An evaluation method for medication compliance based on pharmacodynamic index monitoring

Abstract

The application discloses an evaluation method of medication compliance based on pharmacodynamic index monitoring, comprising the following steps: determining a target user and obtaining a pharmacodynamic model of a target drug; setting a preset medication behavior and a corresponding prior probability; simulating a pharmacodynamic index theoretical distribution corresponding to the preset medication behavior according to the pharmacodynamic model; obtaining pharmacodynamic index monitoring data of the target user taking the target drug; obtaining a posterior probability of being in the preset medication behavior based on a Bayesian inference model according to the pharmacodynamic index monitoring data, the pharmacodynamic index theoretical distribution and the prior probability; and obtaining an evaluation result of medication compliance according to the posterior probability. The technical scheme directly reflects the pharmacodynamic index of the disease state as the core input evidence of Bayesian inference, unifies the evaluation conclusion with the clinical treatment target, and enhances the clinical relevance and decision support value of the evaluation result.

Description

Technical Field

[0001] This application relates to the field of biomedical technology, and in particular to a method for assessing medication adherence based on pharmacodynamic index monitoring. Background Technology

[0002] In long-term drug treatment of chronic diseases, medication adherence is a core variable determining efficacy, influencing prognosis, and impacting healthcare resource consumption. Poor adherence is prevalent in multiple treatment areas, including epilepsy, hypertension, diabetes, organ transplant rejection, and mental illnesses, often leading to disease control failure, symptom recurrence, increased risk of complications, increased emergency room and hospitalization rates, and a heavy socioeconomic burden. Therefore, objective, accurate, and routinely applicable medication adherence assessment is a crucial prerequisite for effective patient management and optimized treatment plans.

[0003] Currently, adherence assessment methods used in clinical practice and research each have their own emphasis. Patient self-reporting or structured questionnaires are low-cost but highly subjective and susceptible to bias. Direct behavioral recording methods such as electronic pillbox monitoring provide accurate timestamps of medication actions, but they rely on dedicated hardware, posing challenges in terms of cost, accessibility, and long-term feasibility. Therapeutic Drug Monitoring (TDM) quantitatively analyzes drug concentrations in biological samples, comparing individual concentrations with treatment reference ranges based on population data to determine whether a dose has been missed or insufficient.

[0004] However, the key evidence upon which existing technologies are based, whether it is blood drug concentration in pharmacokinetics or physical action signals captured by sensors, are all intermediate variables or surrogate markers in the treatment process, rather than the final pharmacodynamic endpoint indicators that directly reflect the disease state and treatment response, such as symptom frequency, physiological function indicators or quality of life scores. This makes it difficult for clinicians to translate the assessment results into clear clinical decisions, thus limiting the in-depth application of medication adherence in real clinical scenarios. Summary of the Invention

[0005] The main objective of this invention is to provide a method for assessing medication adherence based on pharmacodynamic index monitoring.

[0006] To achieve the above objectives, this application proposes a method for assessing medication adherence based on pharmacodynamic index monitoring, comprising the following steps: Identify target users and obtain pharmacodynamic models of the target drug; Set preset medication behaviors and corresponding prior probabilities; Based on the pharmacodynamic model, the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior is simulated. Obtain pharmacodynamic data of target users taking the target drug; Based on the pharmacodynamic index monitoring data, the theoretical distribution of the pharmacodynamic index, and the prior probability, the posterior probability of engaging in the preset medication behavior is obtained using a Bayesian inference model. Based on the posterior probability, the assessment results of medication adherence are obtained.

[0007] Optionally, obtaining the pharmacodynamic model of the target drug includes: Obtain the population pharmacodynamic model of the target drug; Based on the individual characteristic data of the target users, the population pharmacodynamic parameters of the population pharmacodynamic model are adjusted to obtain the pharmacodynamic model.

[0008] Optionally, adjusting the population pharmacodynamic parameters of the population pharmacodynamic model based on the individual characteristic data of the target user to obtain the pharmacodynamic model includes: Obtain the individual characteristic data of the target user; Based on the relationship between pharmacodynamic parameters and feature data in the population pharmacodynamic model, the individual pharmacodynamic parameters corresponding to the individual feature data are determined; The pharmacodynamic model is obtained by replacing the corresponding pharmacodynamic parameters in the population pharmacodynamic model with the individual pharmacodynamic parameters.

[0009] Optionally, the step of simulating the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior based on the pharmacodynamic model includes: Pharmacodynamic parameters are extracted based on the parameter distribution characteristics of the pharmacodynamic model. Based on the extracted pharmacodynamic parameters, a curve of the pharmacodynamic index changing over time under the preset medication behavior assumption is generated through simulation. The parameter extraction and pharmacodynamic index-time curve simulation process is repeated N times to obtain N curves showing the change of pharmacodynamic index over time under the preset medication behavior. Based on the N curves, the theoretical distribution of pharmacodynamic indicators at each time point is statistically analyzed, and the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior is obtained.

[0010] Optionally, obtaining the posterior probability of engaging in the preset medication behavior based on the pharmacodynamic index monitoring data, the theoretical distribution of the pharmacodynamic index, and the prior probability using a Bayesian inference model includes: Obtain the probability value distribution of the pharmacodynamic index corresponding to the theoretical distribution of the pharmacodynamic index; Based on the pharmacodynamic index monitoring data, the probability value distribution, and the prior probability, the posterior probability of being in the preset medication behavior is obtained using a Bayesian inference model.

[0011] Optionally, the probability value distribution is approximated by kernel density estimation or histogram estimation of the theoretical distribution of the pharmacodynamic index.

[0012] Optionally, obtaining the assessment result of medication adherence based on the posterior probability includes: When multiple posterior probabilities corresponding to various preset medication behaviors are obtained, the medication adherence assessment result is obtained based on the preset medication behavior corresponding to the largest posterior probability.

[0013] Optionally, the steps also include: The posterior probabilities of multiple preset medication behaviors are output graphically, and the preset medication behavior corresponding to the highest posterior probability is prominently displayed; and / or Generate a graph showing the relationship between the posterior probabilities of multiple preset medication behaviors and pharmacodynamic index monitoring data, output the graph graphically, and prominently display the pharmacodynamic index monitoring data, as well as the preset medication behavior corresponding to the maximum posterior probability; and / or Based on the assessment results, a tiered warning and recommendations will be issued.

[0014] This application also provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, it implements any of the methods described above.

[0015] This application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements any of the methods described above.

[0016] This application presents a method for assessing medication adherence based on pharmacodynamic indicator monitoring. The method involves: obtaining a pharmacodynamic model of the target drug; setting preset medication behaviors and corresponding prior probabilities; simulating the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behaviors based on the pharmacodynamic model; acquiring pharmacodynamic indicator monitoring data of the target user taking the target drug; obtaining the posterior probability of being in the preset medication behavior based on the pharmacodynamic indicator monitoring data, the theoretical distribution of pharmacodynamic indicators, and the prior probabilities using a Bayesian inference model; and obtaining the medication adherence assessment result based on the posterior probability. The technical solution of this application uses pharmacodynamic indicators that directly reflect disease status as the core input evidence for Bayesian inference, aligning the assessment conclusions with clinical treatment goals and enhancing the clinical relevance and decision support value of the assessment results. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the medication adherence assessment method based on pharmacodynamic index monitoring provided in the embodiments of this application.

[0019] Figure 2 for Figure 1 A schematic diagram of the simulation process of step S3 in the illustrated embodiment.

[0020] Figure 3 for Figure 1 The illustrated embodiment shows a schematic diagram of the process of estimating the probability value distribution in step S5.

[0021] Figure 4 This is a settings interface provided for an embodiment of this application.

[0022] Figure 5 This application provides an example of a posterior probability output table.

[0023] Figure 6 This is a schematic diagram illustrating the relationship between posterior probability and pharmacodynamic index monitoring data, provided as an embodiment of this application.

[0024] Figure 7 This is a schematic diagram illustrating the distribution of FXa activity under different medication behaviors, as provided in the embodiments of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0026] Figure 1 This is a flowchart illustrating the medication adherence assessment method based on pharmacodynamic index monitoring provided in this application embodiment. Figure 1 As shown, this application discloses a method for assessing medication adherence based on pharmacodynamic index monitoring, comprising the following steps: S1, Identify the target users and obtain the pharmacodynamic model of the target drug; Among them, the pharmacodynamic model of the target drug is used to mathematically describe the dynamic effect of the drug on specific clinical endpoints (i.e., pharmacodynamic indicators, such as seizure frequency, blood glucose level, blood pressure value, and depression score) under standard treatment conditions. It is usually represented as a functional relationship of pharmacodynamic indicators over time and includes key influencing factors (such as baseline disease severity and demographic characteristics) as model covariates.

[0027] For situations where a drug involves multiple key efficacy evaluation dimensions (e.g., antiepileptic drugs simultaneously focus on seizure frequency and severity), multiple parallel pharmacodynamic models can be obtained or constructed (e.g., multiple pharmacodynamic indicators can be used simultaneously). Weights can be assigned to different pharmacodynamic indicators to form a more comprehensive efficacy assessment matrix. The pharmacodynamic models used must be constructed based on large-sample clinical data and validated internally or externally to ensure their predictive performance and biological rationality.

[0028] S2, set the preset medication behavior and the corresponding prior probability; This involves pre-defining a set of mutually exclusive discrete medication behaviors to be evaluated. These behaviors are pre-defined based on clinical concerns and potential adherence deviation patterns. Examples include: medication behavior A (complete adherence, i.e., the patient strictly follows the prescribed dosage and time), medication behavior B (missing the most recent dose), medication behavior C (missing the two most recent doses consecutively), and medication behavior D (missing the penultimate dose). In practice, this medication behavior space can be expanded or adjusted according to specific drugs, dosing frequency, and clinical needs to cover more complex behavioral patterns, such as missing three consecutive doses.

[0029] Each pre-defined medication behavior is assigned a prior probability. A prior probability is an initial probability estimate of the likelihood that a patient is currently engaged in a certain medication behavior, prior to the acquisition of new pharmacodynamic monitoring data. Prior probabilities can be determined from multiple sources of prior information and can be selected based on data availability and reliability, for example, following a priority order: Method 1: Based on the patient's (target user's) historical adherence records. If the patient's adherence assessment results over a past period are available (e.g., records from electronic monitoring devices or pharmacy data), the frequency of each preset medication behavior during that historical period can be directly calculated and used as the prior probability for the current assessment moment. For example, if, based on historical data, the patient was judged to have "complete adherence" for 80% of the past 30 days, missed the most recent dose for 15% of the days, and missed two consecutive doses for 5% of the days, then the prior probabilities can be set as: P(complete adherence) = 0.8, P(missed dose once) = 0.15, P(missed two consecutive doses) = 0.05.

[0030] Method 2: Based on population epidemiological data. When individual patient historical data is lacking, epidemiological statistics of the target population matched with patient characteristics (such as disease type, age, disease course, medication regimen, etc.) can be used as priors. These data can come from publicly published literature, disease registry reports, or real data analysis from various medical institutions. For example, if statistics from a hospital show that among similar patients treated at that hospital, the proportion of complete adherence is 50%, the proportion of occasional missed doses (such as missing the most recent dose) is 30%, and the proportion of frequent missed doses (such as missing two consecutive doses) is 20%, then these frequencies can be directly used as prior probabilities.

[0031] Method 3: Based on the clinician's experience. When the above two types of data are unavailable, the clinician's subjective experience can be used. Based on a comprehensive understanding of the patient (such as face-to-face consultation, symptom presentation, and past communication), combined with their own clinical experience, the doctor can provide a subjective probability estimate of the patient's possible medication behavior. For example, if the patient's symptoms are poorly controlled at this follow-up visit and they report "occasionally forgetting to take their medication," the doctor can initially determine that the probability of complete adherence is 20%, the probability of missing the most recent dose is 60%, and the probability of missing two consecutive doses is 20%.

[0032] In practical applications, one or more of the above methods can be flexibly selected based on the availability of data (if multiple sources exist simultaneously, weighted average or Bayesian model averaging can be used for fusion), but the priority order of prior information quality from high to low must be followed to ensure the objectivity and accuracy of the inference results.

[0033] S3, based on the pharmacodynamic model, simulates the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior; Among them, the theoretical distribution of pharmacodynamic indicators represents the expected theoretical change curve of the patient's pharmacodynamic indicators from a certain historical reference time point (usually the period when the patient's compliance was confirmed as complete compliance, such as the treatment start point or the period when the patient was judged to be fully compliant in the last assessment) to the assessment time (i.e. the time point for compliance inference). It reflects the probability range of possible values ​​of the pharmacodynamic indicators after considering all known sources of variation under a given assumption of medication behavior.

[0034] Given the uncertainties in parameters contained in pharmacodynamic models (such as inter-individual variability and standard errors in parameter estimation) and residual variability in model predictions (measurement errors, model errors, etc.), Monte Carlo simulation can be used to quantify these uncertainties and obtain the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior.

[0035] S4, Obtain pharmacodynamic data of target users taking the target drug; During the target users' daily treatment and follow-up, time-series data of their pharmacodynamic indicators are continuously collected in a standardized and structured manner. Data sources may include: daily symptom logs (such as seizure diaries, visual analog scales for pain) voluntarily recorded by the target users through electronic patient reporting outcome systems; physiological parameter measurements automatically synchronized with certified home medical devices (such as blood glucose meters, smart blood pressure monitors, and wearable ECG monitors); and relevant laboratory test results (such as glycated hemoglobin and heart failure biomarkers) periodically extracted from the electronic health record systems of medical institutions. The time relationship between the data collection point and the dosing regimen is known; that is, the collection point of each monitoring data point corresponds to a specific position within the dosing interval, and based on this, the monitoring data can be mapped to the corresponding time on the time axis corresponding to the theoretical distribution of pharmacodynamic indicators.

[0036] S5. Based on the pharmacodynamic index monitoring data, the theoretical distribution of pharmacodynamic indexes, and prior probabilities, the posterior probability of being in the preset medication behavior is obtained based on the Bayesian inference model. In this process, Bayes' theorem is applied, and the current pharmacodynamic index monitoring data is used as actual pharmacodynamic evidence. Based on the corresponding theoretical distribution of pharmacodynamic indexes and prior probabilities, the posterior probability of each preset medication behavior is calculated, that is, the probability of each preset medication behavior occurring when the current pharmacodynamic index monitoring data is observed.

[0037] S6. Based on the posterior probability, obtain the assessment results of medication adherence.

[0038] Since posterior probability represents the probability of various pre-defined medication behaviors occurring when current pharmacodynamic monitoring data are observed, medication adherence assessment results can be provided based on posterior probability. In some embodiments, when multiple posterior probabilities corresponding to various pre-defined medication behaviors are obtained, the medication adherence assessment result is obtained based on the pre-defined medication behavior corresponding to the highest posterior probability. For example, when the posterior probability of complete adherence is the highest, the medication adherence assessment result is good adherence; or, when the posterior probability of missing two consecutive doses is the highest, the medication adherence assessment result is poor adherence. In some embodiments, while providing the assessment result, the pre-defined medication behavior with the highest posterior probability can be suggested for clinical decision-making reference.

[0039] Based on the above technical solution, this application has the following technical effects: 1. The alignment of assessment conclusions with clinical treatment goals enhances the clinical relevance and decision support value of the assessment results. Instead of relying on indirect pathways such as pharmacokinetic intermediate variables or physical actions as evidence, it uses pharmacodynamic endpoints (e.g., symptom frequency, physiological function indicators) that directly reflect the disease state as the core input evidence for Bayesian inference. Since the evidence upon which the assessment is based is itself a direct measure of clinical efficacy, the posterior probability of medication behavior, analyzed through the Bayesian model, naturally and directly reflects the specific impact of medication behavior on the patient's actual treatment benefit or risk. Furthermore, this allows clinicians to intuitively interpret clear clinical significance from the assessment results, fundamentally bridging the gap between behavioral assessment and treatment decision-making. The assessment conclusions can be directly used to guide whether and how to adjust treatment strategies (e.g., strengthening education, adjusting dosage, or switching treatments), thereby deeply integrating adherence assessment into the treatment management loop and greatly enhancing the tool's clinical practical value. 2. By eliminating reliance on specific monitoring conditions, the technical approach becomes more universal, significantly expanding the range of applicable drugs and clinical scenarios. It does not depend on specific drug-based population pharmacokinetic models, nor does it require mandatory TDM (Therapeutic Drug Monitoring). Instead, it is based on dynamic pharmacodynamic models. Since efficacy data can be obtained in a diverse and routine manner through symptom diaries, routine examinations, or home devices, it does not require meeting the stringent prerequisites of "the existence of a validated population pharmacokinetic model" and "the ability to routinely perform TDM." This makes this method applicable to any drug treatment with quantifiable and monitorable clinical endpoints. Furthermore, the technical solution of this application can be widely applied to many chronic disease areas where routine monitoring of therapeutic drugs is unnecessary or impossible, such as long-term medication management for common hypertension, type 2 diabetes, and many neuropsychiatric disorders, significantly enhancing its universality. 3. Based on widely accessible clinical data streams, the implementation threshold and cost are significantly reduced, while data reliability and application sustainability are improved. Patients do not need to use dedicated, additional electronic monitoring hardware. The core data originates from patients' existing disease monitoring activities (such as diaries, self-monitoring of blood pressure and blood sugar) or efficacy records generated during routine medical procedures. This eliminates the purchase and maintenance costs of dedicated hardware and the associated operational complexity, reducing the economic burden and usage barriers for patients, and improving patient acceptance and long-term sustainability. Furthermore, because the data originates from objective or subjective clinical reports reflecting the true disease state, rather than potentially erroneous mechanical sensor signals, its directional and authentic nature as evidence for adherence inference is stronger. Therefore, this application can overcome the promotional difficulties caused by cost, operation, and reliability issues, enabling large-scale, long-term dynamic monitoring of patient medication adherence with lower overall cost, less invasiveness, and higher data ecosystem integration. This generates more reliable assessment results to support clinical decision-making and avoids additional management burdens caused by equipment problems.

[0040] In one embodiment, step S1, obtaining the pharmacodynamic model of the target drug, includes the following steps: S11, Obtain the population pharmacodynamic model of the target drug; S12, Based on the individual characteristic data of the target users, adjust the population pharmacodynamic parameters of the population pharmacodynamic model to obtain the pharmacodynamic model.

[0041] Among these features, the system can retrieve relevant previous clinical studies, literature, and public databases related to the target drug to obtain a fully evaluated and validated population pharmacodynamic model.

[0042] Alternatively, when a sufficiently large patient cohort is available, containing high-quality medication behavior records (such as electronic monitoring data as the gold standard) and concurrent longitudinal data on pharmacodynamic indicators, a population pharmacodynamic model can be constructed using machine learning algorithms. Machine learning algorithms are supervised learning models capable of learning the mapping relationship between medication behavior records and pharmacodynamic indicators from patient cohort data. The model is trained in a supervised manner using medication behavior records and concurrent longitudinal data on pharmacodynamic indicators from the patient cohort as training samples, and its predictive performance and generalization ability are evaluated using cross-validation. Its advantage lies in its ability to uncover potential, complex relationships from real-world data that are not described by traditional pharmacological models, making it particularly suitable for situations where the pharmacodynamic mechanism is not yet clear or the influencing factors are complex.

[0043] Based on the population pharmacodynamic model, parameters are adjusted according to the individual characteristic data of the target users to obtain the pharmacodynamic model, so that the compliance assessment results have a higher degree of matching with the actual situation of patients.

[0044] Individual characteristic data are confirmed covariates that have a significant impact on drug efficacy and are included as key parameters in the pharmacodynamic model, including but not limited to demographic characteristics (such as age, sex, weight, and body surface area) and important pathophysiological indicators (such as estimated glomerular filtration rate (eGFR) reflecting renal function, indicators reflecting liver function, baseline disease severity scores, and genetic polymorphism information).

[0045] Alternatively, when the target drug does not require high individual variability, a population pharmacodynamic model can be directly used as the pharmacodynamic model for the target drug.

[0046] In one embodiment, step S12 involves adjusting the population pharmacodynamic parameters of the population pharmacodynamic model based on the individual characteristic data of the target user to obtain the pharmacodynamic model, including: Obtain individual characteristic data of the target users; Based on the relationship between pharmacodynamic parameters and characteristic data in the population pharmacodynamic model, determine the individual pharmacodynamic parameters corresponding to individual characteristic data; The pharmacodynamic model is obtained by replacing the corresponding pharmacodynamic parameters in the population pharmacodynamic model with individual pharmacodynamic parameters.

[0047] In the pharmacodynamic model, pharmacodynamic parameters (such as baseline effect E0, maximum effect E0) are described in the form of mathematical functions. max EC50, the drug concentration that produces 50% of the maximum effect 50 The quantitative relationship between the data (such as the effect decay rate constant) and the characteristic data (covariates) can be expressed as a linear function, power function, exponential function, or piecewise constant of categorical variables, and has been statistically validated during the model building process.

[0048] The collected individual characteristic data of the target users are substituted one by one into the formula relating pharmacodynamic parameters and characteristic data to calculate the individual pharmacodynamic parameters of the target users. For example, if the typical value of a certain parameter P in the pharmacodynamic model is P... pop Its relationship with the feature data cov is as follows: , Among them, cov ref The covariate values ​​for the reference population and θ are the influence coefficients; both are reported values ​​in the population pharmacodynamic model. Therefore, the covariate values ​​for the target users will be... i Substituting the values ​​will yield the individual parameter P. i For categorical individual feature data, the corresponding parameter subgroup value is selected based on the category to which the target user belongs.

[0049] After determining the individual pharmacodynamic parameters, these parameters replace the typical values ​​in the population pharmacodynamic model, thereby constructing a mathematical model of pharmacodynamic dynamic changes specific to the target user. This individualized pharmacodynamic model forms the basis for subsequent simulations of the theoretical distribution of pharmacodynamic indicators under different medication behaviors, ensuring a higher degree of agreement between the simulation results and the patient's actual condition.

[0050] In one embodiment, step S3, based on a pharmacodynamic model, simulates the theoretical distribution of pharmacodynamic indicators corresponding to a preset medication behavior, including: Pharmacodynamic parameters are extracted based on the parameter distribution characteristics of the pharmacodynamic model; Based on the extracted pharmacodynamic parameters, a curve of pharmacodynamic indicators changing over time under the preset assumptions of medication behavior is generated through simulation. The process of repeated parameter extraction and pharmacodynamic index-time curve simulation was performed N times to obtain N curves showing the change of pharmacodynamic index over time under the preset medication behavior. Based on N curves, the theoretical distribution of pharmacodynamic indicators at each time point is statistically analyzed, and the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior is obtained.

[0051] Among them, the Monte Carlo simulation method is used. First, based on the typical values ​​of the parameters in the pharmacodynamic model and the distribution characteristics of inter-individual variability (usually log-normal distribution), parameters containing inter-individual variability (such as baseline effect E0, maximum effect E) are analyzed. max EC50, the drug concentration that produces 50% of the maximum effect 50 The parameters (such as the effect decay rate constant, etc.) and residual variation are randomly sampled, and each sampled combination of parameters represents the pharmacodynamic characteristics of a possible "virtual patient".

[0052] Next, based on the extracted pharmacodynamic parameters, a curve showing the change of a pharmacodynamic index over time under predefined assumptions about medication behavior is generated, for example, such as... Figure 2 As shown in the left-middle figure. The above parameter extraction and pharmacodynamic index-time curve simulation process is repeated N times (usually N≥1000 times). Finally, N curves showing the changes of pharmacodynamic indicators over time under the preset medication behavior are obtained. Then, the predicted distribution of the pharmacodynamic indicators at each time point is statistically calculated, i.e., the theoretical distribution of the pharmacodynamic indicators. For example, ... Figure 2 As shown in the right-middle figure, this distribution fully reflects the probability range of possible values ​​for the pharmacodynamic index, considering all known sources of variation, given the behavioral assumptions.

[0053] In one embodiment, step S5, based on pharmacodynamic index monitoring data, theoretical distribution of pharmacodynamic indexes, and prior probabilities, obtains the posterior probability of engaging in a preset medication behavior using a Bayesian inference model, including: Obtain the probability distribution of pharmacodynamic indicators corresponding to the theoretical distribution of pharmacodynamic indicators; Based on pharmacodynamic index monitoring data, probability value distribution, and prior probability, the posterior probability of being in a preset medication behavior is obtained using a Bayesian inference model.

[0054] In step S3, each preset medication behavior ω has been addressed. i Monte Carlo simulations were used to obtain the theoretical predicted distribution of pharmacodynamic indicators from a historical reference time point to the evaluation time. This distribution is an empirical distribution approximated by a large number of simulation samples. Furthermore, the corresponding probability distribution can be approximated by kernel density estimation or histogram estimation of the theoretical distribution of pharmacodynamic indicators. For example, Figure 3 The middle right figure illustrates the results of likelihood estimation of data 48 hours after reaching steady state, obtaining the probabilities corresponding to different pharmacodynamic indicators at that time point, i.e., forming a probability value distribution.

[0055] Next, applying Bayes' theorem, the prior probability of each medication behavior is multiplied by its corresponding probability value, and then normalized to calculate the posterior probability of each medication behavior state under the current actual efficacy evidence. Its mathematical expression is: , Where x represents the pharmacodynamic index monitoring data; ω represents medication use; P(ω i ) represents the prior probability of a certain medication behavior to be examined; P(x|ω i () represents the probability of observing x when the medication behavior occurs; P(ω i |x) represents the probability of this medication behavior occurring when x is observed.

[0056] The calculated posterior probability P(ω) i |x) represents a quantitative assessment of the patient's most likely medication behavior state under current pharmacodynamic evidence. For the N medication behavior states to be examined, each state has a posterior probability P(ω). i |x), and the sum of the posterior probabilities is 1. In N P(ω) i In |x), the medication behavior corresponding to the maximum value is the most likely medication behavior to occur.

[0057] In one embodiment, the method of this application further includes the step of: The posterior probabilities of multiple preset medication behaviors are output graphically, and the preset medication behavior corresponding to the highest posterior probability is prominently displayed; and / or Generate a graph showing the relationship between the posterior probabilities of multiple preset medication behaviors and pharmacodynamic index monitoring data. Output the graph graphically, clearly displaying the pharmacodynamic index monitoring data and the preset medication behavior corresponding to the maximum posterior probability; and / or Based on the assessment results, tiered warnings and recommendations will be issued.

[0058] Among them, the following can be adopted Figure 4 The settings interface shown allows you to input pharmacodynamic monitoring data and corresponding individual characteristic data.

[0059] After obtaining the evaluation results, bar charts or pie charts can be used to visually display the posterior probabilities of various preset medication behavior states at the current evaluation time (or recent period), highlighting the most likely behavior.

[0060] Or, such as Figure 5 As shown in the figure, the posterior probabilities of various preset medication behavior states under the pharmacodynamic index x at the current evaluation time are displayed in tabular form, where ω 00 This indicates that two doses have been missed consecutively, ω 01 This indicates that the last dose was missed. ω 10 This indicates that the second-to-last dose was missed. ω 11 It indicates complete compliance.

[0061] Or, such as Figure 6 As shown, taking FXa activity as a pharmacodynamic indicator as an example, a graph showing the relationship between the posterior probabilities of multiple preset medication behaviors and the pharmacodynamic indicator monitoring data is generated based on the evaluation results. This graph is output graphically, clearly displaying the pharmacodynamic indicator monitoring data and the preset medication behavior corresponding to the maximum posterior probability. By using curves, the posterior probabilities corresponding to more pharmacodynamic indicator monitoring data can be shown, providing a more continuous and comprehensive reference.

[0062] Furthermore, based on the assessment results, tiered warnings and recommendations can be generated, such as "Highly suspect recent missed doses; medication review and education recommended" or "Efficacy fluctuations deviate little from the model's predicted complete adherence; current adherence is likely good; other influencing factors should be monitored." These tiered warnings and recommendations make the assessment results more intuitive and easier to understand, which is conducive to improving adherence.

[0063] The following section uses rivaroxaban as the target drug and a population pharmacokinetic-pharmacodynamic model based on FXa factor activity as an example to illustrate the evaluation method of this application. FXa activity is a direct biomarker of the anticoagulant effect of rivaroxaban inhibiting coagulation factor Xa. Its value is driven by the drug-effect relationship of rivaroxaban concentration in the blood and is highly correlated with blood drug concentration. In this case, it serves as a direct observation value of the pharmacodynamic monitoring indicator and is mapped to the posterior probability of medication behavior through the population pharmacokinetic-pharmacodynamic model.

[0064] S1: Identify target users and obtain the pharmacodynamic model of the target drug. Target user: A male patient, weighing 75kg, 60 years old, with a serum creatinine level (Scr) of 1mg / dL and a body mass index (lbm) of 56.5kg, who is taking rivaroxaban 20mg once daily (at 8:00 AM) for atrial fibrillation.

[0065] Rivaroxaban was selected as the target drug, and a population pharmacokinetic-pharmacodynamic model based on FXa factor activity was adopted. The model parameters were obtained from the phase III clinical trial literature (J Clin Pharmacol. 2014 Aug;54(8):917-27. doi:10.1002 / jcph.288.), and the structure was a direct effect model. The pharmacokinetic formula and parameters included: , in: CL(L / h)=6.1×[1-0.011×(age-65)-0.914×(scr-1.09)]×e eta.CL (Apparent clearance rate); V (L)=79.7×[1-0.00133×(age-65)+0.0118×(lbm-57.5)]e eta.V(Apparent volume of distribution); k a (h -1 = 1.16 (absorption rate constant); eta.CL=0.1239 (inter-individual variation in clearance rate); eta.V = 0.0310 (coefficient of variation of clearance rate among individuals); dose is the dosage (mg) used for administration. k e =CL / V; t represents the time (in hours) after drug administration; τ represents the dosing interval (in hours); age refers to age (in years); SCR represents serum creatinine levels (mg / dL). lbm is lean body mass (kg).

[0066] The pharmacodynamic indicator is factor Xa activity (FXa activity), and its mathematical expression is as follows: , in, EFF0 = 104% (FXa activity baseline value); E max =107% (maximum FXa activity); EC 50 (μg / L) = 760 (half-maximum effective concentration); ω EFF0 =16.61% (inter-individual coefficient of variation of baseline FXa activity). ω EC50 =5.97% (inter-individual coefficient of variation at half maximum effect concentration).

[0067] S2: Set preset medication behaviors and corresponding prior probabilities Four mutually exclusive preset medication behaviors are defined (based on a once-daily dosing regimen, with the evaluation point being before the next dose): state Full compliance, with no missed doses; state The assessment point is the most recent missed dose, while previously doses were taken regularly; state The second-to-last dose not taken before the assessment point; the most recent dose taken. state Two consecutive missed doses before the assessment point.

[0068] Prior probability setting: Based on the patient's previous medical records and epidemiological data of the population, P( )=0.85,P( )=0.05, P( )=0.05, P( =0.05.

[0069] S3: Based on the pharmacodynamic model, simulate the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior.

[0070] Among them, the probability value distribution of pharmacodynamic indicators corresponding to the theoretical distribution of pharmacodynamic indicators was analyzed to obtain the following: , , , The distribution of FXa activity under four drug administration behaviors is as follows: Figure 7 As shown.

[0071] S4: Obtain pharmacodynamic data of target users taking the target drug. The patient had a blood sample taken at 8:00 a.m. on the morning of the follow-up visit (on an empty stomach, 24 hours after the last dose of medication), and the FXa activity was measured to be 76%.

[0072] S5: Based on pharmacodynamic index monitoring data, theoretical distribution of pharmacodynamic indexes, and prior probabilities, obtain the posterior probability of engaging in the preset medication behavior using a Bayesian inference model. Calculate the likelihood of each dosing behavior assumption when FXa activity is 76%: state Likelihood =0.0331; state Likelihood =0.0195; state Likelihood =0.0302; state Likelihood =0.0173.

[0073] Calculate the posterior probability of each state using Bayes' theorem: P( |FXa)=0.8935, meaning the posterior probability is approximately 89.35%; P( |FXa)=0.0310, meaning the posterior probability is approximately 3.10%; P( |FXa)=0.0480, meaning the posterior probability is approximately 4.80%; P( |FXa)=0.0275, meaning the posterior probability is approximately 2.75%.

[0074] S6: Obtain the assessment results of medication adherence based on the posterior probability. state The posterior probability is highest for patients with complete compliance. Note: If the patient has good compliance and there are currently no drug-related adverse reactions, the current dosing regimen can be maintained. Regular follow-up examinations are recommended.

[0075] The medication adherence assessment method proposed in this application uses pharmacodynamic indicators that directly reflect the disease state as the core input evidence for Bayesian inference, which aligns the assessment conclusions with clinical treatment goals and enhances the clinical relevance and decision support value of the assessment results.

[0076] This embodiment also provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory and executed by the processor to implement the steps of the above-described evaluation method. Since this embodiment includes all the technical features of the above embodiments, it also possesses all the technical effects of the above embodiments, and will not be repeated here.

[0077] This embodiment also provides a readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the above-described evaluation method. Since this embodiment includes all the technical features of the above embodiments, it also possesses all the technical effects of the above embodiments, and will not be repeated here.

[0078] It is understood that although the steps in the above flowcharts are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated in this embodiment, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the above flowcharts may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.

[0079] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “including,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process or method that includes a series of steps is not necessarily limited to those explicitly listed, but may include other steps not explicitly listed or inherent to such processes or methods.

[0080] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0081] It should be understood that in the description of the embodiments of this application, "multiple" means two or more, "greater than", "less than", "exceeding" etc. are understood to exclude the number itself, and "above", "below", "within" etc. are understood to include the number itself.

[0082] In the several embodiments provided in this application, it should be understood that the disclosed methods can be implemented in other ways.

[0083] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.

[0084] The above is a detailed description of the embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for assessing medication adherence based on pharmacodynamic index monitoring, characterized in that, Includes the following steps: Identify target users and obtain pharmacodynamic models of the target drug; Set preset medication behaviors and corresponding prior probabilities; Based on the pharmacodynamic model, the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior is simulated. Obtain pharmacodynamic data of target users taking the target drug; Based on the pharmacodynamic index monitoring data, the theoretical distribution of the pharmacodynamic index, and the prior probability, the posterior probability of engaging in the preset medication behavior is obtained using a Bayesian inference model. Based on the posterior probability, the assessment results of medication adherence are obtained.

2. The evaluation method as described in claim 1, characterized in that, The method for obtaining the pharmacodynamic model of the target drug includes: Obtain the population pharmacodynamic model of the target drug; Based on the individual characteristic data of the target users, the population pharmacodynamic parameters of the population pharmacodynamic model are adjusted to obtain the pharmacodynamic model.

3. The evaluation method as described in claim 2, characterized in that, The step of adjusting the population pharmacodynamic parameters of the population pharmacodynamic model based on the individual characteristic data of the target user to obtain the pharmacodynamic model includes: Obtain the individual characteristic data of the target user; Based on the relationship between pharmacodynamic parameters and feature data in the population pharmacodynamic model, the individual pharmacodynamic parameters corresponding to the individual feature data are determined. The pharmacodynamic model is obtained by replacing the corresponding pharmacodynamic parameters in the population pharmacodynamic model with the individual pharmacodynamic parameters.

4. The evaluation method as described in claim 2, characterized in that, The step of simulating the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior based on the pharmacodynamic model includes: Pharmacodynamic parameters are extracted based on the parameter distribution characteristics of the pharmacodynamic model. Based on the extracted pharmacodynamic parameters, a curve of the pharmacodynamic index changing over time under the preset medication behavior assumption is generated. The parameter extraction and pharmacodynamic index-time curve simulation process is repeated N times to obtain N curves of pharmacodynamic index changes over time under the preset medication behavior, where N is an integer greater than or equal to 10000. Based on the N curves, the theoretical distribution of pharmacodynamic indicators at each time point is statistically calculated, thus obtaining the theoretical distribution of pharmacodynamic indicators corresponding to the preset medication behavior.

5. The evaluation method as described in claim 1, characterized in that, The step of obtaining the posterior probability of engaging in the preset medication behavior based on the pharmacodynamic index monitoring data, the theoretical distribution of the pharmacodynamic index, and the prior probability using a Bayesian inference model includes: Obtain the probability value distribution of the pharmacodynamic index corresponding to the theoretical distribution of the pharmacodynamic index; Based on the pharmacodynamic index monitoring data, the probability value distribution, and the prior probability, the posterior probability of being in the preset medication behavior is obtained using a Bayesian inference model.

6. The evaluation method as described in claim 5, characterized in that, The probability value distribution is approximated by kernel density estimation or histogram estimation of the theoretical distribution of the pharmacodynamic index.

7. The evaluation method as described in claim 1, characterized in that, The process of obtaining the medication adherence assessment result based on the posterior probability includes: When multiple posterior probabilities corresponding to various preset medication behaviors are obtained, the medication adherence assessment result is obtained based on the preset medication behavior corresponding to the largest posterior probability.

8. The evaluation method according to any one of claims 1 to 7, characterized in that, It also includes the following steps: The posterior probabilities of multiple preset medication behaviors are output graphically, and the preset medication behavior corresponding to the maximum posterior probability is prominently displayed. and / or Generate a graph showing the relationship between the posterior probabilities of multiple preset medication behaviors and pharmacodynamic index monitoring data, output the graph graphically, and prominently display the pharmacodynamic index monitoring data and the preset medication behavior corresponding to the maximum posterior probability. and / or Based on the assessment results, a tiered warning and recommendations will be issued.

9. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, which, when executed by the processor, implements the method of any one of claims 1 to 8.

10. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, implements the method of any one of claims 1 to 8.

Images