Treatment benefit evaluation and treatment scheme recommendation method and system based on causal inference

Abstract

The invention relates to a treatment benefit evaluation and treatment scheme recommendation method and system based on causal inference, and the method comprises the steps: obtaining a prognosis risk index of a patient and a treatment benefit index possibly obtained by the patient from a corresponding treatment method through a trained DML model according to an input electronic medical record of the patient and different treatment schemes; by estimating a progress risk index and a treatment benefit index, electronic medical record data of a patient is converted into two dimensions of risk and benefit, and a targeted treatment recommendation scheme is provided for the patient according to a multi-level recommendation model. Not only is the working process of doctors optimized, but also reasonable distribution of medical resources is promoted. The development of precise medical treatment is facilitated, an intelligent tool is introduced for the medical industry, the diagnosis and treatment efficiency is greatly improved, and the cost is reduced. And for patients, the control feeling of the patients on own illness states is enhanced, the anxiety mood is relieved, meanwhile, the medical experience is improved, and the method has far-reaching significance in promoting the modernization of clinical medical treatment and public health systems.

Description

Technical Field

[0001] The present invention belongs to the medical field, and particularly relates to a method and system for evaluating treatment benefits and recommending treatment plans based on causal inference. Background Art

[0002] For some diseases with slow progression but difficult to cure, understanding the prognosis of patients and key indicators such as progression-free survival (PFS) and overall survival (OS) is crucial for formulating effective personalized treatment plans. Reasonably evaluating the prognosis of patients can help doctors select the most suitable treatment strategies and avoid over-treatment or missed treatment.

[0003] After first-line treatment, many diseases may require maintenance treatment to extend the survival period of patients or delay the recurrence of the disease. Although maintenance treatment has significant potential, such as extending disease-free survival and overall survival, it may also bring some side effects and health risks. For example, some maintenance treatments may increase the risk of infection in patients or cause other adverse reactions. In addition, maintenance treatment usually requires patients to adhere to it for a long time, which will surely increase the physical and economic burdens of patients, especially in cases where frequent monitoring and medication are required.

[0004] Therefore, how to accurately evaluate the benefit index of patients receiving maintenance treatment has become a key issue in clinical treatment. Summary of the Invention

[0005] Aiming at the defects existing in the prior art, the purpose of the present invention is to provide a method and system for evaluating treatment benefits and recommending treatment plans based on causal inference, which can, based on AI technology, by using the electronic medical record data of patients, help doctors predict the survival period, recurrence risk and possible side effects of patients after receiving a certain treatment plan, so as to optimize the treatment plan. In addition, it can also assist in screening out those patient groups who benefit the most from maintenance treatment, reduce unnecessary treatments and side effects, and help patients and doctors make more scientific and reasonable treatment decisions.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is:

[0007] In a first aspect, a method for evaluating treatment benefits based on causal inference, the method obtains the prognosis progression risk index of a patient and the treatment benefit index that the patient may obtain from the corresponding treatment method through a trained DML model according to the input electronic medical record of the patient and different treatment plans.

[0008] Further, the electronic medical record of the patient includes the basic information of the patient and the physiological indicators of each visit of the patient.

[0009] Further, for the same treatment plan, by inputting the same feature in the electronic medical records of different patients and deeply analyzing the risk index of the patient's prognosis progression and the treatment benefit index output, the correlation between the risk index of the patient's prognosis progression and the treatment benefit index and each index in the patient's electronic medical record can be obtained.

[0010] In a second aspect, a treatment plan recommendation method based on causal inference, the method includes the following steps:

[0011] S1. According to the input electronic medical records of the patient and different treatment plans, obtain the risk index of the patient's prognosis progression and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model;

[0012] S2. Based on the risk index of the patient's prognosis progression output by the DML model and the treatment benefit index that the patient may obtain from the corresponding treatment method, recommend a multi-level treatment plan;

[0013] S3. Based on the recommended multi-level treatment plan, combined with the input treatment preferences, comprehensively recommend a most suitable treatment plan.

[0014] Further, in step S2, recommend a multi-level treatment plan based on the SVM model.

[0015] Further, the recommended levels of the multi-level treatment plan in step S2 include four levels: high risk and high benefit, high risk and low benefit, low risk and high benefit, and low risk and low benefit.

[0016] Further, for the high-risk and high-benefit treatment plan in step S2, recommend active treatment; for the high-risk and low-benefit treatment plan, recommend cautious treatment or choose an alternative plan; for the low-risk and high-benefit treatment plan, recommend moderate treatment; for the low-risk and low-benefit treatment plan, recommend not to treat or choose a non-invasive treatment method.

[0017] In a third aspect, a treatment benefit evaluation system based on causal inference, the system adopts a treatment benefit evaluation method based on causal inference described in the first aspect of the present invention and any optional implementation manner thereof. The system obtains the risk index of the patient's prognosis progression and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model according to the input electronic medical records of the patient and different treatment plans.

[0018] In a fourth aspect, a treatment benefit evaluation system based on causal inference, the system adopts a treatment plan recommendation method based on causal inference described in the second aspect of the present invention and any optional implementation manner thereof. The system includes:

[0019] The treatment benefit index acquisition module is used to obtain the prognosis progression risk index of the patient and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model according to the input electronic medical record of the patient and different treatment plans.

[0020] The multi-level treatment plan recommendation module is used to generate recommended treatment plans at multiple levels based on the prognosis progression risk index of the patient output by the DML model and the treatment benefit index that the patient may obtain from the corresponding treatment method.

[0021] The comprehensive treatment plan recommendation module is used to comprehensively recommend a most suitable treatment plan based on the generated recommended treatment plans at multiple levels and in combination with the input treatment preferences.

[0022] Furthermore, the recommended levels for generating recommended treatment plans at multiple levels in the multi-level treatment plan recommendation module include four levels: high risk and high benefit, high risk and low benefit, low risk and high benefit, and low risk and low benefit.

[0023] The beneficial technical effects of the present invention are as follows: The method and system for evaluating treatment benefits and recommending treatment plans based on causal inference disclosed by the present invention have mined more valuable and reference information from the electronic medical record data of patients in the hospital to help doctors make diagnoses.

[0024] First, by estimating the progression risk index and the treatment benefit index, the electronic medical record data of the patient is transformed into two dimensions of risk and benefit, and targeted treatment recommendation plans are proposed for the patient according to the multi-level recommendation model (SVM model), so as to provide a reference plan for doctors to formulate treatment plans, and provide patients with a basic prediction and judgment of their own condition, so as to make a decision on whether to receive treatment together with the doctor.

[0025] Secondly, by deeply analyzing the patient's characteristics, it is possible to give early warnings to high-risk patients, prompt the physiological indicators that need to be focused on, and give corresponding medical suggestions, guiding doctors to conduct simulated interventions on them and observing the possible future conditions of the patients, so as to control the trend of the patient's physical condition to develop in a good direction as much as possible, which has great practical value.

[0026] From a macroscopic perspective, it not only optimizes the work process of doctors, but also promotes the rational allocation of medical resources. It contributes to the development of precision medicine, introduces intelligent tools into the medical industry, greatly improves the diagnosis and treatment efficiency, and reduces costs. For patients, it enhances their sense of control over their own conditions, reduces anxiety, and improves the medical experience at the same time. Therefore, it has far-reaching significance in promoting the modernization of clinical medicine and the public health system. Brief Description of the Drawings

[0027] Figure 1Flowchart of a treatment benefit evaluation method based on causal inference shown in Embodiment 1 of the present invention;

[0028] Figure 2 Flowchart of a treatment plan recommendation method based on causal inference shown in Embodiment 2 of the present invention;

[0029] Figure 3 Flowchart of another treatment plan recommendation method based on causal inference shown in Embodiment 2 of the present invention. Detailed implementation manners

[0030] The present invention will be further described below in conjunction with the accompanying drawings and detailed implementation manners.

[0031] Term explanations

[0032] Electronic medical record

[0033] An electronic medical record (EMR, Electronic Medical Record) is also called a computerized medical record system or a computer-based patient record. It is a digital medical record saved, managed, transmitted, and reproduced by electronic devices (computers, health cards, etc.) to replace handwritten paper medical records. Its content includes all the information in the paper medical record.

[0034] Progression risk index

[0035] The progression risk index is an indicator used to evaluate the progression risk of a patient in the next year (range 0 - 1). The larger the index, the more severe the patient's condition and the more likely the condition will progress in the next year.

[0036] Treatment benefit index

[0037] The treatment benefit index is an indicator used to evaluate the benefit level of a patient after receiving a certain medical treatment, specifically reflected in the reduction of the progression risk. The larger the index, the lower the benefit level of the patient receiving the treatment.

[0038] Causal inference uplift model

[0039] Causal inference attempts to answer the question of what will happen if something is done. In the embodiments of the present invention, causal inference is implemented in the field of diagnosis and treatment recommendation, attempting to explore different diagnosis and treatment plans and ultimately how much they can affect the progression probability of patients. Specifically, in the embodiments of the present invention, through algorithms such as DML (Double Machine Learning) and SVM (Support Vector Machine), information affecting the benefits that patients can obtain from treatment is discovered from the data.

[0040] Embodiment 1

[0041] like Figure 1 As shown, an embodiment of the present invention provides a treatment benefit assessment method based on causal inference, which obtains the patient's prognosis progression risk index and the treatment benefit index that the patient may obtain from the corresponding treatment method through a trained DML model according to the input patient electronic medical records and different treatment plans.

[0042] Electronic medical records include basic patient information and physiological indicator data for each visit.

[0043] Based on the patient's physiological data, clinical characteristics and other information, the patient's prognosis progression risk index and treatment benefit index are predicted. These two indexes represent the risk of disease progression and the possible treatment effect after treatment.

[0044] By inputting the patient's electronic medical records into the DML model, the DML model can calculate the potential benefit of each patient from the corresponding treatment method, which provides data support for clinical decision-making.

[0045] In addition, for the same treatment plan, by inputting the same features in the electronic medical records of different patients, the output patient's prognosis progression risk index and treatment benefit index are deeply analyzed, and the correlation between the patient's prognosis progression risk index and treatment benefit index and various indicators in the patient's electronic medical record can be obtained. It helps to dig out more important and representative key data and provide a scientific basis for optimizing model algorithms and doctors' judgments. For example, by studying the relationship between indicators such as the patient's age, past medical history, genetic characteristics and treatment effects, the personalized design of treatment plans can be further improved to ensure the rational allocation of resources.

[0046] Embodiment 2

[0047] like Figure 2 As shown, an embodiment of the present invention provides a treatment plan recommendation method based on causal inference, the method comprising the following steps:

[0048] S1. Based on the input patient electronic medical records and different treatment plans, the trained DML model is used to obtain the patient's prognostic progression risk index and the treatment benefit index that the patient may obtain from the corresponding treatment method.

[0049] Electronic medical records include basic patient information and physiological indicator data for each visit.

[0050] Based on the patient's physiological data, clinical characteristics and other information, the patient's prognosis progression risk index and treatment benefit index are predicted. These two indexes represent the risk of disease progression and the possible treatment effect after treatment.

[0051] By inputting the electronic medical records of patients into the DML model, the DML model can calculate the potential benefits of each patient for the corresponding treatment methods, which provides data support for clinical decision-making.

[0052] In addition, for the same treatment plan, by deeply analyzing the same feature in the electronic medical records of different patients, as well as the risk index of the patient's prognosis progression and the treatment benefit index output, the correlation between the risk index of the patient's prognosis progression and the treatment benefit index and each index in the patient's electronic medical records can be obtained. This helps to uncover more important and representative key data, providing a scientific basis for optimizing the model algorithm and the doctor's judgment. For example, by studying the relationship between indicators such as the patient's age, past medical history, and genetic characteristics and the treatment effect, the personalized design of the treatment plan can be further improved to ensure the rational allocation of resources.

[0053] S2. Generate multiple levels of recommended treatment plans based on the risk index of the patient's prognosis progression output by the DML model and the treatment benefit index that the patient may obtain from the corresponding treatment method.

[0054] As Figure 3 shown, for the electronic medical records of a certain patient, according to the DML model, the risk index of the patient's prognosis progression and the treatment benefit index that the patient may obtain from the corresponding treatment method can be obtained. Combining these two indicators, multiple levels of medical advice can be generated for the patient.

[0055] Step S2 includes the following sub-steps:

[0056] Input the risk index of the patient's prognosis progression and the treatment benefit index that the patient may obtain from the corresponding treatment method into the SVM model to obtain the recommended level.

[0057] Take the risk index of the prognosis progression and the treatment benefit index as inputs and input them into the SVM model for processing. The SVM model will calculate the treatment level suitable for the patient based on these features, and based on the SVM model, judge the priority (i.e., the recommended level) of the patient applying various treatment plans according to the patient's individual situation. Different SVM models trained during the research process can generate multiple recommended levels for different treatment plans to ensure that the plan most in line with the actual needs of the patient is proposed.

[0058] Comprehensively consider the outputs of different SVM models for comprehensive recommendation: Since different treatment focuses are concerned and multiple SVM models are trained to obtain corresponding recommendation levels, the output results of these SVM models can be comprehensively considered according to specific medical scenarios and patient needs. The recommendations at different levels can be prioritized according to the actual situation. For example, a treatment plan with high efficiency and fewer side effects can be recommended first, or according to the patient's risk preference, the corresponding treatment intensity can be recommended. Finally, the system will automatically obtain the recommended treatment plan that best meets the patient's needs.

[0059] In the embodiment of the present invention, using the existing electronic medical record data, the SVM model is trained from two dimensions of risk and benefit, so as to divide the recommendation levels output by the SVM model into four grades, namely, high risk and high benefit, high risk and low benefit, low risk and high benefit, and low risk and low benefit.

[0060] High risk and high benefit: It means that the patient has a relatively high risk of disease progression, but this treatment may bring significant benefits, and active treatment is recommended.

[0061] High risk and low benefit: It means that the patient has a high disease risk, but the treatment effect is poor, and cautious treatment or alternative options are recommended.

[0062] Low risk and high benefit: It means that the patient has a relatively low disease risk, but this treatment may bring good effects, and moderate treatment is recommended.

[0063] Low risk and low benefit: It means that both the patient's disease risk and treatment benefit are relatively low, and no treatment or non-invasive treatment methods may be recommended.

[0064] S3. Based on the recommended treatment plans at multiple levels generated by the SVM model, combined with the treatment preferences of doctors and patients, comprehensively recommend a most suitable treatment plan.

[0065] Customized treatment plan: In the embodiment of the present invention, by comprehensively considering the prediction results of multiple models, the treatment preferences of doctors and patients, and the individual differences of patients, a treatment plan for each patient is customized. The system will propose the most suitable treatment method based on different treatment levels and in combination with the actual situation of the patient. For example, the system will provide personalized medication recommendations for the patient by combining factors such as the patient's health status, economic burden, and treatment side effects.

[0066] Auxiliary decision-making support: In the embodiment of the present invention, through in-depth data mining, decision-making support is provided for doctors. Doctors can make treatment decisions that best meet the individual needs of patients based on the recommended plans provided by the system and combined with clinical experience. The system can not only improve the accuracy of treatment but also improve the execution efficiency of the treatment plan, thus providing more comprehensive and personalized medical services for patients.

[0067] Example 3

[0068] An embodiment of the present invention provides a treatment benefit evaluation system based on causal inference. The system adopts a treatment benefit evaluation method based on causal inference described in Embodiment 1 of the present invention and any optional implementation manner thereof. The system obtains the prognosis progression risk index of the patient and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model according to the input patient electronic medical record and different treatment plans.

[0069] Example 4

[0070] An embodiment of the present invention provides a treatment plan recommendation system based on causal inference. The system adopts a treatment plan recommendation method based on causal inference described in Embodiment 2 of the present invention and any optional implementation manner thereof. The system includes:

[0071] A treatment benefit index acquisition module, configured to obtain the prognosis progression risk index of the patient and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model according to the input patient electronic medical record and different treatment plans.

[0072] A multi-level treatment plan recommendation module, configured to generate recommended treatment plans at multiple levels based on the prognosis progression risk index of the patient output by the DML model and the treatment benefit index that the patient may obtain from the corresponding treatment method.

[0073] A comprehensive treatment plan recommendation module, configured to comprehensively recommend a most suitable treatment plan by combining the recommended treatment plans at multiple levels generated by the SVM model with the treatment preferences of doctors and patients.

[0074] It can be seen from the above embodiments that the treatment benefit evaluation and treatment plan recommendation method and system based on causal inference disclosed in the present invention are based on the DML model, comprehensively evaluate the risks and benefits of patients by combining the electronic medical record information of patients, and then give personalized medical advice. It can accurately evaluate the disease progression risk and treatment benefit of patients, and provide customized treatment suggestions for patients according to the comprehensive recommendation algorithm. On the one hand, a method for prognostic risk assessment through the DML model is proposed. Using the electronic medical record data of patients, the risk of future disease progression of patients is predicted, and a risk index is generated to help doctors identify high-risk patients. Through a comprehensive analysis of the electronic medical record data of these patients, the model can also evaluate the benefit of a specific treatment to the patient, form a treatment benefit index, so as to provide specific decision-making support for doctors.

[0075] On the other hand, a medication recommendation method is proposed. This method is based on the patient's electronic medical record data, and by comprehensively evaluating the patient's progression risk and treatment benefit, it recommends the most suitable treatment plan for each patient. According to the patient's specific needs (such as avoiding side effects, prolonging survival, etc.) and the characteristics of the disease, personalized treatment suggestions are provided. It not only considers the effectiveness of drug treatment, but also takes into account the patient's physical and mental state, truly achieving a treatment plan recommendation that is tailored to the disease and varies from person to person.

[0076] The method and system described in the present invention are not limited to the embodiments described in the specific implementation manners. Those skilled in the art can obtain other implementation manners based on the technical solutions of the present invention, which also belong to the scope of the technical innovation of the present invention.

Claims

1. A treatment benefit assessment method based on causal inference, which obtains the patient's prognosis progression risk index and the treatment benefit index that the patient may obtain from the corresponding treatment method through a trained DML model based on the patient's electronic medical records and different treatment plans.

2. A method for evaluating treatment benefit based on causal inference according to claim 1, characterized in that: The patient's electronic medical record includes the patient's basic information and the patient's physiological indicators at each visit.

3. A method for evaluating treatment benefit based on causal inference according to claim 1, characterized in that: For the same treatment plan, by inputting the same features in the electronic medical records of different patients, an in-depth analysis is conducted on the output patient's prognostic progression risk index and treatment benefit index, which can obtain the correlation between the patient's prognostic progression risk index and treatment benefit index and various indicators in the patient's electronic medical record.

4. A method for recommending a treatment plan based on causal inference, the method comprising the following steps: S1. Based on the input patient electronic medical records and different treatment plans, the trained DML model is used to obtain the patient's prognostic progression risk index and the treatment benefit index that the patient may obtain from the corresponding treatment method; S2. Recommend multi-level treatment plans based on the patient's prognostic progression risk index output by the DML model and the treatment benefit index that the patient may obtain from the corresponding treatment methods; S3. Based on the recommended multi-level treatment plans and combined with the input treatment preferences, a comprehensive recommendation of the most suitable treatment plan is made.

5. A method for recommending a treatment plan based on causal inference as claimed in claim 4, characterized in that: In step S2, a multi-level treatment plan is recommended based on the SVM model.

6. A method for recommending a treatment plan based on causal inference as claimed in claim 4, characterized in that: The recommended levels of the multi-level treatment plan recommended in step S2 include four levels: high risk and high benefit, high risk and low benefit, low risk and high benefit, and low risk and low benefit.

7. A method for recommending a treatment plan based on causal inference as claimed in claim 6, characterized in that: In step S2, active treatment is recommended for high-risk, high-benefit treatment options; cautious treatment or selection of alternative options is recommended for high-risk, low-benefit treatment options; For low-risk, high-benefit treatment options, moderate treatment is recommended; For low-risk, low-benefit treatment options, it is recommended not to treat or to choose non-invasive treatment methods.

8. A treatment benefit assessment system based on causal inference, the system adopting a treatment benefit assessment method based on causal inference according to any one of claims 1 to 3, characterized in that: The system obtains the patient's prognosis progression risk index and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model based on the input patient electronic medical records and different treatment plans.

9. A treatment plan recommendation system based on causal inference, the system adopts a treatment plan recommendation method based on causal inference as claimed in any one of claims 4 to 7, characterized in that: The system comprises: The treatment benefit index acquisition module is used to obtain the patient's prognosis progression risk index and the treatment benefit index that the patient may obtain from the corresponding treatment method through the trained DML model based on the input patient electronic medical records and different treatment plans. The multi-level treatment plan recommendation module is used to generate multiple levels of recommended treatment plans based on the patient's prognostic progression risk index output by the DML model and the treatment benefit index that the patient may obtain from the corresponding treatment method. The comprehensive treatment plan recommendation module is used to comprehensively recommend a most suitable treatment plan based on the generated multiple levels of recommended treatment plans and the input treatment preferences.

10. A treatment plan recommendation system based on causal inference as claimed in claim 9, characterized in that: The multi-level treatment plan recommendation module generates multiple levels of recommended treatment plans, including four levels: high risk and high benefit, high risk and low benefit, low risk and high benefit, and low risk and low benefit.

Images