Analysis method and system for illegal charging in medical industry

Abstract

The invention relates to a method and system for analyzing illegal charging in the medical industry, and relates to the technical field of data analysis, and the method comprises the steps: obtaining the medical record and charging information of a target user, and extracting charging elements and charging values; randomly combining the charging elements, and analyzing the integration rate and the occurrence rate to obtain an integration rate set and an occurrence rate set; performing charge prediction according to the medical record information to obtain a predicted charge element set and a predicted charge value set, and performing matching analysis with actual charge to obtain an element and value matching rate; and calculating a first violation coefficient by adopting the element matching rate, the integration rate set and the occurrence rate set, calculating a second violation coefficient according to the numerical value matching rate, and calculating to obtain a violation coefficient as an analysis result. According to the invention, for the problems of medical service hidden illegal charging and low accuracy of a traditional analysis means, illegal behaviors such as charging are accurately identified and decomposed through charging element combination analysis, historical data comparison and multi-dimensional coefficient calculation, and the accuracy and pertinence of medical charging analysis are improved.

Description

Technical Field

[0001] This application relates to the field of data analysis, and in particular to analytical methods and systems for identifying illegal charges in the healthcare industry. Background Technology

[0002] With the development of intelligent management of medical charges, accurate identification of illegal charges by hospitals has become crucial for improving industry compliance. Currently, traditional medical charge violation analysis mostly relies on manual verification, which suffers from inefficiency and poor accuracy, making it difficult to meet the demand for accurate identification of hidden illegal charges in medical services.

[0003] Existing analysis methods rely solely on manual experience to compare billing items, resulting in insufficient accuracy in identifying "fragmented" illegal charges. This increases audit costs and is difficult to meet the requirements of real-time monitoring and accurate identification of illegal charges in the digital transformation of the healthcare industry. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides an analysis method and system for illegal charges in the medical industry, which improves the accuracy and efficiency of identifying and analyzing hidden illegal activities such as breaking down medical services into smaller, manageable charges.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a method for analyzing illegal charges in the medical industry, the method comprising:

[0007] Obtain the target user's medical record information and billing information, and extract multiple billing elements and multiple billing values ​​from the billing information;

[0008] Randomly combine the multiple charging elements, perform rounding rate analysis and occurrence rate analysis to obtain rounding rate set and occurrence rate set;

[0009] Based on the medical record information, a billing prediction is performed to obtain a set of predicted billing elements and a set of predicted billing values. These are then matched with the multiple billing elements and multiple billing values ​​to obtain the element matching rate and the value matching rate.

[0010] The first violation coefficient is obtained by processing the element matching rate, the rounding rate set, and the occurrence rate set. The second violation coefficient is obtained by processing the numerical matching rate. The violation coefficient is calculated and used as the analysis result.

[0011] Secondly, embodiments of this application provide an analysis system for illegal charges in the medical industry, the system comprising:

[0012] The data information extraction module is used to obtain the target user's medical record information and billing information, and extract multiple billing elements and multiple billing values ​​from the billing information;

[0013] The element combination analysis module is used to randomly combine the multiple charging elements, perform rounding rate analysis and occurrence rate analysis, and obtain rounding rate set and occurrence rate set;

[0014] The billing prediction and matching module is used to predict billing based on the medical record information, obtain a set of predicted billing elements and a set of predicted billing values, and perform matching analysis with the multiple billing elements and multiple billing values ​​to obtain the element matching rate and the value matching rate.

[0015] The violation coefficient calculation module is used to process the element matching rate, rounding rate set and occurrence rate set to obtain the first violation coefficient, process the numerical matching rate to obtain the second violation coefficient, and calculate the violation coefficient as the analysis result.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] This application proposes a method and system for analyzing illegal charges in the medical industry. By acquiring the target user's medical record and billing information, multiple billing elements and values ​​are extracted from the billing information. Features such as rounding rate, occurrence rate, element matching rate, and value matching rate are comprehensively analyzed to perform illegal charge analysis on multi-dimensional billing data. Simultaneously, multiple billing elements are randomly combined, and rounding rate and occurrence rate analyses are performed to obtain rounding rate sets and occurrence rate sets. Based on medical record information, billing prediction is performed to obtain a predicted billing element set and a predicted billing value set. These are then matched with the actual billing elements and values, effectively improving the accuracy of illegal charge analysis and avoiding the problem of difficult-to-detect illegal charges due to the highly specialized nature of billing items. Through multi-dimensional indicator analysis, predictive matching, and hierarchical coefficient calculation, a two-dimensional indicator of billing element combination and value matching is integrated, effectively solving the problem of inaccurate identification of illegal charges caused by the single nature of traditional analysis methods.

[0018] The technical solution of this application achieves accurate analysis of illegal charges in the medical industry by integrating rounding rate analysis, occurrence rate analysis and predictive matching analysis. It solves the problem of incomplete identification of illegal charges caused by the single analysis method and insufficient accuracy in traditional analysis, improves the reliability of medical charge supervision, and avoids the problem of illegal charges being difficult to detect due to inaccurate charge analysis. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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 effort.

[0020] Figure 1 A flowchart illustrating the method for analyzing illegal charges in the medical industry provided in this application embodiment;

[0021] Figure 2 A schematic diagram of the structure of the medical industry illegal charging analysis system provided in the embodiments of this application;

[0022] The components represented by each number in the attached diagram are explained below:

[0023] Data information extraction module 01, element combination analysis module 02, fee prediction and matching module 03, violation coefficient calculation module 04. Detailed Implementation

[0024] This application provides an analysis method and system for illegal charges in the medical industry, which solves the technical problems of the traditional manual verification process for medical charges being cumbersome and inefficient, having insufficient accuracy in identifying hidden illegal charges, and having the accuracy of analysis being greatly affected by human experience.

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0028] Example 1, as shown in the appendix Figure 1 As shown, this application provides a method for analyzing illegal charges in the medical industry, the method comprising the following steps:

[0029] S100: Obtain the target user's medical record information and billing information, and extract multiple billing elements and multiple billing values ​​from the billing information;

[0030] In this embodiment of the application, in order to achieve accurate analysis of the charging data during the analysis of illegal charges in the medical industry, it is necessary to obtain the electronic medical record information (which includes key contents such as diagnosis results and treatment plans) and the corresponding charge list information of the target user through the hospital information system interface or data interaction interface.

[0031] Furthermore, the billing information is structured and parsed to extract multiple billing elements (such as nursing fees, anesthesia fees, surgical instrument fees, etc.) and their corresponding multiple billing values ​​(such as the specific amount of each fee category), forming a dataset of corresponding billing elements and values.

[0032] This process, through standardized data acquisition and parsing logic, ensures the integrity of medical record information and the accuracy of billing data, providing a reliable data foundation for subsequent analysis of illegal charges.

[0033] Step S100 in the method provided in this application embodiment includes:

[0034] Obtain the target user's medical records and billing information;

[0035] Extract the charging elements and charging values ​​from the charging information to obtain multiple charging elements and multiple charging values, wherein the charging elements and charging values ​​correspond to each other.

[0036] In this embodiment of the application, in order to accurately obtain the basic data for the analysis of illegal charges in the medical industry, it is necessary to obtain the electronic medical record information and corresponding charge list information of the target user through the data interface of the hospital information system.

[0037] Specifically, by calling the API interface of the hospital's HIS (Hospital Information System), data acquisition parameters such as patient name and consultation time range can be set to automatically acquire the target user's medical record information (including diagnosis results, treatment items, etc.) and billing information.

[0038] For example, if the target user is a hospitalized patient, their electronic medical record during hospitalization can be obtained, in which the diagnosis is "acute appendicitis" and the corresponding hospitalization bill.

[0039] Furthermore, the obtained fee list information is structured and automatically retrieved and matched with the item name and amount fields in the fee list through a preset fee item keyword library and rule templates, and the fee elements and fee values ​​are extracted.

[0040] Specifically, the hospital's HIS system scans and matches the billing list text based on a pre-set keyword database of billing items (such as standard terms and synonyms like "anesthesia fee" and "nursing fee") to identify the billing elements.

[0041] At the same time, through preset amount extraction rules, the system automatically locates and extracts the amount corresponding to each charging element, realizing the automatic extraction of charging elements and values ​​and the construction of corresponding relationships.

[0042] The rules for extracting amounts are based on the common formats and positions of fees in the fee list. The rules stipulate that the amount usually appears after specific keywords (such as "amount" or "fee") and has a specific numerical format (such as including the "yuan" symbol, the number of digits, and the position of the decimal point). The rules then automatically locate and extract the amount corresponding to each fee element according to these preset rules.

[0043] For example, if a hospital bill for an inpatient with "acute appendicitis" contains items such as "appendectomy 5000 yuan", "intravenous infusion 50 yuan", and "level 3 nursing care fee 100 yuan / day", the hospital's HIS system will automatically scan and match the bill text based on a preset keyword database of billing items (containing standard terms such as "appendectomy" and "intravenous infusion") to identify the corresponding billing elements.

[0044] Simultaneously, based on the fee extraction rules, "5000 yuan" is located from "appendectomy 5000 yuan" as the value of this fee element, "50 yuan" is extracted from "intravenous infusion 50 yuan", and "100 yuan / day" is extracted from "level 3 nursing care fee 100 yuan / day", thus realizing the automated extraction of fee elements and the construction of corresponding relationships.

[0045] In this way, the unstructured fee list is converted into multiple structured fee elements and fee values, with each fee element having a corresponding fee value, thus forming a correspondence between fee elements and values.

[0046] S200: Randomly combine the multiple charging elements, perform rounding rate analysis and occurrence rate analysis, and obtain a rounding rate set and an occurrence rate set;

[0047] In this embodiment of the application, in order to accurately identify illegal charges such as splitting charges, it is necessary to randomly combine multiple charging elements and conduct analysis on the rounding rate and occurrence rate.

[0048] Specifically, one or more elements need to be randomly selected from multiple charging elements to form the first charging element group. At the same time, historical charging information sets with the same medical record information are extracted from historical charging data, and charging elements are extracted from them to form multiple historical charging element sets.

[0049] Furthermore, based on multiple historical fee element sets, the proportion of historical fee element sets containing the first fee element group within other fee elements is analyzed and calculated to obtain the first rounding rate. Simultaneously, the proportion of historical fee element sets containing the first fee element group is calculated to obtain the first occurrence rate.

[0050] By continuously and randomly combining multiple charging elements and repeating the above-mentioned rounding rate and occurrence rate analysis process, a set of rounding rates and occurrence rates are finally obtained, providing key data support for subsequent calculation of violation coefficients. This process utilizes historical data comparison and analysis to effectively identify abnormal situations in the combination of charging elements, improving the detection accuracy of violations such as decomposing charges.

[0051] Step S200 in the method provided in this application embodiment includes:

[0052] One or more charging elements are randomly selected from the plurality of charging elements as the first charging element group;

[0053] Within the historical billing data, extract the historical billing information set with the same medical record information, and extract the billing elements to obtain multiple historical billing element sets;

[0054] Based on the multiple historical charging element sets, analyze and obtain the first rounding rate and the first occurrence rate of the first charging element group;

[0055] The multiple charging elements are then randomly combined to perform rounding rate analysis and occurrence rate analysis, resulting in rounding rate set and occurrence rate set.

[0056] In this embodiment of the application, in order to identify illegal charges such as splitting charges, it is necessary to randomly select a combination from multiple charging elements. For example, from a set of elements including "nursing fee" and "blood pressure measurement fee", "blood pressure measurement fee" is randomly selected as the first charging element group.

[0057] At the same time, extract historical billing information sets with the same medical record information (such as patients diagnosed with "hypertension") from historical billing data, and extract billing elements from them to form multiple historical billing element sets.

[0058] Furthermore, based on the acquired sets of multiple historical charging elements, an integrated analysis is performed to obtain the first rounding rate and the first occurrence rate of the first charging element group.

[0059] The step of "analyzing and obtaining the rounding rate and occurrence rate of the first charging element group based on the plurality of historical charging element sets" in the method provided in this application includes:

[0060] Within the multiple sets of historical charging elements, the proportion of the first charging element group included in the other charging elements is extracted and calculated to obtain the first rounding rate;

[0061] Within the multiple historical charging element sets, the proportion of historical charging element sets that appear in the first charging element group is extracted and calculated to obtain the first occurrence rate.

[0062] In this embodiment of the application, in the process of analyzing illegal charges in the medical industry, in order to quantify and break down the possibility of illegal charges and other behaviors, it is necessary to conduct statistical analysis based on historical charging data.

[0063] Specifically, taking a hospitalized patient's billing element set as an example, if the first billing element group is "blood pressure measurement fee", it is necessary to extract 100 historical billing element sets containing "hospitalization nursing fee" and "blood pressure measurement fee" from the historical billing data of the same medical record information (such as "hypertension" diagnosis).

[0064] Within multiple historical fee sets, calculate the percentage of historical fee sets where "blood pressure measurement fee" is included within "nursing fee" (i.e., the number of historical fee sets with this combination / the total number of historical fee sets). If 80 out of 100 historical data sets show that "blood pressure measurement fee" is included within "nursing fee", then the first rounding rate is 80% (80 / 100).

[0065] At the same time, the percentage of historical fee element sets that include "blood pressure measurement fee" as an independent fee element is calculated (i.e., the number of historical fee element sets that include this combination / the total number of historical fee element sets). If "blood pressure measurement fee" is listed separately in 20 historical data sets, then the first occurrence rate is 20%.

[0066] Furthermore, multiple charging elements are continuously and randomly combined (such as selecting combinations like "anesthesia fee" and "instrument disinfection fee"), and the above-mentioned rounding rate and occurrence rate analysis process is repeated to finally obtain the rounding rate set and occurrence rate set, providing key data support for subsequent violation coefficient calculation.

[0067] S300: Based on the medical record information, perform charge prediction to obtain a set of predicted charge elements and a set of predicted charge values, and perform matching analysis with the multiple charge elements and multiple charge values ​​to obtain the element matching rate and the value matching rate.

[0068] In this embodiment of the application, in the process of analyzing illegal charges in the medical industry, in order to determine the reasonableness of the actual charges, it is necessary to predict reasonable charge elements and values ​​based on medical record information and match and analyze them with the actual charge data.

[0069] Specifically, the target user's medical record information (such as diagnosis results, treatment plans, etc.) is input into a pre-trained billing predictor. This predictor includes a billing element prediction branch and a billing value prediction branch, and can output a set of predicted billing elements and a set of predicted billing values ​​that match the medical record information.

[0070] Furthermore, the overlap ratio between the predicted charging element set and the actual multiple charging elements is calculated, i.e., the element matching rate. For example, if the predicted charging element set includes 5 elements such as "appendectomy" and "intravenous infusion", and the actual charging elements include 4 of them, then the element matching rate is 80% (4 / 5 × 100%).

[0071] At the same time, the similarity between the predicted charge value set and the actual multiple charge values ​​(i.e. the degree of difference between the total charge value) is calculated. For example, if the predicted total cost is 5,500 yuan and the actual total cost is 5,000 yuan, the numerical matching rate is calculated to be 90% (5,000 / 5,500×100%).

[0072] This process transforms violation judgments into quantifiable matching rate indicators by predictively matching medical record information with billing data, providing data basis for subsequent violation coefficient calculations and effectively improving the accuracy and efficiency of violation billing analysis.

[0073] Step S300 in the method provided in this application embodiment includes:

[0074] The medical record information is input into the billing predictor, and the prediction output obtains a billing element set and a predicted billing value set. The billing prediction network includes a billing element prediction branch and a billing value prediction branch.

[0075] Calculate the overlap ratio between the predicted charging element set and the multiple charging elements to obtain the element matching rate;

[0076] Calculate the similarity between the predicted charge value set and the multiple charge values ​​to obtain the value matching rate.

[0077] In this embodiment of the application, during the analysis of illegal charges in the medical industry, in order to determine the reasonableness of the actual charges, medical record information needs to be input into the charge prediction device. The prediction device analyzes the case information and outputs the corresponding set of predicted charge elements and set of predicted charge values ​​by setting charge element prediction branches and charge value prediction branches.

[0078] The construction steps of the "charge predictor" in the method provided in this application embodiment include:

[0079] Based on historical billing data, a set of sample medical record information is collected, and the set of billing elements and billing values ​​that appear most frequently in different sample medical record information are collected. Multiple sample predicted billing element sets and multiple sample predicted billing value sets are obtained by labeling.

[0080] Based on machine learning, a branch for predicting toll elements and a branch for predicting toll values ​​are constructed.

[0081] Using the sample medical record information set as input features, and using the multiple sample predicted charging element sets and multiple sample predicted charging value sets as output features, the charging element prediction branch and the charging value prediction branch are respectively trained under supervision. After the training converges, the charging predictor is obtained.

[0082] In this embodiment of the application, in order to solve the problem of traditional billing prediction methods relying on human experience and having low prediction accuracy, it is necessary to build a billing predictor trained on historical data in order to achieve accurate prediction of medical bills.

[0083] Specifically, the first step is to collect billing data from the hospital information system (HIS) over the past 10 years, covering medical record information from different departments, disease types, and treatment plans, to form a sample medical record information set.

[0084] Furthermore, for each sample medical record, the set of most frequently occurring chargeable elements is statistically analyzed. For example, for a "pneumonia" medical record, high-frequency chargeable elements may include "intravenous infusion", "antibiotic treatment fee", "chest X-ray examination", etc. At the same time, the numerical sets corresponding to these chargeable elements are collected, such as "intravenous infusion 50 yuan / time" and "antibiotic treatment fee 200 yuan / day".

[0085] Furthermore, through annotation by professional medical personnel, the sample prediction fee element set and value set corresponding to each sample medical record information are clarified, forming a structured training dataset that covers scenarios such as different disease severity, treatment cycle and medical insurance type.

[0086] Based on this, a billing prediction model is constructed using machine learning technology. Considering the complex relationship between medical record information and billing items, a deep learning architecture is adopted to construct billing element prediction branches and billing value prediction branches to achieve accurate prediction of billing items and amounts.

[0087] Specifically, the fee element prediction branch adopts a bidirectional LSTM network. The bidirectional LSTM can process medical record text sequences from both the forward and backward directions simultaneously, capturing long-distance dependencies in texts such as diagnostic descriptions and treatment records. Meanwhile, the attention mechanism can focus on key semantic segments, improving the targeting of feature extraction.

[0088] For example, for the diagnostic text "acute appendicitis with perforation", the model extracts key terms such as "acute", "appendicitis" and "perforation" through bidirectional LSTM, and then strengthens the weight of core words such as "appendicitis" and "perforation" through attention mechanism, and finally maps them to the corresponding charging element "appendectomy".

[0089] The billing prediction branch employs a combination of CNN and fully connected networks. CNN effectively extracts local patterns from the features of billing elements, while the fully connected network integrates global information. This branch fuses the semantic features of billing elements (such as the encoding of "appendectomy") with the numerical features in the medical record (such as patient age, length of hospital stay, and surgical grade), and predicts the specific amount of each billing element through multi-layer nonlinear transformations.

[0090] For example, for "appendectomy", the model combines features such as the patient's age (e.g., 30 years old) and length of hospital stay (5 days) to output a predicted cost of 5,000 yuan for the procedure.

[0091] During the model training phase, the sample medical record information set is used as the input feature. The 1000 sets of data are first split into training set and validation set in an 8:2 ratio. Then, the labeled sample prediction fee element set and numerical set are used as the output features to supervise the training of the two branches.

[0092] Specifically, the cross-entropy loss function is used to optimize the classification accuracy of the charging element prediction, ensuring that the model can accurately distinguish different charging items; the mean squared error loss function is used to improve the accuracy of the charging value prediction and reduce the deviation between the predicted amount and the actual amount; the Adam algorithm is selected for optimization, with an initial learning rate of 0.001 and a batch size of 64, to balance training speed and convergence effect.

[0093] Meanwhile, an early stopping mechanism is set up during the training process. If the loss on the validation set decreases by less than 0.5% for 80 consecutive rounds, the model training is considered to have stalled. To avoid resource consumption, early stopping is automatically triggered and the model is judged to have converged.

[0094] The resulting billing predictor can accurately predict the corresponding billing element set and value set based on the input medical record information, providing reliable predictive data support for the analysis of medical overcharging violations.

[0095] For example, by inputting medical record information for "acute appendicitis", the predictor can output charges and amounts such as "appendectomy 5,000 yuan" and "general anesthesia 2,000 yuan". By comparing with the actual charge data, it can effectively identify violations such as splitting charges and charging repeatedly.

[0096] After training the toll prediction engine, the prediction results need to be quantitatively compared with actual toll data to evaluate the accuracy of the prediction.

[0097] Specifically, the calculation of the element matching rate requires first standardizing the predicted charging element set and the actual multiple charging elements (such as unifying terminology and removing redundant items), and then calculating the overlap ratio through set intersection operation.

[0098] For example, if the predicted charge element set is ("appendectomy", "general anesthesia", "intravenous infusion"), and the actual charge element is ("appendectomy", "anesthesia fee", "intravenous infusion", "gauze fee"), then the overlapping elements are ("appendectomy" and "intravenous infusion"). Using the formula "element matching rate = overlapping elements / actual charge element", the element matching rate can be obtained as 2 / 4 × 100% = 50%.

[0099] Furthermore, the calculation of the numerical matching rate needs to comprehensively consider both the absolute and relative differences in the charging amounts. That is, the predicted charging data set and the actual charging data are normalized, and then the similarity is calculated using cosine similarity or Euclidean distance algorithms.

[0100] For example, if the predicted total cost is 7,500 yuan and the actual total cost is 8,000 yuan, the similarity calculated by Euclidean distance is 1-|7,500-8,000| / 8,000=0.9375, which means the numerical matching rate is 93.75%.

[0101] Quantitative analysis of element matching rate and numerical matching rate can intuitively reflect the degree of deviation between predicted and actual charges. When the element matching rate is lower than the industry threshold (e.g., 70%) and the numerical matching rate continues to decline, it may indicate the existence of charge breakdown (e.g., the addition of unpredicted items such as "instrument disinfection fee" to the actual charges) or price anomalies (e.g., the actual total cost exceeds the predicted range by more than 10%), providing key indicators for subsequent calculation of violation coefficients.

[0102] S400: Using the aforementioned element matching rate, rounding rate set, and occurrence rate set, a first violation coefficient is obtained; based on the aforementioned numerical matching rate, a second violation coefficient is obtained; and the violation coefficient is calculated as the analysis result.

[0103] In this embodiment of the application, in the process of analyzing illegal charges in the medical industry, in order to quantify the degree of illegality of the charging behavior, it is necessary to calculate the violation coefficient by comprehensively considering multiple dimensions of indicators.

[0104] Specifically, the set of rounding rates and the set of occurrence rates are first statistically processed, and the mean of the rounding rates is calculated as the first element correction coefficient. At the same time, the second element correction coefficient is obtained by "1 - mean of occurrence rate".

[0105] Furthermore, in conjunction with the element matching rate, the first violation coefficient is calculated, which is equal to the first element correction coefficient multiplied by (1 - element matching rate). This coefficient is used to reflect the likelihood of violations in the combination of chargeable elements.

[0106] Similarly, the second violation coefficient is calculated based on the element matching rate, that is, the second violation coefficient = the second element correction coefficient multiplied by (1 - element matching rate), which is used to reflect the degree of deviation of the fee amount.

[0107] Finally, based on the obtained first and second violation coefficients, a violation coefficient is calculated using a weighted formula, namely "Violation coefficient = a × first violation coefficient + b × second violation coefficient" (where a + b = 1, and a and b represent the weights of the element direction and numerical direction, respectively). This is used as the analysis result to determine the risk level of illegal charges, thereby providing a quantitative basis for the identification and supervision of illegal charges in the medical industry.

[0108] Step S400 in the method provided in this application embodiment includes:

[0109] Calculate the mean value of the rounding rate set to obtain the first element correction coefficient; calculate the mean value of the occurrence rate set; and subtract the mean value of the occurrence rate from 1 to obtain the second element correction coefficient.

[0110] The first violation coefficient is calculated based on the first element correction coefficient, the second element correction coefficient, and the element matching rate.

[0111] The second violation coefficient is calculated based on the numerical matching rate.

[0112] Based on the first violation coefficient and the second violation coefficient, the violation coefficient is calculated and used as the analysis result.

[0113] In this embodiment of the application, when analyzing illegal charges in the medical industry, in order to quantify the degree of illegality of the charging behavior, it is necessary to integrate indicators from multiple dimensions and perform hierarchical calculations to obtain the analysis results.

[0114] Specifically, the set of rounding rates and the set of occurrence rates are first statistically processed, and the mean of the rounding rate is calculated by the arithmetic mean method. This mean is then used as the first element correction coefficient, which can reflect the degree of integration of the charging element combination in historical data.

[0115] For example, if the rounding rates of a certain set of charging elements (such as "anesthesia fee" and "instrument disinfection fee") in 10 sets of historical charging data are 0.6, 0.7, 0.5, 0.8, 0.6, 0.7, 0.5, 0.6, 0.7, and 0.6 respectively, then the average rounding rate is (0.6+0.7+0.5+0.8+0.6+0.7+0.5+0.6+0.7+0.6) / 10 = 0.63, that is, the first element correction coefficient is 0.63. This indicates that in about 63% of the historical data, this combination of charging elements was integrated into other charging items.

[0116] Meanwhile, the second element correction coefficient is calculated by "Second Element Correction Coefficient = 1 - Average Occurrence Rate". This coefficient is used to reflect the abnormal occurrence of the combination of charging elements in historical charging scenarios.

[0117] For example, if the occurrence rates of the above-mentioned combination of charging elements in 10 sets of historical charging data are 0.3, 0.4, 0.3, 0.2, 0.3, 0.4, 0.3, 0.2, 0.3, and 0.4 respectively, and the average occurrence rate is (0.3+0.4+0.3+0.2+0.3+0.4+0.3+0.2+0.3+0.4) / 10 = 0.32, then the correction coefficient for the second element is 1-0.32 = 0.68, indicating that the combination appears frequently in historical scenarios and there is a possibility of anomalies.

[0118] Furthermore, the two correction coefficients mentioned above are correlated with the element matching rate, and the first violation coefficient is calculated using the formula "First violation coefficient = First element correction coefficient × (1 - element matching rate)".

[0119] For example, if the first element correction coefficient is 0.63 and the element matching rate is 70% (i.e., the overlap ratio between the predicted charging element and the actual charging element is 70%), then the first violation coefficient = 0.63 × (1 - 70%) = 0.63 × 0.3 = 0.189.

[0120] The higher the value, the greater the deviation between the actual charging elements and the predicted results, which may indicate that the complete medical service is broken down into multiple sub-items and charged repeatedly.

[0121] Furthermore, based on the numerical matching rate, the second violation coefficient is calculated by "second violation coefficient = 1 - numerical matching rate". This coefficient is used to quantify the degree of deviation between the charging amount and the predicted value.

[0122] For example, if the predicted total cost is 7500 yuan and the actual total cost is 8000 yuan, the numerical matching rate is 7500 / 8000 = 0.9375 (i.e., 93.75%), then the second violation coefficient = 1 - 0.9375 = 0.0625. The larger this value is, the greater the deviation of the charging amount from the predicted value, and the more likely there is an abnormal pricing violation.

[0123] Finally, based on the obtained first and second violation coefficients, the two-dimensional results are integrated using the weighted calculation formula "Violation Coefficient = a × First Violation Coefficient + b × Second Violation Coefficient" to calculate the overall violation coefficient. This coefficient will be used as the analysis result to quantify the degree of violation risk in the overall charging behavior.

[0124] In this equation, a + b = 1, where a and b represent the weights of the element direction and the numerical direction, respectively, and their values ​​depend on the regulatory focus. For example, when regulators are more concerned about whether there are violations in the combination of elements such as splitting charges or double billing in medical service fees, they tend to increase the weight of the element direction, a; conversely, if the regulatory focus is on the reasonableness of medical service prices and preventing abnormal deviations in charges, they will increase the weight of the numerical direction, b.

[0125] For example, in the scenario of medical service price supervision, since more attention is paid to the reasonableness of the charging amount, more emphasis is placed on the risk of violations in the numerical direction, and the weights of the two are set as a = 0.4 and b = 0.6. If the first violation coefficient is 0.189 and the second violation coefficient is 0.0625, then the comprehensive violation coefficient = 0.4 × 0.189 + 0.6 × 0.0625 = 0.0756 + 0.0375 = 0.1131, indicating that the risk of violations in this charging scenario is low.

[0126] If the first violation coefficient is 0.5 and the second violation coefficient is 0.4, then the comprehensive violation coefficient = 0.4 × 0.5 + 0.6 × 0.4 = 0.2 + 0.24 = 0.44, reaching the industry warning threshold (the value is determined based on industry standards and historical data statistics, such as 0.4). This indicates that a thorough investigation is needed to verify whether there are violations such as splitting charges or double billing. This coefficient allows for a direct comparison of the severity of violations in different billing cases, providing a quantitative basis for medical supervision decisions.

[0127] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0128] This application provides a method for analyzing illegal charges in the medical industry. First, it acquires the target user's medical record information and billing information, extracting multiple billing elements and values ​​from the billing information to provide multi-dimensional data support for subsequent analysis. Second, it randomly combines multiple billing elements and performs rounding rate and occurrence rate analysis to obtain rounding rate sets and occurrence rate sets, thereby identifying illegal behaviors such as billing decomposition. Then, it predicts charges based on medical record information and matches the prediction results with actual billing data to obtain element matching rate and value matching rate, quantifying the deviation between actual and reasonable charges. Finally, it calculates a violation coefficient using multi-dimensional indicators as the analysis result, providing a quantitative basis for identifying illegal charges. This method effectively integrates multi-dimensional indicators, achieving accurate analysis of illegal medical charges and improving the efficiency and reliability of medical charge supervision.

[0129] The method provided in this application solves the problems of low efficiency and poor accuracy of traditional manual verification. Through the steps of "data acquisition - combined analysis - prediction matching - coefficient calculation", it improves the accuracy of identifying hidden violations such as splitting charges, avoids the problem that illegal charges are difficult to detect due to the high professionalism of medical services, and provides a scientific and quantitative solution for medical charge supervision.

[0130] Example 2, as shown in the appendix Figure 2 As shown, based on the inventive concept of the method for analyzing illegal charges in the medical industry provided in Embodiment 1, this application also provides an analysis system for illegal charges in the medical industry, specifically including:

[0131] Data information extraction module 01 is used to obtain the medical record information and billing information of the target user, and extract multiple billing elements and multiple billing values ​​from the billing information;

[0132] The element combination analysis module 02 is used to randomly combine the multiple charging elements, perform rounding rate analysis and occurrence rate analysis, and obtain rounding rate set and occurrence rate set;

[0133] The billing prediction and matching module 03 is used to predict the billing based on the medical record information, obtain a set of predicted billing elements and a set of predicted billing values, and perform matching analysis with the multiple billing elements and multiple billing values ​​to obtain the element matching rate and the value matching rate.

[0134] The violation coefficient calculation module 04 is used to process the element matching rate, rounding rate set, and occurrence rate set to obtain the first violation coefficient, process the numerical matching rate to obtain the second violation coefficient, and calculate the violation coefficient as the analysis result.

[0135] In one embodiment, the data information extraction module 01 is further configured to:

[0136] Obtain the target user's medical records and billing information;

[0137] Extract the charging elements and charging values ​​from the charging information to obtain multiple charging elements and multiple charging values, wherein the charging elements and charging values ​​correspond to each other.

[0138] In one embodiment, the element combination analysis module 02 is also used for:

[0139] One or more charging elements are randomly selected from the plurality of charging elements as the first charging element group;

[0140] Within the historical billing data, extract the historical billing information set with the same medical record information, and extract the billing elements to obtain multiple historical billing element sets;

[0141] Based on the multiple historical charging element sets, analyze and obtain the first rounding rate and the first occurrence rate of the first charging element group;

[0142] The multiple charging elements are then randomly combined to perform rounding rate analysis and occurrence rate analysis, resulting in rounding rate set and occurrence rate set.

[0143] In one embodiment, the charge prediction matching module 03 is also used for:

[0144] The medical record information is input into the billing predictor, and the prediction output obtains a billing element set and a predicted billing value set. The billing prediction network includes a billing element prediction branch and a billing value prediction branch.

[0145] Calculate the overlap ratio between the predicted charging element set and the multiple charging elements to obtain the element matching rate;

[0146] Calculate the similarity between the predicted charge value set and the multiple charge values ​​to obtain the value matching rate.

[0147] In one embodiment, the violation coefficient calculation module 04 is also used for:

[0148] The matching rate, rounding rate set, and occurrence rate set are processed to obtain the first violation coefficient. The numerical matching rate is then processed to obtain the second violation coefficient. The final violation coefficient is calculated and presented as the analysis result, including:

[0149] Calculate the mean value of the rounding rate set to obtain the first element correction coefficient; calculate the mean value of the occurrence rate set; and subtract the mean value of the occurrence rate from 1 to obtain the second element correction coefficient.

[0150] The first violation coefficient is calculated based on the first element correction coefficient, the second element correction coefficient, and the element matching rate.

[0151] The second violation coefficient is calculated based on the numerical matching rate.

[0152] Based on the first violation coefficient and the second violation coefficient, the violation coefficient is calculated and used as the analysis result.

[0153] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0154] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0155] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method of analyzing medical industry violation charges, characterized by, The method comprises: obtaining medical record information and charging information of a target user, extracting a plurality of charging elements and a plurality of charging values in the charging information; randomly combining the plurality of charging elements, performing integer ratio analysis and occurrence rate analysis, and obtaining an integer ratio set and an occurrence rate set; performing charging prediction according to the medical record information, obtaining a predicted charging element set and a predicted charging value set, and performing matching analysis with the plurality of charging elements and the plurality of charging values to obtain an element matching rate and a value matching rate; using the element matching rate, the integer ratio set and the occurrence rate set to obtain a first violation coefficient, using the value matching rate to obtain a second violation coefficient, and calculating a violation coefficient as an analysis result.

2. The method of claim 1, wherein, obtaining medical record information and charging information of a target user, extracting a plurality of charging elements and a plurality of charging values in the charging information, comprising: obtaining medical record information and charging information of a target user; extracting charging elements and charging values in the charging information to obtain a plurality of charging elements and a plurality of charging values, wherein the charging elements and the charging values correspond to each other.

3. The method of claim 1, wherein, randomly combining the plurality of charging elements, performing integer ratio analysis and occurrence rate analysis, and obtaining an integer ratio set and an occurrence rate set, comprising: randomly selecting one or more charging elements in the plurality of charging elements as a first charging element group; extracting a historical charging information set with the same medical record information from historical charging data, and extracting charging elements to obtain a plurality of historical charging element sets; obtaining a first integer ratio and a first occurrence rate of the first charging element group according to the plurality of historical charging element sets; continuing to randomly combine the plurality of charging elements, performing integer ratio analysis and occurrence rate analysis, and obtaining an integer ratio set and an occurrence rate set.

4. The method of claim 3, wherein, obtaining an integer ratio and an occurrence rate of the first charging element group according to the plurality of historical charging element sets, comprising: extracting the proportion of historical charging element sets including the first charging element group in other charging elements from the plurality of historical charging element sets to obtain a first integer ratio; extracting the proportion of historical charging element sets in which the first charging element group appears from the plurality of historical charging element sets to obtain a first occurrence rate.

5. The method of claim 1, wherein, performing charging prediction according to the medical record information, obtaining a predicted charging element set and a predicted charging value set, and performing matching analysis with the plurality of charging elements and the plurality of charging values to obtain an element matching rate and a value matching rate, comprising: inputting the medical record information into a charging predictor to obtain a charging element set and a predicted charging value set, wherein the charging prediction network comprises a charging element prediction branch and a charging value prediction branch; calculating the coincidence rate of the predicted charging element set and the plurality of charging elements to obtain an element matching rate; calculating the similarity of the predicted charging value set and the plurality of charging values to obtain a value matching rate.

6. The method of claim 5, wherein, The construction steps of the charging predictor comprise: According to the charging data in the historical time, a sample medical record information set is collected, and a charging element set and a charging value set with the highest occurrence frequency in different sample medical record information are collected, and a plurality of sample predicted charging element sets and a plurality of sample predicted charging value sets are labeled and obtained; Based on machine learning, a charge element prediction branch and a charge value prediction branch are constructed; The sample medical record information set is used as input features, and the multiple sample predicted charge element sets and multiple sample predicted charge value sets are used as output features, respectively. The charge element prediction branch and the charge value prediction branch are supervised trained, and the charge predictor is obtained after training convergence.

7. The method of claim 1, wherein, The element matching rate, the set of rounding rates, and the set of occurrence rates are used to obtain a first violation coefficient. The second violation coefficient is obtained according to the numerical matching rate. The violation coefficient is calculated and obtained as an analysis result, including: The rounding rate mean of the set of rounding rates is calculated to obtain a first element correction coefficient. The occurrence rate mean of the set of occurrence rates is calculated, and the second element correction coefficient is obtained by subtracting the occurrence rate mean from 1. The first element correction coefficient, the second element correction coefficient, and the element matching rate are used to calculate and obtain a first violation coefficient. The second violation coefficient is obtained according to the numerical matching rate. The violation coefficient is calculated and obtained according to the first violation coefficient and the second violation coefficient as an analysis result.

8. An analysis system for medical industry violation charges, characterized by, The system is used to execute the analysis method of the medical industry violation charge of any one of claims 1-7. The system comprises: A data information extraction module is used to obtain medical record information and charge information of a target user, and extract multiple charge elements and multiple charge values in the charge information; An element combination analysis module is used to randomly combine the multiple charge elements, perform rounding rate analysis and occurrence rate analysis, and obtain a set of rounding rates and a set of occurrence rates; A charge prediction matching module is used to perform charge prediction according to the medical record information, obtain a predicted charge element set and a predicted charge value set, and perform matching analysis with the multiple charge elements and the multiple charge values to obtain an element matching rate and a numerical matching rate; A violation coefficient calculation module is used to obtain a first violation coefficient by processing the element matching rate, the set of rounding rates, and the set of occurrence rates. The second violation coefficient is obtained according to the numerical matching rate. The violation coefficient is calculated and obtained as an analysis result.

Images