Medical quality supervision method, system and equipment based on anti-fact cost

Through a counterfactual cost-based method, using predictive models and potential results theoretical frameworks, the problem of low medical quality is solved, effective supervision and monitoring of medical quality is achieved, and the accuracy and stability of supervision are improved.

CN119005399AInactive Publication Date: 2024-11-22363 HOSPITAL
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411017160.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing medical quality supervision has low quality problems caused by the quality of medical staff, hospital management and external factors, and lacks effective supervision methods.

Method used

Medical quality supervision methods based on counterfactual expenses are adopted to obtain patient hospitalization data, predict counterfactual expenses using prediction models, and calculate the cost deviation index to conduct medical quality supervision. The method includes data feature extraction, data transformation, data augmentation of potential results theory frameworks and the application of a variety of machine learning models, and constructing predictive sub-models to accurately predict costs.

Benefits of technology

Quantitative monitoring of medical quality has been achieved, the reliability and stability of supervision has been improved, and medical behaviors with abnormally expensive expenses can be identified, helping to improve medical quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119005399A_ABST
    Figure CN119005399A_ABST
Patent Text Reader

Abstract

The invention relates to the field of intelligent medical treatment, in particular to a medical quality supervision method, system and device based on anti-fact cost. The method includes: acquiring hospitalization data and cost of a patient; transmitting the hospitalization data of the patient to a prediction model for prediction to obtain anti-fact cost; calculating a deviation index based on the anti-fact cost and the cost to obtain a deviation result; according to the method, the medical behavior with the abnormal cost is monitored by adopting the anti-fact result of the potential result theory framework, and the method has good application value and research significance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent medical care, and specifically to a medical quality supervision method, system, device and computer-readable storage medium based on counterfactual costs. Background Art

[0002] Medical quality supervision is an important measure to ensure the quality of medical and health services. With the continuous development of medical technology and the continuous increase in medical needs, the importance of medical quality supervision has become increasingly prominent. Governments of various countries have introduced relevant policies to strengthen the supervision and management of the quality of medical and health services in order to improve the level of medical and health services and public trust. Medical quality standards are an important indicator for measuring the quality of medical and health services. Governments and medical institutions of various countries have formulated corresponding medical quality standards based on actual conditions, including diagnostic accuracy, treatment effectiveness, patient satisfaction and other aspects. These standards provide a basis and reference for medical quality supervision. However, there are still some problems with the current medical quality that lead to low medical quality. The following are some common factors:

[0003] 1. Personal qualities of medical staff: lack of professional attitude, professional ethics, knowledge and skills of medical staff, such as weak legal awareness, weak sense of responsibility, violation of operating procedures, failure to strictly implement and enforce rules and regulations, etc.; 2. Hospital management factors: lax hospital management, imperfect supervision and inspection mechanism, poor management of medical resources (human resources and equipment resources), and inadequate safety precautions; 3. External factors: atypical individual characteristics of patients, low degree of cooperation, imperfect medical care system, and lax administrative law enforcement of medical institutions.

[0004] Medical quality is directly related to the health rights and interests of the people and their personal experience of medical services. Continuously improving quality and ensuring medical safety are important contents and foundations for the reform and development of the health industry, and are of great significance to the implementation of reform measures such as the construction of a tiered diagnosis and treatment system and the realization of medical reform goals. Summary of the invention

[0005] With the development of information technology, medical quality supervision is also gradually developing towards information and intelligence, such as the application of electronic medical records, medical big data analysis and other technologies. Aiming at the problem of low medical quality, the present invention uses information technology to propose a medical quality supervision method based on counterfactual results, which specifically includes:

[0006] Obtain patient hospitalization data and costs;

[0007] Inputting the patient's hospitalization data into a prediction model to predict counterfactual costs;

[0008] Calculating a deviation index based on the counterfactual cost and the cost to obtain a deviation result;

[0009] Medical quality supervision based on deviation results.

[0010] Furthermore, the prediction is performed using a pre-trained prediction model to obtain a prediction result, and the training process of the pre-trained prediction model includes:

[0011] Get the dataset S and labels;

[0012] The data set S and labels are input into the model for training to obtain a trained prediction model;

[0013] Preferably, the prediction model adopts one or more of the following: lightgbm, XGBoost, GBDT, decision tree, random forest, BP neural network, support vector machine, naive Bayes;

[0014] Preferably, the training process further includes data conversion, performing data conversion on the data set S and the label to obtain the data set D and the label;

[0015] Preferably, the total disease cost is used as a label; the data set D and the label are input into the lightgbm model for training to obtain a prediction model.

[0016] Furthermore, the data conversion includes two steps: counting the data features in the data set S, calculating the costs under different features based on the data features, and then combining the costs under different features with other features to form a data set D;

[0017] Preferably, the data features in the data set S include one or more of the following: hospital identifier, age, gender, department, hospital type, hospital grade, disease primary diagnosis ICD code, disease secondary diagnosis ICD code, surgery code, total disease cost;

[0018] Preferably, the costs with different characteristics include one or more of the following: disease primary diagnosis ICD code - surgery - average cost, disease primary diagnosis ICD code - age - average cost, disease primary diagnosis ICD code - hospital grade - average cost;

[0019] Preferably, the other features include: disease primary diagnosis code, surgery, age;

[0020] Preferably, based on the data characteristics of the inpatient data set S, the disease primary diagnosis ICD code-surgery-average cost, disease primary diagnosis ICD code-age-average cost, disease primary diagnosis ICD code-hospital grade-average cost are obtained, which are recorded as data set T. The data set T and the disease primary diagnosis code, surgery, and age together constitute data set D.

[0021] The method also includes matching a prediction sub-model based on the department, hospital type, and hospital grade in the patient hospitalization data, and inputting the patient hospitalization data into the prediction sub-model to predict the counterfactual cost.

[0022] Furthermore, the construction process of the prediction sub-model is as follows:

[0023] Get the dataset S and labels,

[0024] Based on the department / hospital type / hospital grade of the data set S, the data set is divided to obtain a category data set;

[0025] Inputting the category data set into each sub-model for training to obtain a trained prediction sub-model;

[0026] Preferably, the construction process of the prediction sub-model further includes data expansion, the data set S contains n pieces of data, n is a natural number greater than 1; each piece of data in the data set S is expanded into a data set;

[0027] Preferably, each piece of data expansion adopts the potential result theory framework to screen the department / hospital type / hospital grade of data d0 for a set i of hospital identifiers that meet the requirements in the data set S, and replaces each identifier in the set i with the identifier in data d0 in turn, while the remaining data remain unchanged, to form a data set d1;

[0028] Preferably, the construction process of the prediction sub-model also includes data conversion: converting the data set S into a data set D, using the total disease cost as the label, and dividing the data set D into departments / hospital types / hospital grades to obtain a category data set.

[0029] The prediction model predicts the total disease cost f of the potential result set, calculates the average value m of the total cost f and uses the average value m as the counterfactual cost m1 of the hospitalization data d;

[0030] Preferably, the average value of the total cost is calculated after removing the highest value and the lowest value of the total cost f;

[0031] Preferably, the cost ratios of different hospitals are calculated to obtain coefficients, and the counterfactual costs are adjusted based on the coefficients.

[0032] The deviation index is obtained by the ratio of the total cost df of the hospitalization data d to the counterfactual cost m1 of the hospitalization data d, and the expression formula is:

[0033] O=df / m1

[0034] Wherein, O represents the deviation index.

[0035] The object of the present invention is to provide a medical quality supervision system based on counterfactual costs, comprising:

[0036] Acquisition unit: obtains patient hospitalization data and expenses;

[0037] Conversion unit: inputting the patient hospitalization data into the prediction model to predict the counterfactual cost;

[0038] A calculation unit: calculating a deviation index based on the counterfactual cost and the cost to obtain a deviation result;

[0039] Supervision unit: Medical quality supervision based on deviation results.

[0040] The object of the present invention is to provide a medical quality supervision device based on counterfactual costs, comprising:

[0041] A memory and a processor, wherein the memory is used to store program instructions; the processor is used to call the program instructions, and when the program instructions are executed, any one of the above-mentioned medical quality supervision methods based on counterfactual costs is implemented.

[0042] The object of the present invention is to provide a computer-readable storage medium having a computer program thereon, comprising:

[0043] When the computer program is executed by a processor, it implements any one of the above-mentioned medical quality supervision methods based on counterfactual costs.

[0044] Advantages of the present invention:

[0045] 1. Starting from the expenses incurred by patients in the hospital, a model is built to predict counterfactual costs. The medical quality is monitored based on the counterfactual costs. The medical big data information processing technology is combined to quantify the level of medical quality, which is convenient for monitoring medical quality. It has good practical significance and research value.

[0046] 2. The average cost of patients is obtained based on the dimensions of disease primary diagnosis ICD code, surgery, and age, namely, disease primary diagnosis ICD code-surgery-average cost, disease primary diagnosis ICD code-age-average cost, disease primary diagnosis ICD code-hospital grade-average cost, and other data features constitute the data set; this data set calculates costs from multiple dimensions, enriches data features, and is conducive to the reliability and stability of medical quality supervision.

[0047] 3. Use the potential outcome theory framework to expand the data and obtain the potential outcomes of patients in various hospitals. Use the prediction model to predict the total cost and reverse outcome cost of patients in various hospitals. Calculate the deviation between the reverse outcome cost and the actual cost to obtain the deviation result, and then monitor the medical quality of highly abnormal deviation results.

[0048] 4. Training prediction sub-models based on departments, hospital grades, and hospital types makes the cost prediction clearer and the division more precise, which is conducive to improving the accuracy of the deviation between the predicted costs and the actual costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 A schematic flow chart of a medical quality supervision method based on counterfactual costs provided by an embodiment of the present invention;

[0051] Figure 2 A schematic diagram of a medical quality supervision system based on counterfactual costs provided by an embodiment of the present invention;

[0052] Figure 3 A schematic diagram of a medical quality supervision device based on counterfactual costs provided by an embodiment of the present invention;

[0053] Figure 4 A graph of the correlation between the cost of the disease surgery group and age provided in an embodiment of the present invention;

[0054] Figure 5 A distribution diagram of mean square error of actual costs and counterfactual prediction values ​​provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0056] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as S101, S102, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0057] Figure 1A schematic diagram of a medical quality supervision method based on counterfactual costs provided by an embodiment of the present invention specifically includes:

[0058] S101: Obtain patient hospitalization data and costs;

[0059] S102: Inputting the patient's hospitalization data into a prediction model to predict counterfactual costs;

[0060] In one embodiment, in the theoretical system of causal inference, the potential outcomes framework is one of the most important theoretical models. The potential outcomes model was proposed by Rubin, a well-known statistician at Harvard University, so the model is also called the Rubin causal model. The core of the potential outcomes model is to compare the effects of the same research subject with and without the intervention. For the research subject who receives the intervention, not accepting the intervention is a "counterfactual" state, and for the research subject who does not accept the intervention, accepting the intervention is a "counterfactual" state.

[0061] The potential outcome framework has three basic components: potential outcomes, stability assumptions, and allocation mechanisms. In causal inference, there must be intervention. Without intervention, there is no causality. The intervention here can be a policy, a measure, or an activity. Each intervention state will have a corresponding (potential) outcome. Before the intervention state is realized, there are several potential outcomes for several intervention states, and after the intervention state is realized, only one potential outcome is observable. The potential outcome can be regarded as a constant. For each specific individual, its potential outcome under the two intervention states is given and does not depend on the final realized intervention state. When the intervention state is realized, only the potential outcome in the realized state can be observed, and the potential outcome in the unrealized state cannot be observed. Unobservable potential outcomes are usually called counterfactual outcomes. The phenomenon that all potential outcomes of an individual cannot be observed at the same time is called the basic problem of causal inference.

[0062] The stability assumption has two meanings:

[0063] First, there will be no interaction between potential outcomes for different individuals;

[0064] Second, the level of intervention was the same for all individuals.

[0065] The allocation mechanism is a mechanism that describes why some people are in the intervention group and some are in the control group, or the allocation mechanism is a mechanism that describes which potential results can be observed. When making causal inferences, counterfactual results must be estimated or predicted, and such predictions are often difficult and involve understanding the allocation mechanism or selection mechanism. In order to understand the allocation mechanism, some pretreatment variables or covariates are often needed.

[0066] There are two types of covariates: one is individual attributes, which do not change with the intervention status, such as gender, ethnicity, etc. The other is variables that are taken before the intervention is implemented. For example, when studying the role of training, the income level and economic and social characteristics before training are all covariates. The basic characteristic of covariates is that these variables are not affected by the intervention variables. On the contrary, these variables often determine the individual's intervention status (somewhat similar to control variables).

[0067] Depending on whether the allocation mechanism is known or not, the allocation mechanism can be divided into two categories: randomized experiments and observational studies. These two types of mechanisms have the common feature of studying the causal effect of an intervention.

[0068] In a randomized experiment, the allocation mechanism is controlled by the experimenter or researcher and is known.

[0069] In observational studies, the distribution mechanism is unknown. The purpose of observational studies is to find a way to identify the unknown distribution mechanism and estimate the causal effect.

[0070] Non-confounding is also called conditional independence or selection on observables. It means that after controlling the observed variable Xi, the allocation of individual intervention status does not depend on the potential outcome. Whether the individual is in the intervention group or the control group is independent of the potential outcome (their potential outcome will not affect the outcome).

[0071] In one embodiment, the Potential Outcome Framework is a classic framework for studying causal relationships, which is used to estimate the impact of an intervention on an outcome. Here, intervention refers to medical treatment in different medical institutions, so data expansion is a theoretical requirement of the causal framework. This framework can reflect the advancement of conventional data analysis that relies solely on correlation relationships.

[0072] In one embodiment, the counterfactual cost is the potential cost after intervention. For example, the actual cost of A's medical treatment at Hospital A is the counterfactual cost through the assumption (intervention) that A may receive medical treatment at Hospital B or other medical institutions during that visit. This is a counterfactual result because it is actually impossible for A to receive medical treatment at Hospital B or other medical institutions during that visit.

[0073] In one embodiment, LightGBM (Light Gradient Boosting Machine) is a gradient boosting machine (GBM) algorithm for solving classification and regression problems. LightGBM is a tree-based ensemble learning method that uses gradient boosting technology to combine multiple weak learners (usually decision trees) into a powerful model. The principles are as follows: Gradient boosting: LightGBM uses gradient boosting technology to continuously fit the residual of the model (i.e., the difference between the actual value and the predicted value) in an iterative manner, thereby gradually improving the prediction ability of the model. Tree-based model: LightGBM uses a tree-based model as a weak learner, in which each tree consists of multiple nodes, each node contains a feature and a threshold, and by comparing the feature value of the sample with the size of the threshold, the sample is assigned to the left subtree or the right subtree, thereby achieving classification or regression of the sample. Efficient feature splitting strategy: LightGBM adopts a feature splitting strategy called "Leaf-wise" (leaf node splitting). Different from the traditional "Level-wise" (level splitting), it can select the features and thresholds that contribute the most to the current sample set at each split, thereby reducing the depth of the tree and improving the training speed. Efficient parallel computing: LightGBM also supports multi-threading and parallel computing, which can make full use of the computing power of multi-core CPUs and accelerate the training process of the model.

[0074] In one embodiment, GBDT (Gradient Boosting Decision Tree) is a boosting method that uses an additive model (i.e., a linear combination of basis functions) and a forward step-by-step algorithm with a decision tree as the basis function. In layman's terms, GBDT is an iterative decision tree algorithm that consists of multiple decision trees, and the conclusions of all trees are accumulated as the final answer. It constructs a set of weak learners (trees) and accumulates the results of multiple decision trees as the final prediction output. The algorithm effectively combines decision trees with ensemble ideas.

[0075] In one embodiment, XGBoost is an additive model based on boosted trees, which is a type of GBDT. Its basic idea is the same as GBDT, but many optimizations are added. Each base classifier used is a regression tree, and the training method is to use a forward step-by-step algorithm to gradually optimize each base learner inside. XGBoost is an improvement on the gradient boosting algorithm. When solving the extreme value of the loss function, Newton's method is used, and the loss function Taylor is expanded to the second order. In addition, a regularization term is added to the loss function. The objective function during training consists of two parts, the first part is the gradient boosting algorithm loss, and the second part is the regularization term.

[0076] In one embodiment, a decision tree is a decision analysis method that constructs a decision tree to obtain the probability that the expected value of the net present value is greater than or equal to zero based on the known probability of occurrence of various situations, evaluates project risks, and determines its feasibility. It is a graphical method that intuitively uses probability analysis. Since this decision branch is drawn into a graph that looks like a tree branch, it is called a decision tree. In machine learning, a decision tree is a prediction model that represents a mapping relationship between object attributes and object values.

[0077] In the process of constructing a decision tree, information gain (or information gain rate, Gini index, etc.) is usually used to help determine the node splitting of the tree. Each internal node of the decision tree represents a test on an attribute, each branch represents a test output, and each leaf node represents a category.

[0078] The main advantages of decision trees include:

[0079] 1. Simple and intuitive, the generated decision tree is very intuitive.

[0080] 2. Basically no preprocessing is required, no advance normalization is required, and no missing value processing is required.

[0081] 3. The cost of using decision tree prediction is O(log2m). m is the number of samples.

[0082] 4. Can process both discrete and continuous values. Many algorithms only focus on discrete values ​​or continuous values.

[0083] 5. Can handle classification problems with multi-dimensional output.

[0084] 6. Compared with black-box classification models such as neural networks, decision trees can be well explained logically.

[0085] 7. Cross-validated pruning can be used to select models to improve generalization capabilities.

[0086] 8. Good fault tolerance for outliers and high robustness.

[0087] In one embodiment, the prediction is performed using a pre-trained prediction model to obtain a prediction result, and the training process of the pre-trained prediction model includes:

[0088] Get the dataset S and labels;

[0089] The data set S and labels are input into the model for training to obtain a trained prediction model;

[0090] Preferably, the prediction model adopts one or more of the following: lightgbm, XGBoost, GBDT, decision tree, random forest, BP neural network, support vector machine, naive Bayes;

[0091] Preferably, the training process further includes data conversion, performing data conversion on the data set S and the label to obtain the data set D and the label;

[0092] Preferably, the total disease cost is used as a label; the data set D and the label are input into the lightgbm model for training to obtain a prediction model.

[0093] In one embodiment, the data conversion includes two steps: counting the data features in the data set S, calculating the costs under different features based on the data features, and then combining the costs under different features with other features to form a data set D;

[0094] Preferably, the data features in the data set S include one or more of the following: hospital identifier, age, gender, department, hospital type, hospital grade, disease primary diagnosis ICD code, disease secondary diagnosis ICD code, surgery code, total disease cost;

[0095] Preferably, the costs with different characteristics include one or more of the following: disease primary diagnosis ICD code - surgery - average cost, disease primary diagnosis ICD code - age - average cost, disease primary diagnosis ICD code - hospital grade - average cost;

[0096] Preferably, the other features include: disease primary diagnosis code, surgery, age;

[0097] Preferably, based on the data characteristics of the inpatient data set S, the disease primary diagnosis ICD code-surgery-average cost, disease primary diagnosis ICD code-age-average cost, disease primary diagnosis ICD code-hospital grade-average cost are obtained, which are recorded as data set T. The data set T and the disease primary diagnosis code, surgery, and age together constitute data set D.

[0098] In one embodiment, the method further includes matching a prediction sub-model based on the department, hospital type, and hospital grade in the patient hospitalization data, and inputting the patient hospitalization data into the prediction sub-model to predict the counterfactual costs.

[0099] In one embodiment, the construction process of the prediction sub-model is:

[0100] Get the dataset S and labels,

[0101] Based on the department / hospital type / hospital grade of the data set S, the data set is divided to obtain a category data set;

[0102] Inputting the category data set into each sub-model for training to obtain a trained prediction sub-model;

[0103] Preferably, the construction process of the prediction sub-model further includes data expansion, the data set S contains n pieces of data, n is a natural number greater than 1; each piece of data in the data set S is expanded into a data set;

[0104] Preferably, each piece of data expansion adopts the potential result theory framework to screen the department / hospital type / hospital grade of data d0 for a set i of hospital identifiers that meet the requirements in the data set S, and replaces each identifier in the set i with the identifier in data d0 in turn, while the remaining data remain unchanged, to form a data set d1;

[0105] Preferably, the construction process of the prediction sub-model also includes data conversion: converting the data set S into a data set D, using the total disease cost as the label, and dividing the data set D into departments / hospital types / hospital grades to obtain a category data set.

[0106] In one embodiment, the prediction model predicts the total disease cost f of the potential result set, calculates the average value m of the total cost f and uses the average value m as the counterfactual cost m1 of the hospitalization data d;

[0107] Preferably, the average value of the total cost is calculated after removing the highest value and the lowest value of the total cost f;

[0108] Preferably, the cost ratios of different hospitals are calculated to obtain coefficients, and the counterfactual costs are adjusted based on the coefficients.

[0109] In a specific embodiment, low medical quality is usually reflected in medical expenses that are significantly higher than normal expenses. Therefore, the medical behavior with abnormally high expenses can be found through the counterfactual results in the potential result theory framework. Therefore, the core lies in calculating the counterfactual results through the model method. The specific method is as follows:

[0110] 1. The data used are hospitalization data from multiple hospitals. The data dimensions include hospital unique identifier, patient age, patient gender, department, hospital type, hospital grade, disease primary diagnosis ICD code, disease secondary diagnosis ICD code, surgery code, and total disease cost. This data is referred to as D in the following.

[0111] 2. Based on data set D, statistical data of three dimensions were obtained: "ICD code of disease diagnosis - surgery - average cost", "ICD code of disease diagnosis - age - average cost", and "ICD code of disease diagnosis - hospital grade - average cost", which will be referred to as D1 in the following.

[0112] 3. According to the disease main diagnosis ICD code, surgery, and age in data set D, connect data set D1 to form data set D2;

[0113] 4. On the dataset D2, with the total disease cost as the label and other dimensions as features, the model M is trained using lightgbm;

[0114] 5. For each data item (i.e., one hospitalization) in data set D2, referred to as d, and the total cost of d, referred to as df, according to the requirements of the potential outcome theory framework, all hospital unique identifier sets i that meet the requirements are selected according to the departments, hospital types, and hospital grades in d, and the unique identifiers in set i are used to replace the hospital unique identifiers in d, while keeping other dimensional data unchanged, to form data set d1;

[0115] 6. Use model M to predict the total disease cost f of the d1 data set, so as to obtain the counterfactual result of the total cost of hospitalization in other hospitals, that is, if the hospitalization d is spent in other hospitals, the average value m of the total cost f is calculated as the final counterfactual cost m1 of the hospitalization d;

[0116] In a specific embodiment, the correlation between disease surgery cost and age is as follows: Figure 4 As shown, as the cost of disease surgery increases, the relevance of age also increases.

[0117] S103: Calculating a deviation index based on the counterfactual cost and the cost to obtain a deviation result;

[0118] In one embodiment, the deviation index is obtained by the ratio of the total cost df of the hospitalization data d to the counterfactual cost m1 of the hospitalization data d, and the expression formula is:

[0119] O=df / m1

[0120] Wherein, o represents the deviation index.

[0121] In a specific embodiment, 7. Calculate the total cost deviation index o=df / m1 for a hospitalization d, which is also called blame and regret in the potential outcome framework;

[0122] 8. Calculate the deviation index of each data (all hospitalizations) in data set D2 according to the logic of steps 5, 6, and 7;

[0123] In a specific embodiment, the mean square error between the actual cost and the counterfactual predicted value is as follows: Figure 5 shown.

[0124] In a specific embodiment, the standard deviation, mean, and maximum value calculation results of the deviation index in the D2 data set are shown in Table 1:

[0125] Table 1 Deviation index calculation results

[0126] Model Results Deviation index standard deviation 0.553 Deviation from the mean index 0.969 Deviation index maximum 27.7

[0127] In one embodiment, the quality supervision through the deviation index of billing costs in the present invention also includes cost prediction for patients with a type of disease, such as the total cost of patients with cardiovascular disease, the total cost of patients with diabetes, and the total cost of patients with liver cancer.

[0128] S104: Conduct medical quality supervision based on deviation results.

[0129] In a specific embodiment, 9. A portion of hospitalizations with a higher deviation index o in D2 are suspected cases of poor medical quality and can be used for subsequent key supervision.

[0130] In a specific embodiment, medical supervision usually uses cost as a standard. If the actual cost exceeds expectations by a large margin, the medical quality is considered to be low. Good medical quality consumes fewer medical resources to achieve the expected results. In actual applications, the deviation index is used to sort (or the mean of the deviation index + 3 times the standard deviation), and supervision is carried out selectively according to actual audit requirements.

[0131] Figure 2 A schematic diagram of a medical quality supervision system based on counterfactual costs provided by an embodiment of the present invention specifically includes:

[0132] Acquisition unit: obtains patient hospitalization data and expenses;

[0133] Conversion unit: inputting the patient hospitalization data into the prediction model to predict the counterfactual cost;

[0134] A calculation unit: calculating a deviation index based on the counterfactual cost and the cost to obtain a deviation result;

[0135] Supervision unit: Medical quality supervision based on deviation results.

[0136] Figure 3 A schematic diagram of a medical quality supervision device based on counterfactual costs provided by an embodiment of the present invention specifically includes:

[0137] A memory and a processor; the memory is used to store program instructions; the processor is used to call program instructions, and when the program instructions are executed, any one of the above-mentioned medical quality supervision methods based on counterfactual costs is implemented.

[0138] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, is any one of the above-mentioned medical quality supervision methods based on counterfactual costs.

[0139] The verification results of this verification example show that assigning inherent weights to indications can improve the performance of the method relative to the default settings. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. In several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways.

[0140] For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, each functional unit in each embodiment of the present invention may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. A person skilled in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0141] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be implemented by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. The above-mentioned medium storage can be a read-only memory, a disk or an optical disk, etc.

[0142] The above is a detailed introduction to a computer device provided by the present invention. For a person skilled in the art, according to the concept of the embodiments of the present invention, there may be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A medical quality supervision method based on counterfactual costs, characterized in that: The method specifically includes: obtaining patient hospitalization data and expenses; Inputting the patient's hospitalization data into a prediction model to predict counterfactual costs; Calculating a deviation index based on the counterfactual cost and the cost to obtain a deviation result; Medical quality supervision based on deviation results.

2. The medical quality supervision method based on counterfactual costs according to claim 1 is characterized in that: The prediction uses a pre-trained prediction model to predict and obtain a prediction result. The training process of the pre-trained prediction model includes: Get the dataset S and labels; The data set S and labels are input into the model for training to obtain a trained prediction model; Preferably, the prediction model adopts one or more of the following: lightgbm, XGBoost, GBDT, decision tree, random forest, BP neural network, support vector machine, naive Bayes; Preferably, the training process further includes data conversion, performing data conversion on the data set S and the label to obtain the data set D and the label; Preferably, the total disease cost is used as a label; the data set D and the label are input into the lightgbm model for training to obtain a prediction model.

3. The medical quality supervision method based on counterfactual costs according to claim 2 is characterized in that: The data conversion includes two steps: counting the data features in the data set S, calculating the costs under different features based on the data features, and then combining the costs under different features with other features to form a data set D; Preferably, the data features in the data set S include one or more of the following: hospital identifier, age, gender, department, hospital type, hospital grade, disease primary diagnosis ICD code, disease secondary diagnosis ICD code, surgery code, total disease cost; Preferably, the costs with different characteristics include one or more of the following: disease primary diagnosis ICD code - surgery - average cost, disease primary diagnosis ICD code - age - average cost, disease primary diagnosis ICD code - hospital grade - average cost; Preferably, the other features include: disease primary diagnosis code, surgery, age; Preferably, based on the data characteristics of the inpatient data set S, the disease primary diagnosis ICD code-surgery-average cost, disease primary diagnosis ICD code-age-average cost, disease primary diagnosis ICD code-hospital grade-average cost are obtained, which are recorded as data set T. The data set T and the disease primary diagnosis code, surgery, and age together constitute data set D.

4. The medical quality supervision method based on counterfactual costs according to claim 1 is characterized in that: The method also includes matching a prediction sub-model based on the department, hospital type, and hospital grade in the patient hospitalization data, and inputting the patient hospitalization data into the prediction sub-model to predict the counterfactual cost.

5. The medical quality supervision method based on counterfactual costs according to claim 4 is characterized in that: The construction process of the prediction sub-model is as follows: Get the dataset S and labels, Based on the department / hospital type / hospital grade of the data set S, the data set is divided to obtain a category data set; Inputting the category data set into each sub-model for training to obtain a trained prediction sub-model; Preferably, the construction process of the prediction sub-model further includes data expansion, the data set S contains n pieces of data, n is a natural number greater than 1; each piece of data in the data set S is expanded into a data set; Preferably, each piece of data expansion adopts the potential result theory framework to screen the department / hospital type / hospital grade of data d0 for a set i of hospital identifiers that meet the requirements in the data set S, and replaces each identifier in the set i with the identifier in data d0 in turn, while the remaining data remain unchanged, to form a data set d1; Preferably, the construction process of the prediction sub-model also includes data conversion: converting the data set S into a data set D, using the total disease cost as the label, and dividing the data set D into departments / hospital types / hospital grades to obtain a category data set.

6. The medical quality supervision method based on counterfactual costs according to claim 1 is characterized in that: The prediction model predicts the total disease cost f of the potential result set, calculates the average value m of the total cost f and uses the average value m as the counterfactual cost m1 of the hospitalization data d; Preferably, the average value of the total cost is calculated after removing the highest value and the lowest value of the total cost f; Preferably, the cost ratios of different hospitals are calculated to obtain coefficients, and the counterfactual costs are adjusted based on the coefficients.

7. The medical quality supervision method based on counterfactual costs according to claim 1 is characterized in that: The deviation index is obtained by the ratio of the total cost df of the hospitalization data d to the counterfactual cost m1 of the hospitalization data d, and the expression formula is: O=df / m1 Wherein, O represents the deviation index.

8. A medical quality supervision system based on counterfactual costs, characterized in that: include: Acquisition unit: obtains patient hospitalization data and expenses; Conversion unit: inputting the patient hospitalization data into the prediction model to predict the counterfactual cost; A calculation unit: calculating a deviation index based on the counterfactual cost and the cost to obtain a deviation result; Supervision unit: Medical quality supervision based on deviation results.

9. A medical quality supervision device based on counterfactual costs, characterized in that: include: A memory and a processor, wherein the memory is used to store program instructions; the processor is used to call program instructions, and when the program instructions are executed, the medical quality supervision method based on counterfactual costs as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program thereon, characterized in that: include: When the computer program is executed by a processor, the medical quality supervision method based on counterfactual costs described in any one of claims 1 to 7 is implemented.

Citation Information

Cited By

  • Medical Consumables Anomaly Analysis System

    TWI940449B