A fee charging model and method based on artificial intelligence surgery cost monitoring

Abstract

This invention discloses a charging model and method for surgical cost monitoring based on artificial intelligence, belonging to the field of surgical cost monitoring technology. The charging model and method based on artificial intelligence for surgical cost monitoring include: constructing a multi-source data knowledge base related to surgical charges; acquiring basic information related to the patient's surgery; generating a pre-operative prediction list; collecting relevant data during surgery in real time; generating abnormal early warning information; automatically generating an accurate settlement list; and constructing a cost detail traceability link. The model includes: a knowledge base construction module, a prediction module, a real-time monitoring module, and a settlement traceability module. This invention solves the problems of poor real-time performance of surgical monitoring, low efficiency of post-operative settlement, and large errors in existing technologies. This invention can ensure settlement accuracy and achieve full-link traceability of cost details from the final settlement items to the pre-operative prediction basis, surgical usage records, and charging standard basis.

Description

Technical Field

[0001] This invention relates to the field of surgical cost monitoring technology, specifically to a charging model and method for surgical cost monitoring based on artificial intelligence. Background Technology

[0002] In the current medical service system, the settlement and management of surgical fees is one of the core links in medical fee management. Its accuracy and compliance are directly related to the operational efficiency of medical institutions, the safety of medical insurance funds, and the vital interests of patients.

[0003] The current surgical fee management model relies primarily on manual processes. The specific workflow is as follows: Before surgery, medical staff, based on the patient's diagnosis, determine the surgical and anesthesia methods, estimate the required consumables and medications, and generate a preliminary cost list; during surgery, operating room nurses manually record the actual consumables, medications, and surgical procedures used; after surgery, medical staff integrate the intraoperative records with the preoperative list, verify them, generate a billing slip and anesthesia billing slip, and submit them for settlement. This model has several shortcomings: (1) Poor real-time monitoring during surgery: Dynamic information such as the use of consumables, medications, and changes in surgical procedures during surgery needs to be recorded manually in real time. Due to busy surgery schedules and human negligence, records are easily missed or misrecorded. Furthermore, it is impossible to compare the actual usage information with the pre-operative prediction list in real time, making it difficult to detect potential abnormal charges in a timely manner.

[0004] (2) Low efficiency and large error in postoperative settlement: After surgery, it is necessary to manually integrate multiple sources of information such as the preoperative prediction list, the intraoperative record, and the billing catalog. The verification process is cumbersome and time-consuming. Furthermore, manual verification is prone to problems such as omissions, errors, and duplicate charges in the settlement list due to the complexity of information and incorrect coding associations, which affects the accuracy of settlement.

[0005] Therefore, since it does not meet the existing needs, we propose a charging model and method based on artificial intelligence for surgical cost monitoring. Summary of the Invention

[0006] The purpose of this invention is to provide a billing model and method based on artificial intelligence for surgical cost monitoring. This model can accurately identify the list of medical consumables and drugs required for surgery, generate an accurate preliminary cost budget, quickly detect potential abnormal charges and provide timely warnings, facilitating timely handling by medical staff. It can also quickly generate an accurate settlement list including surgical and anesthesia bills, significantly reducing the workload of manual integration and verification, and improving settlement efficiency. Furthermore, by accurately identifying cases of misuse of medication and overcharging and marking disputed items, it ensures the accuracy of settlement, protects the legitimate rights and interests of patients and medical institutions, and solves the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a billing method based on artificial intelligence for surgical cost monitoring, comprising the following steps: S1: Construct a multi-source data knowledge base related to surgical fees; S2: Obtain basic information related to the patient's surgery and perform standardized preprocessing on the basic information; S3: Based on the preprocessed basic information and multi-source data knowledge base, generate a preoperative prediction list; S4: Real-time data collection of surgical consumables usage, drug usage, and surgical operation dynamics during surgery; comparison and verification of real-time data with the pre-operative prediction list; and generation of abnormal warning information. S5: After the surgery, integrate the actual data confirmed by the monitoring during the surgery and the surgical operation record to automatically generate an accurate settlement list, which includes the surgical fee bill and the anesthesia fee bill. S6: Store data records for the entire process before, during, and after surgery to build a cost detail traceability chain.

[0008] Preferably, the construction of the multi-source data knowledge base related to surgical fees specifically includes: Collect basic medical information, surgical-related coding information, surgical fee catalogs for various regions, and dynamic medical service information during surgery for surgical patients; Remove duplicate, missing, and abnormal data from the collected data; complete missing coding information and charging standard information; and standardize and convert date and amount data with inconsistent formats. Establish a mapping table that associates diagnostic codes with codes for common surgical procedures and anesthesia methods; Establish a mapping table linking surgical and anesthesia method codes with corresponding charge items in the surgical charge catalogs of various regions; Determine the standard fee item list corresponding to different code combinations, classify and store the standardized data in the knowledge base, and build an associated index.

[0009] Preferably, the S3 implementation process specifically includes: Feature extraction was performed on the patient's preoperative basic medical information to obtain the core feature vector; Using historical surgical fee case data in a standardized knowledge base as training samples, the core feature vectors in the samples are taken as input, and the corresponding standard fee item list and cost budget are taken as output to train the deep learning model. The model parameters are adjusted through cross-validation to optimize the prediction accuracy of the model and obtain the prediction model. Input the patient's preoperative core feature vector into the trained prediction model, call the encoding mapping table in the standardized knowledge base and the corresponding regional charging standards, and output the preoperative prediction charging item list, the estimated cost of each item and the total cost budget.

[0010] Preferably, the S4 implementation process specifically includes: Based on the coding mapping rules and charging standards in the standardized knowledge base, the real-time data is compared and verified with the preoperative predicted charging item list; A deviation threshold is set, which is determined based on statistical analysis of historical surgical data and clinical guidelines. The deviation value between real-time data and the prediction list is calculated. When the deviation between real-time data and the pre-judgment list exceeds the threshold, or when consumables, drugs, or procedures not included in the pre-operative pre-judgment list are found, it is judged as abnormal, and a graded warning is immediately triggered and pushed to the operating room nurses, anesthesiologists, and billing management terminals in real time. For any reasonable additional consumables, medications, or procedures confirmed during surgery, medical staff will review and confirm them before including them in the real-time usage list and updating the corresponding cost information.

[0011] Preferably, the comparison and verification of real-time data with the pre-operative predicted charge list specifically includes: Determine whether the consumables, drugs, surgical procedures, anesthesia procedures, and blood products actually used during the operation are on the pre-operative prediction list. If new items are added outside the prediction list, mark them as new items to be verified. If items on the pre-operative prediction list do not appear in the real-time data during the operation, mark them as prediction items not used and to be verified. Verify that the diagnostic codes, surgical codes, and anesthesia codes corresponding to the actual items used in the surgery are consistent with the preoperative prediction and comply with the coding requirements of the charging catalogs of each region. If there is a coding mismatch or coding error, mark it as a coding error to be corrected. By combining the patient's surgical type, weight, surgical duration, and historical case data in the standardized knowledge base, it is determined whether the actual usage of consumables, drugs, and blood products during the surgery is within a reasonable range. If it exceeds the reasonable usage threshold, it is marked as abnormal usage pending verification. Verify whether the charges for the actual procedures used during the surgery conform to the surgical fee catalog of the corresponding region, and whether there are any cases of overcharging or double charging. If there are any illegal charges, mark them as illegal charges pending correction.

[0012] Preferably, the S5 implementation process specifically includes: Collect the actual list of consumables used, the actual list of drugs used, the actual surgical operation record, and the actual anesthesia operation record as confirmed by monitoring during the operation. At the same time, retrieve the preliminary cost budget before the operation, the regional fee catalog database, and the basic coding database information. Based on the fee standards in the regional fee catalog, the costs of consumables, medicines, surgical procedures and anesthesia procedures actually used are calculated, and the detailed costs of each item and the total cost are calculated. Based on the basic coding library, clinical association rule library and regional charging catalog library, the reasonableness of the calculated drug usage and charging items is verified; The system automatically generates accurate settlement lists that include surgical and anesthesia bills. These lists are then sent to medical staff and billing administrators for manual review. Once approved, the final settlement list is generated.

[0013] Preferably, the construction method of the cost detail tracing link specifically includes: The data for the entire process is categorized and collected according to the time sequence of pre-operative, intra-operative, and post-operative periods, as well as the dimensions of patients, surgery, and billing. A unique surgical cost traceability code is generated for each surgical patient, and the generated traceability code is linked and bound one by one with the collected full-process data; Based on the associated full-process data, a two-way structured traceability link is constructed, which includes forward time-series tracing and reverse cost tracing. The entire process data, after being linked and coded, is stored in an encrypted distributed database in accordance with medical data management standards.

[0014] Preferably, the step of comparing and verifying real-time data with the pre-operative predicted charge list includes the step of generating an abnormal risk index for surgical costs using a dynamic calculation model, specifically including: A real-time monitoring time window is activated, and the sampling frequency is adaptively adjusted according to the surgical level. The calculation parameters are captured in parallel through a high-speed data bus and then standardized. Obtain the cumulative surgical cost deviation value within the current monitoring time window, the current surgical duration, the standard estimated total duration of similar surgeries, and the surgical complexity coefficient of the patient; The process of obtaining the cumulative surgical cost deviation value includes: obtaining the actual cost incurred at the current moment, and executing an outlier filtering mechanism to remove instantaneous jump data where the amount of a single transaction exceeds a set threshold; and extracting the budgeted cost from the cost budget curve that evolves over time based on the current percentage of surgical progress using an interpolation algorithm. The process of obtaining the standard estimated total time for the same type of surgery includes: parsing the patient's diagnosis and surgical code, retrieving historical records from the knowledge base, and using the truncated average method to remove the longest and shortest preset proportion samples in the time distribution and then calculating the arithmetic mean to obtain the standard estimated total time for the same type of surgery. The process of obtaining the surgical complexity coefficient of the patient includes: using natural language processing technology to parse electronic medical records, extracting feature vectors including BMI, ASA classification, CCI index and previous surgical history, and inputting them into a logistic regression scoring model for calculation; Based on the multi-source data knowledge base, retrieve the historical cost fluctuation variance of the same type of surgery at the current time. The abnormal risk index of the surgical cost is calculated based on the following formula: in, The dimensionless surgical cost abnormality risk index; The actual cost incurred is included in the cumulative cost deviation value of the surgery. The budgeted cost corresponding to the preoperative estimated charges list; The duration of the current surgery. The standard estimated total duration for the same type of surgery; The surgical complexity coefficient is used to characterize the degree to which the patient's individual physical signs and underlying diseases interfere with the standardized surgical procedure. For the normalized time progression factor, and ; As the cost deviation weighting factor, As the time deviation weighting factor, This is the time sensitivity decay constant, used to adjust the tolerance for cost deviations in the later stages of surgery; To prevent the minimum compensation constant where the denominator is zero, is the base of the natural logarithm. For natural logarithm operations; Calculate dynamic alarm thresholds ,in This is the historical average. For historical variance, Confidence coefficient; The calculated abnormal risk index of surgical costs With the dynamic alarm threshold Real-time comparison is performed; if the surgical cost is abnormal, the risk index is considered. Greater than the dynamic alarm threshold The system determines that the current cost increase is due to a non-technical anomaly and generates the anomaly warning information.

[0015] Preferably, before the S5 automatically generates the accurate settlement list, a logical consistency verification step based on the semantic topology graph is further included, specifically including: Based on the clinical pathway rules and pricing standards in the aforementioned multi-source data knowledge base, a directed attribute graph is constructed. A topological diagram of surgical medical entity behavior; in which, For a set of nodes, The set of edges; the set of nodes Includes operational entity nodes representing specific medical actions. , representing the physical nodes of chargeable medical consumables and the physical nodes representing chargeable drugs edge set This includes directed edges that represent strong dependencies that indicate the existence of prerequisite relationships between entities. and mutually exclusive undirected edges representing entities that cannot appear simultaneously. ; Using a pre-trained BERT-Medical deep language model, word segmentation and named entity recognition were performed on the surgical operation records, anesthesia records, and nursing records. The unstructured text was then transformed into standardized action feature vectors through a semantic encoding layer. ; and encode each charge item in the accurate settlement list to generate a charge item feature vector. ; The extracted action feature vectors With the feature vectors of each of the aforementioned fee items Mapped to the surgical medical entity behavior topology graph, the semantic similarity between the vector and the corresponding node is calculated. When the semantic similarity exceeds a preset threshold, the corresponding node state is marked as active, thereby forming the current surgical actual execution subgraph. ; The current surgical procedure execution subgraph Then, run a depth-first search algorithm to perform connectivity analysis and isolated node detection, and execute the following decision logic: If a consumable entity node is active in the current surgical execution subgraph, and the algorithm traces back along all its strong dependent directed edges to find that all its predecessor operation entity nodes are inactive, it is determined that the corresponding billing item has a logical violation without operational basis. If two nodes are active in the current surgical execution subgraph, and there is a mutually exclusive undirected edge between the two nodes in the surgical medical entity behavior topology graph, it is determined that the corresponding charging item has a logical violation of mutual exclusion conflict. Check the timestamp attributes of the active nodes at both ends of the strongly dependent edge. If the timestamp of the predecessor operation entity node is later than the timestamp of the successor node, it is determined to be a timing logic violation. The charging items that are determined to be logically irregular are marked as high-risk items to be verified, and a logical traceability view is loaded into the generated accurate settlement list. The irregular logical links are identified by highlighting the irregular items and the broken dependency relationships.

[0016] A billing model based on artificial intelligence-based surgical cost monitoring, applied in a billing method based on artificial intelligence-based surgical cost monitoring, includes: The knowledge base construction module is used to standardize various types of data collected by the data acquisition module and build a multi-dimensional standardized knowledge base. The prediction module uses deep learning algorithms to build a model for predicting charges and costs, and combines this with the patient's preoperative diagnosis and treatment information to output accurate preoperative charge prediction results. The real-time monitoring module is used to collect dynamic medical service information during surgery in real time, compare and verify it with the pre-operative prediction results, and realize real-time early warning and correction of billing risks. The settlement traceability module is used to integrate the actual data verified during the operation, generate an accurate settlement list, and realize full-process data traceability.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention uses a deep learning model to construct a preoperative prediction model. Combined with feature engineering and clinical association rule constraints, it can accurately identify the list of medical consumables and drugs required for surgery, generate an accurate preliminary cost budget, effectively reduce prediction deviations caused by differences in human experience, reduce the problem of consumable redundancy or shortage, and lay the foundation for subsequent identification of overcharging and misuse of drugs. Through multi-source intelligent devices, it collects dynamic data on the use of consumables and drugs and surgical operations in real time, and compares the real-time data with the preoperative prediction list in real time. It can quickly detect abnormal charging risks and issue timely warnings, which is convenient for medical staff to deal with them in a timely manner. 2. This invention automatically integrates actual surgical data and combines it with regional fee schedules for cost calculation and verification of medication and charge reasonableness. It can quickly generate accurate settlement lists including surgical and anesthesia bills, significantly reducing the workload of manual integration and verification, and improving settlement efficiency. At the same time, by accurately identifying cases of misuse of medication and overcharging and marking disputed items, it ensures the accuracy of settlement and protects the legitimate rights and interests of patients and medical institutions. By retaining data records of the entire process before, during, and after surgery, and assigning a unique traceability code to each patient, it achieves full-chain traceability of cost details from the final settlement items to the basis for preoperative prediction, surgical usage records, and charging standards, facilitating the handling of disputes over misuse of medication and overcharging and clinical quality control. Attached Figure Description

[0018] Fig. 1 This is a schematic diagram of a billing method based on artificial intelligence-based surgical cost monitoring according to the present invention; Fig. 2 This is a schematic diagram of a charging model based on artificial intelligence for surgical cost monitoring according to the present invention; Fig. 3 This is a diagram of the settlement list generation method for the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To address the issues of poor real-time performance of surgical monitoring, low post-operative settlement efficiency, and large errors in existing technologies, please refer to [link / reference]. Figs. 1-3 This embodiment provides the following technical solution: A billing method based on artificial intelligence for surgical cost monitoring includes the following steps: S1: Construct a multi-source data knowledge base related to surgical charges, which includes a basic coding library, a regional charging catalog library, a clinical association rule library, and a historical surgical data warehouse; S2: Obtain basic information related to the patient's surgery and perform standardized preprocessing on the basic information; S3: Based on the preprocessed basic information and multi-source data knowledge base, generate a preoperative prediction list; S4: Real-time data collection of surgical consumables usage, drug usage, and surgical operation dynamics during surgery; comparison and verification of real-time data with the pre-operative prediction list; and generation of abnormal warning information. S5: After the surgery, integrate the actual data confirmed by the monitoring during the surgery and the surgical operation record to automatically generate an accurate settlement list, which includes the surgical fee bill and the anesthesia fee bill. S6: Store data records for the entire process before, during, and after surgery to build a detailed cost traceability chain and ensure the accuracy and compliance of settlement.

[0021] Constructing a multi-source data knowledge base related to surgical fees, specifically including: Collect basic medical information, surgical-related coding information, surgical fee catalogs for various regions, and dynamic medical service information during surgery for surgical patients; Remove duplicate, missing, and abnormal data from the collected data; complete missing coding information and charging standard information; and standardize and convert date and amount data with inconsistent formats. Establish a mapping table between the diagnostic codes of the Chinese medical insurance version of ICD-10 and the codes of common surgical procedures and anesthesia methods; Establish a mapping table between ICD-9-CM-3 surgical and anesthesia method codes and corresponding charge items in the surgical charge catalogs of various regions; Determine the standard fee item list corresponding to different code combinations, and classify and store the standardized diagnostic code information, surgical and anesthesia code information, fee standard information of various regions, and historical surgical fee case data into the knowledge base, and build an association index containing diagnosis, surgical method, anesthesia and fee items.

[0022] The implementation process of S3 specifically includes: Feature extraction was performed on the patient's preoperative basic diagnostic and treatment information to obtain a core feature vector. The core feature vector includes ICD-10 diagnostic coding features, ICD-9-CM-3 coding features of the proposed surgery, ICD-9-CM-3 coding features of the proposed anesthesia method, patient age, type of underlying disease, and surgical risk level. Using historical surgical fee case data from a standardized knowledge base as training samples, the core feature vectors in the samples are taken as input, and the corresponding standard fee item list (including surgical consumables, drugs, surgical procedures, anesthesia procedures, blood products, etc.) and cost budget are taken as output. The deep learning model is trained, and the model parameters are adjusted through cross-validation to optimize the prediction accuracy of the model and obtain the prediction model. The deep learning model adopts a fusion model of convolutional neural network and long short-term memory network. The convolutional neural network is used to extract key encoded association features in the feature vector, and the long short-term memory network is used to mine the temporal association pattern between fee items and patient features in historical cases. The patient's preoperative core feature vector is input into the trained prediction model. The model calls the coding mapping table in the standardized knowledge base and the corresponding regional charging standards to output a list of preoperative predicted charges, the estimated cost of each item, and the total cost budget. The list of preoperative predicted charges includes the ICD code, the basis for the charging standard, and the estimated usage for each item. At the same time, the prediction result description is generated, and high-risk charges (items that are easy to miss or miscalculate) are marked.

[0023] The S4 implementation process specifically includes: Based on the coding mapping rules and charging standards in the standardized knowledge base, the real-time data is compared and verified with the preoperative predicted charging item list; A deviation threshold is set, which is determined based on statistical analysis of historical surgical data and clinical guidelines. The deviation value between real-time data and the prediction list is calculated. When the deviation between real-time data and the pre-judgment list exceeds the threshold, or when consumables, drugs, or procedures not included in the pre-operative pre-judgment list are found, it is judged as abnormal, and a graded warning is immediately triggered and pushed to the operating room nurses, anesthesiologists, and billing management terminals in real time. For any reasonable additional consumables, medications, or procedures confirmed during surgery, medical staff will review and confirm them before including them in the real-time usage list and updating the corresponding cost information.

[0024] The tiered early warning system includes: Level 1 (minor anomalies, such as unused predicted items) alerts operating room nurses or anesthesiologists via a system pop-up; Level 2 (moderate anomalies, such as coding errors or newly added items) alerts via pop-up and simultaneously notifies the department's billing administrator; Level 3 (serious anomalies, such as large overcharges or unrecorded misuse of high-value consumables) immediately triggers an audible and visual alarm and pushes the warning information to the hospital's medical insurance management department. After receiving the warning, medical staff can upload verification explanations or correction suggestions through the system. The corrected information is automatically synchronized to the module, completing the closed-loop processing of abnormal information. The real-time data is compared and verified with the pre-operative predicted charge list, specifically including: Determine whether the consumables, drugs, surgical procedures, anesthesia procedures, and blood products actually used during the operation are on the pre-operative prediction list. If new items are added outside the prediction list, mark them as new items pending verification. If items on the pre-operative prediction list do not appear in the real-time data during the operation, mark them as predicted items not used pending verification. Verify that the ICD-10 diagnostic code, ICD-9-CM-3 surgical and anesthesia code corresponding to the actual items used in the surgery are consistent with the preoperative prediction and meet the coding requirements of the charging catalog of each region. If there is a coding mismatch or coding error, mark it as a coding error to be corrected. By combining the patient's surgical type, weight, surgical duration, and historical case data in the standardized knowledge base, it is determined whether the actual usage of consumables, drugs, and blood products during the surgery is within a reasonable range. If it exceeds the reasonable usage threshold, it is marked as abnormal usage pending verification. Verify whether the charges for the actual procedures used during the surgery conform to the surgical fee catalog of the corresponding region, and whether there are any cases of overcharging or double charging. If there are any illegal charges, mark them as illegal charges pending correction.

[0025] The S5 implementation process specifically includes: Collect the actual list of consumables used, the actual list of drugs used, the actual surgical operation record, and the actual anesthesia operation record as confirmed by monitoring during the operation. At the same time, retrieve the preliminary cost budget before the operation, the regional fee catalog database, and the basic coding database information. Based on the fee standards in the regional fee catalog, the costs of consumables, drugs, surgical procedures, and anesthesia procedures actually used are calculated, including detailed costs and total costs for each item. Specifically, for surgical procedure fees and anesthesia procedure fees, the corresponding fee standards are matched according to the ICD-9-CM-3 code; for consumables and drug costs, the costs are calculated based on the actual quantity used and the unit price; for blood products and other special items, the costs are calculated according to the regional fee standards. Based on the basic coding library, clinical association rule library, and regional fee catalog library, the reasonableness of the calculated drug usage and fee items is verified, including: verifying the correlation between drug use and patient diagnosis and surgical procedure, and determining whether there is any misuse of drugs such as off-label use or overdosing; verifying whether there is any overcharge of fee items such as duplicate charges, overcharge of standard charges, or charges without actual usage basis; and marking disputed items suspected of misuse of drugs or overcharge. It automatically generates an accurate settlement list containing surgical and anesthesia bills. The settlement list includes patient information, surgical information, details of each charge item, total cost, medication, and explanation of the reasonableness of the charges. The generated accurate settlement list is pushed to medical staff and billing management personnel for manual review. The focus is on reviewing and processing the marked items of misuse of drugs and disputed charges. Once the review is approved, a final settlement list is generated. If the review fails, the problematic link is located and the calculation is recalculated based on the located link.

[0026] The construction method for the cost detail traceability chain specifically includes: The entire process data is categorized and collected according to the time sequence of preoperative, intraoperative, and postoperative periods, as well as the dimensions of patient, surgical, and billing operations. Preoperative data collection includes basic patient information, standardized pre-processed diagnostic / surgical / anesthesia coding information, a preoperative predicted consumables list, a preoperative predicted medication list, and a preliminary cost budget. Intraoperative data collection includes real-time records of consumable requisition / usage / remaining stock, medication requisition / dispensing / dosage records, surgical operation logs, anesthesia operation logs, comparison and verification records between real-time data and the predicted list, abnormal warning triggering and handling records, and medical staff review and confirmation records. Postoperative data collection includes an integrated list of actual consumable usage, an actual medication usage list, detailed cost accounting, medication and billing rationality verification records, a precise settlement list, and manual review records. During the collection process, the system automatically adds a timestamp, module identifier, and patient association identifier to each data entry to ensure data location. A unique surgical cost traceability code is generated for each surgical patient. The coding rule adopts a combination of hospital identifier, surgery date, patient unique identifier, and random check code to ensure the uniqueness and readability of the code. The generated traceability code is linked and bound to the collected full-process data one by one. The specific association methods include: embedding a traceability code field in each data record; establishing a mapping index table between the traceability code and each data category; and determining the data storage path for each stage of the pre-operative / intra-operative / post-operative process corresponding to the traceability code. Through this association mechanism, a data association foundation is achieved where one patient, one code, and one code connects the entire process.

[0027] Based on the associated full-process data, a bidirectional structured traceability link is constructed, consisting of forward time-series tracing and reverse cost tracing. The forward time-series tracing link uses the operation time as the axis and connects key data from each stage in the order of preoperative prediction, operation, and postoperative settlement, clearly presenting the complete process of surgical cost formation. The reverse cost tracing link starts with the specific chargeable items in the final settlement list and reversely associates the corresponding actual usage records during the operation, the basis for preoperative prediction, the basis for charging standards, and the coding association mapping relationship, realizing full-link traceability from chargeable items, usage records, prediction basis to standard basis.

[0028] The entire process data, after being linked and coded, is stored in an encrypted distributed database in accordance with medical data management standards.

[0029] In a preferred embodiment of the present invention, regarding the process of real-time collection of surgical consumable usage data, drug usage data, and dynamic surgical operation data during surgery as described in step S4, and comparing and verifying the real-time collected data with the pre-operative prediction list to generate abnormal warning information, in order to overcome the technical defects of low sensitivity, strong lag, and inability to adapt to individual differences among patients caused by the use of static fixed thresholds in the prior art, the present invention introduces a nonlinear dynamic calculation model based on multidimensional parameter coupling. This model is not only for numerical monitoring of costs, but also for digital modeling and real-time calculation of surgical progress, patient physiological characteristics, and medical behavior norms.

[0030] Specifically, the step of generating an abnormal risk index for surgical costs using a dynamic calculation model relies on an underlying distributed data acquisition architecture and a high-performance computing engine. Its specific execution logic covers four cascaded processing stages: high-frequency data acquisition, standardized parameter parsing, nonlinear calculation of the core model, and closed-loop risk determination.

[0031] The system first initiates a real-time monitoring time window, the sampling frequency of which can be adaptively adjusted according to the surgical level (e.g., once every 30 seconds for level IV surgery, and once every 5 minutes for level I surgery). Within each monitoring time window, the system concurrently captures the following four sets of core computational parameters via the hospital's internal high-speed data bus (ESB), and performs rigorous cleaning and standardization processing: First, regarding the deviation value of the cumulative surgical cost ( The acquisition of ).

[0032] The system does not simply read the total amount from the financial system; instead, it performs a dynamic process alignment. The system first interacts with the HIS system via an IoT interface to obtain the current time. Actual costs incurred The data is compiled in real time from three parts: consumable costs generated by the intelligent cabinet for high-value consumables in the operating room (RFID), drug costs automatically uploaded by the Anesthesia Clinical Information System (AIMS), and operating room service fees automatically accumulated based on the operation duration.

[0033] Simultaneously, the system retrieves the pre-operative cost estimate list generated in step S3. It's worth noting that this list is not a static table, but rather a cost budget curve that evolves over time. Based on the current percentage of the surgery's progress (e.g., 45% complete), the system uses either a linear interpolation algorithm or a non-linear fitting algorithm based on historical big data to extract the theoretically expected cost at that point in time. .

[0034] The system then calculates the difference between the two. To eliminate the effect of positive and negative offsetting, this model takes the absolute value. It should be noted that when obtaining the actual costs incurred... At the same time, the system has a built-in outlier filtering mechanism. For a single transaction with an amount exceeding a set threshold (such as 50,000 yuan), the system will immediately trigger a data verification bit. Only after confirming that it is not a system entry error can it be entered, thus ensuring the accuracy of the source data.

[0035] Second, regarding the acquisition of time dimension parameters.

[0036] The current surgery has lasted for [duration]. The data source is directly taken from the operating room control panel or the synchronization timing interface of the anesthesia monitor, accurate to the second. To eliminate errors caused by network latency, the system uses NTP (Network Time Protocol) for end-to-end clock calibration.

[0037] Standard estimated total time for similar surgeries The determination of the standard value relies on in-depth mining of a multi-source knowledge base related to surgical fees. The system first parses the current patient's ICD-9-CM-3 surgical operation code and ICD-10 diagnostic code. Then, it searches the knowledge base for historical records of similar surgeries performed by doctors of the same level (e.g., associate chief physician or above) within the past 36 months at the same hospital. To eliminate the influence of extreme data caused by medical accidents or special emergency treatments on the standard value, the system uses a truncated average method. This involves removing the longest and shortest 5% of samples from the duration distribution and calculating the arithmetic mean of the remaining samples as the standard estimated total duration for similar surgeries. This ensures the objectivity, scientific rigor, and comparability of the time benchmark.

[0038] Third, regarding the patient's surgical complexity coefficient ( Quantization calculation of ).

[0039] These are key parameters that demonstrate the patient-centered approach and tolerance for individual differences in this model. The system uses Natural Language Processing (NLP) technology to deeply analyze the patient's electronic medical record (EMR) and extract key feature vectors, including: patient age, body mass index (BMI), ASA classification (I-VI), CCI index (Charlson Comorbidity Index), and whether there is a history of abdominal surgery (which affects the degree of adhesion).

[0040] The system inputs these feature vectors into a pre-set logistic regression scoring model and outputs a dimensionless floating-point coefficient, which is the surgical complexity coefficient. The baseline value for this coefficient is 1.0. For example, the coefficient increases by 0.1 for each additional serious comorbidity; and by 0.15 for each increase of one level in the ASA classification. Therefore, for a complex patient who is elderly, obese, and has cardiopulmonary insufficiency, the surgical complexity coefficient... It could be as high as 1.8. In subsequent formulas, this coefficient, as a denominator, will automatically dilute the high cost bias caused by complex patients, thereby avoiding unintended interference from the system in the critical care process.

[0041] Fourth, regarding the dynamic retrieval of historical cost fluctuation variance.

[0042] The system is based on the current surgical time progress factor. The system queries the historical database for the standard deviation of the cost growth rate for similar surgeries at that specific stage. For example, costs typically fluctuate significantly during the core lesion resection phase, while they should stabilize during the abdominal closure and suturing phase. This parameter will be used to dynamically adjust the alarm threshold.

[0043] After accurately obtaining the aforementioned physical parameters, the system invokes an algorithm engine deployed on a high-performance computing unit (such as a GPU cluster) to calculate the abnormal risk index of surgical costs in real time based on the following rigorously mathematically derived and clinically validated nonlinear coupling formula. : in, The dimensionless surgical cost abnormality risk index; The actual cost incurred is included in the cumulative cost deviation value of the surgery. The budgeted cost corresponding to the preoperative estimated charges list; The duration of the current surgery. The standard estimated total duration for the same type of surgery; The surgical complexity coefficient is used to characterize the degree to which the patient's individual physical signs and underlying diseases interfere with the standardized surgical procedure. For the normalized time progression factor, and ; As the cost deviation weighting factor, As the time deviation weighting factor, This is the time sensitivity decay constant, used to adjust the tolerance for cost deviations in the later stages of surgery; To prevent the minimum compensation constant where the denominator is zero, is the base of the natural logarithm. For natural logarithm operations; The following provides a detailed explanation of each parameter in the formula and its physical / clinical significance: Cost Deviation Rate and its minimum compensation constant This aspect aims to quantify the relative degree of deviation in costs. Here, the invention introduces a minimal compensation constant. (For example, a value of 1.0 yuan). The technical advantage of this design lies in solving the cold start and singularity problems: in the preparation stage at the very beginning of surgery, or in certain procedures involving only extremely low costs, the budgeted cost... It may be close to or equal to zero. If there is no minimal compensation constant... If the denominator is zero, the calculation program will crash or the result will overflow. A minimal compensation constant is introduced. This not only ensures the stability of numerical calculations but also smooths out the drastic fluctuations in the deviation rate caused by the small base number in the early stages of surgery. The cost deviation weighting factor is usually set at 0.6-0.8, indicating that cost overruns are the core source of risk.

[0044] Time Deviation Item Logarithmic adjustment mechanism: This invention uses the natural logarithm function. This addresses time skewness. It's an innovative application of the Weber-Fechner Law in healthcare management. Clinical experience shows that the contribution of surgical delays to risk perception exhibits diminishing marginal returns. For example, a surgery expected to take one hour that extends by one hour (100% overtime) often signifies an unexpected event and extremely high risk; however, if the surgery has already lasted 10 hours (such as a complex organ transplant), an additional one-hour extension (10% overtime) does not significantly increase the risk. Using a linear function would overemphasize the risk of long surgeries. Utilizing the smoothing properties of a logarithmic function effectively compresses the impact of extreme overtime data on the overall index, preventing anomalies in the cost dimension from being masked by purely time-related factors. This is the time deviation weighting factor, used to adjust the weight of time factors.

[0045] Dynamic adjustment factor The timing control mechanism: First, the surgical complexity coefficient Located in the denominator. The more complex the patient's condition (surgical complexity coefficient)... The larger the value, the higher the risk index of abnormal surgical costs calculated under the same deviation. The smaller.

[0046] Secondly It is an exponentially decaying term based on the time process. Wherein... This is a normalized time progression factor, representing the relative progress of the surgery; The time sensitivity decay constant is preferably in the range of 0.5 to 1.5. The duration of the current surgery. The standard estimated total duration for the same type of surgery.

[0047] As the surgery progresses (normalized time progression factor) Increase), exponential term The value of this term will decrease monotonically (from 1 to 0). Since this term is in the denominator, this means that the overall value of the fraction (adjustment factor) will increase exponentially as the operation progresses.

[0048] The physical correspondence of this mathematical property is: in the early stages of surgery (normalized time progression factor) The system is relatively tolerant of cost deviations, allowing doctors to perform preoperative preparations and various trial procedures, especially as the surgery nears its end (normalized time progression factor). (As the scale increases), the adjustment factor rapidly increases, making the system extremely sensitive to any new cost deviations. This design creates a gradually tightening regulatory field, precisely targeting issues such as hastily entering false costs at the end of the procedure or deliberately delaying surgery to consume consumables, forcing medical staff to maintain standardized cost entry throughout the entire process.

[0049] The abnormal risk index of surgical costs was calculated. Then, the system performs the comparison steps: The system reads the dynamic alarm threshold calculated based on the department's historical data for the same period. (in This is the historical average. For historical variance, (This is the confidence coefficient, usually taken as 2 or 3).

[0050] If the surgical cost is abnormal, the risk index is high. The system determined that the current cost increase was due to a non-technical anomaly.

[0051] The system not only generates alarm signals, but also executes attribution algorithms to calculate the contribution of cost items and time items to the total risk index, indicating whether the main cause is overspending on consumables or inefficiency. This detailed warning information is pushed in real time to the PDA terminal of the operating room head nurse and the monitoring screen of the medical insurance management department, requiring manual verification and feedback.

[0052] In step S5, before the surgery ends and a precise settlement statement is automatically generated, this invention introduces the aforementioned logical consistency verification step based on semantic topology graphs. This step utilizes graph theory and natural language processing (NLP) techniques to address deep compliance issues where numerical values ​​are correct but logical fallacies are not, which traditional systems cannot identify.

[0053] The construction of the surgical medical entity behavior topology graph is typically completed during the system initialization phase and dynamically maintained as the knowledge base is updated. The system constructs a directed attribute graph based on clinical pathway rules, pricing standards, and medical expert consensus from a multi-source data knowledge base. .in, A set of nodes; Let it be the set of edges.

[0054] Specifically, node set The definitions and classifications include: Operate entity nodes This refers to specific medical procedures or actions. Examples include: intravenous puncture, endotracheal intubation, dissection of the Triangle of Gallbladder, ultrasonic scalpel hemostasis, and skin suturing.

[0055] Consumables physical nodes This category represents chargeable medical consumables. Examples include: indwelling intravenous catheters, endotracheal tubes, ultrasonic scalpel tips, absorbable sutures, and titanium clips.

[0056] Drug physical nodes This indicates a drug that is charged for. Examples include: propofol emulsion injection, remifentanil, and rocuronium bromide.

[0057] Furthermore, edge set The definitions and attributes include: Strongly dependent edges This is a type of directed edge ( This indicates that the existence of entity B is contingent upon the existence of entity A. For example, an edge pointing from "Consumables: Anastomosing Device" to "Operation: Anastomosis" means that without an anastomosis operation, the legitimate consumption of the anastomosing device is impossible.

[0058] Mutual Exclusive Edges This is a type of undirected edge ( This indicates that two entities cannot appear simultaneously within a single surgical timeline. For example, the relationship between the nodes "Anesthesia: General Anesthesia" and "Anesthesia: Spinal Anesthesia"; or between "Nursing: Level 1 Nursing" and "Nursing: Special Level Nursing".

[0059] When the surgery is completed and the system is ready to generate the settlement statement, the NLP engine is activated to extract features from the multimodal data, mapping the real-world surgery onto a topological graph. The specific process is as follows: The system acquires surgical records (unstructured text), anesthesia records (semi-structured text), and nursing records. Using a pre-trained BERT-Medical deep language model, it performs word segmentation, part-of-speech tagging, and named entity recognition on the text. Through a semantic encoding layer, the system transforms natural language descriptions in the text (e.g., "gangrenous appendix was observed during surgery, and the root was ligated and removed") into standardized action feature vectors. This process supports fuzzy matching and synonym generalization, ensuring that doctors' different writing habits can be correctly identified.

[0060] The system encodes each charge item in the proposed accurate settlement list, generating a charge vector. .

[0061] System traversal of the entire graph If a certain toll vector With the consumables node in the diagram If the semantic similarity exceeds a preset threshold (e.g., 0.95), then the consumable node will be... The state is marked as active. Similarly, if an action vector matches an operation node, the corresponding operation node is activated. All activated nodes and their connected edges constitute a subgraph reflecting the current surgical situation. .

[0062] The system in the subgraph Then, run a graph traversal algorithm (such as Depth-First Search, DFS) and perform the following verification steps: The system traverses all active consumable entity nodes in the subgraph. For each consumable node, the algorithm traces back along all its strong dependency edges, checking the state of the predecessor operation entity node it points to.

[0063] If, in a subgraph, a certain anastomosis node is active (i.e., it has been charged), but all the predecessor operation nodes it points to (such as gastrointestinal anastomosis, stump closure) are inactive (i.e., no related operation is written in the medical record), then the consumable node logically becomes an isolated node with a suspended source point.

[0064] The system determined that the chargeable item contained a logical violation without any operational basis. This typically corresponds precisely to violations such as charging without performing any operation, overcharging, or substituting consumables.

[0065] Furthermore, the system traverses all existing mutually exclusive edges in the subgraph. .in, and The nodes at both ends of a mutual exclusion edge.

[0066] If the nodes at both ends of a mutual exclusion edge are found and It is currently active.

[0067] For example, the list may contain both Level 1 nursing care fees and Special Level nursing care fees (with overlapping time periods), or both laparoscopic surgery kits and open surgery kits may appear without any transfer records.

[0068] The system determines that the corresponding fee items have a logical violation of mutual exclusion.

[0069] In addition, the system uses timestamp attributes to check the temporal order of strongly dependent edges. For example, anesthesia induction (operation) must precede surgical incision (operation). If the timestamp is reversed, even though all nodes are activated, it is still considered a logical violation.

[0070] For billing items deemed logically irregular, the system marks them as high-risk items requiring verification. On the precise settlement list interface generated in step S5, the system loads a logical tracing view: irregular items are highlighted in red, and broken dependencies or conflicting mutually exclusive nodes are connected by dotted lines. This visual feedback mechanism allows medical staff or billing administrators to easily pinpoint problems and quickly supplement medical records or remove charges, thereby ensuring the compliance and accuracy of the final settlement list.

[0071] A billing model based on artificial intelligence-based surgical cost monitoring, applied in a billing method based on artificial intelligence-based surgical cost monitoring, includes: The knowledge base construction module is used to standardize various types of data collected by the data acquisition module and build a multi-dimensional standardized knowledge base. The prediction module uses deep learning algorithms to build a model for predicting charges and costs, and combines this with the patient's preoperative diagnosis and treatment information to output accurate preoperative charge prediction results. The real-time monitoring module is used to collect dynamic medical service information during surgery in real time, compare and verify it with the pre-operative prediction results, and realize real-time early warning and correction of billing risks. The settlement traceability module is used to integrate the actual data verified during the operation, generate an accurate settlement list, and realize full-process data traceability.

[0072] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not specifically listed, or elements inherent to such process, method, article, or apparatus.

[0073] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A billing method based on artificial intelligence-based surgical cost monitoring, characterized in that, Includes the following steps: S1: Construct a multi-source data knowledge base related to surgical fees; S2: Obtain basic information related to the patient's surgery and perform standardized preprocessing on the basic information; S3: Based on the preprocessed basic information and multi-source data knowledge base, generate a preoperative prediction list; S4: Real-time data collection of surgical consumables usage, drug usage, and surgical operation dynamics during surgery; comparison and verification of real-time data with the pre-operative prediction list; and generation of abnormal warning information. S5: After the surgery, integrate the actual data confirmed by the monitoring during the surgery and the surgical operation record to automatically generate an accurate settlement list, which includes the surgical fee bill and the anesthesia fee bill. S6: Store data records for the entire process before, during, and after surgery to build a cost detail traceability chain.

2. The charging method based on artificial intelligence surgical cost monitoring according to claim 1, characterized in that, The construction of the multi-source data knowledge base related to surgical fees specifically includes: Collect basic medical information, surgical-related coding information, surgical fee catalogs for various regions, and dynamic medical service information during surgery for surgical patients; Remove duplicate, missing, and abnormal data from the collected data; complete missing coding information and charging standard information; and standardize and convert date and amount data with inconsistent formats. Establish a mapping table that associates diagnostic codes with codes for common surgical procedures and anesthesia methods; Establish a mapping table linking surgical and anesthesia method codes with corresponding charge items in the surgical charge catalogs of various regions; Determine the standard fee item list corresponding to different code combinations, classify and store the standardized data in the knowledge base, and build an associated index.

3. The charging method based on artificial intelligence surgical cost monitoring according to claim 1, characterized in that, The S3 implementation process specifically includes: Feature extraction was performed on the patient's preoperative basic medical information to obtain the core feature vector; Using historical surgical fee case data in a standardized knowledge base as training samples, the core feature vectors in the samples are taken as input, and the corresponding standard fee item list and cost budget are taken as output to train the deep learning model. The model parameters are adjusted through cross-validation to optimize the prediction accuracy of the model and obtain the prediction model. Input the patient's preoperative core feature vector into the trained prediction model, call the encoding mapping table in the standardized knowledge base and the corresponding regional charging standards, and output the preoperative prediction charging item list, the estimated cost of each item and the total cost budget.

4. The charging method based on artificial intelligence surgical cost monitoring according to claim 1, characterized in that, The S4 implementation process specifically includes: Based on the coding mapping rules and charging standards in the standardized knowledge base, the real-time data is compared and verified with the preoperative predicted charging item list; A deviation threshold is set, which is determined based on statistical analysis of historical surgical data and clinical guidelines. The deviation value between real-time data and the prediction list is calculated. When the deviation between real-time data and the pre-judgment list exceeds the threshold, or when consumables, drugs, or procedures not included in the pre-operative pre-judgment list are found, it is judged as abnormal, and a graded warning is immediately triggered and pushed to the operating room nurses, anesthesiologists, and billing management terminals in real time. For any reasonable additional consumables, medications, or procedures confirmed during surgery, medical staff will review and confirm them before including them in the real-time usage list and updating the corresponding cost information.

5. A billing method based on artificial intelligence-based surgical cost monitoring according to claim 4, characterized in that, The comparison and verification of real-time data with the pre-operative predicted charge list specifically includes: Determine whether the consumables, drugs, surgical procedures, anesthesia procedures, and blood products actually used during the operation are on the pre-operative prediction list. If new items are added outside the prediction list, mark them as new items to be verified. If items on the pre-operative prediction list do not appear in the real-time data during the operation, mark them as prediction items not used and to be verified. Verify that the diagnostic codes, surgical codes, and anesthesia codes corresponding to the actual items used in the surgery are consistent with the preoperative prediction and comply with the coding requirements of the charging catalogs of each region. If there is a coding mismatch or coding error, mark it as a coding error to be corrected. By combining the patient's surgical type, weight, surgical duration, and historical case data in the standardized knowledge base, it is determined whether the actual usage of consumables, drugs, and blood products during the surgery is within a reasonable range. If it exceeds the reasonable usage threshold, it is marked as abnormal usage pending verification. Verify whether the charges for the actual procedures used during the surgery conform to the surgical fee catalog of the corresponding region, and whether there are any cases of overcharging or double charging. If there are any illegal charges, mark them as illegal charges pending correction.

6. The charging method based on artificial intelligence surgical cost monitoring according to claim 1, characterized in that, The S5 implementation process specifically includes: Collect the actual list of consumables used, the actual list of drugs used, the actual surgical operation record, and the actual anesthesia operation record as confirmed by monitoring during the operation. At the same time, retrieve the preliminary cost budget before the operation, the regional fee catalog database, and the basic coding database information. Based on the fee standards in the regional fee catalog, the costs of consumables, medicines, surgical procedures and anesthesia procedures actually used are calculated, and the detailed costs of each item and the total cost are calculated. Based on the basic coding library, clinical association rule library and regional charging catalog library, the reasonableness of the calculated drug usage and charging items is verified; The system automatically generates accurate settlement lists that include surgical and anesthesia bills. These lists are then sent to medical staff and billing administrators for manual review. Once approved, the final settlement list is generated.

7. The charging method based on artificial intelligence surgical cost monitoring according to claim 1, characterized in that, The construction method of the aforementioned cost detail tracing link specifically includes: The data for the entire process is categorized and collected according to the time sequence of pre-operative, intra-operative, and post-operative periods, as well as the dimensions of patients, surgery, and billing. A unique surgical cost traceability code is generated for each surgical patient, and the generated traceability code is linked and bound one by one with the collected full-process data; Based on the associated full-process data, a two-way structured traceability link is constructed, which includes forward time-series tracing and reverse cost tracing. The entire process data, after being linked and coded, is stored in an encrypted distributed database in accordance with medical data management standards.

8. A billing method based on artificial intelligence-based surgical cost monitoring according to claim 4, characterized in that, The comparison and verification of real-time data with the pre-operative predicted charge list includes the step of generating an abnormal surgical cost risk index using a dynamic calculation model, specifically including: A real-time monitoring time window is activated, and the sampling frequency is adaptively adjusted according to the surgical level. The calculation parameters are captured in parallel through a high-speed data bus and then standardized. The system acquires the cumulative surgical cost deviation value, the current surgical duration, the standard estimated total duration of similar surgeries, and the patient's surgical complexity coefficient within the current monitoring time window. The acquisition of the cumulative surgical cost deviation value includes: obtaining the actual cost incurred at the current moment and implementing an outlier filtering mechanism to remove instantaneous data jumps where the single amount exceeds a set threshold; extracting the budgeted cost from the cost budget curve evolving over time using an interpolation algorithm based on the current surgical progress percentage. The acquisition of the standard estimated total duration of similar surgeries includes: parsing the patient's diagnosis and surgical code, retrieving historical records from the knowledge base, and calculating the arithmetic mean after removing the longest and shortest samples from the duration distribution using a truncated average method to obtain the standard estimated total duration of the similar surgeries. The acquisition of the patient's surgical complexity coefficient includes: parsing electronic medical records using natural language processing technology, extracting feature vectors including BMI, ASA classification, CCI index, and past surgical history, and inputting them into a logistic regression scoring model for calculation. Based on the multi-source data knowledge base, retrieve the historical cost fluctuation variance of the same type of surgery at the current time. The abnormal risk index of the surgical cost is calculated based on the following formula: in, The dimensionless surgical cost abnormality risk index; The actual cost incurred is included in the cumulative cost deviation value of the surgery. The budgeted cost corresponding to the preoperative estimated charges list; The duration of the current surgery. The standard estimated total duration for the same type of surgery; The surgical complexity coefficient is used to characterize the degree to which the patient's individual physical signs and underlying diseases interfere with the standardized surgical procedure. For the normalized time progression factor, and ; As the cost deviation weighting factor, As the time deviation weighting factor, This is the time sensitivity decay constant, used to adjust the tolerance for cost deviations in the later stages of surgery; To prevent the minimum compensation constant where the denominator is zero, is the base of the natural logarithm. For natural logarithm operations; Calculate dynamic alarm thresholds ,in This is the historical average. For historical variance, Confidence coefficient; The calculated abnormal risk index of surgical costs With the dynamic alarm threshold Real-time comparison is performed; if the surgical cost is abnormal, the risk index is considered. Greater than the dynamic alarm threshold The system determines that the current cost increase is due to a non-technical anomaly and generates the anomaly warning information.

9. A billing method based on artificial intelligence-based surgical cost monitoring according to claim 1, characterized in that, Before the S5 automatically generates the accurate settlement list, a logical consistency verification step based on the semantic topology graph is also included, specifically including: Based on the clinical pathway rules and pricing standards in the aforementioned multi-source data knowledge base, a directed attribute graph is constructed. A topological diagram of surgical medical entity behavior; in which, For a set of nodes, The set of edges; the set of nodes Includes operational entity nodes representing specific medical actions. , representing the physical nodes of chargeable medical consumables and the physical nodes representing chargeable drugs edge set This includes directed edges that represent strong dependencies that indicate the existence of prerequisite relationships between entities. and mutually exclusive undirected edges representing entities that cannot appear simultaneously. ; Using a pre-trained BERT-Medical deep language model, word segmentation and named entity recognition were performed on the surgical operation records, anesthesia records, and nursing records. The unstructured text was then transformed into standardized action feature vectors through a semantic encoding layer. ; and encode each charge item in the accurate settlement list to generate a charge item feature vector. ; The extracted action feature vectors With the feature vectors of each of the aforementioned fee items Mapped to the surgical medical entity behavior topology graph, the semantic similarity between the vector and the corresponding node is calculated. When the semantic similarity exceeds a preset threshold, the corresponding node state is marked as active, thereby forming the current surgical actual execution subgraph. ; The current surgical procedure execution subgraph Then, run a depth-first search algorithm to perform connectivity analysis and isolated node detection, and execute the following decision logic: If a consumable entity node is active in the current surgical execution subgraph, and the algorithm traces back along all its strong dependent directed edges to find that all its predecessor operation entity nodes are inactive, it is determined that the corresponding billing item has a logical violation without operational basis. If two nodes are active in the current surgical execution subgraph, and there is a mutually exclusive undirected edge between the two nodes in the surgical medical entity behavior topology graph, it is determined that the corresponding charging item has a logical violation of mutual exclusion conflict. Check the timestamp attributes of the active nodes at both ends of the strongly dependent edge. If the timestamp of the predecessor operation entity node is later than the timestamp of the successor node, it is determined to be a timing logic violation. The charging items that are determined to be logically irregular are marked as high-risk items to be verified, and a logical traceability view is loaded into the generated accurate settlement list. The irregular logical links are identified by highlighting the irregular items and the broken dependency relationships.

10. A charging model based on artificial intelligence-based surgical cost monitoring, applied in the charging method based on artificial intelligence-based surgical cost monitoring as described in claim 7, characterized in that, include: The knowledge base construction module is used to standardize various types of data collected by the data acquisition module and build a multi-dimensional standardized knowledge base. The prediction module uses deep learning algorithms to build a model for predicting charges and costs, and combines this with the patient's preoperative diagnosis and treatment information to output accurate preoperative charge prediction results. The real-time monitoring module is used to collect dynamic medical service information during surgery in real time, compare and verify it with the pre-operative prediction results, and realize real-time early warning and correction of billing risks. The settlement traceability module is used to integrate the actual data verified during the operation, generate an accurate settlement list, and realize full-process data traceability.

Images