A method for predicting the cost of treatment of chronic diseases

Abstract

The application discloses a chronic disease treatment cost prediction method, comprising the following steps: setting a data collection rule, acquiring multi-source chronic disease medical data based on the data collection rule, obtaining a standardized medical data set based on the multi-source chronic disease medical data, obtaining an optimized feature set based on expert opinions and the standardized medical data set, optimizing the feature set based on recursive feature elimination and regularization processing to obtain an optimal feature subset, obtaining a final data set based on the optimal feature subset, training a plurality of candidate models based on the final data set to obtain an optimal model set, and obtaining a chronic disease treatment cost prediction value based on the optimal model set. The application can accurately predict the chronic disease treatment cost, provide strong support for medical insurance cost control and medical decision-making, and has important practical value and social benefits.

Description

Technical Field

[0001] This invention belongs to the field of statistical technology, and in particular relates to a method for predicting the cost of chronic disease treatment. Background Technology

[0002] Predicting the cost of chronic disease treatment is a complex technical problem involving multiple stages and factors. First, it requires acquiring a large amount of medical data from chronic disease patients, including medical history, treatment plans, medication records, and examination results. This data typically comes from hospital electronic medical record systems, but different hospitals' systems may exhibit heterogeneity and incompatibility, making data standardization and integration a major challenge. Second, after obtaining the data, it needs to be preprocessed and feature extracted to identify key factors highly correlated with treatment costs, such as disease severity, complication types, and treatment choices. This requires specialized medical knowledge and data analysis techniques. Third, building a cost prediction model requires comprehensively considering individual patient differences and selecting appropriate algorithms and models. Balancing model complexity and generalization ability to avoid overfitting and underfitting is also a technical challenge. Finally, after model building, it needs to be validated and optimized to evaluate its predictive performance and continuously iterate and update to adapt to changes in medical technology and cost standards. In summary, predicting the cost of chronic disease treatment involves multiple disciplines such as medicine, public health, statistics, and computer science, requiring overcoming a series of technical challenges in data collection, feature selection, and model building. It is a complex and challenging subject. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a method for predicting the treatment costs of chronic diseases, thereby resolving the issues present in the prior art.

[0004] To achieve the above objectives, the present invention provides a method for predicting the cost of chronic disease treatment, comprising:

[0005] Set data collection rules, acquire multi-source chronic disease medical data based on the data collection rules, and obtain a standardized medical dataset based on the multi-source chronic disease medical data;

[0006] An optimized feature set is obtained based on expert opinions and the standardized medical dataset; the optimal feature subset is obtained by recursively eliminating and regularizing the optimized feature set.

[0007] The final dataset is obtained based on the optimal feature subset;

[0008] The final dataset is used to train several candidate models to obtain the optimal model set;

[0009] Predicted costs for chronic disease treatment are obtained based on the optimal set of models.

[0010] Optionally, the process of setting data collection rules includes: designing the data interface for each data collection object and predefining data mapping rules; and obtaining data collection specifications and processes based on data collection attributes.

[0011] Optionally, the process of obtaining a standardized medical dataset includes:

[0012] Association rule mining algorithms are used to analyze multi-source chronic disease medical data to obtain the correlation between chronic diseases and treatment costs; based on the attributes of medical history records, treatment plans, medication use, and examination results in the multi-source chronic disease medical data, patient medical profiles are constructed; based on the attributes of medical data and electronic medical records in the multi-source chronic disease medical data, natural language processing technology is used to obtain structured supplementary data. Among these steps, the multi-source chronic disease medical data is cleaned before analyzing the correlation.

[0013] Optionally, the process of obtaining the optimal feature subset includes:

[0014] A recursive feature elimination algorithm is adopted. By recursively building the model, a preset number of features that do not meet the requirements are eliminated in each recursion to obtain a first feature subset. A regularization method is adopted, which introduces a regularization term into the loss function. By controlling the regularization parameter, feature coefficients that do not meet the preset conditions are eliminated to obtain a second feature subset. If the first feature subset and the second feature subset are consistent, they are determined to be the optimal feature subset. If they are inconsistent, the first feature subset and the second feature subset are merged, and the optimal feature subset is obtained through cross-validation.

[0015] Optionally, the process of obtaining the final dataset includes:

[0016] The initial dataset is obtained by acquiring historical treatment costs and feature datasets of chronic disease patients based on the optimal feature subset; the initial dataset is then cleaned; and the cleaned initial dataset is filtered based on business knowledge and data analysis results to obtain the final dataset.

[0017] Optionally, the process of obtaining the optimal model set includes:

[0018] The final dataset is divided into training, validation, and test sets in a 6:2:2 ratio. Several candidate models are trained using the training set. Hyperparameter optimization is performed on the trained candidate models based on the validation set and grid search. The test set is input into the candidate models after hyperparameter optimization to obtain prediction results. The performance of the candidate models is evaluated using a linear regression model as the benchmark model to determine whether the prediction error threshold is exceeded. If it is exceeded, the candidate models are further optimized.

[0019] Optionally, several candidate models can be constructed based on decision trees, support vector machines, and neural networks.

[0020] Optionally, it also includes: acquiring updated multi-source chronic disease medical data based on a preset time interval and the optimal feature subset as an incremental dataset; processing outliers and missing values ​​in the incremental dataset, converting the format, and inputting it into the optimal model; calculating the error between the model's predicted output and the actual cost through forward propagation; and, based on the calculated error, using a gradient descent algorithm to calculate the gradient direction of the model parameters and updating the model parameters according to a set learning rate until the model's prediction error on the incremental dataset converges to below a preset threshold or reaches the maximum number of iterations.

[0021] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for predicting the cost of chronic disease treatment.

[0022] Compared with the prior art, the present invention has the following advantages and technical effects:

[0023] This invention collects patient medical data from hospital electronic medical record systems, extracts key features through data standardization and feature engineering, and constructs a cost prediction model. Addressing the heterogeneous and dynamically changing nature of medical data, this invention designs a cross-hospital data interface and an incremental learning strategy to achieve generalized application and continuous optimization of the model. Through multidisciplinary integration, this invention continuously improves the prediction technology, forming a standardized and systematic solution. This method can accurately predict the costs of chronic disease treatment, providing strong support for medical insurance cost control and medical decision-making, and has significant practical value and social benefits. Attached Figure Description

[0024] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0025] Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0026] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0027] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0028] Example 1

[0029] like Figure 1 As shown, this embodiment provides a method for predicting the cost of chronic disease treatment, including:

[0030] Set data collection rules, acquire multi-source chronic disease medical data based on the data collection rules, and obtain a standardized medical dataset based on the multi-source chronic disease medical data;

[0031] Furthermore, the process of setting data collection rules includes: designing the data interface for each data collection object and predefining data mapping rules; and obtaining data collection specifications and processes based on data collection attributes.

[0032] Specifically, based on the electronic medical record systems of different hospitals, their database structures and field information are obtained. For each hospital's database, a corresponding data extraction interface is designed to retrieve raw medical data from the hospital's database. The retrieved raw medical data is then converted into a standardized data format using predefined data mapping rules, eliminating the heterogeneity of data formats across different hospitals and obtaining unified, standardized medical data.

[0033] For example, Hospital A's electronic medical record system might use a MySQL database, containing tables for basic patient information, diagnosis information, and medication records; while Hospital B might use an Oracle database, with different field naming and data storage methods. For each hospital's database, a corresponding data extraction interface is designed, such as developing an SQL query script for Hospital A and an API interface for Hospital B, ensuring efficient retrieval of raw medical data from the hospital databases. The retrieved raw medical data is then standardized using predefined data mapping rules. Assuming Hospital A uses ICD-9 encoding for diagnosis codes while Hospital B uses ICD-10 encoding, the mapping rules unify both to ICD-10 encoding, eliminating data format heterogeneity. For instance, Hospital A's code for "heart disease" is 410, which is converted to Hospital B's "I21" through the mapping rules, ensuring data consistency.

[0034] Specifically, based on the data collection attributes, establish data collection standards and procedures to ensure that the acquired medical data is complete, accurate, and timely. Standardize the data collection process, clearly defining the scope, frequency, source, and methods of data collection.

[0035] For example, outpatient and inpatient medical record data should be entered into the system within 24 hours of the patient's visit to ensure data timeliness; data entry templates should be established to standardize field filling and reduce data entry errors. Simultaneously, a data quality monitoring mechanism should be established to regularly check the completeness and accuracy of the data and promptly correct any problems found. Good data collection standards can guarantee data quality and are the foundation for building effective predictive models.

[0036] Furthermore, the process of obtaining a standardized medical dataset includes:

[0037] Association rule mining algorithms are used to analyze multi-source chronic disease medical data to obtain the correlation between chronic diseases and treatment costs; based on the attributes of medical history records, treatment plans, medication use, and examination results in the multi-source chronic disease medical data, patient medical profiles are constructed; based on the attributes of medical data and electronic medical records in the multi-source chronic disease medical data, natural language processing technology is used to obtain structured supplementary data. Among these steps, the multi-source chronic disease medical data is cleaned before analyzing the correlation.

[0038] Specifically, based on the attributes of chronic diseases and treatment costs, data mining algorithms are used to extract cost-related features from standardized data. Patient medical profiles are constructed based on attributes such as medical history records, treatment plans, medication records, and examination results, depicting changes in patient status at different treatment stages. Natural language processing techniques are used to extract structured information and enrich data dimensions based on the attributes of medical data and electronic medical records. Data cleaning and transformation processes are designed based on heterogeneous data and data integration attributes to achieve a unified representation of data from different sources. Data collection standards and processes are formulated based on data collection attributes to ensure the completeness, accuracy, and timeliness of the acquired medical data.

[0039] For example, data mining algorithms can be used to extract cost-related features from chronic disease and treatment cost attributes. Association rule mining algorithms, such as the Apriori algorithm, can be used to analyze the correlation between different chronic diseases and treatment costs. For instance, the data shows that cardiovascular drug expenditures for hypertensive patients are significantly higher than average, insulin and blood glucose monitoring costs account for a large proportion for diabetic patients, and treatment costs for diabetic patients with kidney complications increase significantly, indicating that complications are an important factor leading to increased costs. This method can identify key chronic diseases and their complications that have a significant impact on costs, providing important indicators for cost prediction. This mining can identify which chronic diseases are the main factors leading to high costs and how they influence each other. Based on medical history records, treatment plans, medication records, and examination results, patient medical profiles can be constructed to depict changes in patient status at different treatment stages. Taking diabetes as an example, in the early stages, patients may only need oral hypoglycemic drugs to control blood sugar. At this time, the patient profile mainly shows slightly elevated blood sugar, slightly higher than normal glycated hemoglobin levels, and relatively low treatment costs. As the disease progresses, patients may require insulin injections, and their profile will be updated to indicate insulin dependence. Blood sugar levels will fluctuate significantly, and abnormal test results may be associated with complications such as retinopathy, diabetes, and kidney disease. Treatment costs will also increase significantly due to the increase in the type and dosage of medications. At this stage, the profile will record insulin dosage, frequency of blood sugar fluctuations, and test results for complications. Later, if serious complications such as diabetic foot occur, the patient profile will include records of wound management and anti-infection treatment, further increasing costs. This approach allows for a comprehensive understanding of the patient's condition and cost changes. Based on medical data and electronic medical record attributes, natural language processing technology is used to extract structured information and enrich the data dimensions. For example, named entity recognition technology can be used to extract key information such as disease names, medication names, and surgical procedures from the medical record text. For example, from the text "The patient has a history of coronary heart disease, has been taking aspirin for a long time, and was recently admitted to the hospital due to worsening chest tightness. The electrocardiogram showed ST segment depression, and coronary angiography was performed," entities such as "coronary heart disease," "aspirin," "chest tightness," "electrocardiogram," "ST segment depression," and "coronary angiography" can be extracted. Further contextual analysis reveals that "coronary heart disease" represents past medical history, "aspirin" represents medication information, and "chest tightness" and "ST segment depression" represent symptoms and examination results. This transforms unstructured text information into structured data, enriching the patient information dimension and helping to improve the model's predictive accuracy. Through natural language processing, key information can be mined from massive amounts of medical record text, providing the model with more features.

[0040] An optimized feature set is obtained based on expert opinions and the standardized medical dataset; the optimal feature subset is obtained by recursively eliminating and regularizing the optimized feature set.

[0041] Furthermore, the process of obtaining the optimal feature subset includes:

[0042] A recursive feature elimination algorithm is adopted. By recursively building the model, a preset number of features that do not meet the requirements are eliminated in each recursion to obtain a first feature subset. A regularization method is adopted, which introduces a regularization term into the loss function. By controlling the regularization parameter, feature coefficients that do not meet the preset conditions are eliminated to obtain a second feature subset. If the first feature subset and the second feature subset are consistent, they are determined to be the optimal feature subset. If they are inconsistent, the first feature subset and the second feature subset are merged, and the optimal feature subset is obtained through cross-validation.

[0043] For example, data on a diabetic patient is extracted from an electronic medical record system, including blood glucose levels, medication types, frequency of visits, and complication status. This data constitutes the initial feature set, providing a foundation for subsequent feature selection and model building. Based on professional medical knowledge, highly relevant key features such as disease severity, complication types, and treatment choices are extracted from the initial features to form an optimized feature set. Taking a diabetic patient as an example, disease severity can be determined by glycated hemoglobin (HbA1c) levels, complication types include retinopathy, diabetes, and nephropathy, and treatment choices include oral medications and insulin injections. These key features can more directly reflect the patient's condition and treatment costs. A recursive feature elimination algorithm is used to recursively build the model, eliminating several of the least important features in each recursive step, ultimately obtaining the optimal feature subset affecting cost prediction. Assuming the initial feature set contains 20 features, the recursive feature elimination algorithm eliminates 5 of the least important features each time. After several iterations, it is possible to ultimately retain the 5 most important features: blood glucose levels, HbA1c, diabetes, and nephropathy complications. Regularization methods are employed, introducing a regularization term into the loss function. By controlling the regularization parameter, some feature coefficients are eliminated, thereby achieving feature selection and obtaining the optimal feature subset. For example, using Lasso regression (L1 regularization), by adjusting the regularization parameter λ, the coefficients of some features become 0, ultimately retaining features such as blood glucose level, medication type, diabetes, and kidney complications. The optimal feature subsets obtained by recursive feature elimination and regularization methods are then merged. If the feature subsets selected by the two methods are consistent, they are determined as the final feature subset; otherwise, the two subsets are merged, and cross-validation is used to determine the optimal feature subset. For instance, assuming the feature subset selected by recursive feature elimination is {blood glucose level, HbA1c, kidney complications}, and the feature subset selected by regularization is {blood glucose level, medication type, kidney complications}, the two subsets are merged into {blood glucose level, HbA1c, medication type, kidney complications}, and cross-validation is used to determine the final optimal feature subset.

[0044] The final dataset is obtained based on the optimal feature subset;

[0045] Furthermore, the process of obtaining the final dataset includes:

[0046] The initial dataset is obtained by acquiring historical treatment costs and feature datasets of chronic disease patients based on the optimal feature subset; the initial dataset is then cleaned; and the cleaned initial dataset is filtered based on business knowledge and data analysis results to obtain the final dataset.

[0047] Specifically, historical treatment cost data and related feature data of patients with chronic diseases are obtained. The data is preprocessed, including missing value handling, outlier handling, and data standardization, to obtain a cleaned dataset. Based on business knowledge and data analysis results, features that meet the requirements for correlation with treatment costs are selected from the dataset. Feature selection algorithms such as recursive feature elimination are used to further optimize the feature subset and obtain the final dataset.

[0048] For example, predicting treatment costs for diabetic patients first requires collecting characteristic data such as the patient's age, gender, disease duration, blood glucose level, complications (e.g., retinopathy, nephropathy), medication use (e.g., insulin dosage), and frequency of visits, as well as corresponding historical treatment cost data. Data preprocessing is a crucial step in ensuring model accuracy. Missing value handling can be achieved through imputation methods (e.g., mean imputation, median imputation) or by deleting records with a large number of missing values. Outlier handling can be achieved through box plot analysis to identify and remove outliers. Data standardization normalizes features of different dimensions, such as using Z-score standardization to convert feature values ​​into a standard normal distribution with a mean of 0 and a standard deviation of 1. Based on business knowledge and data analysis results, features with high correlation to treatment costs are selected. For example, correlation analysis reveals that diabetic patients' blood glucose levels, number of complications, and treatment methods (e.g., whether an insulin pump is used) are highly correlated with treatment costs; these features are then used to construct a feature subset. Furthermore, the Recursive Feature Elimination (RFE) method is adopted. By recursively constructing a decision tree model, several least important features are eliminated each time to determine the data contained in the final dataset, such as blood glucose levels, number of complications, insulin dosage, etc.

[0049] The final dataset is used to train several candidate models to obtain the optimal model set;

[0050] Furthermore, several candidate models are constructed based on decision trees, support vector machines, and neural networks.

[0051] Furthermore, the process of obtaining the optimal model set includes:

[0052] The final dataset is divided into training, validation, and test sets in a 6:2:2 ratio. Several candidate models are trained using the training set. Hyperparameter optimization is performed on the trained candidate models based on the validation set and grid search. The test set is input into the candidate models after hyperparameter optimization to obtain prediction results. The performance of the candidate models is evaluated using a linear regression model as the benchmark model to determine whether the prediction error threshold is exceeded. If it is exceeded, the candidate models are further optimized.

[0053] Specifically, predictive models are constructed using machine learning algorithms such as decision trees, support vector machines, and neural networks. Multiple candidate models are trained by setting different model hyperparameters, such as the maximum depth of the decision tree, the kernel function type of the support vector machine, and the number of hidden layers and neurons in the neural network. Hyperparameter optimization methods, such as grid search and random search, are used to fine-tune the candidate models on the validation set. The performance of the models under different hyperparameter combinations is evaluated through cross-validation, and the hyperparameter combination with the best generalization ability is selected as the optimal model. The predictive performance of the optimal model is evaluated on the test set using metrics such as mean squared error and mean absolute error to assess the model's predictive accuracy and robustness. The model is then compared with a benchmark model to verify its effectiveness.

[0054] For example, the dataset is divided into training, validation, and test sets, with a common ratio of 6:2:2. The training set is used for model training, the validation set for model tuning, and the test set for evaluating model performance. For instance, out of 1000 data points, 600 are used for training, 200 for validation, and 200 for testing. For the optimal feature set, predictive models are constructed using algorithms such as decision trees, support vector machines (SVM), and neural networks. Decision tree models explore the optimal structure by setting different maximum depths (e.g., 3, 5, 7); SVM models optimize classification performance by selecting different kernel functions (e.g., linear kernels, radial basis functions); and neural network models find the optimal network structure by adjusting the number of hidden layers and neurons (e.g., 50 neurons in one layer, 30 neurons in two layers). Hyperparameter optimization methods such as grid search and random search are used to tune the candidate models on the validation set. For example, grid search is used to try combinations of the maximum depth and minimum number of sample splits in the decision tree, and cross-validation is used to evaluate the model performance under different combinations, selecting the hyperparameter combination with the smallest mean squared error. Random search randomly samples from the hyperparameter space and finds the optimal solution through multiple iterations. The predictive performance of the optimal model is evaluated on a test set using metrics such as mean squared error (MSE) and mean absolute error (MAE). For example, a decision tree model with an MSE of 5000 and an MAE of 600 on the test set indicates low prediction error and good robustness. Comparison with a baseline model (such as simple linear regression) is also performed; if the new model's MSE and MAE are significantly lower than the baseline model, its effectiveness is validated. The trained optimal model is then deployed to a production environment, and by inputting the feature data of new patients, the model automatically predicts their treatment costs over a certain period.

[0055] Predicted costs for chronic disease treatment are obtained based on the optimal set of models.

[0056] Furthermore, the cost prediction method also includes: acquiring updated multi-source chronic disease medical data based on a preset time interval and an optimal feature subset as an incremental dataset; processing outliers and missing values ​​in the incremental dataset, converting the format, and inputting it into the optimal model; calculating the error between the model's predicted output and the actual cost through forward propagation; and, based on the calculated error, using a gradient descent algorithm to calculate the gradient direction of the model parameters and updating the model parameters according to a set learning rate until the model's prediction error on the incremental dataset converges to below a preset threshold or reaches the maximum number of iterations.

[0057] Specifically, in order to adapt to the dynamic changes in medical technology and cost standards, an incremental learning strategy is adopted to regularly acquire new patient medical data, update and iteratively optimize the cost prediction model online, and fine-tune the model parameters through algorithms such as gradient descent, so that it can continuously learn new cost distribution and change patterns, and maintain the timeliness and accuracy of prediction.

[0058] According to pre-set time intervals, the latest patient medical data and cost data are periodically retrieved from the medical system database and used as incremental datasets. The incremental datasets are preprocessed to clean outliers and missing values, extract feature variables relevant to cost prediction, and construct data conforming to the model input format. The preprocessed incremental data is input into a trained cost prediction model, and the error between the model's predicted output and the actual cost is calculated through forward propagation. Based on the calculated error, the gradient descent algorithm is used to calculate the gradient direction of the model parameters, and the model's weights and biases are updated according to a set learning rate. Multiple rounds of iterative optimization are performed until the model's prediction error on the incremental dataset converges below a preset threshold or reaches the maximum number of iterations. The optimized model parameters are saved, replacing the original model, making it adaptable to the latest medical technologies and cost standards, improving the timeliness of cost prediction. The prediction accuracy of the updated model on new patient data is periodically evaluated; if it is lower than expected, incremental data is retrieved again, triggering the model's online learning and optimization process again.

[0059] For example, in the continuous optimization of medical cost prediction models, regularly retrieving the latest patient medical data and cost data from the medical system database is crucial. Assuming a monthly time interval, at the beginning of each month, newly added patient data and cost records from the previous month are extracted from the database to form an incremental dataset. For instance, in a certain month, medical data for 300 new patients is added, including features such as age, gender, medical history, treatment method, medication, and corresponding actual treatment costs. Preprocessing the incremental dataset is a key step in ensuring data quality. First, outliers are cleaned; for example, if a patient's treatment cost is abnormally high, it may be due to data entry errors and needs to be removed or corrected. Second, missing values ​​are handled; for example, if a patient's medication record is missing, interpolation or using the mean of similar patients can be used to fill in the missing values. Finally, feature variables related to cost prediction, such as patient age, disease duration, and treatment method, are extracted to construct data conforming to the model input format. The preprocessed incremental data is then input into the trained cost prediction model, and the error between the model's predicted output and the actual cost is calculated through forward propagation. Suppose a patient's actual treatment cost is 6000 yuan, the model predicts 5800 yuan, and the error is 200 yuan. Based on the calculated error, a gradient descent algorithm is used to calculate the gradient direction of the model parameters, and the model's weights and biases are updated according to a set learning rate (e.g., 0.01). In this way, the model can gradually adjust to better fit the new data distribution. Repeat the above steps for multiple rounds of iterative optimization until the model's prediction error on the incremental dataset converges below a preset threshold or reaches the maximum number of iterations (e.g., 100). For example, after 50 iterations, the mean squared error of the model on the incremental dataset decreases from the initial 500 to below 100, indicating that the model has adapted well to the new data. The optimized model parameters are saved, replacing the original model, so that it can adapt to the latest medical technologies and cost standards, improving the timeliness of cost prediction. For example, with the use of new drugs and the promotion of new treatment technologies, the original model may not be able to accurately predict new cost trends. By regularly updating the model parameters, the accuracy of the prediction results can be ensured. Regularly evaluating the prediction accuracy of the updated model on new patient data is a crucial step in ensuring model performance. Suppose an evaluation is conducted quarterly, and it's found that the model's predictive accuracy on the latest data has dropped from 90% to 85%, below expectations. In this case, incremental data needs to be acquired again, triggering the model's online learning and optimization process once more. In this way, the model can continuously learn and adapt to changes in data, maintaining high predictive accuracy. This mechanism of regular updates and optimization not only improves the model's predictive performance but also allows for the timely identification and resolution of problems in practical applications. For example, if the model's cost prediction error for a certain type of disease is large for a period of time, analysis of incremental data might reveal changes in the treatment methods for that disease, leading to alterations in the cost structure.Timely model updates enable rapid adjustments to prediction strategies, improving accuracy. Furthermore, regular model updates help address changes in data distribution. Over time, patient age structures and disease patterns may shift, rendering existing models inapplicable. Regularly introducing new data allows the model to learn and adapt to these changes, maintaining high generalization ability. In practice, various techniques can be combined to further enhance model performance. For example, ensemble learning combines predictions from multiple models to improve robustness; deep learning models uncover deeper features in the data, enhancing nonlinear fitting capabilities. This combined approach allows for the construction of an efficient and accurate healthcare cost prediction model, providing strong support for resource planning and cost control in healthcare institutions. In conclusion, a series of steps—including regular acquisition of incremental data, preprocessing, model updates, and evaluation—ensure the continuous optimization and efficient operation of healthcare cost prediction models, providing a scientific basis for healthcare management decisions and improving the quality and efficiency of healthcare services.

[0060] Based on interdisciplinary integration, and comprehensively considering medical knowledge and statistical methods, the prediction model is further optimized to improve its interpretability and generalization ability, forming a standardized scheme for predicting chronic disease treatment costs. For example, by incorporating the opinions of medical chronic disease management experts, the weights of certain features in the model are adjusted to make the prediction results more consistent with clinical reality. Simultaneously, the bootstrap sampling method is used to enhance the model's robustness under different data distributions. Integrating the standardized scheme into the medical information system enables real-time prediction and monitoring of chronic disease treatment costs, and allows for dynamic adjustments to medical insurance policies and medical resource allocation based on the prediction results, achieving the goal of medical insurance cost control. For example, a municipal medical insurance bureau integrated a diabetes treatment cost prediction model into its medical insurance information system to monitor patients' treatment costs in real time. When abnormal cost increases are detected, timely intervention is made to adjust medication plans or hospital stays, effectively controlling expenditures. Interdisciplinary integration further enhances the model's practicality and generalization ability. After integration into the medical information system, the real-time prediction and monitoring functions can promptly detect cost anomalies, helping hospitals and medical insurance departments to accurately control costs. Continuous iterative optimization is a crucial guarantee for ensuring the long-term effectiveness of the model. By comprehensively applying these steps, the chronic disease treatment cost prediction model can not only provide accurate prediction results, but also provide strong support for the rational allocation of medical resources and the scientific formulation of medical insurance policies, ultimately promoting the improvement of the chronic disease prevention and control system and the enhancement of medical service levels.

[0061] This embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method for predicting the cost of chronic disease treatment.

[0062] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting the cost of chronic disease treatment, characterized in that, Includes the following steps: Set data collection rules, and acquire multi-source chronic disease medical data based on the data collection rules; A standardized medical dataset is obtained based on the aforementioned multi-source chronic disease medical data; The process of obtaining a standardized medical dataset includes: This study employs association rule mining algorithms to analyze multi-source chronic disease medical data and obtain the correlation between chronic diseases and treatment costs. Based on the attributes of medical history records, treatment plans, medication use, and examination results in the multi-source chronic disease medical data, a patient medical profile is constructed to grasp the overall trend of patient condition development and cost changes. Based on the attributes of medical data and electronic medical records in the multi-source chronic disease medical data, natural language processing (NLP) technology is used to obtain structured supplementary data. NLP enables the extraction of key information from massive amounts of medical record text. Prior to analyzing the correlation relationships, the multi-source chronic disease medical data undergoes data cleaning. An optimized feature set is obtained based on expert opinions and the standardized medical dataset; the optimal feature subset is obtained by recursively eliminating and regularizing the optimized feature set. The process of obtaining the optimal feature subset includes: A recursive feature elimination algorithm is employed. The model is constructed recursively, and a preset number of features that do not meet the requirements are eliminated in each iteration to obtain a first feature subset. A regularization method is then used, introducing a regularization term into the loss function. By controlling the regularization parameter, feature coefficients that do not meet preset conditions are eliminated to obtain a second feature subset. If the first feature subset and the second feature subset are identical, they are determined to be the optimal feature subset. If they are not identical, the first feature subset and the second feature subset are merged, and the optimal feature subset is obtained through cross-validation. The final dataset is obtained based on the optimal feature subset, specifically including: The initial dataset is obtained by acquiring historical treatment costs and feature datasets of chronic disease patients based on the optimal feature subset; the initial dataset is then cleaned; the cleaned initial dataset is then filtered according to business knowledge and data analysis results, and feature selection algorithms are used to select features that meet the requirements of correlation with treatment costs, thus obtaining the final dataset. The final dataset is used to train several candidate models to obtain the optimal model set; The process of obtaining the optimal model set includes: The final dataset is divided into training, validation, and test sets in a 6:2:2 ratio. Several candidate models are trained using the training set. Hyperparameter optimization is performed on the trained candidate models based on the validation set and grid search. The test set is input into the hyperparameter-optimized candidate models to obtain prediction results. The performance of the candidate models is evaluated using a linear regression model as the benchmark model to determine whether the prediction error threshold is exceeded. If it is exceeded, the candidate models are further optimized. Predicted costs of chronic disease treatment are obtained based on the optimal set of models. It also includes: acquiring updated multi-source chronic disease medical data based on a preset time interval and the optimal feature subset as an incremental dataset; processing outliers and missing values ​​in the incremental dataset, converting the format, and inputting it into the optimal model; calculating the error between the model's predicted output and the actual cost through forward propagation; and, based on the calculated error, using a gradient descent algorithm to calculate the gradient direction of the model parameters and updating the model parameters according to a set learning rate until the model's prediction error on the incremental dataset converges to below a preset threshold or reaches the maximum number of iterations.

2. The method for predicting chronic disease treatment costs according to claim 1, characterized in that, The process of setting data collection rules includes: designing the data interface for each data collection object and predefining data mapping rules; and obtaining data collection specifications and processes based on data collection attributes.

3. The method for predicting chronic disease treatment costs according to claim 1, characterized in that, Several candidate models were constructed based on decision trees, support vector machines, and neural networks.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of a method for predicting the cost of chronic disease treatment as described in any one of claims 1-3.

Images