evaluating medication adherence of a chronic disease patient according to claim 1, wherein: after obtaining the contribution confidence of each drug through coupling processing, treatment path consistency verification is further included: a drug relationship graph is constructed based on the complete medication regimen of the patient, the nodes in the drug relationship graph representing drugs, and the edges representing the interaction relationship between drugs, the weight of the edge being determined based on the conflict intensity value in the drug interaction database; conflict patterns in the drug relationship graph are detected, the conflict patterns including antagonistic drug rings and repeated treatment combinations; According to the detection result of the conflict mode, an individual consistency influence factor of each drug is calculated, which is used to quantify the degree of negative influence of the drug conflict participated by the drug on the individual contribution confidence of the drug; The individual consistency influence factor of each drug is multiplied by the contribution confidence thereof to obtain an individual correction confidence based on the overall rationality of the treatment plan.

5. The medication adherence prediction and evaluation system for chronic patients according to claim 4, characterized in that: The calculation process of the individual consistency influence factor includes: Identify all conflict modes participated by each drug, including the antagonistic drug ring and the repeated treatment combination thereof; Calculate a conflict penalty score according to each conflict mode participated by the drug, and obtain an individual consistency influence factor based on the conflict penalty score mapping; The calculation formula of the conflict penalty score is: wherein, is a conflict penalty score for the ith drug, denotes a set of antagonistic loops comprising the ith drug, denotes a conflict score for the mth antagonistic loop, denotes an antagonistic weight, denotes a repeated treatment combination comprising the ith drug, denotes an overlap coefficient for the nth repeated treatment combination, denotes a repetition weight.

6. The medication adherence prediction and evaluation system for chronic patients according to claim 5, characterized in that: The process of obtaining the individual consistency influence factor based on the conflict penalty score mapping is realized by using an exponential decay function: The exponential decay function ensures that the higher the conflict penalty score is, the lower the individual consistency influence factor is, and the value range of the individual consistency influence factor is between zero and one; the calculation formula of the mapping process is: wherein, represents an individual consistency influence factor of the i-th drug, represents a conflict penalty score of the i-th drug, represents a preset attenuation coefficient.

7. The medication adherence prediction and evaluation system for chronic patients according to claim 5, characterized in that: The calculation process of the conflict score includes: Identify all closed loops formed in the drug relationship graph, wherein each edge represents the antagonistic action between drugs; For each identified loop, extract the interaction strength of all drug pairs in the loop; Calculate the conflict score of the loop by using a geometric mean method; the calculation formula of the conflict score is: wherein, represents the conflict score of the mth antagonistic loop, L represents the number of drugs in the loop, represents the interaction strength between the ith drug and the jth drug, and takes a negative value to represent antagonism, represents the mth identified antagonistic loop. 8.The system for predicting and evaluating medication adherence of a chronic patient according to claim 5, wherein: The calculation process of the overlap coefficient includes: Classify drugs according to their treatment targets based on a medical knowledge base, and drugs in the same treatment target category enter the repeated treatment detection; the repeated treatment combination refers to a drug combination formed after the drugs are classified according to the treatment target category based on the medical knowledge base; Calculate a quantity overlap component based on the number of drugs in the category, and calculate a strength overlap component based on the treatment intensity similarity of the drugs in the category; Obtain the final overlap coefficient by integrating the quantity overlap component and the strength overlap component; The calculation formula of the overlap coefficient is: The repeated treatment combination refers to a combination of drugs formed by classifying drugs according to their therapeutic goals based on a medical knowledge base. This represents the overlap coefficient of the nth repeated treatment combination, which is also the overlap coefficient of the nth treatment target category. This represents the set of drugs included in the nth treatment target category. This indicates the number of drugs in that treatment target category. This indicates the maximum number of drugs allowed for this predefined treatment target category. This indicates the similarity in therapeutic intensity between drug i and drug j. Indicates from The number of combinations of choosing 2 out of 10 drugs.

9. A method for predicting and evaluating medication adherence for a chronic disease patient, characterized by, The medication compliance prediction and evaluation system for chronic disease patients according to any one of claims 1-8, performs the following steps: Obtain the medication behavior data and treatment effect data of the patient; Calculate the behavior compliance score of each drug according to the medication behavior data, wherein the behavior compliance score represents the degree of compliance of the actual medication behavior with the pre-planned scheme; Calculate the effect compliance score of each drug according to the treatment effect data, wherein the effect compliance score represents the degree of improvement of the treatment effect to the expected improvement target; According to the behavior compliance score and the effect compliance score, the contribution confidence of each drug is obtained through coupling processing, wherein the contribution confidence represents the confidence degree of the contribution of the behavior compliance to the treatment effect; Determine whether the contribution confidence meets a preset high confidence condition, and if yes, calculate the overall compliance of the behavior compliance score with a preset clinical importance weight. Otherwise, it is further judged whether the contribution confidence satisfies a preset low confidence condition. If yes, the corresponding drug is excluded from the overall adherence calculation. Otherwise, the clinical importance weight is adjusted based on the contribution confidence, and then the overall adherence calculation is performed. An output of a medication adherence evaluation result based on the overall adherence calculation is outputted. The overall adherence calculation is based on the behavior adherence score of each drug and the corresponding weight.

Images