Contribution degree evaluation-based health medical data element income distribution method and system

Abstract

The invention discloses a contribution degree evaluation-based health medical data element income distribution method and system, and the method comprises the steps: S1, building a data element contribution evaluation system: building a multi-dimensional index according to the data of a medical institution, evaluating the multi-dimensional indexes based on an XGBoost machine learning model and an SHAP value to obtain contribution degrees of data elements; s2, establishing an income distribution model: calculating the value of the data according to the contribution degree of the data elements and the evaluated contribution degree of each medical institution, and carrying out the distribution of the total income of each medical institution according to an agreed income distribution proportion; and S3, a revenue distribution method under the dynamic evaluation mechanism: recalculating the dynamic influence factor, adjusting the data element contribution degree of each medical institution according to the change of the dynamic influence factor, and performing revenue distribution adjustment.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a method and system for allocating benefits of health care data elements based on contribution evaluation. Background Art

[0002] In the field of health care, this policy provides guiding principles for the distribution of benefits from data elements, namely that benefits should be distributed based on an assessment of contribution.

[0003] In data transactions, medical institutions, as data providers, have access to a large amount of health and medical data, but their standardization, data desensitization level, and data value assessment capabilities are relatively insufficient, resulting in the inability to effectively circulate and share high-quality resources, and the inability to obtain reasonable returns. On the other hand, medical institutions, as data acquirers and users, lack channels to obtain data when conducting medical data analysis and auxiliary treatment, and it is difficult to obtain preferential purchase prices.

[0004] At present, there are few methods for distributing benefits from data elements in the medical field. Existing methods do not fully consider data characteristics, do not fully evaluate the contribution of data value, and cannot flexibly evaluate the contribution and subsequent benefit distribution. The existing benefit distribution methods mainly include Shapley value method, supply chain contract coordination method, and fair entropy method. Although the Shapley value benefit distribution method can ensure the fairness of benefit distribution and find a unique solution, it does not consider other factors affecting benefit distribution such as the risk and public welfare of participating members in actual use. The supply chain contract coordination method can motivate various institutions to trade, but it is difficult to quantify the distribution plan. When using the fair entropy method, data needs to be disclosed to a certain extent, which conflicts with the privacy requirements of medical data. In addition, medical data contains many non-quantitative indicators, such as patient satisfaction and medical service quality, which are difficult to quantify and evaluate using entropy values.

[0005] The existing medical data factor income distribution mechanism is mainly aimed at the blockchain scenario. Due to the decentralized and tamper-proof characteristics of blockchain technology, some medical institutions use this technology to solve data privacy and security issues. However, blockchain technology faces some challenges during use. For example, its tamper-proof nature means that once the data is written into the system, it cannot be modified, and the flexibility of data transactions is relatively poor. Therefore, the existing medical data factor income distribution mechanism has the following shortcomings: First, data value assessment is difficult. The lack of a unified assessment system makes it difficult to quantify the value of data. In addition, the assessment is not flexible enough and the contribution score cannot be updated in a timely manner based on data characteristics, affecting the fairness of profit distribution.

[0006] Second, privacy and security issues. Medical and health data contain sensitive personal information, and privacy protection and data security issues are prominent, which limits data sharing.

[0007] Third, there is a lack of incentive-compatible allocation mechanisms. The current allocation mechanism is not sufficient to motivate data providers to share data, nor is it sufficient to motivate data processors and analysts to invest resources in data value mining.

[0008] In response to the above technical problems, the present invention proposes a method and system for allocating benefits of health care data elements based on contribution evaluation. Summary of the invention

[0009] The purpose of the present invention is to address the defects of the prior art and provide a method and system for allocating benefits of health care data elements based on contribution evaluation.

[0010] In order to achieve the above purpose, the present invention adopts the following technical solutions: The method of allocating benefits of health and medical data elements based on contribution evaluation includes: S1. Establish a data element contribution evaluation system: Establish multi-dimensional indicators based on the data of medical institutions, evaluate the multi-dimensional indicators based on the XGBoost machine learning model and SHAP value, and obtain the contribution of data elements; S2. Establish a profit distribution model: Calculate the value of the data itself based on the contribution of the data elements and the contribution of each assessed medical institution, and distribute the total profit of each medical institution according to the agreed profit distribution ratio; S3. The profit distribution method under the dynamic evaluation mechanism: recalculate the dynamic impact factor, adjust the contribution of the data elements of each medical institution according to the changes in the dynamic impact factor, and adjust the profit distribution.

[0011] Furthermore, the multi-dimensional indicators established in step S1 include data scarcity, data integrity, data consistency, data structuring, data redundancy, data volume, data time span, data quality, and data interpretability.

[0012] Furthermore, the step S1 is specifically as follows: S11. Obtain medical data corresponding to the multi-dimensional indicators, and divide the obtained medical data into a training data set and a test data set; S12. Build an XGBoost model and adjust the parameters of the XGBoost model, input the training data set into the adjusted XGBoost model for training, and optimize the XGBoost model through a cross-validation method to obtain a trained XGBoost model; S13. Use the trained XGBoost model to output the predicted results of the contribution of the data elements; S14. Use the SHAP interpreter to calculate the SHAP value corresponding to the medical data, and interpret the prediction results output by the trained XGBoost model based on the SHAP value.

[0013] Furthermore, the training of the XGBoost model in step S12 also includes training the XGBoost model using federated learning technology.

[0014] Furthermore, the distribution of the total revenue of each medical institution in step S2 is expressed as: ; Among them, R h represents the income allocated to each medical institution as a data provider; C h Indicates the total contribution of the data elements provided by the medical institution; C total Indicates the total contribution of data elements provided by all data providers; C obtain represents the sum of the contribution of each data element provided by the medical institution as the data acquirer; R p Indicates the revenue of data provider.

[0015] Furthermore, the step S3 is specifically as follows: S31. Identify and quantify the dynamic factors that affect the change of data value; S32. Regularly recalculate the dynamic impact factor; S33. Readjust the contribution of the data elements of each medical institution according to the dynamic impact factor calculated in step S32; S34. Adjust the profit distribution according to the contribution of the updated data elements.

[0016] Accordingly, a health and medical data element benefit distribution system based on contribution evaluation is also provided, including: Establish a module for establishing a data element contribution evaluation system: establish multi-dimensional indicators based on the data of medical institutions, evaluate the multi-dimensional indicators based on the XGBoost machine learning model and SHAP value, and obtain the contribution of data elements; The allocation module is used to establish a profit distribution model: the value of the data itself is calculated based on the contribution of the data elements and the contribution of each assessed medical institution, and the total profit of each medical institution is distributed according to the agreed profit distribution ratio; The first adjustment module is used for the profit distribution method under the dynamic evaluation mechanism: recalculating the dynamic impact factor, adjusting the data element contribution of each medical institution according to the change of the dynamic impact factor, and adjusting the profit distribution.

[0017] Furthermore, the establishment module includes: An acquisition module, used to acquire medical data corresponding to the multi-dimensional indicators, and divide the acquired medical data into a training data set and a test data set; A construction module is used to construct an XGBoost model and adjust the parameters of the XGBoost model, input the training data set into the adjusted XGBoost model for training, and optimize the XGBoost model through a cross-validation method to obtain a trained XGBoost model; The output module is used to output the predicted results of the contribution of data elements using the trained XGBoost model; The analysis module is used to calculate the SHAP value corresponding to the medical data using the SHAP interpreter, and interpret the prediction results output by the trained XGBoost model based on the SHAP value.

[0018] Furthermore, the distribution of the total revenue of each medical institution in the distribution module is expressed as: ;

[0019] Among them, R h represents the income allocated to each medical institution as a data provider; C h Indicates the total contribution of the data elements provided by the medical institution; C total Indicates the total contribution of data elements provided by all data providers; C obtain represents the sum of the contribution of each data element provided by the medical institution as the data acquirer; R p Indicates the revenue of data provider.

[0020] Furthermore, the first adjustment module includes: Identification module, used to identify and quantify the dynamic factors that affect the change of data value; A calculation module, used to periodically recalculate the dynamic impact factor; The second adjustment module is used to readjust the contribution of the data elements of each medical institution according to the calculated dynamic impact factor; The third adjustment module is used to adjust the profit distribution according to the contribution of the updated data elements.

[0021] Compared with the prior art, the present invention distributes the benefits of the health and medical data elements provided by participating institutions based on contribution evaluation, which includes three parts: establishing a data element contribution evaluation system, establishing a benefit distribution model, and a benefit distribution method under a dynamic evaluation mechanism. Through the three major modules, the contribution of data elements is quantified, the benefits of participating institutions are reasonably distributed, and the flexibility of data is increased through a dynamic evaluation mechanism. The specific effects are summarized as follows: 1. The present invention establishes a data element contribution evaluation system by comprehensively considering multiple key indicators such as data volume, data quality, scarcity, timeliness, etc., which not only quantifies the value of data, but also promotes the sharing and circulation of different data, strengthens the collaboration between medical institutions and the effective use of data resources.

[0022] 2. The present invention uses advanced data processing technologies including federated learning frameworks and machine learning models. These technologies allow the model to be continuously updated between various participants, and can flexibly adjust the global model based on new data, avoiding the need to re-collect and process data, and improving data processing efficiency and model scalability. In this way, the model can reflect the latest changes in data value, ensuring the timeliness and accuracy of the platform.

[0023] 3. The present invention proposes a fair and reasonable profit distribution mechanism, which aims to encourage medical institutions to accelerate the process of informatization and improve data quality. By evaluating the contribution of each participant and distributing profits, this mechanism promotes fair competition and cooperation among medical institutions, ensures the transparency and fairness of profit distribution, and thus encourages medical institutions to participate more actively in the data sharing platform, promoting the healthy development of the medical data factor market. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a diagram of the architecture of the health care data element benefit distribution method based on contribution evaluation provided in Example 1; Figure 2 It is a data element contribution evaluation flow chart provided in Example 1; Figure 3 This is a schematic diagram of privacy issues and security protection provided in Example 1; Figure 4 This is a health management flow chart provided in Example 2. DETAILED DESCRIPTION

[0025] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0026] The purpose of the present invention is to address the defects of the prior art and provide a method and system for allocating benefits of health care data elements based on contribution evaluation.

[0027] Embodiment 1 This embodiment provides a method for distributing benefits of health care data elements based on contribution evaluation, such as Figure 1 As shown, including: S1. Establish a data element contribution evaluation system: Establish multi-dimensional indicators based on the data of medical institutions, evaluate the multi-dimensional indicators based on the XGBoost machine learning model and SHAP value, and obtain the contribution of data elements; S2. Establish a profit distribution model: Calculate the value of the data itself based on the contribution of the data elements and the contribution of each assessed medical institution, and distribute the total profit of each medical institution according to the agreed profit distribution ratio; S3. The profit distribution method under the dynamic evaluation mechanism: recalculate the dynamic impact factor, adjust the contribution of the data elements of each medical institution according to the changes in the dynamic impact factor, and adjust the profit distribution.

[0028] In step S1, a data element contribution evaluation system is established: multi-dimensional indicators are established based on the data of medical institutions, and the multi-dimensional indicators are evaluated based on the XGBoost machine learning model and SHAP value to obtain the contribution of data elements.

[0029] like Figure 2 As shown in the figure, in the data element contribution evaluation system, the value of data elements is quantified by combining the XGBoost machine learning model and SHAP (Shapley Additive Explanations) value interpretability analysis, thereby obtaining the contribution of each participating institution. The system adopts a multi-dimensional indicator design, comprehensively considering the quantity, quality, diversity, scarcity, timeliness and other characteristics of data, and quantitatively evaluates the value of data based on the XGBoost model, and reflects the changes in data value through a dynamic evaluation mechanism.

[0030] The SHAP value is a method for explaining prediction results. It is based on the Shapley value in cooperative game theory and provides global and local interpretability for the model by quantifying the impact of each feature on the prediction results.

[0031] The establishment of a data element contribution evaluation system in step S1 specifically includes: S11. Obtain medical data corresponding to the multi-dimensional indicators, and divide the acquired medical data into a training data set and a test data set.

[0032] In order to comprehensively and objectively evaluate data characteristics, the design of multidimensional data indicators takes into account a variety of factors, including data scarcity, completeness, consistency, degree of structure, redundancy, data volume, and interpretability. The purpose of designing multidimensional data indicators is to reduce the bias caused by single-dimensional evaluation and ensure the accuracy and reliability of data in the overall evaluation process. Table 1 shows the multidimensional indicators corresponding to the data of medical institutions.

[0033] Table 1 Quantification table of multidimensional features of data elements

[0034] This embodiment generates a feature vector for each organization by defining and cyclically calling functions for calculating various indicators, so as to ensure the representativeness and comprehensiveness of the XGBoost model input data.

[0035] This embodiment needs to ensure the quality and integrity of the data collected and organized for medical institutions: by defining and calculating the functions of various indicators, generating the characteristic vector of each institution, performing data cleaning, processing missing values ​​and outliers, and preprocessing the data, such as normalization, encoding categorical variables, etc., to obtain processed data.

[0036] The collected data set is divided into training set and test set in a ratio of 3:7. The training set is used for the learning process of the model, while the test set is used for model verification and evaluation.

[0037] S12. Construct an XGBoost model and adjust the parameters of the XGBoost model, input the training data set into the adjusted XGBoost model for training, and optimize the XGBoost model through a cross-validation method to obtain a trained XGBoost model.

[0038] Import the processed training set into the constructed XGBoost model, initialize the XGBoost model by adjusting multiple hyperparameters such as booster, eta, min_child_weight, max_depth, gamma, objective and subample, and then adjust the two regularization parameters α and λ to reduce the complexity of the XGBoost model and improve the performance of the XGBoost model. After the XGBoost model is initialized, use the training set data to train the XGBoost model, and use the cv function to perform cross-validation in each iteration, and improve the generalization ability of the model by iteratively building a decision tree. Table 2 shows the parameter settings.

[0039] Table 2 Parameter settings

[0040] S13. Use the trained XGBoost model to output the predicted results of the contribution of data elements.

[0041] The trained XGBoost model automatically provides the importance value of each feature. The feature contribution can be obtained through the feature_importances_ attribute and sorted by importance. In the XGBoost model, the feature_importances_ attribute provides a convenient way to evaluate the contribution of each feature to the model prediction. Through the feature_importances_ attribute, we can identify the features that have the greatest impact on the model prediction results, thereby eliminating those features with less contribution in feature selection and simplifying the model.

[0042] In this embodiment, before step S14, the step also includes: performing performance evaluation on the trained XGBoost model.

[0043] After the XGBoost model prediction is completed, the mean square error (MSE) is used as a performance evaluation indicator to measure the deviation between the XGBoost model prediction value and the actual value. MSE calculates the average of the square of the difference between the predicted value and the true value. The smaller the MSE, the closer the prediction result of the XGBoost model is to the true value, and the better the performance of the model. In the present invention, the MSE indicator is used to evaluate the performance of the XGBoost model.

[0044] S14. Use the SHAP interpreter to calculate the SHAP value corresponding to the medical data, and interpret the prediction results output by the trained XGBoost model based on the SHAP value.

[0045] This embodiment uses SHAP values ​​to explain the prediction results of the machine learning model. The implementation steps of this method are to initialize a SHAP interpreter based on the trained XGBoost model, and use the interpreter to calculate the SHAP values ​​of the data, which represent the contribution of each feature to the model prediction. At the same time, four SHAP value visualization methods, SHAP Summary Plot, SHAP Force Plot, SHAP Bar Plot and SHAP Decision Plot, are used to show the impact of various factors on the model prediction results.

[0046] In this embodiment, training the XGBoost model also includes training the XGBoost model using federated learning technology.

[0047] This embodiment uses federated learning technology to maintain the data sovereignty of each institution by keeping data localized, and adapts to the legal requirements of cross-border transportation. In the face of the massive and dynamically updated nature of medical data, the federated learning framework effectively disperses the computing pressure, eliminates the need for centralized data processing, and supports an efficient and continuous learning process.

[0048] This embodiment uses federated learning technology to train a global XGBoost model to evaluate the contribution of data elements of medical institutions while protecting the privacy and security of the data. Figure 3 shown.

[0049] In this embodiment, a basic global model is first initialized and its parameters are distributed to all participating medical institutions, which then use these parameters to train local models locally without sharing local data. After training, each medical institution returns the updated model parameters to the central server. Then, a parameter aggregation function is used to integrate the model parameters of all institutions, and the aggregated parameters are used to update the global model, and the new parameters are distributed to all participants again. This process will be iterated until the model performance converges or meets the preset number of iterations. Finally, in order to evaluate the overall performance of the model, the model is applied locally in each institution and the evaluation performance indicators are calculated.

[0050] In order to further enhance privacy protection and prevent the central server from seeing the model updates of each participant, the present invention uses differential privacy technology to add noise to the model parameters, ensuring the security and reliability of the entire training process.

[0051] In step S2, a profit distribution model is established: the value of the data itself is calculated based on the contribution of the data elements and the contribution of each assessed medical institution, and the total profit of each medical institution is distributed according to the agreed profit distribution ratio.

[0052] The income of the data element provider mainly depends on the value of the data itself. The more data elements provided, the greater the income; the higher the value of each data element, the greater the income. Therefore, the income distribution of each institution providing medical data is determined by the contribution of the data elements and the value of the unit contribution.

[0053] The value of the data itself can be obtained through the contribution and unit contribution values, which can be converted into a specific amount of income. However, when actually calculating the income, it will be affected by other factors. For example, a medical institution has joined the data element platform and is a provider of medical data, so the institution can enjoy certain preferential policies when obtaining medical data from the platform; if the data provided contains personal information, the data value should be higher; if the data is used for non-profit purposes such as large-scale infectious disease monitoring and prevention, no fees should be charged.

[0054] In summary, the total revenue R total The revenue of the data element platform builder is R c and data provider revenue R p The total income is calculated according to the agreed income distribution ratio t:

[0055] Among them, the total revenue R total The calculation is as follows:

[0056] C pay Indicates the total contribution that needs to be paid, which is equal to the sum of the contribution of the data elements that need to be paid, calculated as follows:

[0057] Among them, C i is the contribution of the ith data element provided, and n is the total number of data elements that require payment.

[0058] A value The evaluation value of unit contribution can be comprehensively considered through a variety of value evaluation methods, such as obtaining other paid medical data from the market, calculating the contribution of these data according to the data element evaluation model, and then calculating the payment amount of unit contribution as a reference.

[0059] F total It represents the weighted coefficient of other influencing factors, which is equal to the product of the weighted coefficients of each other influencing factor. Other influencing factors include public welfare, data type coefficient, and nature of data acquisition, etc. The calculation is as follows:

[0060] Medium, F j represents the weighted coefficient of the jth other factor, and m is the total number of other factors.

[0061] In actual calculations, if the institution that obtains the data is a medical institution in the data factorization platform, then there is no need to pay for the data provided by the institution. Therefore, in the total revenue, the contribution that needs to be paid should be subtracted from the contribution of the data elements provided by the institution. At this time, the actual revenue is:

[0062] Among them, C obtain It represents the sum of the contribution of each data element provided by the medical institution as the data acquirer. If the institution is both the data provider and the data acquirer, then no fee is required for obtaining the data, and therefore the contribution of this part of the data element is not included in the profit distribution.

[0063] The medical data provided by the data provider is provided by various medical institutions. Each medical institution distributes the income as a data provider according to the proportion of its contribution to the data provided. The income of each medical institution is:

[0064] Among them, R h Indicates the income allocated to a medical institution as a data provider; C h represents the total contribution of the data elements provided by the medical institution, which is obtained by adding up the contributions of the data elements of each medical institution; C total Indicates the total contribution of data elements provided by all data providers; C obtain It represents the sum of the contribution of each data element provided by the medical institution as the data acquisition party. h -C obtain ) represents the net contribution of the medical institution after excluding the contribution of the data elements provided by the medical institution as a data acquisition party. The denominator (C total -C obtain ) represents the total contribution of all data providers, excluding the contribution of medical institutions as data acquirers.

[0065] This model ensures that each medical institution reasonably distributes the total revenue based on the relative value of the data elements it provides to the total contribution. It takes into account the situation that a medical institution may serve as both a data provider and a data acquirer, thus ensuring the accuracy and fairness of revenue distribution.

[0066] In the profit distribution method of the present invention, fairness in profit distribution is a core principle, which ensures that medical institutions of different sizes and influences can obtain corresponding profits when providing data with equal contribution. This mechanism encourages all participating institutions, regardless of their size and data provision, to actively participate in the data sharing platform and actively conduct data transactions.

[0067] In step S3, the profit distribution method under the dynamic evaluation mechanism is: recalculating the dynamic impact factor, adjusting the contribution of the data elements of each medical institution according to the change of the dynamic impact factor, and adjusting the profit distribution.

[0068] Evaluate the value of unit contribution: Obtain reference prices for other paid medical data through market research. Calculate the contribution of these data using the data element evaluation model. Calculate the payment amount for unit contribution based on the contribution and market reference price.

[0069] Determine unit contribution: Unit contribution refers to the contribution value of each data element to the total revenue. This value can be evaluated through the above market reference price and model.

[0070] Step S3 is specifically as follows: S31. Identify and quantify the dynamic factors that affect the change of data value; Identify and quantify dynamic factors that affect changes in data value, such as data relevance, data innovation, data coverage, and data growth rate, and then establish dynamic influencing factors.

[0071] S32. Regularly recalculate the dynamic impact factor; Set the calculation period of the dynamic impact factor, for example, once a month; recalculate the dynamic impact factor based on the latest data and market conditions.

[0072] S33. Readjust the contribution of the data elements of each medical institution according to the dynamic impact factor calculated in step S32; If the change in the dynamic impact factor is greater than the preset threshold (for example, 5%), the data element contribution assessment of each medical institution will be readjusted.

[0073] S34. Adjust the profit distribution according to the contribution of the updated data elements.

[0074] Adjustment of revenue distribution: Adjust revenue distribution based on updated contribution assessment; ensure fairness in the revenue distribution process and promote continuous improvement of data quality.

[0075] Calculate total revenue: Calculate total revenue based on adjusted contribution and unit contribution value.

[0076] Distribute revenue: distribute total revenue based on the contribution ratio of the data elements provided by the medical institutions; take into account the situation that medical institutions may serve as both data providers and data acquirers to ensure the accuracy and fairness of revenue distribution.

[0077] The dynamic evaluation mechanism of this embodiment ensures the continuous optimization of medical resources and service quality, and enables the system to identify and respond to changes in medical needs by regularly recalculating dynamic impact factors. These dynamic impact factors include data relevance, data innovation, data coverage, and data growth rate, reflecting the latest changes in data value.

[0078] In this embodiment, the calculation cycle of the dynamic impact factor is set to 1 month. The dynamic impact factor is recalculated every month. If the change of the factor is greater than 5%, the system will readjust the contribution evaluation of the data elements of each medical institution to ensure that the evaluation results reflect the latest data value. While updating the contribution evaluation, the profit distribution will also be adjusted to maintain fairness in the profit distribution process and promote continuous improvement of data quality. Table 3 shows the identification and quantification of dynamic impact factors.

[0079] Table 3 Identification and quantification of dynamic impact factors

[0080] This embodiment is to solve the defects of the existing health and medical data element income distribution method, based on the evaluation of the contribution of data elements provided by each participating institution, to provide an efficient and practical health and medical data element income distribution scheme. In view of the difficulty of data value assessment and the inflexibility of assessment, this embodiment proposes to establish a data element contribution evaluation system and a dynamic evaluation mechanism, design multi-dimensional indicators, integrate the XGBoost machine learning model and SHAP value interpretability analysis, and obtain an objective comprehensive score for the data value contribution, and then distribute the income based on the contribution score. In view of privacy and security issues, the present invention proposes to use federated learning technology to achieve secure data cooperation, ensure the efficiency of model training and the privacy security of data. In view of the lack of a reasonable distribution mechanism, this embodiment proposes a revenue distribution model, based on the contribution of health and medical data elements provided by participating institutions, to fairly distribute the income generated by data use, so that medical institutions can obtain income consistent with their contributions by sharing data, and increase the motivation to participate in data sharing.

[0081] Accordingly, this embodiment also provides a health care data element benefit distribution system based on contribution evaluation, including: Establish a module for establishing a data element contribution evaluation system: establish multi-dimensional indicators based on the data of medical institutions, evaluate the multi-dimensional indicators based on the XGBoost machine learning model and SHAP value, and obtain the contribution of data elements; The allocation module is used to establish a profit distribution model: the value of the data itself is calculated based on the contribution of the data elements and the contribution of each assessed medical institution, and the total profit of each medical institution is distributed according to the agreed profit distribution ratio; The first adjustment module is used for the profit distribution method under the dynamic evaluation mechanism: recalculating the dynamic impact factor, adjusting the data element contribution of each medical institution according to the change of the dynamic impact factor, and adjusting the profit distribution.

[0082] Furthermore, the establishment module includes: An acquisition module, used to acquire medical data corresponding to the multi-dimensional indicators, and divide the acquired medical data into a training data set and a test data set; A construction module is used to construct an XGBoost model and adjust the parameters of the XGBoost model, input the training data set into the adjusted XGBoost model for training, and optimize the XGBoost model through a cross-validation method to obtain a trained XGBoost model; The output module is used to output the predicted results of the contribution of data elements using the trained XGBoost model; The analysis module is used to calculate the SHAP value corresponding to the medical data using the SHAP interpreter, and interpret the prediction results output by the trained XGBoost model based on the SHAP value.

[0083] Furthermore, the distribution of the total revenue of each medical institution in the distribution module is expressed as: ; Among them, R h represents the income allocated to each medical institution as a data provider; C h Indicates the total contribution of the data elements provided by the medical institution; C total Indicates the total contribution of data elements provided by all data providers; C obtain represents the sum of the contribution of each data element provided by the medical institution as the data acquirer; R p Indicates the revenue of data provider.

[0084] Furthermore, the first adjustment module includes: Identification module, used to identify and quantify the dynamic factors that affect the change of data value; A calculation module, used to periodically recalculate the dynamic impact factor; The second adjustment module is used to readjust the contribution of the data elements of each medical institution according to the calculated dynamic impact factor; The third adjustment module is used to adjust the profit distribution according to the contribution of the updated data elements.

[0085] Embodiment 2 The difference between the health care data element benefit distribution method based on contribution evaluation provided in this embodiment and the first embodiment is that: This embodiment applies the revenue distribution method to the health management scenario, establishes a data element revenue distribution mechanism based on the health management platform, and verifies the rationality of this revenue distribution method. In this embodiment, the data element platform provides data elements to the health management platform, and the health management platform pays for the provided data, and distributes the revenue to the data provider and the platform builder.

[0086] The data element platform includes the platform builders and data providers. The platform builders and data providers are the main beneficiaries to distribute the payment amount of the health management platform. The data providers also include various medical institutions. The total income of the data providers will be redistributed to various medical institutions according to the income distribution model.

[0087] The management process of the health management platform is as follows: Figure 4 As shown, taking patient A as an example, medical institutions A, B, and C are the providers of medical data in the data element platform.

[0088] 1. Patient A signed a contract on the health management platform through medical institution A.

[0089] 2. The health management platform provides free disease screening and prediction reports based on the patient's existing information. Based on the patient's information, the health management platform screens and predicts the disease, and finds that patient A has a chronic disease of hypertension, is suspected of having rheumatoid arthritis but has not been diagnosed, and is predicted to have a high probability of suffering from stroke. In this management plan, the data element platform obtains the patient's disease screening and prediction data set, and the patient is scored for disease management.

[0090] 3. The patient skipped personalized diagnosis, and the health management platform recommended chronic disease management plans for hypertension, diabetes, and allergic rhinitis to the patient. According to the patient's condition, multiple management plans for each chronic disease in different price ranges were provided. The patient chose a 1-year chronic disease management plan for hypertension and did not choose other chronic disease management plans. In this management plan, the data element platform obtained a data set of all influencing factors of the patient's hypertension, and the patient was scored for hypertension disease management.

[0091] 4. The health management platform recommended a stroke prevention and management program to the patient, and the patient chose a one-year management program. In this management program, the data element platform obtained a data set of all influencing factors of the patient's stroke disease, and the patient was scored for stroke disease management.

[0092] 5. The health management platform recommends a rheumatology and immunology department number to the patient based on the patient's suspected rheumatoid arthritis disease, and the patient chooses to register and pay.

[0093] Relying on the above health management platform, this method is specifically implemented. In this health management platform, patients undergo disease screening and prediction, hypertension disease management and prediction, and stroke disease management and prediction. The data element platform provides data in the following three scenarios: (1) The data element platform provides patient data for disease screening and prediction. (Step 2) The data elements provided by the three medical institutions are shown in Table 4. There are 15 data element attributes provided for disease screening and prediction.

[0094] Table 4 Data element content

[0095] According to the data provided by the three participating medical institutions, the feature calculation function is defined according to the multidimensional data indicators of this method, and the nine data features of each medical institution, including data scarcity R, data integrity C, data consistency U, data structured degree S, data redundancy Rd, data volume V, data time span T, data quality O, and data interpretability E, are calculated to generate a data set. Then the features and labels are divided, and the XGBoost model is trained. Finally, the feature importance feature_importances_ attribute is obtained and saved for contribution evaluation. Table 5 shows the contribution of data elements of the three medical institutions.

[0096] Table 5 Contribution of data elements

[0097] The total contribution values ​​of the three medical institutions A, B, and C are 3.67, 2.47, and 2.86 respectively. total =3.67+2.47+2.86=9. Specifically in this embodiment, the health management platform belongs to the regional medical center medical institution A, then the contribution of the data elements provided by the data acquisition party medical institution A is C obatin =3.67.

[0098] Table 6 shows the other factors affecting the distribution of income. There are three aspects below. The weighted coefficients of other factors are calculated by formula (6) of the first embodiment: F total =0.8×5×0.8=3.2.

[0099] Table 6 Other influencing factors

[0100] Through comprehensive consideration of various value assessment methods, the value of unit contribution is A value =20. According to formula (7) of the first embodiment, the total revenue is calculated as: R total =(9-3.67)×20×3.2=340.

[0101] Assume t = 0.3, that is, the benefits of the data element platform builder and the data provider are distributed in a ratio of 3:7. Then the data builder's benefits are calculated using formula (2) of Example 1: R c =R total ×t=340×0.3=102, the data provider's income is calculated by formula (3) in Example 1: R p =R total ×(1-t)=340×(1-0.3)=238.

[0102] The income of each medical institution of the data provider is calculated by formula (8) of the first embodiment: The income of medical institution A is RhA =(3.67-3.67) / (9-3.67)×238=0, the income of medical institution B is R hB =(2.47-0) / (9-3.67)×238=110, the income of medical institution C is R hC =(2.86-0) / (9-3.67)×238=128.

[0103] (2) Every quarter, obtain the data of all factors affecting the patient's hypertension through the data element platform and perform disease management scoring. (Step 3) The data elements provided by the three medical institutions are shown in Table 7, providing 20 data element attributes on all influencing factors of hypertension.

[0104] Table 7 Data element content

[0105] According to the calculation method in scenario one, the data feature contributions of the three medical institutions are calculated, as shown in Table 8.

[0106] Table 8 Contribution of data elements

[0107] The total contribution values ​​of the three medical institutions A, B, and C are 3.36, 2.82, and 3.18 respectively. total =3.36+2.82+3.18=9.36. Specifically in this embodiment, among the data providers, medical institutions A and B are also data acquirers, so C obatin-A =3.36, C obatin-B =2.82, C obatin =C obatin-A +C obatin-B =6.18.

[0108] Other factors affecting the distribution of benefits and the value of unit contribution are the same as in scenario 1, i.e., F total =3.2, A value =20. According to formula (7) of the first embodiment, the total revenue is calculated as: R total =(9.36-6.18)×20×3.2=204.

[0109] Assume t = 0.3, and then calculate the data construction party's income through formula (2) of embodiment 1: R c =R total ×t=204×0.3=61, the data provider's income is calculated by formula (3) in Example 1: R p =R total ×(1-t)=204×(1-0.3)=143.

[0110] The income of each medical institution of the data provider is calculated by formula (8) of the first embodiment: The income of medical institution A is R hA =(3.36-3.36) / (9.36-6.18)×143=0, the income of medical institution B is R hB =(2.82-2.82) / (9.36-6.18)×143=0, the income of medical institution C is R hC =(3.18-0) / (9.36-6.18)×143=143.

[0111] (3) Every quarter, obtain all the influencing factors of stroke in patients through the data element platform and perform disease prediction scoring. (Step 4) The data elements provided by the three medical institutions are shown in Table 9, which provides 17 data element attributes on all influencing factors of stroke.

[0112] Table 9 Data element content

[0113] The total contribution values ​​of the three medical institutions A, B, and C are 3, 2.97, and 3.03 respectively. total =3+2.97+3.03=9. Specifically in this embodiment, among the data providers, medical institution B is the data acquirer, so C obatin =2.97.

[0114] Other factors affecting the distribution of benefits and the value of unit contribution are the same as in scenario 1, i.e., F total =3.2, A value =20. According to formula (7) of the first embodiment, the total revenue is calculated as: R total =(9-2.97)×20×3.2=386.

[0115] Assume t = 0.3, and then calculate the data construction party's income through formula (2) of embodiment 1: R c =R total ×t=386×0.3=116, the data provider's income is calculated by formula (3) in Example 1: R p =R total ×(1-t)=386×(1-0.3)=270.

[0116] The income of each medical institution of the data provider is calculated by formula (8) of the first embodiment: The income of medical institution A is R hA =(3-0) / (9-2.97)×270=134, the income of medical institution B is R hB=(2.97-2.97) / (9-2.97)×270=0, the income of medical institution C is R hC =(3.03-0) / (9-2.97)×270=136.

[0117] (4) The summary of the benefits of each beneficiary in this health management platform is shown in Table 10.

[0118] Table 10 Profit Distribution

[0119] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for distributing benefits of health care data elements based on contribution evaluation, characterized in that: include: S1. Establish a data element contribution evaluation system: Establish multi-dimensional indicators based on the data of medical institutions, evaluate the multi-dimensional indicators based on the XGBoost machine learning model and SHAP value, and obtain the contribution of data elements; S2. Establish a profit distribution model: Calculate the value of the data itself based on the contribution of the data elements and the contribution of each assessed medical institution, and distribute the total profit of each medical institution according to the agreed profit distribution ratio; S3. The profit distribution method under the dynamic evaluation mechanism: recalculate the dynamic impact factor, adjust the contribution of the data elements of each medical institution according to the changes in the dynamic impact factor, and adjust the profit distribution.

2. The method for distributing benefits of health care data elements based on contribution evaluation according to claim 1 is characterized in that: The multi-dimensional indicators established in step S1 include data scarcity, data integrity, data consistency, data structuring, data redundancy, data volume, data time span, data quality, and data interpretability.

3. The method for distributing benefits of health care data elements based on contribution evaluation according to claim 1 is characterized in that: The step S1 is specifically as follows: S11. Obtain medical data corresponding to the multi-dimensional indicators, and divide the obtained medical data into a training data set and a test data set; S12. Build an XGBoost model and adjust the parameters of the XGBoost model, input the training data set into the adjusted XGBoost model for training, and optimize the XGBoost model through a cross-validation method to obtain a trained XGBoost model; S13. Use the trained XGBoost model to output the predicted results of the contribution of the data elements; S14. Use the SHAP interpreter to calculate the SHAP value corresponding to the medical data, and interpret the prediction results output by the trained XGBoost model based on the SHAP value.

4. The method for distributing benefits of health care data elements based on contribution evaluation according to claim 3 is characterized in that: The training of the XGBoost model in step S12 also includes training the XGBoost model using federated learning technology.

5. The method for distributing benefits of health care data elements based on contribution evaluation according to claim 1, characterized in that: The distribution of the total revenue of each medical institution in step S2 is expressed as: ； Among them, R h represents the income allocated to each medical institution as a data provider; C h Indicates the total contribution of the data elements provided by the medical institution; C total Indicates the total contribution of data elements provided by all data providers; C obtain represents the sum of the contribution of each data element provided by the medical institution as the data acquirer; R p Indicates the revenue of data provider.

6. The method for distributing benefits of health care data elements based on contribution evaluation according to claim 1, characterized in that: The step S3 is specifically as follows: S31. Identify and quantify the dynamic factors that affect the change of data value; S32. Regularly recalculate the dynamic impact factor; S33. Readjust the contribution of the data elements of each medical institution according to the dynamic impact factor calculated in step S32; S34. Adjust the profit distribution according to the contribution of the updated data elements.

7. A health care data element profit distribution system based on contribution evaluation, characterized in that: include: Establish a module for establishing a data element contribution evaluation system: establish multi-dimensional indicators based on the data of medical institutions, evaluate the multi-dimensional indicators based on the XGBoost machine learning model and SHAP value, and obtain the contribution of data elements; The allocation module is used to establish a profit distribution model: the value of the data itself is calculated based on the contribution of the data elements and the contribution of each assessed medical institution, and the total profit of each medical institution is distributed according to the agreed profit distribution ratio; The first adjustment module is used for the profit distribution method under the dynamic evaluation mechanism: recalculating the dynamic impact factor, adjusting the data element contribution of each medical institution according to the change of the dynamic impact factor, and adjusting the profit distribution.

8. The health care data element profit distribution system based on contribution evaluation according to claim 7 is characterized in that: The establishment module includes: An acquisition module, used to acquire medical data corresponding to the multi-dimensional indicators, and divide the acquired medical data into a training data set and a test data set; A construction module is used to construct an XGBoost model and adjust the parameters of the XGBoost model, input the training data set into the adjusted XGBoost model for training, and optimize the XGBoost model through a cross-validation method to obtain a trained XGBoost model; The output module is used to output the predicted results of the contribution of data elements using the trained XGBoost model; The analysis module is used to calculate the SHAP value corresponding to the medical data using the SHAP interpreter, and interpret the prediction results output by the trained XGBoost model based on the SHAP value.

9. The health care data element profit distribution system based on contribution evaluation according to claim 7 is characterized in that: The allocation of the total revenue of each medical institution in the allocation module is expressed as: ； Among them, R h represents the income allocated to each medical institution as a data provider; C h Indicates the total contribution of the data elements provided by the medical institution; C total Indicates the total contribution of data elements provided by all data providers; C obtain represents the sum of the contribution of each data element provided by the medical institution as the data acquirer; R p Indicates the revenue of data provider.

10. The health care data element profit distribution system based on contribution evaluation according to claim 7 is characterized in that: The first adjustment module includes: Identification module, used to identify and quantify the dynamic factors that affect the change of data value; A calculation module, used to periodically recalculate the dynamic impact factor; The second adjustment module is used to readjust the contribution of the data elements of each medical institution according to the calculated dynamic impact factor; The third adjustment module is used to adjust the profit distribution according to the contribution of the updated data elements.

Images