Medical service cost estimation method based on machine learning algorithm

Abstract

The invention relates to the technical field of computer service, and discloses a medical service cost estimation method based on a machine learning algorithm, comprising the following steps: S1, data collection and cleaning; s2, performing medical service cost influence factor analysis by adopting a machine learning algorithm, and constructing an influence factor model; s3, performing deep analysis on the medical service cost composition and the influence factors of the influence factor model, performing importance sorting on the factors influencing the medical service cost, and establishing a medical service cost prediction index system; s4, constructing a medical service cost prediction model based on a machine learning algorithm; and S5, referring to the medical standard operation process of each service item, converting the medical standard operation process into data, putting the data into the medical service cost prediction model in the step S4, and predicting the price of the corresponding service item. According to the invention, comprehensive and large-scale objective data and original cost data of key factors obtained through expert interviews are collected, and whether offers of different hospitals are scientific and reasonable can be checked in a large range.

Description

Technical Field

[0001] The present invention relates to the field of computer services, and in particular to a method for estimating medical service costs based on a machine learning algorithm. Background Art

[0002] The existing medical service cost estimation model is a pricing method based on cost-plus. This estimation model is to give appropriate markups on the basis of the measured costs to form the price of medical services. The problems with the existing cost estimation of medical services are mainly manifested in the inability to accurately reflect the labor value of doctors, the unreasonable growth and structure of medical expenses, and the slow and distorted adjustment of medical service prices. The traditional cost-plus pricing method has been seriously lagging behind and cannot adapt to the high-quality development of medical institutions. Corresponding reforms are urgently needed.

[0004] In order to improve the current pricing mechanism, some researchers have proposed corresponding pricing strategies from different perspectives. The price of medical services should reflect the value of complex technology, such as service technology, labor intensity and service risk. The finance and pricing departments should jointly formulate plans to guide medical prices. According to market conditions, the competent departments of medical and health institutions should allow medical institutions to have a certain floating price. According to the principle of combining unified leadership with hierarchical management, the management authority should be gradually decentralized so that health authorities and medical institutions can participate more in pricing research and decision-making. By increasing the service price gap between various technical positions or medical staff at different levels of hospitals, the distribution of patients can be effectively guided and the technical and allocation efficiency of health resources can be improved.

[0005] In summary, the technical defects of existing medical service cost estimation methods are as follows:

[0006] (1) The existing cost estimation method is simple and crude. Due to the particularity of the medical service field, there is a serious information asymmetry problem, which makes it difficult to accurately measure the cost of medical services. Therefore, the pricing method based on cost plus is relatively simple and crude, which deviates from the actual cost of medical service items. Without the support of big data, the value of each doctor's medical service is difficult to estimate and cannot well reflect the actual situation of medical services.

[0007] (2) Existing cost estimation methods do not classify the various attributes of each doctor's different operating procedures, and are unable to obtain potential internal correlation items in the cost data, resulting in the inability to correlate the various attribute data of the statistics. In the cost calculation, contradictory errors often occur and cannot be verified.

[0008] (3) Due to the inability to conduct in-depth analysis of the cost structure and influencing factors of medical services, it has been impossible to establish a comprehensive and systematic medical service cost prediction indicator system that can operate normally and sustainably for a long time. Summary of the invention

[0009] In order to overcome or alleviate one or more of the above technical problems, the purpose of the present invention is to provide a medical service cost estimation method based on a machine learning algorithm. The method gradually constructs a machine learning pricing model based on large-scale historical data. Through the machine learning pricing model, the cost of medical services can be more accurately estimated.

[0010] The present invention provides the following technical solutions:

[0011] A method for estimating medical service costs based on a machine learning algorithm comprises the following steps:

[0012] S1: Data collection and cleaning, which specifically includes:

[0013] S11: Use expert interviews to obtain key factors affecting medical costs;

[0014] S12: Obtain the main factors affecting medical costs through literature;

[0015] S13: Objective data set collection and cleaning, construction of original cost data set;

[0016] S2: Based on the original cost data set, a machine learning algorithm is used to analyze the factors affecting the cost of medical services and to construct an influencing factor model;

[0017] S3: Conduct an in-depth analysis of the medical service cost structure and influencing factors of the influencing factor model, rank the factors influencing the medical service cost in terms of importance, and establish a medical service cost prediction indicator system based on this ranking and combined with expert consultation;

[0018] S4: Construct a medical service cost prediction model based on machine learning algorithm;

[0019] S5: Refer to the medical standard operating procedures of each service item, convert them into data and put them into the medical service cost prediction model of step S4 to predict the price of the corresponding service item.

[0020] According to some implementation methods, step S2 includes the following steps:

[0021] S21. Algorithm selection and evaluation;

[0022] Algorithm selection: selecting a suitable machine learning algorithm based on the characteristics of the medical service cost data of the original cost dataset and the analysis objectives; algorithm evaluation: after the algorithm is selected, the algorithm performance is evaluated using cross-validation;

[0023] S22. Feature engineering;

[0024] Feature selection: extract features related to cost prediction from raw medical data, and select features that have a significant impact on cost prediction through correlation analysis and feature importance assessment; feature processing: perform necessary processing on selected features, such as missing value filling, outlier processing, data normalization or standardization, to ensure the stability and accuracy of model training;

[0025] Feature dimensionality reduction, which is performed through principal component analysis or linear discriminant analysis to reduce the number of features while retaining key information;

[0026] S23. Model training;

[0027] Divide the data set, divide the processed data set into training set, validation set and test set. The training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the final performance of the model; Model training, use the training set data to train the selected machine learning algorithm, and adjust the model parameters through the iterative optimization algorithm to minimize the prediction error; Hyperparameter tuning, use the validation set to tune the model's hyperparameters to find the best model configuration;

[0028] S24. Model evaluation and selection;

[0029] Performance evaluation: Use the test set to evaluate the performance of the trained model, calculate the evaluation indicators of accuracy, recall rate, and F1 score, and draw ROC curves and PR curves to intuitively display the performance of the model;

[0030] Model selection: select the model with the best performance as the final service cost prediction model based on the evaluation results. If multiple models have similar performance, make a comprehensive selection based on model complexity, training time, and interpretability.

[0031] S25. Model optimization and construction;

[0032] Model optimization: further optimize the shortcomings of the model, such as introducing more features, adjusting algorithm parameters, and using ensemble learning methods to improve model performance.

[0033] Model construction: deploy the optimized model into the actual production environment, connect it with the business system of the medical institution, and build an influencing factor model.

[0034] According to some implementation methods, the machine learning algorithm in step S21 is selected from the XGBoost algorithm.

[0035] According to some implementation methods, the algorithm evaluation indicators of step S21 include accuracy, recall, F1 score and AUC value.

[0036] According to some implementation methods, step S3 includes the following steps:

[0037] S31: In-depth analysis of the medical service cost structure and influencing factors based on the influencing factor model of step S2, first identifying the most significant influencing factors;

[0038] S32: Based on the feature importance scores or bar charts, analyze which factors are most important for the target variable medical cost;

[0039] S33: Sort all influencing factors and, combined with expert consultation, build a cost prediction indicator system for medical services.

[0040] According to some implementation methods, step S4 includes the following steps:

[0041] S41: Select the XGboost machine learning algorithm, input the test set data into the model for training, and repeatedly adjust the parameters and train until the error reaches the index range of regression prediction;

[0042] S42: Entering the testing phase of the medical service cost prediction model, after testing the model, fitting the project cost prediction results, observing whether the cost prediction model obtained by training and adjusting parameters can better fit the true value of the actual data, and then comparing the fitting degree of the training sample and the prediction test sample, and testing the reliability of the cost prediction model through the fitting effect of the cost prediction;

[0043] S43: According to the influencing factor variables that have an important impact on medical costs screened in the medical service cost prediction index system of step S3, the indicators of step S3 are put into the XGboost model in step S2 to construct a medical service cost prediction model, where the X variable of the medical service cost prediction model is the influencing factor variable in the medical service cost prediction index system of step S3, and Y is the medical service cost.

[0044] According to some implementation methods, the objective data set of step S13 includes medical-related data from medical institutions, relevant departments, and patients, specifically including patient basic information, diagnosis information, treatment plan, cost details, and doctor's title, education, years of work experience, technical service content, materials used, relevant auxiliary personnel, and operation time data.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The medical service cost estimation method based on machine learning algorithm provided by the present invention collects more comprehensive and large-scale objective data and the original cost data of key factors obtained through expert interviews, which can verify whether the quotations of different hospitals are scientific and reasonable in a large range. Through a large amount of real data, machine learning is used to identify the key factors affecting medical prices, and then the relevant prediction model is constructed, combined with the medical standard operating procedures, to determine the cost price of medical services.

[0047] This medical service cost estimation method can, to a certain extent, more realistically reflect the resource consumption of medical projects, form a medical service project cost management decision-making model, and provide decision-making reference for government management departments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flowchart of a method for estimating medical service costs based on a machine learning algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention is described in detail below in conjunction with the embodiments and drawings, but it should be understood that the embodiments and drawings are only used to exemplify the present invention and do not constitute any limitation on the protection scope of the present invention. All reasonable changes and combinations within the scope of the inventive concept of the present invention fall within the protection scope of the present invention.

[0050] The present invention will be further described below in conjunction with the accompanying drawings.

[0051] Example 1

[0052] like Figure 1 This embodiment takes a hospital in a certain city as the research object and adopts a medical service cost estimation method based on a machine learning algorithm, including the following steps:

[0053] S1: Data collection and cleaning;

[0054] This step collects relevant objective data information, including hospital settlement data, treatment data, and payment data from the Medical Insurance Bureau, and collects expert information through expert interviews;

[0055] S11: Use expert interviews to obtain key factors affecting medical costs.

[0056] Traditional cost analysis methods often rely on historical data and statistical models, which makes it difficult to fully capture the impact of dynamic factors such as market changes and technological innovations on cost structures. Therefore, the introduction of expert interviews has become an effective way to obtain key factors of cost composition by directly obtaining the experience and knowledge of senior professionals such as medical industry experts, financial analysts, and hospital cost accountants. The expert interview method can obtain the main factors of cost composition, and through the design of a scientific and reasonable interview outline, the selection of representative interviewees, the implementation of an effective interview process, and systematic data analysis, it can comprehensively and deeply reveal the inherent laws of cost composition.

[0057] Interview outline design: Based on the needs and objectives of cost management, design an interview outline that includes key issues such as cost classification, cost drivers, and cost control points.

[0058] Selection of interviewees: Select industry experts, financial analysts, cost accountants, etc. with rich experience and expertise as interviewees. Ensure that the interviewees are representative and diverse, and can cover all aspects of cost composition.

[0059] Interview implementation: Conduct interviews according to the interview outline, using various methods such as face-to-face interviews, telephone interviews, video conferences, etc. to ensure the smooth progress of the interview process. Record key information and opinions during the interview process, including experts' direct answers, suggestions, case analysis, etc.

[0060] Interview data analysis: organize, classify and code the interview records, and use the structural equation model to analyze the main factors of cost composition. Conduct in-depth analysis and interpretation of the main factors of cost composition, and put forward targeted cost control suggestions.

[0061] S12: Obtain the main factors affecting medical costs through literature;

[0062] S13: Objective dataset collection and cleaning;

[0063] Responsible for collecting medical-related data from medical institutions, relevant departments, patients, etc., including but not limited to basic patient information, diagnosis information, treatment plans, cost details, etc. For example: doctor's information (including doctor's title, education, years of work, etc.), doctor's technical service content (including operation process, main path, method, steps, etc.), materials used (equipment, materials, consumables, etc.), related auxiliary personnel (in order to participate in the medical activity, the various departments of the hospital need to participate in the cooperation of personnel), operation time (time spent in each operation link), etc., to build the original cost data set;

[0064] The collected data is cleaned, deduplicated, and standardized to obtain the original cost data set, ensuring the quality and consistency of the data and providing a reliable data basis for subsequent analysis. Feature variables related to medical service costs, such as patient age, gender, disease type, treatment method, and drug use, are extracted from the preprocessed data, and feature selection and dimensionality reduction are performed to improve the prediction accuracy and efficiency of the model.

[0065] S2: Analysis of influencing factors of medical service costs based on machine learning XGboost algorithm and construction of influencing factor model;

[0066] Each doctor's operation will have different attributes, and the process will be quite different. Based on the original cost data set constructed in step S1, the machine learning XGboost algorithm is used to mine each medical service item and analyze the factors affecting the cost, and then the data is analyzed in combination with accounting subjects. The potential internal associations in the original cost data are mined through machine learning. Combined with the medical service management model and the characteristics of medical service items, the basic information of the original cost data contains multiple factors and factor associations that may affect the cost. According to the importance of the output features, it is explained which factors have a significant impact on the results.

[0067] The main factors of medical cost composition obtained from expert interviews are taken as X, and the medical service cost submitted by the hospital is taken as Y, and a machine learning model is constructed.

[0068] S21. Algorithm selection and evaluation;

[0069] Algorithm selection, select the appropriate machine learning algorithm according to the characteristics and analysis objectives of the medical service cost data of the original cost data set. Common algorithms include random forest, gradient boosting tree (such as XGBoost, LightGBM), neural network (such as deep neural network DNN, recurrent neural network RNN), etc. Each algorithm has its advantages and disadvantages, which need to be considered comprehensively based on factors such as data volume and complexity, prediction accuracy requirements, and model interpretability requirements. In this embodiment, XGBoost is selected as the final algorithm.

[0070] Algorithm evaluation, after selecting the algorithm, use cross-validation (such as K-fold cross-validation) and other techniques to evaluate the algorithm performance. Evaluation indicators include accuracy, recall, F1 score, AUC value, etc., to ensure the effectiveness and stability of the selected algorithm in the task of medical service cost prediction.

[0071] S22. Feature engineering;

[0072] Feature selection: extract features related to cost prediction from the original medical data, such as basic patient information (age, gender), disease type, treatment plan, drug use, hospital stay, etc. Features that have a significant impact on cost prediction are screened out through correlation analysis, feature importance assessment and other methods.

[0073] Feature processing: perform necessary processing on the selected features, such as filling missing values ​​(using mean, median, mode or interpolation), outlier processing (deletion or replacement), data normalization or standardization, etc., to ensure the stability and accuracy of model training.

[0074] Feature dimensionality reduction: If the number of features is too large, it may lead to model overfitting or increased computational complexity. Feature dimensionality reduction is performed through methods such as principal component analysis (PCA) and linear discriminant analysis (LDA) to reduce the number of features while retaining key information.

[0075] S23. Model training;

[0076] Divide the data set into training set, validation set and test set. The training set is used to train the model, the validation set is used to adjust the model parameters (such as hyperparameter tuning), and the test set is used to evaluate the final performance of the model.

[0077] Model training uses the training set data to train the selected machine learning algorithm, and continuously adjusts the model parameters through iterative optimization algorithms (such as gradient descent) to minimize the prediction error.

[0078] Hyperparameter tuning: Use the validation set to tune the model's hyperparameters (such as the number of trees, learning rate, number of leaf nodes, etc.) to find the best model configuration.

[0079] S24. Model evaluation and selection;

[0080] Performance evaluation: Use the test set to evaluate the performance of the trained model and calculate evaluation indicators such as accuracy, recall, and F1 score. At the same time, you can draw charts such as ROC curves and PR curves to intuitively display the performance of the model.

[0081] Model selection: select the model with the best performance as the final service cost prediction model based on the evaluation results. If multiple models have similar performance, you can consider factors such as model complexity, training time, and interpretability to make a comprehensive selection.

[0082] S25. Model optimization and construction;

[0083] Model optimization: further optimize the shortcomings of the model, such as introducing more features, adjusting algorithm parameters, and using ensemble learning to improve model performance.

[0084] Model construction: deploy the optimized model into the actual production environment, connect it with the business system of the medical institution, and build an influencing factor model.

[0085] S3: Establishment of a medical service cost prediction index system: In-depth analysis of the medical service cost structure and influencing factors of the influencing factor model in step S2, ranking the factors affecting the medical service cost in terms of importance, and based on this ranking, constructing a medical service cost prediction index system. Specifically, the following steps are included:

[0086] S31: In-depth analysis of the medical service cost structure and influencing factors based on the influencing factor model of step S2, first identifying the most significant influencing factors.

[0087] S32: Analyze which factors are most important for the target variable medical cost based on feature importance scores or bar charts (using functions such as feature_importances_ and plot_importance()). Features with higher scores or longer bar charts generally indicate that they contribute more to the model predictions and are therefore the most important factors affecting medical costs.

[0088] S33: Sort all influencing factors and, combined with expert consultation, build a medical service cost prediction indicator system.

[0089] S4: Construct a medical service cost prediction model based on machine learning algorithm;

[0090] S41: Select the XGboost machine learning algorithm, input the test set data into the model for training, and repeatedly adjust the parameters and train until the error reaches the index range of regression prediction;

[0091] S42: Entering the testing phase of the medical service cost prediction model, after testing the model, fitting the project cost prediction results, observing whether the cost prediction model obtained by training and adjusting parameters can better fit the true value of the actual data, and then comparing the fitting degree of the training sample and the prediction test sample, and testing the reliability of the cost prediction model through the fitting effect of the cost prediction;

[0092] S43: According to the influencing factor variables that have an important impact on medical costs screened in the medical service cost prediction index system of step S3, the indicators of step S3 are put into the XGboost model in step S2 to construct a medical service cost prediction model, where the X variable of the medical service cost prediction model is the influencing factor variable in the medical service cost prediction index system of step S3, and Y is the medical service cost.

[0093] S5: Refer to the medical standard operating procedures of each service item, convert them into data and put them into the medical service cost prediction model of step S4 to predict the price of the corresponding service item.

[0094] Medical standard operating procedures (SOP) refer to a series of standardized operating procedures developed by medical institutions to ensure the quality and safety of medical services. These procedures cover the entire process from patient reception, diagnosis, treatment to rehabilitation, including the use of medical equipment, the procurement and management of drugs, the workflow of medical staff and other aspects. By following SOP, medical institutions can ensure the standardization and consistency of medical services and improve service quality and efficiency.

[0095] Refer to the standard operating procedures of different medical service projects, put the data of the standard procedures into the constructed cost prediction model, obtain the standard cost estimate price, then adjust the cost prediction model, and finally use expert consultation and other methods to obtain the final cost estimate price of the project.

[0096] The above embodiments are only preferred implementations of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for estimating medical service costs based on a machine learning algorithm, characterized in that: The following steps are involved: S1: Data collection and cleaning, which specifically includes: S11: Use expert interviews to obtain key factors affecting medical costs; S12: Obtain the main factors affecting medical costs through literature; S13: Objective data set collection and cleaning, construction of original cost data set; S2: Based on the original cost data set, a machine learning algorithm is used to analyze the factors affecting the cost of medical services and to construct an influencing factor model; S3: Conduct an in-depth analysis of the medical service cost structure and influencing factors of the influencing factor model, rank the factors influencing the medical service cost in terms of importance, and establish a medical service cost prediction indicator system based on this ranking and combined with expert consultation; S4: Construct a medical service cost prediction model based on machine learning algorithm; S5: Refer to the medical standard operating procedures of each service item, convert them into data and put them into the medical service cost prediction model of step S4 to predict the price of the corresponding service item.

2. The method for estimating medical service costs based on a machine learning algorithm according to claim 1, characterized in that: Step S2 includes the following steps: S21. Algorithm selection and evaluation; Algorithm selection: selecting a suitable machine learning algorithm based on the characteristics of the medical service cost data of the original cost dataset and the analysis objectives; algorithm evaluation: after the algorithm is selected, the algorithm performance is evaluated using cross-validation; S22. Feature engineering; Feature selection: extract features related to cost prediction from raw medical data, and select features that have a significant impact on cost prediction through correlation analysis and feature importance assessment; feature processing: perform necessary processing on selected features, such as missing value filling, outlier processing, data normalization or standardization, to ensure the stability and accuracy of model training; Feature dimensionality reduction, which is performed through principal component analysis or linear discriminant analysis to reduce the number of features while retaining key information; S23. Model training; Divide the data set, divide the processed data set into training set, validation set and test set. The training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the final performance of the model; Model training, use the training set data to train the selected machine learning algorithm, and adjust the model parameters through the iterative optimization algorithm to minimize the prediction error; Hyperparameter tuning, use the validation set to tune the model's hyperparameters to find the best model configuration; S24. Model evaluation and selection; Performance evaluation: Use the test set to evaluate the performance of the trained model, calculate the evaluation indicators of accuracy, recall rate, and F1 score, and draw ROC curves and PR curves to intuitively display the performance of the model; Model selection: select the model with the best performance as the final service cost prediction model based on the evaluation results. If multiple models have similar performance, make a comprehensive selection based on model complexity, training time, and interpretability. S25. Model optimization and construction; Model optimization: further optimize the shortcomings of the model, such as introducing more features, adjusting algorithm parameters, and using ensemble learning methods to improve model performance. Model construction: deploy the optimized model into the actual production environment, connect it with the business system of the medical institution, and build an influencing factor model.

3. The method for estimating medical service costs based on machine learning algorithm according to claim 2, characterized in that: The machine learning algorithm in step S21 is selected from the XGBoost algorithm.

4. The method for estimating medical service costs based on a machine learning algorithm according to claim 2, characterized in that: The algorithm evaluation indicators of step S21 include accuracy, recall, F1 score and AUC value.

5. The method for estimating medical service costs based on machine learning algorithm according to claim 2, characterized in that: Step S3 includes the following steps: S31: In-depth analysis of the medical service cost structure and influencing factors based on the influencing factor model of step S2, first identifying the most significant influencing factors; S32: Based on the feature importance scores or bar charts, analyze which factors are most important for the target variable medical cost; S33: Sort all influencing factors and, combined with expert consultation, build a cost prediction indicator system for medical services.

6. The method for estimating medical service costs based on a machine learning algorithm according to claim 5, characterized in that: Step S4 includes the following steps: S41: Select the XGboost machine learning algorithm, input the test set data into the model for training, and repeatedly adjust the parameters and train until the error reaches the index range of regression prediction; S42: Entering the testing phase of the medical service cost prediction model, after testing the model, fitting the project cost prediction results, observing whether the cost prediction model obtained by training and adjusting parameters can better fit the true value of the actual data, and then comparing the fitting degree of the training sample and the prediction test sample, and testing the reliability of the cost prediction model through the fitting effect of the cost prediction; S43: According to the influencing factor variables that have an important impact on medical costs screened in the medical service cost prediction index system of step S3, the indicators of step S3 are put into the XGboost model in step S2 to construct a medical service cost prediction model, where the X variable of the medical service cost prediction model is the influencing factor variable in the medical service cost prediction index system of step S3, and Y is the medical service cost.

7. The method for estimating medical service costs based on a machine learning algorithm according to claim 1, characterized in that: The objective data set of step S13 includes medical-related data from medical institutions, relevant departments, and patients, specifically including patient basic information, diagnosis information, treatment plan, cost details, and doctor's title, education, years of work experience, technical service content, materials used, relevant auxiliary personnel, and operation time data.

Images