Medical data maintenance method, device and equipment of multi-source business system of medical industry and medium

Abstract

The application discloses a medical data maintenance method, device and equipment of a multi-source business system in the medical industry and a medium, and relates to the technical field of data processing. The method comprises the following steps: generating a change event fingerprint for uniquely identifying a data change event; generating a minimum affected subgraph based on the change event fingerprint and a pre-constructed index dependency graph, and performing hierarchical verification on the minimum affected subgraph, including structure consistency verification, semantic correlation verification and index formula consistency verification; if any hierarchical verification fails, performing local compensation repair on the minimum affected subgraph according to the priority of a pre-configured trusted repair source, and performing local recalculation on the repaired minimum affected subgraph; performing selective re-verification on the local recalculation result, and generating an audit chain record containing at least the change event fingerprint, the trusted repair source evidence and the re-verification result after the re-verification passes. The application can solve the data consistency problem caused by complex field dependence and variable index caliber in the medical statistical scene.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to methods, apparatus, equipment and media for maintaining medical data in multi-source business systems in the medical industry. Background Technology

[0002] Currently, the data integration, aggregation, and statistical supervision of multi-source medical business systems generally adopt a sequential processing mode of "synchronization first, verification later, and manual repair after anomalies." This approach can only perceive data additions, modifications, and deletions at a coarse-grained level, without establishing a unified change event identifier for end-to-end tracking; it fails to establish dependencies between fields, definitions (statistical constraints of indicators), formulas, and statistical indicators, making it impossible to pinpoint the range of indicator anomalies caused by field changes; verification is limited to format and quantity levels, making it difficult to identify issues where records are synchronized normally but indicator definitions are incorrect; anomaly repair relies on manual intervention, requiring full review after repair, lacking automated and reliable repair and partial recalculation mechanisms. These shortcomings result in high data processing latency, high resource consumption, and difficulty in traceability, and are prone to problems such as duplicate processing, omissions, miscalculations, and insufficient repair basis, failing to meet the requirements of medical business analysis and regulatory reporting for data accuracy, timeliness, and traceability. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a method, apparatus, equipment, and medium for maintaining medical data in multi-source business systems in the medical industry, capable of achieving refined tracking, selective recalculation, and automated reliable repair of integrated medical data. The specific solution is as follows: Firstly, this application discloses a method for maintaining medical data in a multi-source business system in the medical industry, including: Collect data change events from multiple business systems in the medical industry and extract the corresponding change information to generate a change event fingerprint that uniquely identifies the data change event; Based on the change event fingerprint and the pre-built indicator dependency graph, a minimum affected subgraph covering the scope of the data change event is generated, and hierarchical verification is performed on the minimum affected subgraph; the hierarchical verification includes a first-level structural consistency verification, a second-level semantic association verification, and a third-level indicator formula consistency verification. If any level of verification fails, the least affected subgraph is locally compensated and repaired according to the priority of the pre-configured trusted repair source, and after the compensation and repair is completed, the repaired least affected subgraph is locally recalculated based on the index dependency graph. Selective re-verification is performed on the partial recalculation results, and after the re-verification passes, an audit chain record is generated that includes at least the change event fingerprint, trusted repair source evidence, and the re-verification results.

[0004] Optionally, after generating the minimum affected subgraph covering the scope of the data change event, the method further includes: Data related to the least affected subgraph is selectively synchronized to the target storage area, and a snapshot summary of the source data is generated before synchronization and a snapshot summary of the target data is generated after synchronization. Specifically, for unstructured data, the snapshot summary is a content summary or structured index extracted from the data; the snapshot summary is used for data comparison and consistency verification in the hierarchical verification.

[0005] Optionally, the step of collecting data change events from multiple business systems in the medical industry and extracting corresponding change information to generate a change event fingerprint for uniquely identifying the data change event includes: Collect data change events from multiple business systems in the healthcare industry, and extract source system identifiers, business primary keys, transaction sequence numbers, change types, set of change fields, change timestamps, and operator information; The set of changed fields in the changed information is encoded into a changed field bitmap, and combined or hashed with other changed information in the changed information to generate a changed event fingerprint that uniquely identifies the data changed event.

[0006] Optionally, constructing the metric dependency graph includes: Obtain the indicator semantic templates pre-configured for statistical indicators in medical statistics; the indicator semantic templates are used to describe the calculation rules of the statistical indicators; An indicator dependency graph is constructed based on the indicator semantic template; the indicator dependency graph is a directed graph structure composed of nodes and edges that describe data mapping relationships and indicator derivation dependency relationships. Accordingly, generating a minimum affected subgraph covering the scope of the data change event based on the change event fingerprint and a pre-built metric dependency graph includes: The changed fields are determined based on the change event fingerprint, and the changed fields are mapped to target fields on the integration platform side through a preset field mapping table; Starting with the target field as the starting node, traverse along the edges of the index dependency graph to collect nodes that have direct or indirect dependencies on the target field, and use the connected subgraph formed by the edges between the nodes as the least affected subgraph.

[0007] Optionally, the step of performing local compensation repair on the least affected subgraph according to the priority of pre-configured trusted repair sources includes: Multiple candidate repair sources are obtained, and the applicability of the candidate repair sources is verified to filter out invalid candidate values ​​that do not meet the current business rules and applicable scope. The valid candidate values ​​that pass the applicability check are sorted according to the priority of the pre-configured trusted repair sources to determine an ordered candidate sequence from high to low priority; A consistency comparison is performed on the valid candidate values ​​in the ordered candidate sequence. If the target valid candidate value corresponding to the highest priority is unique, or if the highest priority corresponds to multiple target valid candidate values ​​and the contents of the target valid candidate values ​​are consistent, then the target valid candidate value is used as the repair value to perform local compensation repair on the least affected subgraph. If the contents of the multiple valid candidate values ​​of the target corresponding to the highest priority are inconsistent, a comprehensive score is calculated for each valid candidate value of the target based on a preset weighted calculation formula. When the difference between the highest and second-highest scores in the comprehensive score is greater than a preset threshold, the target valid candidate value corresponding to the highest score in the comprehensive score is used as the repair value to perform local compensation repair on the least affected subgraph.

[0008] Optionally, the step of locally recalculating the least affected subgraph after repair based on the index dependency graph includes: If the index dependency graph is a directed acyclic graph, then the repaired least affected subgraph is sorted using a preset topological sorting algorithm to determine the topological order. The repaired minimum affected subgraph is locally recalculated sequentially according to the topological order. If a cycle dependency is detected during the sorting process, an alarm is triggered and the local recalculation is terminated.

[0009] Optionally, the method for maintaining medical data in a multi-source business system in the medical industry further includes: When multiple consecutive data change events occur for the same business primary key, they are sorted according to the transaction sequence number or change timestamp of each data change event. Based on the sorting results, the audit chain record of the current data change event is referenced with the summary of the audit chain record of the previous data change event to generate a chain-like trace record of multiple changes to the same business object.

[0010] Secondly, this application discloses a medical data maintenance device for a multi-source business system in the medical industry, comprising: The change acquisition module is used to collect data change events from multiple business systems in the healthcare industry. The event fingerprint generation module is used to extract the corresponding change information to generate a change event fingerprint that uniquely identifies the data change event; The three-level verification module is used to generate the minimum affected subgraph covering the scope of the data change event based on the change event fingerprint and the pre-built indicator dependency graph, and to perform hierarchical verification on the minimum affected subgraph; the hierarchical verification includes the first level of structural consistency verification, the second level of semantic association verification, and the third level of indicator formula consistency verification. The local compensation and repair module is used to perform local compensation and repair on the least affected subgraph according to the priority of the pre-configured trusted repair source if any level of verification fails. The local recalculation module is used to perform local recalculation of the least affected subgraph after compensation and repair are completed, based on the index dependency graph. The selective re-verification module is used to selectively re-verify the results of local recalculation. The audit chain record module is used to generate an audit chain record that includes at least the change event fingerprint, trusted repair source evidence, and the re-verification result after the re-verification is passed.

[0011] Thirdly, this application discloses an electronic device, which includes a processor and a memory; wherein the memory is used to store a computer program, which is loaded and executed by the processor to implement the aforementioned medical data maintenance method for a multi-source business system in the medical industry.

[0012] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the medical data maintenance method for a multi-source business system in the medical industry as described above.

[0013] Beneficial Effects: This application employs a combination of event-driven and asynchronous processing. Upon receiving a data change event from the source, a unified change event fingerprint is first generated. This ensures that each basic data change possesses a unique and traceable event identifier throughout the synchronization, verification, repair, re-verification, and auditing stages, maintaining the association of the same event. This fundamentally solves the long-standing technical problem of traditional coarse-grained parsing methods being unable to track the flow of the same change event across multiple stages. Secondly, the metric dependency graph is used to represent how changes in source fields affect target metrics along the dependency chain. Therefore, locating the least affected subgraph based on the metric dependency graph ensures that when a field changes, only the local area with which it depends is processed, significantly reducing redundant scanning of irrelevant records and metrics, and lowering latency and computational complexity. The system optimizes resource consumption. Furthermore, the three-level verification system overcomes the limitations of traditional methods that only check field formats, delving into semantic relationships and indicator formulas to address the hidden problem of correct record synchronization but incorrect derived indicators. Subsequently, it only initiates trusted repair source determination and partial compensation recalculation when verification fails. Trusted repair sources are automatically selected based on preset priorities, ensuring clear technical justification for repair actions and preventing secondary errors caused by arbitrary value adjustments. After repair, only the least affected subgraphs undergo partial recalculation and selective re-verification, further reducing the processing scope and ensuring repair efficiency and accuracy. Finally, it generates an append-only audit chain record, forming a closed-loop event-level traceability evidence that meets the compliance and traceability requirements of medical data supervision. Therefore, this application enables efficient, accurate, low-cost, and traceable verification, repair, and recalculation of integrated medical data, suitable for medical business analysis and supervision scenarios with complex indicator definitions and high data throughput.

[0014] Furthermore, the medical data maintenance device, equipment, and storage medium for a multi-source business system in the medical industry provided in this application correspond to the aforementioned medical data maintenance method for a multi-source business system in the medical industry, and have the same effect. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 This application discloses a flowchart of a method for maintaining medical data in a multi-source business system in the medical industry. Figure 2 This application discloses a schematic diagram of a medical data maintenance device for a multi-source business system in the medical industry. Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0017] 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.

[0018] With the advancement of smart hospital construction, intelligent medical insurance supervision, and the interconnection of medical data, hospitals generally need to aggregate business data from outpatient, emergency, inpatient, laboratory, examination, pharmaceutical, billing, and internet hospital systems onto an integrated platform for medical quality analysis, operational monitoring, medical insurance review, research statistics, and external reporting. However, in this process, the current synchronization of multiple source business systems such as hospital HIS (Hospital Information System), LIS (Laboratory Information System), PACS (Picture Archiving and Communication System), electronic medical records, and internet hospitals with the integrated platform faces the following urgent technical bottlenecks: (1) Difficulty in tracking changes: The addition, modification, deletion and withdrawal of source records can often only be perceived as "data changes" in a coarse-grained manner. There is a lack of a unified event identifier that runs through the entire process of synchronization, verification, repair and review, which makes it impossible to track the same change in different processing stages. This can easily lead to duplicate processing, omissions, disordered order and difficulty in tracing. (2) Low efficiency of verification and repair: Currently, most of the field format verification, record count verification or manual spot check are performed after synchronization. It is impossible to determine the truly affected data subset and affected indicator set based on "the field set, business primary key, indicator formula and caliber boundary of this specific change", which leads to the system having to perform large-scale rescan, full review or manual location, resulting in high processing latency and high resource consumption. (3) Lack of semantic association between indicators: For derived indicators in hospital operation analysis and supervision scenarios, there is currently no unified dependency relationship between "data fields - statistical indicators - caliber rules - formula templates". For example, indicators such as total outpatient visits, outpatient visits, average outpatient cost, drug ratio, consumable ratio, and basic drug list cost ratio have differences in caliber, such as whether they include Internet hospitals, whether they exclude cancelled appointments, and whether they exclude physical examination revenue; when the source fields change, it is currently impossible to automatically locate the affected indicators and perform selective recalculation and verification. (4) Insufficient automation: Existing anomaly repairs mostly rely on manual verification and lack an automated mechanism of "trustworthy repair source priority + uniqueness judgment + local compensation recalculation"; after repair, the entire table or batch of data is often re-verified, and it is impossible to achieve re-verification only for the affected subgraphs; at the same time, log records mostly stay at the level of synchronization success / failure, making it difficult to form chain audit evidence for specific change events, repair basis and re-verification results.

[0019] To address the aforementioned issues, understanding the technological background and existing solutions is crucial. Current integration solutions typically employ interface synchronization, message buses, ETL (Extract-Transform-Load) incremental synchronization, or database log collection to achieve multi-source aggregation. However, the core of most solutions remains a sequential processing model of "synchronization first, verification later, and manual repair after anomalies." This model is more suitable for general data exchange but not for medical data supervision scenarios with complex definitions, numerous field dependencies, and strong indicator derivation. Existing technologies generally fall into the following categories: first, data synchronization solutions based on timed incremental synchronization or full data patching; second, static verification solutions based on field format, non-empty, uniqueness, or record quantity; third, manual repair solutions based on anomaly lists; and fourth, auditing solutions that use log recording as an auxiliary function.

[0020] The reasons for the above technical problems are as follows: (1) The current technology adopts the "timed synchronization or coarse-grained incremental synchronization" mechanism, which makes it impossible to map the source changes to the target processing actions one by one at the event level, resulting in the problem that the synchronization is successful but the verification object is inaccurate and it is difficult to confirm whether all affected indicators are covered after the repair; (2) The current technology only performs structural verification or quantity verification on the record itself, and does not include the scope boundary, formula expression and dimension filtering conditions of statistical indicators in the verification model, which means that although the source field changes have been synchronized, the derived indicators may still be wrong; (3) The existing repair methods usually directly modify the target side error value or re-trigger synchronization, and lack the judgment process of selecting a reliable repair source from the current valid value of the source business system, the master data index mapping value, the standard coding dictionary value, the medical insurance catalog rule value, and the associated value that has passed the previous verification, so the repair action lacks technical basis stability; (4) The existing review methods usually re-scan the full table, full batch or whole day data, and do not determine the minimum review scope according to the dependency path reached by this change, resulting in large amount of calculation, poor timeliness, and easy to cover the real anomaly due to irrelevant data noise.

[0021] It is evident that while the aforementioned solutions can resolve some data transmission or formatting errors, they share common shortcomings: they lack the generation of a unified change event fingerprint that spans all stages; they fail to establish a dependency graph of "source field—target field—indicator formula—caliber boundary"; they lack selective recalculation and re-verification around the least affected subgraph; and they lack a reliable repair source determination mechanism centered on the repair basis. Therefore, this application provides a medical data maintenance solution for multi-source business systems in the healthcare industry, capable of achieving: (1) Generate a unified change event fingerprint so that every source change can maintain the same event association in the synchronization, verification, repair, re-verification and auditing stages; (2) Establish an indicator dependency graph consisting of field mapping relationships, indicator formula templates and caliber boundary rules, and automatically determine the affected record set, affected field set and affected indicator set based on the changed field set to form the minimum affected subgraph; (3) Perform structural consistency verification, semantic association verification and index formula consistency verification on the least affected subgraph, and implement compensation repair according to the priority of trusted repair source when an anomaly occurs; (4) After the repair, selective recalculation and re-verification are performed only on the affected subgraphs according to the topological order of the indicator dependency graph to avoid full rescanning, and an append-type audit chain with previous and subsequent links is constructed.

[0022] This invention discloses a method for maintaining medical data in a multi-source business system in the medical industry. (See also...) Figure 1 As shown, the method includes: Step S11: Collect data change events from multiple business systems in the medical industry and extract the corresponding change information to generate a change event fingerprint that uniquely identifies the data change event.

[0023] In this embodiment, the change event fingerprint is used as a unified technical object that runs through synchronization, verification, repair, re-verification and auditing to solve the problem of "how to uniquely track the same change in multiple stages".

[0024] In one specific embodiment of the present invention, this step is performed according to the following process: First, data change events from multiple source business systems in the medical industry are collected, and the source business systems are monitored in real time through optional methods (such as parsing logs from the source database, subscribing to source message queues, etc.). When a data change event is detected, the following change information is extracted: source system identifier, business primary key, transaction sequence number, change type, set of changed fields, change timestamp, and operator information. To ensure that the change event fingerprint can uniquely identify a valid change at the source and directly participate in the subsequent location of affected subgraphs, the present invention encodes the key fields of the source-side indicators that can be affected by bitmaps. That is, the set of changed fields in the change information is encoded into a change field bitmap. This bitmap can directly participate in the rapid filtering of subsequent indicator dependency graphs.

[0025] Taking the outpatient and emergency department operation supervision scenario as an example, the following field numbers can be pre-established: b0=register_status, b1=pay_status, b2=refund_flag, b3=online_flag, b4=visit_type, b5=doctor_type, b6=item_category, b7=drug_category, b8=antibiotic_flag, b9=essential_drug_flag, b10=income_type, b11=catalog_mapping_id; if a change only involves the three fields pay_status, refund_flag, and online_flag, then its field bitmap can be encoded as 0000 0000 0000 1110 (binary, low bit on the right); if a change in drug_category is also added, then the encoding is 0000 0000 1000 1110. The system does not need to compare the full log field by field. It can determine whether the event may affect a certain indicator template by performing a bitwise AND operation on the bitmap: when "Change Field Bitmap AND Indicator Template Focus Bitmap" is not 0, it means that there is an impact relationship and the sub-graph should be located along the dependency graph.

[0026] For example, the attention bitmap for the total outpatient visits template can be set to four bits: register_status, pay_status, refund_flag, and online_flag. The outpatient visits template is similar, but the "online_flag=0" filter condition is added to the caliber template. The attention bitmap for the drug ratio template mainly includes bits such as item_category, drug_category, and income_type. In this way, a very low-cost impact screening can be completed at the log access layer.

[0027] Furthermore, the changed field bitmap is combined with other changed information extracted above to generate a change event fingerprint that uniquely identifies the data change event. In an optional implementation, to further improve uniqueness and security, a hash digest can be calculated after concatenating the above elements to serve as the change event fingerprint.

[0028] Through the above steps, a change in source data is transformed into a unique and traceable event fingerprint, solving the problems of difficulty in tracing changes, repeated processing, and disordered sequence in traditional solutions.

[0029] Step S12: Based on the change event fingerprint and the pre-built indicator dependency graph, generate the minimum affected subgraph covering the scope of the data change event, and perform hierarchical verification on the minimum affected subgraph; the hierarchical verification includes the first level of structural consistency verification, the second level of semantic association verification, and the third level of indicator formula consistency verification.

[0030] In this embodiment of the application, an indicator dependency graph is used instead of a static rule list to describe the influence relationship of "field-caliber-formula-indicator" in order to solve the problem of "which records, which indicators, and which formulas are affected by this change".

[0031] Specifically, using the unique change event fingerprint generated by the above steps and the pre-built metric dependency graph, the entire data range affected by this change is precisely located, forming the least affected subgraph. This subgraph is the sole scope of all subsequent synchronization, verification, repair, and recalculation operations. This location mechanism avoids scanning the entire table, batch, or all metrics during subsequent verification and repair processes. It should be noted that the metric dependency graph is used to represent how changes in source fields affect target metrics along the dependency chain. Nodes include: field nodes representing source or target business fields, fact record nodes representing specific business details, dimension nodes representing business dimension information, metric nodes representing statistical metrics, and formula nodes representing the calculation logic or caliber rules of the metrics. Edges in the metric dependency graph represent logical relationships between nodes, including: field mapping edges representing mapping relationships between fields, dependency edges representing node dependencies, filtering edges representing data filtering, aggregation edges representing data aggregation, and derivation edges representing the derivation of new metrics from basic metrics.

[0032] It should be noted that, in one feasible implementation, after determining the least affected subgraph, selective synchronization and snapshot caching are also included. Specifically, only data related to the least affected subgraph is selectively synchronized to the target storage area (including the cache area or target temporary storage area) to avoid resource consumption and increased latency caused by full table or full data synchronization. To ensure the traceability of subsequent verification and repair, a snapshot summary of the source data is generated before synchronization, and a snapshot summary of the target data is generated after synchronization. For unstructured data, the snapshot summary is a content summary or structured index extracted from the data. The snapshot summary is used for data comparison and consistency verification at the hierarchical level, and also provides an immutable before-and-after comparison basis for subsequent trusted repair, re-verification, and audit traceability, ensuring that the synchronization, verification, and repair process is verifiable and traceable.

[0033] After determining the least affected subgraph, the data within that subgraph undergoes hierarchical and progressive verification. It is understandable that for commonly used indicators in hospital operations and supervision, a change in one field of the same business record often affects multiple derived indicators. For example, a change in registration status may affect total outpatient visits, offline outpatient visits, and specialist visits; a change in the category of charges may simultaneously affect outpatient revenue, average outpatient cost, drug ratio, consumable ratio, non-drug and non-consumable ratio, and the proportion of essential drug list costs. Therefore, the consistency problem of integrated medical data is not only a record-level synchronization problem, but also an indicator-level semantic consistency problem. In this embodiment, unlike traditional solutions that only check field format or record quantity, three progressive verification levels are designed, corresponding to different dimensions of data correctness: structural level (first level), semantic level (second level), and statistical indicator level (third level).

[0034] The validation process consists of three levels: First, structural consistency verification, which checks the integrity of basic data record attributes (e.g., checking for null values, type, format, uniqueness, and primary / foreign key integrity); second, semantic association verification, which verifies the logical relationships between business records (e.g., checking for visit type, fee category, diagnosis code, directory mapping, time sequence, and record association); and third, indicator formula consistency verification, which checks whether derived indicators meet the defined template and formula expression. Introducing indicator formula consistency verification during the validation phase allows for the direct detection of errors in derived indicator definitions. Through these three progressive levels of validation, not only field-level errors can be detected, but also hidden anomalies such as "records appearing correct but indicator definitions being incorrect" can be uncovered.

[0035] Step S13: If any level of verification fails, then according to the priority of the pre-configured trusted repair source, perform local compensation repair on the least affected subgraph, and after the compensation repair is completed, perform local recalculation on the repaired least affected subgraph based on the index dependency graph.

[0036] After the three-level verification process described above, if any verification fails within the least affected subgraph, the system enters the repair phase. In this phase, for items that fail verification, candidate repair values ​​are automatically selected and adapted based on the priority of pre-configured trusted repair sources. These values ​​are then written to the corresponding data location in the least affected subgraph to perform local compensation repair. After the compensation repair is executed, the entire database is not recalculated. Instead, the indicator nodes within the least affected subgraph are locally recalculated according to the topological order based on the indicator dependency graph. If an indicator depends on multiple intermediate nodes, the intermediate nodes are recalculated first, followed by the terminal indicator, ensuring that all derived indicators remain consistent with the repaired basic data.

[0037] In one feasible implementation, if the index dependency graph is a directed acyclic graph, once the least affected subgraph is determined, the entire graph is not sorted; only the subgraph is topologically sorted. Specifically, a preset topological sorting algorithm is used to sort the repaired least affected subgraph to determine the topological order. Local recalculation is then performed on the repaired least affected subgraph according to the topological order. For example, the Kahn algorithm can be used: first, the in-degree of each node in the affected subgraph is counted; nodes with an in-degree of 0 are added to a queue; each time a node is taken out, local calculation is performed, and the in-degree of its successor node is decremented by 1; when the in-degree of a successor node becomes 0, it is added to the queue; this continues until the queue is empty. It should be noted that if the final output number of nodes is less than the number of nodes in the affected subgraph, it indicates a cyclic dependency, and a rule configuration anomaly alarm should be triggered, terminating the local recalculation.

[0038] In healthcare metrics scenarios, the priority is typically as follows: field node / mapping node → caliber filtering node → revenue aggregation node or patient aggregation node → intermediate ratio node → terminal indicator node. For example, the calculation of the drug ratio depends on the drug revenue aggregation node and the hospital total revenue aggregation node; only after these two intermediate aggregation nodes are completed can the terminal ratio node be calculated. This sorting mechanism ensures the determinism of local recalculation results.

[0039] Step S14: Perform selective re-verification on the local recalculation results, and after the re-verification passes, generate an audit chain record that includes at least the change event fingerprint, trusted repair source evidence, and the re-verification results.

[0040] After local recalculation, only the least affected subgraph is re-verified at the corresponding level (i.e., re-verification). If the re-verification passes, the repair result is submitted to the official area; if the re-verification fails, the automatic repair result is rolled back and a record awaiting manual handling is generated. This step avoids secondary errors caused by relying on manual random value adjustments or simple resynchronization in traditional solutions, while ensuring that the repair scope is always limited to the least affected subgraph and does not spread to irrelevant data.

[0041] Finally, an audit chain record is generated that includes at least the change event fingerprint, trusted repair source evidence, and verification results. Furthermore, the audit chain record may also include the least affected subgraph identifier, failure rule identifier, repair basis, and snapshot summaries before and after repair. This append-type audit chain with linked before-and-after summaries forms event-level traceability evidence, enabling subsequent audits to fully trace events along the event dimension. The trusted repair source evidence characterizes candidate repair values ​​and their sources, including current values ​​from the source business system, master data index values, standard coding dictionary values, medical insurance catalog rule values, related record reference values, historical verified snapshot values, and their priorities, confidence levels, and applicable scopes.

[0042] In one feasible implementation, for multiple consecutive data change events generated for the same business primary key, they can be sorted according to the transaction sequence number or change timestamp of each data change event; subsequent changes can be generated by referencing the summary of the audit chain record of the previous data change event to the audit chain record of the current data change event based on the sorting result, so as to generate a chain-like traceability record of multiple changes to the same business object.

[0043] Beneficial Effects: This application employs a combination of event-driven and asynchronous processing. Upon receiving a data change event from the source, a unified change event fingerprint is first generated. This ensures that each basic data change possesses a unique and traceable event identifier throughout the synchronization, verification, repair, re-verification, and auditing stages, maintaining the association of the same event. This fundamentally solves the long-standing technical problem of traditional coarse-grained parsing methods being unable to track the flow of the same change event across multiple stages. Secondly, the metric dependency graph is used to represent how changes in source fields affect target metrics along the dependency chain. Therefore, locating the least affected subgraph based on the metric dependency graph ensures that when a field changes, only the local area with which it depends is processed, significantly reducing redundant scanning of irrelevant records and metrics, and lowering latency and computational complexity. The system optimizes resource consumption. Furthermore, the three-level verification system overcomes the limitations of traditional methods that only check field formats, delving into semantic relationships and indicator formulas to address the hidden problem of correct record synchronization but incorrect derived indicators. Subsequently, it only initiates trusted repair source determination and partial compensation recalculation when verification fails. Trusted repair sources are automatically selected based on preset priorities, ensuring clear technical justification for repair actions and preventing secondary errors caused by arbitrary value adjustments. After repair, only the least affected subgraphs undergo partial recalculation and selective re-verification, further reducing the processing scope and ensuring repair efficiency and accuracy. Finally, it generates an append-only audit chain record, forming a closed-loop event-level traceability evidence that meets the compliance and traceability requirements of medical data supervision. Therefore, this application enables efficient, accurate, low-cost, and traceable verification, repair, and recalculation of integrated medical data, suitable for medical business analysis and supervision scenarios with complex indicator definitions and high data throughput.

[0044] Healthcare operation and regulatory indicators differ from ordinary business statistics. The difficulty lies not in the complexity of the formulas themselves, but in the simultaneous involvement of "inclusion criteria, exclusion criteria, field mapping relationships, and directory merging rules" in the calculation, with only minor differences in caliber between different indicators. If the general industry approach of full synchronization plus format validation is still used, the problem of "underlying records being synchronized but indicator calibers being misaligned" often goes undetected. Taking the outpatient and emergency department indicators in Table 1 as an example, although "Total Outpatient Visits" and "Outpatient Visits" have similar names, the former includes visits to internet hospitals, while the latter only counts offline outpatient visits and removes the internet hospital label; the difference between "Total Outpatient Revenue" and "Total Outpatient Revenue (excluding internet hospitals)" reflects whether internet hospital revenue is included; and there is also a boundary difference between "Outpatient Revenue" and "Total Outpatient Revenue" regarding whether physical examination revenue is included. In other words, a change in a single field, `online_flag`, can have different directions and ranges of impact on multiple indicators. For example, indicators such as "drug ratio," "consumables ratio," "non-drug and non-consumables ratio," "antibiotic ratio," and "essential drug list cost ratio," although all use revenue or expenses as the numerator and denominator, their classification logic relies on multi-dimensional rules including drug category, consumable category, chargeable item category, catalog mapping, whether it is traditional Chinese medicine / prepared traditional Chinese medicine, whether it is an antibiotic, whether it belongs to the essential drug list, and whether it is included in the hospital's total revenue. Simply performing field non-empty checks or total amount checks cannot confirm whether these catalog merging and caliber filtering are correct.

[0045] Therefore, the reason why current technologies are difficult to apply in medical scenarios is not because they cannot synchronize data, but because they do not explicitly model the chain relationship of "field-directory mapping-caliber boundary-formula template-indicator result". Once a change in an underlying field affects multiple similar but different indicators, traditional technologies can only recalculate the entire batch or rely on manual verification, making it difficult to automatically locate the minimum impact range.

[0046] Table 1 Examples of Outpatient and Emergency Department Indicators

[0047] Based on the above embodiments, this application describes the construction of the indicator dependency graph and the construction of the least affected subgraph. The process of constructing the indicator dependency graph includes the following steps: Obtain the indicator semantic templates pre-configured for statistical indicators in medical statistics; the indicator semantic templates are used to describe the calculation rules of the statistical indicators; An indicator dependency graph is constructed based on the indicator semantic template; the indicator dependency graph is a directed graph structure composed of nodes and edges that describe data mapping relationships and indicator derivation dependency relationships.

[0048] In this embodiment, to avoid the problem of "correct record synchronization but incorrect derived indicators," commonly used hospital operational and regulatory indicators are abstracted into configurable indicator semantic templates. These templates describe the data collection objects, inclusion criteria, exclusion criteria, filtering dimensions, summarization dimensions, calculation formulas, and result constraints for a statistical indicator. For example, whether to include internet hospitals, whether to exclude cancelled appointments, whether to exclude physical examination revenue, and whether to group by fee category, etc.

[0049] For typical validation indicators, templates can be configured as shown in Tables 2 and 3. Each template should include at least the indicator name, inclusion scope, exclusion scope, data retrieval conditions, grouping criteria, calculation formula, target data table, and field mapping relationships.

[0050] Table 2 Typical Template Example 1

[0051] Table 3 Typical Template Example 2

[0052] Through the template-based modeling described above, an indicator dependency graph is then constructed based on this template. In this way, when the registration status, cancellation status, fee classification, internet hospital identification, drug classification identification, or catalog mapping relationship changes, the system can directly locate the affected indicators through the indicator dependency graph without having to recalculate all indicators.

[0053] In a preferred embodiment, the metric dependency graph is preferably constructed as a directed acyclic graph. The nodes in the graph include at least field nodes, filter nodes, aggregation nodes, intermediate metric nodes, and terminal metric nodes; edges indicate that "the result of the current node is dependent on the next node".

[0054] Furthermore, the process of locating the minimum affected subgraph covering the scope of the data change event includes: First, determining the changed field based on the change event fingerprint, and mapping the changed field to the target field on the integration platform side through a preset field mapping table; then, starting from the target field, traversing along the edges of the indicator dependency graph to collect nodes that have direct or indirect dependencies on the target field (such as affected fact record nodes, dimension nodes, indicator nodes, and formula nodes); finally, forming the connected subgraph composed of the edges between the nodes to create the minimum affected subgraph that only covers the scope of this change. The size of this subgraph precisely matches the scope of influence of the changed field: if the change involves only a single field and that field only affects a few indicators, the subgraph contains only a small number of nodes; if the change involves a key field (such as a charging category), the subgraph may cover multiple intermediate and terminal indicators.

[0055] Based on the above embodiments, the trusted repair source includes at least one or more of the following: the current valid value of the source business system, the master data index mapping value, the standard encoding dictionary value, the medical insurance catalog rule value, the associated record reference value, and the historical verified snapshot value. In this embodiment, when a verification item in the three-level verification fails, the system retrieves possible correct values ​​from multiple candidate repair sources according to the type of failure rule, and performs uniqueness determination according to the pre-configured priority; automatic compensation repair is only implemented when the candidate value is unique and the applicable scope matches.

[0056] In this embodiment, when the same failure item can obtain candidate values ​​from multiple repair sources, a conflict resolution strategy of "applicability verification + priority sorting + consistency comparison + difference threshold judgment" is adopted, rather than simply taking the nearest value. Suppose that the consistency verification of the third-level indicator formula finds that the "charge category" field of a certain outpatient billing record is empty or invalid, causing the record to be incorrectly classified into outpatient revenue, thus affecting the derived calculation result of the average outpatient fee. At this time, the system performs the following operations: Step 1: Obtain multiple candidate repair sources and perform applicability verification on the candidate repair sources to filter out invalid candidate values ​​that do not meet the current business rules and applicable scope; In this step, the system retrieves the correct category for the chargeable item from multiple sources, obtaining multiple candidate repair sources. First, an applicability check is performed, filtering out candidate values ​​that do not meet the current rules, such as those not belonging to the current chargeable item catalog, mismatched encoding versions, expired time validity periods, or those not applicable to the hospital area.

[0057] Step 2: Sort the valid candidate values ​​that pass the applicability check according to the priority of the pre-configured trusted repair sources to determine an ordered candidate sequence from high to low priority; In this step, the valid candidate values ​​that pass the applicability check are sorted in the following order: the current valid value of the source business system, the master data index mapping value, the standard coding dictionary value, the medical insurance catalog rule value, the related record reference value, and the historical verified snapshot value, to obtain an ordered candidate sequence with high priority to low priority.

[0058] Step 3: Perform a consistency comparison on the valid candidate values ​​in the ordered candidate sequence. If the target valid candidate value corresponding to the highest priority is unique, or if the highest priority corresponds to multiple target valid candidate values ​​and the contents of the target valid candidate values ​​are consistent, then use the target valid candidate value as the repair value to perform local compensation repair on the least affected subgraph. In this step, it is checked whether the valid candidate values ​​in the ordered candidate sequence are consistent: if the valid candidate value corresponding to the highest priority is unique, it is directly used as the repair value to perform local compensation repair on the least affected subgraph; if the valid candidate value corresponding to the highest priority is not unique, but the contents of the multiple candidate values ​​ranked first are consistent, it can also be directly regarded as the unique credible candidate; if the contents are inconsistent, it proceeds to the scoring judgment in step 4.

[0059] Step 4: If the contents of the multiple valid candidate values ​​of the target corresponding to the highest priority are inconsistent, a comprehensive score is calculated for each valid candidate value of the target based on a preset weighted calculation formula. In this step, a comprehensive score is calculated for each candidate value: Score = Wp × P + Wf × F + Wc × C + Wr × R; where P represents the priority score, F represents the time freshness score, C represents the consistency score with other candidate values, and R represents the matching score with the failure rule; Wp, Wf, Wc, and Wr are weights.

[0060] Step 5: When the difference between the highest score and the second highest score in the comprehensive score is greater than a preset threshold, the target valid candidate value corresponding to the highest score in the comprehensive score is used as the repair value to perform local compensation repair on the least affected sub-image. In this step, automatic compensation and repair are allowed only when the difference between the highest and second-highest candidate scores is greater than a preset threshold Δ and the highest candidate score does not violate strong constraint rules, and the highest candidate score is used as the repair value; otherwise, the process is switched to manual processing.

[0061] For example, a manual processing flow is required, with manual intervention addressing the following situations: Insufficient score difference: The score difference between the highest and second-highest candidate values ​​is less than a preset threshold; Strong constraint conflict: If the code corresponding to the highest candidate value is invalid, such as being discontinued in the medical insurance catalog, the system will not adopt this value even if it has the highest score, and manual processing will be initiated; Missing key candidate sources: If none of the candidate repair sources provide valid values, the system will also initiate manual processing, generating a record pending manual processing. Simultaneously, all candidate values ​​and their supporting evidence tuples are written into the audit chain for manual confirmation before repair.

[0062] It is evident that the technical significance of the aforementioned conflict resolution strategy lies in the fact that automatic repair is not simply "changing every value that exists." Instead, it first filters out common issues in medical data scenarios, such as catalog version conflicts, identical names with different codes, differences between hospitals, and outdated historical snapshots, through rule-based filtering and candidate consistency verification, thereby improving the stability of the repair results.

[0063] To enhance the technical effectiveness of this solution, this section provides exemplary estimation data based on hospital operational analysis and regulatory scenarios. The following data is used to illustrate the technical effectiveness formation mechanism of the present invention's "minimum affected subgraph + selective recalculation" compared to "full recalculation / batch review," and is not equivalent to the final measured values; before formal submission, it is advisable to replace it with stress test or control experiment data in a specific hospital environment.

[0064] The estimated scenario is as follows: Taking the daily operational monitoring data link of a tertiary hospital as an example, the daily increase in factual records such as outpatient and emergency visits, billing, prescriptions, and item details is approximately 1.2 million; the core indicators configured on the operational monitoring side are approximately 220, of which approximately 68 are directly related to outpatient and emergency visits and outpatient billing. It is assumed that a single source-side change mainly occurs in fields such as registration status, payment status, cancellation flag, internet hospital flag, billing category, drug category, and item catalog mapping.

[0065] In the traditional full-scale recalculation model, if any of the above fields changes, a full batch scan of the relevant fact table and its associated indicators for that day is required. If we estimate based on 68 relevant indicators and 1.2 million daily incremental fact records, then a single verification task needs to cover approximately 81.6 million "record-indicator judgment units".

[0066] In this invention, the changed fields are first located using change event fingerprints, and then the least affected subgraph is determined using an indicator dependency graph. Based on a typical scenario of "1 outpatient record + 12 billing details + 8 related dimension records + 11 affected indicators," a single selective recalculation only requires processing approximately 231 "record-indicator judgment units." The computational load is reduced from 81.6 million to 231 compared to a full recalculation, corresponding to a 99.9997% reduction in recalculation coverage. Even after considering system overhead such as cache hits, index scans, log writing, and topology sorting, and using a conservative approach, the actual reduction in recalculated data volume can typically still reach over 95%. Table 1 shows a comparison of the effects of the two solutions.

[0067] Table 4 Comparison of computational efficiency under different modes

[0068] The end-to-end processing latency can be estimated as follows: Assume that in the conventional mode, the time consumed for fact table scanning, indicator aggregation and verification is Tscan, Tagg, and Tcheck respectively, then the total latency is approximately Tfull=Tscan+Tagg+Tcheck. Assume that in the mode of the present invention, the time consumed for affected subgraph extraction, local recalculation and selective re-verification is Tsub, Tlocal, and Trecheck respectively, then the total latency is approximately Tsel=Tsub+Tlocal+Trecheck. Since Tsub mainly depends on index and dependency graph traversal, and Tlocal and Trecheck only act on the local subgraph, Tsel<<Tfull is usually satisfied in typical medical operation indicator scenarios. Estimated with an example configuration of 16-core CPU, 64GB memory, daily increment of 1.2 million records, and 11 indicators affected by a single change, the single exception processing latency of the conventional mode can be between 8 seconds and 25 seconds, while the mode of the present invention ranges from 0.2 seconds to 1.5 seconds, with the latency reduction ratio being approximately 81.25% to 98.13%.

[0069] The core of the above technical effect is not simply "faster algorithm", but that the present invention shrinks the original processing for the entire table, the entire batch and the entire indicator set to deterministic processing oriented to the local influence range through the event-level field bitmap, indicator dependency graph and minimum affected subgraph, thereby reducing irrelevant scanning, irrelevant aggregation and irrelevant verification from the architecture level.

[0070] Correspondingly, the embodiments of the present application further disclose a medical data maintenance apparatus for multi-source business systems in the medical industry, see Figure 2 as shown in the figure, the apparatus includes: A change collection module 11, configured to collect data change events of multi-source business systems in the medical industry; An event fingerprint generation module 12, configured to extract corresponding change information to generate a change event fingerprint for uniquely identifying the data change event; A three-level verification module 13, configured to generate a minimum affected subgraph covering the influence range of the data change event based on the change event fingerprint and a pre-constructed indicator dependency graph, and perform hierarchical verification on the minimum affected subgraph; the hierarchical verification includes structure consistency verification of the first level, semantic association verification of the second level, and indicator formula consistency verification of the third level; A local compensation repair module 14, configured to, if any level of verification fails, perform local compensation repair on the minimum affected subgraph according to the preconfigured priority of trusted repair sources; A local recalculation module 15, configured to perform local recalculation on the repaired minimum affected subgraph based on the indicator dependency graph after the compensation repair is completed; A selective re-verification module 16, configured to perform selective re-verification on the local recalculation result; The audit chain record module 17 is used to generate an audit chain record that includes at least the change event fingerprint, trusted repair source evidence, and the re-verification result after the re-verification is passed.

[0071] For more detailed information on the working process of each of the above modules, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0072] Therefore, the above-described scheme in this embodiment combines event-driven and asynchronous processing. When a data change event is collected from the source, a unified change event fingerprint is first generated. This ensures that each basic data change has a unique and traceable event identifier in each stage of synchronization, verification, repair, re-verification, and auditing, maintaining the association of the same event. This fundamentally solves the long-standing technical problem that traditional coarse-grained parsing methods cannot track the flow of the same change event across multiple stages. Secondly, the indicator dependency graph is used to represent how changes in the source field affect the target indicator along the dependency chain. Therefore, by locating the least affected subgraph based on the indicator dependency graph, it can be ensured that when a field changes, only the local range that has a dependency on it is processed. This significantly reduces the repeated scanning of irrelevant records and indicators, and reduces latency and computation. The system optimizes resource consumption. Furthermore, the three-level verification system overcomes the limitations of traditional methods that only check field formats, delving into semantic relationships and indicator formulas to address the hidden problem of correct record synchronization but incorrect derived indicators. Subsequently, it only initiates trusted repair source determination and partial compensation recalculation when verification fails. Trusted repair sources are automatically selected based on preset priorities, ensuring clear technical justification for repair actions and preventing secondary errors caused by arbitrary value adjustments. After repair, only the least affected subgraphs undergo partial recalculation and selective re-verification, further reducing the processing scope and ensuring repair efficiency and accuracy. Finally, it generates an append-only audit chain record, forming a closed-loop event-level traceability evidence that meets the compliance and traceability requirements of medical data supervision. Therefore, this application enables efficient, accurate, low-cost, and traceable verification, repair, and recalculation of integrated medical data, suitable for medical business analysis and supervision scenarios with complex indicator definitions and high data throughput.

[0073] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0074] Figure 3This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the medical data maintenance method of the multi-source business system in the medical industry disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment can be deployed on a hospital integration platform server, a data middleware node, or a dedicated monitoring node, and is not limited to a specific chip model, interface model, or operating system version.

[0075] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0076] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored on it can include an operating system 221, computer programs 222, and data 223, etc. The data 223 can include various types of data. The storage method can be temporary storage or permanent storage.

[0077] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the medical data maintenance method of the multi-source business system in the medical industry, which is executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0078] Furthermore, this application also discloses a computer-readable storage medium, which includes random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, magnetic disks, optical disks, or any other form of storage medium known in the art. When the computer program is executed by a processor, it implements the aforementioned medical data maintenance method for a multi-source business system in the medical industry. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0079] Furthermore, embodiments of this application also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements any of the above-described methods for maintaining medical data in a multi-source business system of the medical industry.

[0080] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0081] The steps of the medical data maintenance method or algorithm for a multi-source business system in the healthcare industry described in conjunction with the embodiments disclosed herein can be implemented directly using hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0082] Finally, 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 expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0083] The above provides a detailed description of the medical data maintenance method, apparatus, equipment, and medium for a multi-source business system in the medical industry provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only intended to help understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for maintaining medical data in a multi-source business system in the medical industry, characterized in that, include: Collect data change events from multiple business systems in the medical industry and extract the corresponding change information to generate a change event fingerprint that uniquely identifies the data change event; Based on the change event fingerprint and the pre-built indicator dependency graph, a minimum affected subgraph covering the scope of the data change event is generated, and hierarchical verification is performed on the minimum affected subgraph. The hierarchical verification includes a first-level structural consistency verification to verify the integrity of the basic attributes of data records, a second-level semantic association verification to verify the rationality of the logical association between business records, and a third-level indicator formula consistency verification to detect whether derived indicators meet the caliber template and formula expression. The indicator dependency graph is a directed graph structure composed of nodes and edges describing data mapping relationships and indicator derivation dependency relationships; it is used to describe the influence relationship between fields, calibers, formulas, and indicators, and to represent how changes in source fields affect target indicators along the dependency chain. If any level of verification fails, the least affected subgraph is locally compensated and repaired according to the priority of the pre-configured trusted repair source, and after the compensation and repair is completed, the repaired least affected subgraph is locally recalculated based on the index dependency graph. Selective re-verification is performed on the partial recalculation results, and after the re-verification passes, an audit chain record is generated that includes at least the change event fingerprint, trusted repair source evidence, and the re-verification results.

2. The method for maintaining medical data in a multi-source business system in the medical industry according to claim 1, characterized in that, After generating the minimum affected subgraph covering the scope of the data change event, the method further includes: Data related to the least affected subgraph is selectively synchronized to the target storage area, and a snapshot summary of the source data is generated before synchronization and a snapshot summary of the target data is generated after synchronization. Specifically, for unstructured data, the snapshot summary is a content summary or structured index extracted from the data; the snapshot summary is used for data comparison and consistency verification in the hierarchical verification.

3. The method for maintaining medical data in a multi-source business system in the medical industry according to claim 1, characterized in that, The process of collecting data change events from multiple business systems in the medical industry and extracting corresponding change information to generate a change event fingerprint for uniquely identifying the data change events includes: Collect data change events from multiple business systems in the healthcare industry, and extract source system identifiers, business primary keys, transaction sequence numbers, change types, set of change fields, change timestamps, and operator information; The set of changed fields in the changed information is encoded into a changed field bitmap, and combined or hashed with other changed information in the changed information to generate a changed event fingerprint that uniquely identifies the data changed event.

4. The method for maintaining medical data in a multi-source business system in the medical industry according to claim 1, characterized in that, Constructing the metric dependency graph includes: Obtain the indicator semantic templates pre-configured for statistical indicators in medical statistics; the indicator semantic templates are used to describe the calculation rules of the statistical indicators; Construct an indicator dependency graph based on the indicator semantic template; Accordingly, generating a minimum affected subgraph covering the scope of the data change event based on the change event fingerprint and a pre-built metric dependency graph includes: The changed fields are determined based on the change event fingerprint, and the changed fields are mapped to target fields on the integration platform side through a preset field mapping table; Starting with the target field as the starting node, traverse along the edges of the index dependency graph to collect nodes that have direct or indirect dependencies on the target field, and use the connected subgraph formed by the edges between the nodes as the least affected subgraph.

5. The method for maintaining medical data in a multi-source business system in the medical industry according to claim 1, characterized in that, The step of performing local compensation repair on the least affected subgraph according to the priority of the pre-configured trusted repair sources includes: Multiple candidate repair sources are obtained, and the applicability of the candidate repair sources is verified to filter out invalid candidate values ​​that do not meet the current business rules and applicable scope. The valid candidate values ​​that pass the applicability check are sorted according to the priority of the pre-configured trusted repair sources to determine an ordered candidate sequence from high to low priority; A consistency comparison is performed on the valid candidate values ​​in the ordered candidate sequence. If the target valid candidate value corresponding to the highest priority is unique, or if the highest priority corresponds to multiple target valid candidate values ​​and the contents of the target valid candidate values ​​are consistent, then the target valid candidate value is used as the repair value to perform local compensation repair on the least affected subgraph. If the contents of the multiple valid candidate values ​​of the target corresponding to the highest priority are inconsistent, a comprehensive score is calculated for each valid candidate value of the target based on a preset weighted calculation formula. When the difference between the highest score and the second highest score in the comprehensive score is greater than a preset threshold, the target valid candidate value corresponding to the highest score in the comprehensive score is used as the repair value to perform local compensation repair on the least affected subgraph.

6. The method for maintaining medical data in a multi-source business system in the medical industry according to claim 1, characterized in that, The step of locally recalculating the least affected subgraph based on the index dependency graph includes: If the index dependency graph is a directed acyclic graph, then the repaired least affected subgraph is sorted using a preset topological sorting algorithm to determine the topological order. The repaired minimum affected subgraph is locally recalculated sequentially according to the topological order. If a cycle dependency is detected during the sorting process, an alarm is triggered and the local recalculation is terminated.

7. The method for maintaining medical data in a multi-source business system in the medical industry according to any one of claims 1 to 6, characterized in that, Also includes: When multiple consecutive data change events occur for the same business primary key, they are sorted according to the transaction sequence number or change timestamp of each data change event. Based on the sorting results, the audit chain record of the current data change event is referenced with the summary of the audit chain record of the previous data change event to generate a chain-like trace record of multiple changes to the same business object.

8. A medical data maintenance device for a multi-source business system in the medical industry, characterized in that, include: The change acquisition module is used to collect data change events from multiple business systems in the healthcare industry. The event fingerprint generation module is used to extract the corresponding change information to generate a change event fingerprint that uniquely identifies the data change event; The three-level verification module is used to generate a minimum affected subgraph covering the scope of the data change event based on the change event fingerprint and a pre-built indicator dependency graph, and to perform hierarchical verification on the minimum affected subgraph. The hierarchical verification includes a first-level structural consistency verification to verify the integrity of the basic attributes of data records, a second-level semantic association verification to verify the rationality of the logical association between business records, and a third-level indicator formula consistency verification to detect whether derived indicators meet the caliber template and formula expression. The indicator dependency graph is a directed graph structure composed of nodes and edges describing data mapping relationships and indicator derivation dependency relationships; it is used to describe the influence relationship between fields, calibers, formulas, and indicators, and to represent how changes in source fields affect target indicators along the dependency chain. The local compensation and repair module is used to perform local compensation and repair on the least affected subgraph according to the priority of the pre-configured trusted repair source if any level of verification fails. The local recalculation module is used to perform local recalculation of the least affected subgraph after compensation and repair are completed, based on the index dependency graph. The selective re-verification module is used to selectively re-verify the results of local recalculation. The audit chain record module is used to generate an audit chain record that includes at least the change event fingerprint, trusted repair source evidence, and the re-verification result after the re-verification is passed.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory; wherein the memory is used to store a computer program, which is loaded and executed by the processor to implement the medical data maintenance method for a multi-source business system in the medical industry as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer programs; wherein the computer programs, when executed by a processor, implement the medical data maintenance method for a multi-source business system in the medical industry as described in any one of claims 1 to 7.

Images