An intelligent financial analysis system based on multi-source heterogeneous data fusion

Abstract

This invention relates to the field of financial management and discloses an intelligent financial analysis system based on multi-source heterogeneous data fusion. It includes: a data source adaptation module, an event-based transformation module, a business-finance-medical insurance mapping module, a time-series status verification module, an event-level cost aggregation module, a composite benchmark deviation identification module, an anomaly attribution module, and a closed-loop feedback module. The data source adaptation module accesses multi-source business data from the hospital and adds source, time, status, responsible party, and field version identifiers; the event-based transformation module generates hospital operational financial events and event chains; the mapping module establishes the relationships between objects, events, and indicators; the time-series status verification module determines the analysis status threshold; the cost aggregation module generates cost results; the composite benchmark deviation identification module generates financial anomaly objects; and the anomaly attribution module and closed-loop feedback module complete the generation and retesting of the cause chain.

Description

Technical Field

[0001] This invention relates to the field of financial management, and more specifically to an intelligent financial analysis system based on the fusion of multi-source heterogeneous data. Background Technology

[0002] As hospitals become increasingly sophisticated in their operational management, they generate a large amount of business data, financial data, medical insurance settlement data, medical record data, material and consumable data, equipment and asset data, staff scheduling data, and budget execution data in their daily operations. This data typically originates from hospital business systems, electronic medical record systems, hospital resource planning systems, medical insurance settlement systems, medical record front page systems, material and consumable systems, equipment and asset systems, staff scheduling and performance systems, and budget management systems. Existing hospital operational analysis systems generally extract data from these business systems through data interfaces. After cleaning, summarizing, and calculating indicators, this data forms hospital-level dashboards, financial analysis reports, cost analysis reports, medical insurance analysis reports, departmental operational dashboards, and budget execution dashboards. These dashboards allow management personnel to understand the hospital's revenue structure, cost composition, disease-specific profit and loss, medical insurance payments, departmental performance, and resource utilization.

[0003] However, existing hospital financial analysis systems mostly focus on financial categories, departmental summaries, billing items, medical insurance settlement results, or budget items, typically emphasizing report display, indicator statistics, and individual threshold warnings. Due to differences in source systems, field definitions, item codes, generation times, and confirmation status among hospital business data, financial data, medical insurance data, and cost data, the execution of medical orders, confirmation of charges, cost accumulation, medical insurance settlement, finalization of medical records, and budget consumption within the same treatment process often cannot be correlated within the same event chain. When medical orders have been executed but charges have not been confirmed, consumables have been issued but costs have not been accumulated, medical insurance has been submitted but settlement has not been completed, and the medical record cover sheet has not been finalized, existing systems easily confuse delayed data confirmation with actual financial anomalies, leading to inconsistencies in the analysis results of departmental costs, disease-specific costs, medical insurance refusal, disease group profit and loss, and budget execution, including inconsistencies in definitions, timing, and unclear responsible parties.

[0004] Meanwhile, hospital financial anomalies are often not caused by fluctuations in a single financial item, but are related to multiple business processes such as treatment pathways, consumable usage structure, bed and equipment occupancy, staff scheduling, medical record coding, medical insurance disease groups, and budget allocation. Even if existing systems can identify an indicator exceeding a threshold, it is difficult to trace the anomaly back to specific patient visits, medical order execution events, billing settlement events, cost incurrence events, medical insurance payment events, resource occupancy events, and budget consumption events. Consequently, it is difficult to establish a causal chain corresponding to departments, diseases, medical insurance disease groups, medical teams, doctors, consumables, equipment, and budget items. Therefore, subsequent rectification and retesting still rely on manual experience-based judgment, making it difficult to ensure that anomaly identification, cause localization, and closed-loop feedback are within the same data scope and responsible parties, affecting the accuracy and traceability of the hospital's intelligent financial analysis results. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent financial analysis system based on the fusion of multi-source heterogeneous data, thereby resolving the technical problems existing in the prior art.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: An intelligent financial analysis system based on multi-source heterogeneous data fusion includes: The module includes a data source adaptation module, an event-based conversion module, a business, finance, and medical insurance mapping module, a time-series status verification module, an event-level cost aggregation module, a composite benchmark deviation identification module, an anomaly attribution module, and a closed-loop feedback module. The data source adaptation module is used to access multi-source business data from the hospital, obtain raw data, and add source system identifier, time identifier, data status identifier, responsible object identifier, and field version identifier to the raw data; The event-based transformation module connects with the data source adaptation module to convert raw data into hospital operation and financial events according to patient visit identifiers, responsible party identifiers, and project identifiers, and merges related hospital operation and financial events into an event chain. The business, finance, and medical insurance mapping graph module is connected to the event-based transformation module. It is used to establish a mapping graph including object nodes, event nodes, and indicator nodes based on the event chain, and to generate association edges with time identifiers, data status identifiers, responsible object identifiers, and field version identifiers between object nodes, event nodes, and indicator nodes. The timing status verification module is connected to the business, finance and medical insurance mapping graph module. It is used to determine the analysis status water level based on the data status identifier of each hospital's operational financial event in the event chain, and to set a delayed confirmation identifier for hospital operational financial events that have not reached the required analysis status water level of the corresponding indicator node. The event-level cost aggregation module is connected to the business, finance, and medical insurance mapping module and the time-series status verification module. It is used to aggregate event-level costs for hospital operation and financial events that have not been marked with delayed confirmation, and to generate cost aggregation results corresponding to the responsible objects. The composite benchmark deviation identification module is connected to the event-level cost aggregation module. It is used to generate objects to be analyzed based on the cost aggregation results, and to identify deviations of the objects to be analyzed based on the budget benchmark, medical insurance payment benchmark and similar benchmarks within the hospital. When the objects to be analyzed meet the deviation conditions corresponding to at least two types of benchmarks, financial anomaly objects are generated. The anomaly attribution module is connected to the composite benchmark deviation identification module and the business finance and medical insurance mapping graph module. It is used to trace back the associated hospital operation and financial events and responsible parties along the mapping graph, starting from the financial anomaly object, and generate a cause chain including anomaly indicators, responsible parties, related events and evidence data. The closed-loop feedback module is connected to the anomaly attribution module. It is used to generate anomaly feedback records based on the cause chain and lock the retest samples in subsequent analysis cycles according to the same field version identifier, the same analysis status level, and the same scope of responsible objects. It updates the hospital's similar benchmarks and cause chain ranking weights based on the retest results.

[0007] Preferably, the multi-source business data of the hospital accessed by the data source adaptation module includes data from the hospital business system, electronic medical record system, hospital resource planning system, medical insurance settlement system, medical record front page system, material and consumable system, equipment and asset system, shift scheduling and performance system, and budget management system. The time identifiers include data generation time, business occurrence time, accounting confirmation time, cost collection time, and medical insurance settlement time; The data status identifiers include cost confirmation status, cost collection status, medical insurance settlement status, and medical record finalization status.

[0008] Preferably, the data source adaptation module includes a field registration unit, a source verification unit, and a caliber version unit; The field registration unit is used to establish a field registration table, which includes field name, field type, source system identifier, business meaning, value range, effective time and expiration time; The source verification unit is used to perform mapping verification on the medical order item identifier, charge item identifier, medical insurance settlement item identifier and cost item identifier under the same patient visit identifier according to the preset item mapping table, and set an item mismatch identifier for the original data that does not match the corresponding relationship; The caliber version unit is used to generate a field version identifier when the field name, field type, business meaning, and value range of the field registration table change.

[0009] Preferably, the event-based conversion module includes an event template library and an event merging unit; The event template library stores templates for diagnosis and treatment events, medical order execution events, billing and settlement events, cost occurrence events, medical insurance payment events, resource occupancy events, and budget consumption events. Each event template includes an event type, a main object identifier, a responsible object identifier, a business time field, an amount field, a quantity field, a status field, and a related item field. The event merging unit is used to merge hospital operation and financial events with the same patient visit identifier, the same responsible object identifier, a mapping relationship between related project fields, and business occurrence time within the start and end time fields of the corresponding event template into the same event chain.

[0010] Preferably, the business finance medical insurance mapping graph module includes an object relationship generation unit, an item relationship generation unit, and an indicator relationship generation unit; The object relationship generation unit is used to establish the affiliation relationships between patients, departments, wards, medical groups, doctors, diseases, and medical insurance disease groups; The project relationship generation unit is used to establish mapping relationships between medical order items, billing items, medical insurance settlement items, consumable items, equipment items, cost items, and budget items; The indicator relationship generation unit is used to associate revenue structure indicators, department cost indicators, disease cost indicators, project cost indicators, medical insurance refusal indicators, disease group profit and loss indicators, and budget execution indicators to the corresponding object nodes and event nodes, respectively.

[0011] Preferably, the timing state verification module includes a state node configuration unit and a water level determination unit; The status node configuration unit is used to configure status nodes for the same event chain, such as "business has occurred", "medical order has been executed", "fees have been confirmed", "costs have been collected", "medical insurance has been settled", and "medical record has been finalized". The water level determination unit is used to read the completion time of each state node in the same event chain, and to determine the state node that has been completed consecutively and is the last in the order in the same event chain as the analysis state water level of the event chain. When the analysis status level of the event chain is lower than the lowest status node of the corresponding indicator node in the indicator status requirement table, the level determination unit sets a delayed confirmation flag for the hospital operation and financial event corresponding to the event chain.

[0012] Preferably, the event-level cost aggregation module includes a project cost aggregation unit, a resource cost aggregation unit, and a subject cost aggregation unit; The project cost collection unit is used to collect the cost of billing items and the cost of consumable items according to the mapping relationship between medical order execution events, billing settlement events and consumable items. The resource cost collection unit is used to collect bed occupancy costs, equipment occupancy costs, operating room occupancy costs, and staff scheduling costs based on resource occupancy events; The cost collection unit is used to collect cost items and budget items in the hospital resource planning system based on cost occurrence events, and generate cost collection results according to department, disease type, medical insurance disease group and medical group.

[0013] Preferably, the composite benchmark deviation identification module includes a budget benchmark generation unit, a medical insurance payment benchmark generation unit, and an in-hospital benchmark generation unit; The budget baseline generation unit is used to generate budget baselines according to budget items, responsible object identifiers, and analysis cycles; The medical insurance payment benchmark generation unit is used to generate medical insurance payment benchmarks according to medical insurance disease groups, medical insurance settlement items, cost consumption index, time consumption index and medical insurance settlement status; The in-hospital benchmark generation unit is used to generate in-hospital benchmarks based on the same disease type, the same medical insurance disease group, the same surgical procedure, the same department, the same field version identifier, and the same analysis status water level.

[0014] Preferably, the composite benchmark deviation identification module calculates the budget deviation between the object to be analyzed and the budget benchmark, the medical insurance payment deviation between the object to be analyzed and the medical insurance payment benchmark, and the intra-hospital similar deviation between the object to be analyzed and the intra-hospital similar benchmark, and reads the threshold table according to the budget item, medical insurance disease group and responsible object identifier; When at least two of the following—budget deviation, medical insurance payment deviation, and similar deviation within the hospital—exceed the corresponding threshold in the threshold table, the object to be analyzed will be marked as a financially abnormal object, and the deviation type, deviation value, responsible object identifier, and associated event identifier that triggered the financially abnormal object will be recorded.

[0015] Preferably, the anomaly attribution module includes a candidate cause chain generation unit and a cause chain sorting unit; The candidate cause chain generation unit is used to generate candidate cause chains by starting with the indicator node corresponding to the financial anomaly object, tracing back along the mapping graph to the event node and object node in sequence. The candidate cause chains include the treatment path deviation chain, the consumable usage structure anomaly chain, the medical record coding deviation chain, the bed day cost increase chain, the examination waiting time chain, the medical insurance refusal chain, and the budget allocation deviation chain. The cause chain sorting unit is used to sort candidate cause chains according to the contribution of events, chronological order, consistency of responsible parties, and degree of rule matching, and write the sorting position, candidate cause chain type, and corresponding evidence data into the cause chain; The closed-loop feedback module includes an anomaly feedback record generation unit, a retest sample locking unit, and a weight update unit. The abnormal feedback record generation unit is used to generate abnormal feedback records based on the cause chain, including abnormal indicators, responsible parties, related events, evidence data, retest trigger time, and retest cycle. The retest sample locking unit is used to lock retest samples in subsequent analysis cycles according to the same field version identifier, the same analysis status level, the same scope of responsible objects, and the same indicator nodes. The weight update unit is used to update the ranking weights of similar benchmarks and causal chains within the institute based on the deviation changes of financially abnormal objects in the retest samples.

[0016] In summary, the present invention has the following main beneficial effects: The system utilizes a data source adaptation module to uniformly access raw data from hospital business systems, electronic medical record systems, hospital resource planning systems, medical insurance settlement systems, medical record front page systems, material and consumable systems, equipment asset systems, shift scheduling and performance systems, and budget management systems. It adds source system identifiers, time identifiers, data status identifiers, responsible party identifiers, and field version identifiers to the raw data. Then, an event-based conversion module transforms business records from different systems into hospital operational financial events and merges them into an event chain. This achieves the effect of unifying data originally scattered across business, finance, cost, medical insurance, and resource systems under a single patient visit process and a single responsible party for analysis. Therefore, when performing financial analysis, the system no longer relies solely on departmental summary tables, financial statements, or medical insurance settlement tables for post-hoc statistical analysis. Instead, it establishes an analytical foundation based on the actual business sequence between patient visits, medical order execution, billing settlement, cost incurrence, medical insurance payment, resource usage, and budget consumption. This reduces analytical biases caused by inconsistent field definitions, incomplete project mappings, and unclear data sources, ensuring that the identification of financial anomalies has clear data sources, responsible parties, and event evidence.

[0017] The system establishes relationships between object nodes, event nodes, and indicator nodes through a business-finance-medical insurance mapping graph module. Time identifiers, data status identifiers, responsible object identifiers, and field version identifiers are written into the associated edges. Simultaneously, a time-series status verification module determines the analysis status watermark based on the data status of each hospital's operational financial event in the event chain. Hospital operational financial events that do not reach the required analysis status watermark for the corresponding indicator node are marked with a delayed confirmation flag, achieving the effect of distinguishing between confirmed and unconfirmed data under the same analytical caliber. For situations such as medical orders that have been executed but costs not yet confirmed, consumables that have been issued but costs not yet collected, medical insurance submissions that have not yet been settled, and medical record front pages that have not yet been finalized, the system will not directly use these as the basis for judging final financial anomalies, thus avoiding misjudging delayed data confirmation as genuine operational anomalies. Furthermore, the system can trace back indicators such as revenue structure, departmental costs, disease costs, project costs, medical insurance refusal, disease group profit and loss, and budget execution along the mapping graph to the corresponding departments, diseases, medical insurance disease groups, medical groups, doctors, consumables, equipment, and budget items, providing traceable event chain support for abnormal indicators.

[0018] The event-level cost aggregation module aggregates costs for hospital operational financial events without delayed confirmation flags. A composite benchmark deviation identification module identifies deviations in the analyzed object based on budget benchmarks, medical insurance payment benchmarks, and similar internal benchmarks. Financial anomalies are generated when deviation conditions corresponding to at least two benchmarks are triggered. An anomaly attribution module then generates a causal chain along a mapping graph, including anomaly indicators, responsible parties, related events, and evidence data. A closed-loop feedback module locks down retest samples based on the same field version identifier, the same analysis status level, and the same responsible party scope, achieving a closed-loop processing effect from financial anomaly identification and attribution to retest feedback. This approach avoids the insufficient adaptability of single-threshold alarms to complex hospital operational scenarios. It further links anomalies in patient group profit and loss, departmental cost anomalies, medical insurance refusal anomalies, and budget execution anomalies to specific causal chains such as treatment pathways, consumable usage structures, medical record coding, bed-day costs, examination waiting times, reasons for medical insurance refusal, and budget allocation. This ensures that subsequent retests are based on the same criteria, guaranteeing a continuous and consistent data foundation between anomaly analysis results, responsibility identification results, and feedback results. Attached Figure Description

[0019] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0020] 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.

[0021] Example 1 refer to Figure 1 An intelligent financial analysis system based on multi-source heterogeneous data fusion includes: The module includes a data source adaptation module, an event-based conversion module, a business, finance, and medical insurance mapping module, a time-series status verification module, an event-level cost aggregation module, a composite benchmark deviation identification module, an anomaly attribution module, and a closed-loop feedback module. The data source adaptation module is used to access multi-source business data from the hospital, obtain raw data, and add source system identifier, time identifier, data status identifier, responsible object identifier, and field version identifier to the raw data; The event-based transformation module connects with the data source adaptation module to convert raw data into hospital operation and financial events according to patient visit identifiers, responsible party identifiers, and project identifiers, and merges related hospital operation and financial events into an event chain. The business, finance, and medical insurance mapping graph module is connected to the event-based transformation module. It is used to establish a mapping graph including object nodes, event nodes, and indicator nodes based on the event chain, and to generate association edges with time identifiers, data status identifiers, responsible object identifiers, and field version identifiers between object nodes, event nodes, and indicator nodes. The timing status verification module is connected to the business, finance and medical insurance mapping graph module. It is used to determine the analysis status water level based on the data status identifier of each hospital's operational financial event in the event chain, and to set a delayed confirmation identifier for hospital operational financial events that have not reached the required analysis status water level of the corresponding indicator node. The event-level cost aggregation module is connected to the business, finance, and medical insurance mapping module and the time-series status verification module. It is used to aggregate event-level costs for hospital operation and financial events that have not been marked with delayed confirmation, and to generate cost aggregation results corresponding to the responsible objects. The composite benchmark deviation identification module is connected to the event-level cost aggregation module. It is used to generate objects to be analyzed based on the cost aggregation results, and to identify deviations of the objects to be analyzed based on the budget benchmark, medical insurance payment benchmark and similar benchmarks within the hospital. When the objects to be analyzed meet the deviation conditions corresponding to at least two types of benchmarks, financial anomaly objects are generated. The anomaly attribution module is connected to the composite benchmark deviation identification module and the business finance and medical insurance mapping graph module. It is used to trace back the associated hospital operation and financial events and responsible parties along the mapping graph, starting from the financial anomaly object, and generate a cause chain including anomaly indicators, responsible parties, related events and evidence data. The closed-loop feedback module is connected to the anomaly attribution module. It is used to generate anomaly feedback records based on the cause chain and lock the retest samples in subsequent analysis cycles according to the same field version identifier, the same analysis status level, and the same scope of responsible objects. It updates the hospital's similar benchmarks and cause chain ranking weights based on the retest results.

[0022] The system is deployed on the hospital's intranet server, hospital private cloud server, or hospital data center server. The system includes a data source adaptation module, an event-based transformation module, a business, finance, and medical insurance mapping graph module, a time-series status verification module, an event-level cost aggregation module, a composite benchmark deviation identification module, an anomaly attribution module, and a closed-loop feedback module. The server includes a processor, memory, database, and data interface. The memory stores a computer program, and the processor executes the computer program to complete multi-source data access, event-based transformation, graph construction, status verification, cost aggregation, deviation identification, anomaly attribution, and closed-loop retesting. The database stores the following tables: field registration table, project mapping table, event template table, event linked list, graph node table, graph edge table, status node table, indicator status requirement table, indicator benchmark table, threshold table, benchmark sample rule table, anomaly object table, cause rule table, cause linked list, retesting rule table, and anomaly feedback record table.

[0023] In this embodiment, the hospital business system is used to record data on registration, outpatient services, inpatient services, billing, medical order execution, admission and discharge, and surgery; the electronic medical record system is used to record medical records, treatment processes, diagnoses, surgical procedures, clinical pathways, and medical order content; the hospital resource planning system is used to record cost items, budget items, asset depreciation, personnel costs, and departmental expenses; the medical insurance settlement system is used to record medical insurance disease groups, medical insurance settlement items, medical insurance payment amounts, medical insurance refusal amounts, cost consumption index, time consumption index, and medical insurance settlement status; the medical record cover page system is used to record the primary diagnosis, other diagnoses, surgical procedures, medical record codes, final medical record status, and medical insurance disease group results; the materials and consumables system is used to record the requisition, issuance, return, and consumption of medicines and consumables; the equipment asset system is used to record equipment numbers, equipment departments, equipment depreciation, equipment maintenance costs, and equipment usage time; the shift scheduling performance system is used to record information on doctors, nurses, medical technicians, operating room personnel, and shifts; and the budget management system is used to record budget items, responsible parties, budget amounts, budget control rules, and budget execution cycles.

[0024] The data source adaptation module connects to the hospital's business system, electronic medical record system, hospital resource planning system, medical insurance settlement system, medical record front page system, material and consumable system, equipment and asset system, shift scheduling and performance system, and budget management system to obtain raw data. It then adds source system identifiers, data generation time, business occurrence time, accounting confirmation time, cost collection time, medical insurance settlement status, responsible object identifier, and field version identifiers to the raw data. The data source adaptation module includes a field registration unit, a source verification unit, and a caliber version unit.

[0025] The field registration unit establishes a field registration table for raw data from different systems. The field registration table includes the field name, field type, source system identifier, business meaning, value range, effective time, and expiration time. The value range in the field registration table is determined by the field dictionary, hospital billing item catalog, medical insurance settlement item catalog, hospital resource planning system cost account table, budget item table, drug and consumable catalog, and equipment asset catalog of the corresponding source system. The data source adaptation module does not manually estimate the field value range; when the source system catalog changes, the caliber version unit generates a new field version identifier based on the catalog change record and writes the field version identifier into the subsequently generated hospital operation financial events. For field values ​​not registered in the source system catalog, the source verification unit sets a field unregistered identifier and excludes the raw data from the final financial anomaly object generation process.

[0026] The source verification unit performs project-level verification on data from different systems according to a pre-set project mapping table. The pre-set project mapping table includes the source system identifier, source project identifier, target project identifier, mapping type, effective time, expiration time, and field version identifier. The source verification unit retrieves the billing item identifier, medical order item identifier, medical insurance settlement item identifier, and cost item identifier for the same medical visit process based on the patient's visit identifier, and matches these item identifiers with the pre-set project mapping table. Raw data that matches a corresponding relationship enters the event-based transformation module; raw data that does not match a corresponding relationship is marked with a "project not matched" flag and enters the mapping table maintenance queue. Raw data with a "project not matched" flag is not used to generate the final financial anomaly object.

[0027] The caliber version unit is used to generate a field version identifier when there are changes to the field name, field type, business meaning, or value range in the field registration table. The field version identifier is stored together with the source system identifier, effective time, and expiration time. Subsequent hospital operation and financial events, event chains, graph edges, benchmark values, thresholds, cause chains, and anomaly feedback records are all written to the field version identifier to prevent data generated under different field calibers from being mixed into the same analysis results.

[0028] The event-based transformation module connects to the data source adaptation module and is used to convert raw data into hospital operational financial events according to patient visit identifiers, department identifiers, disease identifiers, billing item identifiers, consumable item identifiers, equipment resource identifiers, and budget item identifiers. Hospital operational financial events include treatment occurrence events, medical order execution events, billing settlement events, cost occurrence events, medical insurance payment events, resource occupancy events, and budget consumption events. The event-based transformation module includes an event template library and an event merging unit.

[0029] The event template library stores templates for diagnosis and treatment events, medical order execution events, billing and settlement events, cost occurrence events, medical insurance payment events, resource occupancy events, and budget consumption events. Each event template includes an event type, a main object identifier, a responsible object identifier, a business time field, an amount field, a quantity field, a status field, and a related item field.

[0030] The primary identifier for the medical treatment event template is the patient's visit identifier. Responsible identifiers include department, ward, medical team, and doctor identifiers. Business time fields include visit start time, visit end time, admission time, discharge time, surgery start time, and surgery end time. The associated fields for the medical order execution event template include medical order item identifier, billing item identifier, and consumable item identifier. Status fields include medical order creation, medical order execution, medical order cessation, and medical order cancellation. The amount fields for the billing settlement event template include amount receivable, amount received, amount reduced / waived, and amount refunded. Status fields include billed, confirmed, refunded, and written off. The associated fields for the cost occurrence event template include cost item identifier, budget item identifier, department identifier, and cost occurrence time. The associated fields for the medical insurance payment event template include medical insurance disease group, medical insurance settlement item, medical insurance payment amount, medical insurance refusal amount, cost consumption index, time consumption index, and medical insurance settlement status. The associated fields for the resource occupancy event template include bed identifier, equipment resource identifier, operating room identifier, personnel shift identifier, occupancy start time, and occupancy end time. The associated fields for the budget consumption event template include budget item identifier, responsible object identifier, budget period, budget amount, and budget consumption amount.

[0031] The start and end time fields corresponding to the event templates are determined according to the event type. For diagnosis and treatment events, the business time range is from the start time of the consultation to the end time; for medical order execution events, it's from the start time of medical order execution to the end time of medical order execution; for billing and settlement events, it's from the time of cost generation to the time of cost confirmation; for cost occurrence events, it's from the time of cost occurrence to the time of cost accumulation; for medical insurance payment events, it's from the time of medical insurance settlement submission to the time of medical insurance settlement completion; for resource occupancy events, it's from the start time of resource occupancy to the end time of resource occupancy; and for budget consumption events, it's from the start time of the budget period to the end time of the budget period. If the business occurrence time of the original data does not fall into the corresponding start and end time fields, the event merging unit will not merge the original data into the corresponding event chain.

[0032] The event merging unit merges raw data according to patient visit identifier, responsible object identifier, project mapping relationship, and business occurrence time. For raw data with the same patient visit identifier, the same responsible object identifier, a mapping relationship between related project fields, and business occurrence times within the start and end time fields of the same event template, the event merging unit merges them into the same event chain and generates an event chain identifier for the event chain. The event chain identifier is generated by combining the patient visit identifier, event template identifier, responsible object identifier, target project identifier, business occurrence time period, and field version identifier. The event chain list records the event chain identifier, patient visit identifier, department identifier, disease identifier, medical insurance disease group, medical group identifier, doctor identifier, charge item identifier, consumable item identifier, equipment resource identifier, budget item identifier, and field version identifier.

[0033] The business, finance, and medical insurance mapping module connects with the event-based transformation module to establish a mapping graph that includes object nodes, event nodes, and indicator nodes. Object nodes include departments, wards, medical teams, doctors, patients, diseases, medical insurance disease groups, beds, equipment, medicines and consumables, and budget items. Event nodes correspond to events such as treatment occurrence, medical order execution, billing settlement, cost occurrence, medical insurance payment, resource occupancy, and budget consumption. Indicator nodes include revenue structure indicators, departmental cost indicators, disease cost indicators, project cost indicators, medical insurance refusal indicators, disease group profit and loss indicators, and budget execution indicators.

[0034] The Business, Finance, and Medical Insurance Mapping Graph module includes an object relationship generation unit, a project relationship generation unit, and an indicator relationship generation unit. The object relationship generation unit establishes the attribution relationships between patients, departments, wards, medical groups, doctors, diseases, and medical insurance disease groups. The project relationship generation unit establishes the mapping relationships between medical order items, billing items, medical insurance settlement items, consumable items, equipment items, cost items, and budget items. The indicator relationship generation unit associates revenue structure indicators, department cost indicators, disease cost indicators, project cost indicators, medical insurance refusal indicators, disease group profit and loss indicators, and budget execution indicators to their corresponding object nodes and event nodes.

[0035] The association edges in the mapping graph include attribution edges, occurrence edges, consumption edges, settlement edges, aggregation edges, and evaluation edges. Attribution edges represent the attribution relationships between patients, departments, wards, medical groups, doctors, disease types, and medical insurance disease groups; occurrence edges represent the hospital operation financial events generated by object nodes; consumption edges represent the consumption of beds, equipment, operating rooms, staff shifts, drugs, and consumables by hospital operation financial events; settlement edges represent the settlement relationship between billing settlement events and medical insurance payment events; aggregation edges represent the aggregation of cost occurrence events to cost accounts, budget items, departments, disease types, and medical insurance disease groups; and evaluation edges represent the evaluation of object nodes and event nodes by indicator nodes. The graph edge table records the edge identifier, starting node, ending node, edge type, business occurrence time, accounting confirmation time, cost aggregation time, medical insurance settlement status, responsible object identifier, and field version identifier.

[0036] The time-series status verification module is connected to the business, finance, and medical insurance mapping module. It is used to determine the analysis status thresholds for each hospital's operational financial events based on the business occurrence time, accounting confirmation time, cost collection time, and medical insurance settlement status. For hospital operational financial events that have not reached the same analysis status threshold, a delayed confirmation flag is set. The time-series status verification module includes a status node configuration unit and a threshold determination unit.

[0037] The status node configuration unit configures six status nodes for the same event chain: Business Occurred, Medical Order Executed, Cost Confirmed, Cost Accumulated, Medical Insurance Settled, and Medical Record Finalized. Each status node records a status code, completion time, source system identifier, and status source field. The "Business Occurred" status node originates from the hospital's business system or electronic medical record system; the "Medical Order Executed" status node originates from the hospital's business system or electronic medical record system; the "Cost Confirmed" status node originates from the billing and settlement data; the "Cost Accumulated" status node originates from the hospital resource planning system; the "Medical Insurance Settled" status node originates from the medical insurance settlement system; and the "Medical Record Finalized" status node originates from the medical record cover page system.

[0038] The water level determination unit reads the completion time of each state node in the same event chain, and takes the cutoff time of the current analysis period as the analysis cutoff time. The state nodes are sequentially ordered as: business occurred, medical order executed, expense confirmed, cost collected, medical insurance settled, and medical record finalized. For any event chain, the water level determination unit determines the analysis state water level according to the following formula: wherein, represents the event chain identifier; represents the cutoff time of the current analysis period; represents the sequence number of the state node; represents the serial number of the consecutively completed state nodes; represents the event chain the th state node of, which takes a value of 1 if completed and 0 if not completed; represents the event chain the th state node completion time of; represents the event chain the analysis state water level corresponding to the cutoff time of the current analysis period. When there are no consecutively completed state nodes in the event chain, the analysis state water level is set to 0.

[0039] The indicator state requirement table is used to record the minimum analysis state water level corresponding to different indicator nodes. The indicator state requirement table includes indicator node identifier, minimum state node, source rule, field version identifier, effective time and expiration time. The minimum state node of the income structure indicator is expense confirmed; the minimum state nodes of the department cost indicator, disease type cost indicator and item cost indicator are cost collected; the minimum state nodes of the medical insurance denial indicator and disease group profit and loss indicator are medical insurance settled; the minimum state node of the indicator nodes related to medical insurance disease groups and medical record coding is medical record finalized. When the analysis state water level of the current event chain is lower than the corresponding minimum state node in the indicator state requirement table, the time sequence state verification module sets a delay confirmation identifier. Hospital operation financial events with the delay confirmation identifier do not participate in the final financial abnormal object generation process, but are retained in the process monitoring data.

[0040] The event-level cost collection module is connected with the business-finance-medical insurance mapping graph module and the time sequence state verification module, and is used for performing cost collection on hospital operation financial events that reach the same analysis state water level, and generating cost collection results by department, disease type, medical insurance disease group and medical team. The event-level cost collection module includes an item cost collection unit, a resource cost collection unit and a account cost collection unit.

[0041] The project cost aggregation unit aggregates the costs of billing items and consumable items based on the mapping relationship between medical order execution events, billing settlement events, and consumable items. Consumable item costs are derived from the outbound unit price, outbound quantity, returned quantity, and cost confirmation status in the materials and consumables system; drug costs are derived from drug requisition, return, and expense confirmation records; and billing item costs are derived from the mapping relationship between billing settlement events and cost accounts in the hospital resource planning system.

[0042] The resource cost aggregation unit aggregates bed occupancy costs, equipment occupancy costs, operating room occupancy costs, and personnel scheduling costs based on resource occupancy events. The unit cost of resources is jointly determined by the cost items in the hospital resource planning system, depreciation and maintenance records in the equipment asset system, bed cost allocation records, and personnel cost records in the scheduling performance system. Bed occupancy costs are aggregated according to the cost allocation records of the ward to which the bed belongs and the duration of bed occupancy; equipment occupancy costs are aggregated according to equipment resource identifiers, equipment depreciation records, maintenance cost records, and equipment usage duration; operating room occupancy costs are aggregated according to operating room cost items and operating room occupancy duration; and personnel scheduling costs are aggregated according to personnel shift identifiers, job cost records, and shift duration. All the above cost data comes from the hospital's existing business and management systems. The event-level cost aggregation module only aggregates costs according to event chains and mapping diagrams, and does not generate fixed cost values ​​that are separate from the source systems.

[0043] The cost collection unit collects cost items and budget items from the hospital resource planning system based on cost occurrence events, and generates cost collection results separately according to department, disease type, medical insurance disease group, and medical group.

[0044] The cost aggregation for a single hospital operational financial event is calculated using the following method: ; in, This indicates a financial event in the hospital's operations; Indicates an event The corresponding event cost; Indicates an event The corresponding cost of the fee items; Indicates an event The corresponding costs of medicines and consumables; Indicates an event The corresponding resource usage cost; Indicates an event The corresponding cost items and their accumulated amounts. The above cost data are sourced from the billing and settlement system, the materials and consumables system, the equipment and asset system, the scheduling and performance system, and the hospital resource planning system, respectively.

[0045] Resource occupancy cost is calculated based on the unit cost of the resource and the duration of occupancy, using the following formula: ; in, Indicates an event The set of resources occupied; This represents a resource object among beds, equipment, operating rooms, and staff shifts; Represents resource objects The unit cost corresponding to the current field version and the current analysis period; Indicates an event Resource-consuming objects Duration or quantity; Indicates an event The corresponding resource occupancy cost. The unit cost of resources is provided by the hospital resource planning system, equipment asset system, and scheduling performance system, which aggregate the costs according to the resource occupancy relationships in the mapping diagram.

[0046] The composite benchmark deviation identification module is connected to the event-level cost aggregation module. It generates objects to be analyzed based on cost aggregation results and identifies deviations in these objects based on budget benchmarks, medical insurance payment benchmarks, and similar benchmarks within the hospital. When an object simultaneously meets two or more of the following conditions, a financial anomaly object is generated. Objects to be analyzed include department-level objects, disease-level objects, medical insurance disease group-level objects, medical group-level objects, project-level objects, and budget project-level objects. Each object contains an object identifier, indicator nodes, analysis period, field version identifier, analysis status level, responsible object identifier, actual indicator value, and a set of associated event identifiers.

[0047] The composite benchmark deviation identification module includes a budget benchmark generation unit, a medical insurance payment benchmark generation unit, and an intra-hospital benchmark generation unit. The budget benchmark generation unit generates budget benchmarks according to budget items, responsible object identifiers, and analysis cycles. Budget benchmarks are derived from the budget amounts, budget control rules, budget decomposition rules, and budget cycles already in effect in the budget management system. The medical insurance payment benchmark generation unit generates medical insurance payment benchmarks according to medical insurance disease groups, medical insurance settlement items, cost consumption index, time consumption index, and medical insurance settlement status. Medical insurance payment benchmarks are derived from the medical insurance settlement system and the medical record homepage system, including medical insurance disease groups, medical insurance payment amounts, settlement status, cost consumption index, and time consumption index. The intra-hospital benchmark generation unit generates intra-hospital benchmarks based on the same disease type, the same medical insurance disease group, the same surgical procedure, the same department, and the same analysis status watermark. Intra-hospital benchmarks use historical event chains that have reached the same analysis status watermark and have consistent field version identifiers.

[0048] Before generating an internal benchmark, the intra-hospital benchmark generation unit reads the benchmark sample number threshold. This threshold is configured in the hospital's data governance rules table and saved along with the field version identifier. If the number of historical objects to be analyzed that meet the criteria of the same disease type, same medical insurance group, same surgical procedure, same department, same field version identifier, and same analysis status level does not reach the benchmark sample number threshold, the intra-hospital benchmark generation unit will not generate an intra-hospital benchmark and will set a benchmark sample insufficiency flag for the current object to be analyzed. Objects to be analyzed with a benchmark sample insufficiency flag will not enter the composite benchmark deviation identification process.

[0049] Within the institute, similar benchmarks are generated using the median, calculated as follows: ; in, Indicates the object currently being analyzed; Indicates the historical object to be analyzed; Indicates the historical object to be analyzed. The actual indicator value; Indicates the object to be analyzed. A collection of historical subjects who share the same disease type, the same medical insurance disease group, the same surgical procedure, and the same department; Indicates the historical object to be analyzed. The field is the version identifier; Indicates the object to be analyzed. The field is the version identifier; Indicates the historical object to be analyzed. Analysis of the water level status; Indicates the object to be analyzed. Analysis of the water level status; Indicates the object to be analyzed. The same benchmark within the hospital.

[0050] The composite benchmark deviation identification module calculates the deviations between the object under analysis and the budget benchmark, the medical insurance payment benchmark, and similar benchmarks within the hospital. The deviation calculations use a standardized format: ; in, Indicates the object to be analyzed; Indicates the reference type, The time represents the budget benchmark. The time represents the medical insurance payment benchmark. The time indicates the benchmark of the same type within the institute; Indicates the object to be analyzed The actual indicator value; Indicates the object to be analyzed The corresponding number Class benchmark value; Indicates the object to be analyzed Compared to the first Deviation rate of the benchmark; This indicates a positive number that prevents the denominator from being zero. Recorded in the system parameter table, the minimum measurement precision is determined according to the corresponding indicator amount field in the database. This is used to avoid the denominator being zero and is not used as a threshold for judging financial anomalies.

[0051] Budget deviation thresholds are determined by the budget control rules in the budget management system or the budget threshold table corresponding to the hospital's budget management system; medical insurance payment deviation thresholds are determined by the medical insurance settlement system, the medical insurance cost control rule table, and the medical insurance settlement rules maintained by the hospital's medical insurance management department; intra-hospital similar deviation thresholds are determined by the indicator distribution results of similar objects in the archived historical analysis period. Similar objects refer to historical objects to be analyzed that have the same field version identifier, the same analysis status water level, and share the same disease type, the same medical insurance disease group, the same surgical procedure, and the same department. The threshold table includes indicator nodes, responsible object type, benchmark type, threshold source, threshold value, effective time, expiration time, and field version identifier. When the threshold table changes, the system generates a new threshold version record and does not rewrite the archived historical analysis results with the changed threshold.

[0052] The determination of financial anomalies adopts a composite triggering method: ; in, Indicates the object to be analyzed The exception flag has a value of 1 indicating that a financial exception object is generated, and a value of 0 indicating that a financial exception object is not generated. Indicates the budget deviation rate; Indicates the medical insurance payment deviation rate; This indicates the deviation rate of the same type within the hospital; Indicates the object to be analyzed The corresponding budget deviation threshold; Indicates the object to be analyzed The corresponding medical insurance payment deviation threshold; Indicates the object to be analyzed The corresponding deviation threshold for the same type within the hospital; This indicates an indicator function; the value is 1 when the condition within the parentheses is true, and 0 when it is false. The composite benchmark deviation identification module reads the threshold table according to budget items, medical insurance disease groups, and responsible object identifiers. When at least two of the budget deviation, medical insurance payment deviation, and similar intra-hospital deviations exceed the corresponding thresholds in the threshold table, the object to be analyzed is marked as a financially abnormal object, and the deviation type, deviation value, responsible object identifier, and associated event identifier that triggered the financially abnormal object are recorded.

[0053] The anomaly attribution module is connected to the composite benchmark deviation identification module and the business-finance-medical insurance mapping module. It is used to trace back related hospital operational and financial events along the mapping map, starting with anomaly objects in the financial data, and generate causal chains based on event contribution, chronological order, consistency of responsible parties, and rule matching. The causal chain includes anomaly indicators, responsible parties, related events, and evidentiary data. The anomaly attribution module includes a candidate causal chain generation unit and a causal chain sorting unit.

[0054] The candidate cause chain generation unit starts with the indicator node corresponding to the financial anomaly object, and traces back along the mapping graph to the event node and object node to generate a candidate cause chain. The candidate cause chains include the treatment path deviation chain, the consumable usage structure anomaly chain, the medical record coding deviation chain, the bed day cost increase chain, the examination waiting time chain, the medical insurance refusal chain, and the budget allocation deviation chain. The treatment pathway deviation chain is generated based on clinical pathway nodes, medical order execution events, and actual resource occupancy events; the consumable usage structure anomaly chain is generated based on consumable items, consumable costs, and historical consumable structures for the same disease and medical insurance group; the medical record coding deviation chain is generated based on diagnosis codes, surgical codes, finalized medical record status, and medical insurance group change records in the medical record homepage system; the bed day cost increase chain is generated based on admission time, discharge time, bed occupancy cost, and the benchmark for the same disease; the examination waiting time chain is generated based on medical technology examination application time, report completion time, length of stay, and resource occupancy events; the medical insurance refusal chain is generated based on the medical insurance refusal amount, refusal reason code, and billing settlement events; and the budget allocation deviation chain is generated based on budget items, cost items, and responsible object identifiers.

[0055] Event contribution, chronological order, consistency of responsible parties, and rule matching degree all use normalized values ​​between 0 and 1. Event contribution is determined by the proportion of the amount of related events in the candidate cause chain to the deviation amount of the financial anomaly object; chronological order is determined by whether the occurrence time of related events in the candidate cause chain is earlier than the formation time of the financial anomaly object; consistency of responsible parties is determined by whether the responsible party identifier in the candidate cause chain is consistent with the responsible party identifier in the financial anomaly object; rule matching degree is determined by whether the candidate cause chain type matches the cause type, triggering condition, and evidence field in the cause rule table. The cause rule table includes cause chain type, triggering condition, evidence field, applicable indicator node, applicable responsible party type, field version identifier, and effective time. The cause chain sorting unit generates a sorting value according to the cause rule table and writes the sorting position, candidate cause chain type, and evidence data into the cause chain list.

[0056] The cause chain sort value is calculated using the following formula: ; in, Indicates the first One candidate cause chain; Indicates the first The sorting value of the candidate cause chains; Indicates the first The contribution of the event amount or cost corresponding to each candidate cause chain to the deviation amount of the financial anomaly object; Indicates the first The degree of chronological match between the occurrence time of events and the formation time of financial anomalies in the candidate cause chain; Indicates the first The degree of consistency between the responsible party identifier in the candidate cause chain and the responsible party identifier of the financial anomaly object; Indicates the first The degree of matching between each candidate cause chain and the cause type in the cause rule table; , , , These represent the ranking weights, all four of which are recorded in the cause chain weight table and satisfy the following conditions: When the cause chain weight table changes, the closed-loop feedback module records the weight version and writes the weight version into the cause chain list and the exception feedback record table.

[0057] The closed-loop feedback module is connected to the anomaly attribution module. It generates anomaly feedback records based on the causal chain and retests the corresponding financial anomalies within subsequent analysis cycles according to the same field version identifier, the same analysis status level, and the same scope of responsible parties. Based on the retest results, it updates the ranking weights of similar benchmarks and causal chains within the institute. The closed-loop feedback module includes an anomaly feedback record generation unit, a retest sample locking unit, and a weight update unit.

[0058] The anomaly feedback record generation unit generates anomaly feedback records based on the cause chain. Anomaly feedback records include anomaly object identifier, anomaly indicator, responsible party, related events, evidence data, candidate cause chain type, sorting position, field version identifier, analysis status level, retest trigger time, and retest cycle. The retest trigger time and retest cycle are determined by the retest rule table. The retest rule table includes anomaly indicator, responsible party type, retest trigger conditions, retest cycle, data status requirements, field version identifier, and effective time. Anomaly feedback records related to budget execution are retested according to the budget analysis cycle; anomaly feedback records related to medical insurance payment are retested according to the medical insurance settlement cycle; and anomaly feedback records related to cost collection are retested according to the cost collection cycle.

[0059] The retest sample locking unit locks retest samples within subsequent analysis periods based on the same field version identifier, the same analysis status level, the same scope of responsible objects, and the same indicator nodes. The retest sample locking unit does not include data whose field version, analysis status level, or scope of responsible objects has changed in the same retest sample. The weight update unit updates the ranking weights of similar benchmarks and causal chains within the institute based on the deviation changes of financial anomalies in the retest samples.

[0060] During retesting, the system reads the deviation rate of financial anomalies before rectification and the deviation rate of retested samples after rectification, forming the deviation change. For cause chains that rank high and whose corresponding deviations decrease after retesting, the ranking weight of the rule matching item corresponding to that cause chain is increased; for cause chains that rank high but whose deviations do not decrease after retesting, the ranking weight of the rule matching item corresponding to that cause chain is decreased. The weight update uses the following formula: ; in, Indicates the first The weight of the cause chain before the update; Indicates the first The updated weight of the cause chain; This indicates the update step size, which is recorded in the cause chain weight update parameter table and is used to limit the adjustment range of the sorting weight in a single retest. Indicates the first The decrease in bias of the retested samples corresponding to the causal chain; Indicates the category number of the reason chain that participated in this update; This indicates the number of causal chain categories involved in this update. Deviation decrease. It is obtained by subtracting the deviation rate after rectification from the deviation rate before rectification; when the deviation does not decrease after rectification. Less than or equal to zero. This formula is used to ensure that the weights of each cause chain remain normalized after the update.

[0061] In a specific operational process, the system analyzes the profit and loss of a particular medical insurance disease group. The data source adaptation module reads patient admission, discharge, billing, and medical order execution data for that medical insurance disease group from the hospital's business system; diagnostic, surgical, and clinical pathway data from the electronic medical record system; the medical insurance disease group, diagnosis code, surgical code, and finalized medical record status from the medical record homepage system; medical insurance payment amount, rejected payment amount, cost consumption index, time consumption index, and medical insurance settlement status from the medical insurance settlement system; consumables issuance and return data from the materials and supplies system; equipment usage time from the equipment asset system; personnel shifts from the scheduling performance system; and cost items and budget items from the hospital resource planning system. The field registration unit confirms the field version identifiers corresponding to the above data, and the source verification unit performs mapping verification of medical order items, billing items, medical insurance settlement items, and cost items.

[0062] The event-based transformation module converts the above data into events such as diagnosis and treatment occurrence, medical order execution, billing settlement, cost occurrence, medical insurance payment, resource occupancy, and budget consumption, and merges them into an event chain with patient visit identifiers and medical insurance disease groups as the main lines. The business, finance, and medical insurance mapping module associates this event chain with patients, diseases, medical insurance disease groups, departments, medical groups, doctors, consumables, beds, equipment, budget items, and disease group profit and loss indicators. The time-series status verification module determines whether the event chain has reached the status of medical insurance settlement and medical record finalization; if medical insurance settlement has not been completed or the medical record front page has not been finalized, the event chain is set with a delayed confirmation mark and will not enter the final disease group profit and loss anomaly object generation process; if medical insurance settlement and medical record finalization have both been completed, the event-level cost aggregation module performs cost aggregation on the event chain.

[0063] The composite benchmark deviation identification module reads the budget benchmark, medical insurance payment benchmark, and intra-hospital benchmark corresponding to the medical insurance disease group. The budget benchmark comes from the budget items of the medical insurance disease group or the relevant responsible department in the budget management system; the medical insurance payment benchmark comes from the medical insurance disease group payment rules, cost consumption index, and time consumption index in the medical insurance settlement system; the intra-hospital benchmark comes from the historical event chain of the same disease, the same medical insurance disease group, the same surgical procedure, the same department, the same field version, and the same analysis status level. If the object to be analyzed exceeds the corresponding threshold in the threshold table in at least two of the budget deviation, medical insurance payment deviation, and intra-hospital benchmark deviation, the system generates a financial anomaly object and records the triggered deviation type, deviation value, responsible object identifier, and associated event identifier.

[0064] The anomaly attribution module starts with the disease group profit and loss indicator node and traces back along the mapping graph to medical insurance payment events, billing settlement events, medical order execution events, consumable consumption events, resource occupancy events, and medical record cover page events. When the backtracking results show that the cost consumption index is too high, the consumable cost ratio exceeds the hospital's benchmark, or there are records of changes in medical insurance disease groups in the medical record coding, the system generates anomaly chains for consumable usage structure, medical record coding deviation chains, and medical insurance payment deviation chains, respectively, and sorts them according to the contribution of the events, their temporal relationship, the consistency of the responsible parties, and the degree of rule matching. The closed-loop feedback module writes the sorted cause chains into the anomaly feedback record and locks the retest samples in subsequent analysis cycles according to the same field version, the same analysis status water level, and the same scope of responsible parties.

[0065] In another specific operational process, the system analyzes anomalies in the departmental costs of a particular clinical department. The data source adaptation module reads the corresponding cost items and budget items for the department from the hospital resource planning system, outpatient, inpatient, surgical, and billing data from the hospital business system, drug and consumable consumption data from the materials and supplies system, equipment occupancy data from the equipment asset system, and personnel shift data from the scheduling and performance system. The event-based transformation module converts the above data into cost occurrence events, medical order execution events, billing settlement events, resource occupancy events, and budget consumption events. The business-finance-medical insurance mapping module associates these events with departments, wards, medical teams, doctors, diseases, equipment, drugs and consumables, and budget items.

[0066] The timing status verification module determines whether the event chain related to the department's cost indicators has reached the cost aggregation state. Event chains that have not reached the cost aggregation state are marked with a delayed confirmation flag and do not enter the final abnormal object generation process. The event-level cost aggregation module aggregates the event chains that have reached the cost aggregation state, forming departmental costs, disease costs, project costs, and resource occupancy costs. The composite benchmark deviation identification module compares the department's costs with the budget benchmark, medical insurance payment benchmark, and similar benchmarks within the hospital; when at least two benchmarks show deviations, a financial abnormal object is generated. The abnormal attribution module further traces back consumable items, equipment resources, staff scheduling, and budget items, forming abnormal chains for consumable usage structure, equipment occupancy costs, bed-day cost increases, and budget allocation deviations. The closed-loop feedback module retests the department's cost indicators in the next analysis cycle using the same field version and the same scope of responsibility, and updates the ranking weights of similar benchmarks within the hospital and the cause chains based on the retest results.

[0067] In this embodiment, the data flow between modules is as follows: the data source adaptation module outputs raw data with field version identifiers and project mapping results; the event-based transformation module outputs hospital operation financial events and event chain identifiers; the business finance medical insurance mapping graph module outputs object nodes, event nodes, indicator nodes, and related edges; the time-series status verification module outputs analysis status water level and delay confirmation identifiers; the event-level cost aggregation module outputs cost aggregation results; the composite benchmark deviation identification module outputs financial anomaly objects; the anomaly attribution module outputs cause chains; and the closed-loop feedback module outputs anomaly feedback records, retest results, and updated hospital-wide benchmarks and cause chain sorting weights.

[0068] 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, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent financial analysis system based on multi-source heterogeneous data fusion, characterized in that, include: The module includes a data source adaptation module, an event-based conversion module, a business, finance, and medical insurance mapping module, a time-series status verification module, an event-level cost aggregation module, a composite benchmark deviation identification module, an anomaly attribution module, and a closed-loop feedback module. The data source adaptation module is used to access multi-source business data from the hospital, obtain raw data, and add source system identifier, time identifier, data status identifier, responsible object identifier, and field version identifier to the raw data; The event-based transformation module connects with the data source adaptation module to convert raw data into hospital operation and financial events according to patient visit identifiers, responsible party identifiers, and project identifiers, and merges related hospital operation and financial events into an event chain. The business, finance, and medical insurance mapping graph module is connected to the event-based transformation module. It is used to establish a mapping graph including object nodes, event nodes, and indicator nodes based on the event chain, and to generate association edges with time identifiers, data status identifiers, responsible object identifiers, and field version identifiers between object nodes, event nodes, and indicator nodes. The timing status verification module is connected to the business, finance and medical insurance mapping graph module. It is used to determine the analysis status water level based on the data status identifier of each hospital's operational financial event in the event chain, and to set a delayed confirmation identifier for hospital operational financial events that have not reached the required analysis status water level of the corresponding indicator node. The event-level cost aggregation module is connected to the business, finance, and medical insurance mapping module and the time-series status verification module. It is used to aggregate event-level costs for hospital operation and financial events that have not been marked with delayed confirmation, and to generate cost aggregation results corresponding to the responsible objects. The composite benchmark deviation identification module is connected to the event-level cost aggregation module. It is used to generate objects to be analyzed based on the cost aggregation results, and to identify deviations of the objects to be analyzed based on the budget benchmark, medical insurance payment benchmark and similar benchmarks within the hospital. When the objects to be analyzed meet the deviation conditions corresponding to at least two types of benchmarks, financial anomaly objects are generated. The anomaly attribution module is connected to the composite benchmark deviation identification module and the business finance and medical insurance mapping graph module. It is used to trace back the associated hospital operation and financial events and responsible parties along the mapping graph, starting from the financial anomaly object, and generate a cause chain including anomaly indicators, responsible parties, related events and evidence data. The closed-loop feedback module is connected to the anomaly attribution module. It is used to generate anomaly feedback records based on the cause chain and lock the retest samples in subsequent analysis cycles according to the same field version identifier, the same analysis status level, and the same scope of responsible objects. It updates the hospital's similar benchmarks and cause chain ranking weights based on the retest results.

2. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The data source adaptation module accesses multi-source business data from hospitals, including data from hospital business systems, electronic medical record systems, hospital resource planning systems, medical insurance settlement systems, medical record front page systems, material and consumable systems, equipment and asset systems, shift scheduling and performance systems, and budget management systems. The time identifiers include data generation time, business occurrence time, accounting confirmation time, cost collection time, and medical insurance settlement time; The data status identifiers include cost confirmation status, cost collection status, medical insurance settlement status, and medical record finalization status.

3. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 2, characterized in that, The data source adaptation module includes a field registration unit, a source verification unit, and a caliber version unit. The field registration unit is used to establish a field registration table, which includes field name, field type, source system identifier, business meaning, value range, effective time and expiration time; The source verification unit is used to perform mapping verification on the medical order item identifier, charge item identifier, medical insurance settlement item identifier and cost item identifier under the same patient visit identifier according to the preset item mapping table, and set an item mismatch identifier for the original data that does not match the corresponding relationship; The caliber version unit is used to generate a field version identifier when the field name, field type, business meaning, and value range of the field registration table change.

4. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 3, characterized in that, The event-based conversion module includes an event template library and an event merging unit; The event template library stores templates for diagnosis and treatment events, medical order execution events, billing and settlement events, cost occurrence events, medical insurance payment events, resource occupancy events, and budget consumption events. Each event template includes an event type, a main object identifier, a responsible object identifier, a business time field, an amount field, a quantity field, a status field, and a related item field. The event merging unit is used to merge hospital operation and financial events with the same patient visit identifier, the same responsible object identifier, a mapping relationship between related project fields, and business occurrence time within the start and end time fields of the corresponding event template into the same event chain.

5. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 4, characterized in that, The business finance and medical insurance mapping graph module includes an object relationship generation unit, a project relationship generation unit, and an indicator relationship generation unit. The object relationship generation unit is used to establish the affiliation relationships between patients, departments, wards, medical groups, doctors, diseases, and medical insurance disease groups; The project relationship generation unit is used to establish mapping relationships between medical order items, billing items, medical insurance settlement items, consumable items, equipment items, cost items, and budget items; The indicator relationship generation unit is used to associate revenue structure indicators, department cost indicators, disease cost indicators, project cost indicators, medical insurance refusal indicators, disease group profit and loss indicators, and budget execution indicators to the corresponding object nodes and event nodes, respectively.

6. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 5, characterized in that, The timing state verification module includes a state node configuration unit and a water level determination unit; The status node configuration unit is used to configure status nodes for the same event chain, such as "business has occurred", "medical order has been executed", "fees have been confirmed", "costs have been collected", "medical insurance has been settled", and "medical record has been finalized". The water level determination unit is used to read the completion time of each state node in the same event chain, and to determine the state node that has been completed consecutively and is the last in the order in the same event chain as the analysis state water level of the event chain. When the analysis status level of the event chain is lower than the lowest status node of the corresponding indicator node in the indicator status requirement table, the level determination unit sets a delayed confirmation flag for the hospital operation and financial event corresponding to the event chain.

7. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 6, characterized in that, The event-level cost aggregation module includes a project cost aggregation unit, a resource cost aggregation unit, and a subject cost aggregation unit; The project cost collection unit is used to collect the cost of billing items and the cost of consumable items according to the mapping relationship between medical order execution events, billing settlement events and consumable items. The resource cost collection unit is used to collect bed occupancy costs, equipment occupancy costs, operating room occupancy costs, and staff scheduling costs based on resource occupancy events; The cost collection unit is used to collect cost items and budget items in the hospital resource planning system based on cost occurrence events, and generate cost collection results according to department, disease type, medical insurance disease group and medical group.

8. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 7, characterized in that, The composite benchmark deviation identification module includes a budget benchmark generation unit, a medical insurance payment benchmark generation unit, and an in-hospital benchmark generation unit; The budget baseline generation unit is used to generate budget baselines according to budget items, responsible object identifiers, and analysis cycles; The medical insurance payment benchmark generation unit is used to generate medical insurance payment benchmarks according to medical insurance disease groups, medical insurance settlement items, cost consumption index, time consumption index and medical insurance settlement status; The in-hospital benchmark generation unit is used to generate in-hospital benchmarks based on the same disease type, the same medical insurance disease group, the same surgical procedure, the same department, the same field version identifier, and the same analysis status water level.

9. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 8, characterized in that, The composite benchmark deviation identification module calculates the budget deviation between the object to be analyzed and the budget benchmark, the medical insurance payment deviation between the object to be analyzed and the medical insurance payment benchmark, and the intra-hospital similar deviation between the object to be analyzed and the same benchmark within the hospital, and reads the threshold table according to the budget item, medical insurance disease group and responsible object identifier; When at least two of the following—budget deviation, medical insurance payment deviation, and similar deviation within the hospital—exceed the corresponding threshold in the threshold table, the object to be analyzed will be marked as a financially abnormal object, and the deviation type, deviation value, responsible object identifier, and associated event identifier that triggered the financially abnormal object will be recorded.

10. The intelligent financial analysis system based on multi-source heterogeneous data fusion according to claim 9, characterized in that, The anomaly attribution module includes a candidate cause chain generation unit and a cause chain sorting unit; The candidate cause chain generation unit is used to generate candidate cause chains by starting with the indicator node corresponding to the financial anomaly object, tracing back along the mapping graph to the event node and object node in sequence. The candidate cause chains include the treatment path deviation chain, the consumable usage structure anomaly chain, the medical record coding deviation chain, the bed day cost increase chain, the examination waiting time chain, the medical insurance refusal chain, and the budget allocation deviation chain. The cause chain sorting unit is used to sort candidate cause chains according to the contribution of events, chronological order, consistency of responsible parties, and degree of rule matching, and write the sorting position, candidate cause chain type, and corresponding evidence data into the cause chain; The closed-loop feedback module includes an anomaly feedback record generation unit, a retest sample locking unit, and a weight update unit. The abnormal feedback record generation unit is used to generate abnormal feedback records based on the cause chain, including abnormal indicators, responsible parties, related events, evidence data, retest trigger time, and retest cycle. The retest sample locking unit is used to lock retest samples in subsequent analysis cycles according to the same field version identifier, the same analysis status level, the same scope of responsible objects, and the same indicator nodes. The weight update unit is used to update the ranking weights of similar benchmarks and causal chains within the institute based on the deviation changes of financially abnormal objects in the retest samples.

Images