Health information quality management and control method and system

Abstract

The application provides a health information quality control method and system. When accessing a health data platform through an online interface, based on rule shunting results, online sub-rules of online quality control rules and / or online-offline mixed quality control rules are called for target health data accessed to perform real-time data quality verification to obtain real-time quality control results. Within a preset offline scheduling period, based on rule shunting results, offline sub-rules of offline quality control rules and / or online-offline mixed quality control rules are called for historical health data stored in the health data platform to perform offline data quality reevaluation to obtain offline quality control results. Then, according to the real-time quality control results and the offline quality control results, the health data is quality controlled, so that the real-time quality verification and the periodic offline quality reevaluation of the health data are coordinated without increasing the unacceptable response delay of the front-end business system.

Description

Technical Field

[0001] This application relates to data processing technology, and more particularly to a method and system for quality control of health information. Background Technology

[0002] With the construction of the maternal and child health information system and the promotion and application of the provincial regional maternal and child health information platform, maternal and child health data from multiple sources such as county-level maternal and child health care institutions, midwifery institutions, and community health service institutions are continuously reported online via interfaces.

[0003] To ensure data quality, existing systems typically employ rule-driven data quality control to check the completeness, legality, and logical consistency of reported data. However, in traditional models of single real-time checks or single offline batch processing, on the one hand, configuring a large number of complex data quality rules for real-time execution can easily lead to a significant increase in online interface response latency, severely impacting the user experience of front-end business systems such as outpatient reception, registration, and follow-up data entry; on the other hand, if most rules are postponed to the offline batch processing stage, obvious errors cannot be detected and corrected in time during the front-end data entry process, and erroneous data has already been persisted and flowed to subsequent business analysis and decision-making stages, resulting in high remedial costs and difficulties in determining responsibility.

[0004] Therefore, in the context of a hybrid online / offline processing architecture for health data, how to rationally divide and schedule data quality rules has become one of the key bottlenecks restricting the data management effectiveness and system performance of existing maternal and child health information systems. Summary of the Invention

[0005] This application provides a method and system for health information quality control, which addresses the technical problem of how to reasonably divide and schedule data quality rules under a hybrid online / offline health data processing architecture.

[0006] Firstly, this application provides a method for health information quality control, applied to a health information quality control system with an online / offline hybrid architecture including an online quality control engine and an offline quality control engine; the method includes: When health data is accessed to the health data platform through an online interface, based on the rule-based traffic distribution results, the online sub-rules of the online quality control rules and / or the online-offline hybrid quality control rules are invoked on the accessed target health data to perform real-time data quality verification and obtain real-time quality control results. Within a preset offline scheduling period, based on the rule-based traffic distribution results, the offline quality control rules and / or the offline sub-rules of the online-offline hybrid quality control rules are invoked on the historical health data stored in the health data platform to perform offline data quality re-evaluation and obtain offline quality control results. The health data is subject to quality control based on the real-time quality control results and the offline quality control results.

[0007] Secondly, this application provides a health information quality control system, comprising: an online / offline hybrid architecture of an online quality control engine and an offline quality control engine; wherein the hybrid architecture includes: The quality verification module is used to perform real-time data quality verification on the target health data when health data is accessed through the online interface, based on the rule-based traffic distribution results, by calling the online quality control rules and / or the online sub-rules of the online-offline hybrid quality control rules, and obtaining real-time quality control results. The quality assessment module is used to, within a preset offline scheduling period, based on the rule-based traffic distribution results, call the offline quality control rules and / or the offline sub-rules of the online-offline hybrid quality control rules to reassess the offline data quality of historical health data stored in the health data platform, and obtain the offline quality control results. The quality control system is used to perform quality control on the health data based on the real-time quality control results and the offline quality control results. Attached Figure Description

[0008] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0009] Figure 1 This is a flowchart illustrating a health information quality control method according to an example embodiment of this application; Figure 2 This is a flowchart illustrating a health information quality control method according to another example embodiment of this application; Figure 3 This is a schematic diagram of the structure of a health information quality control system according to an example embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application.

[0010] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0011] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0012] Many maternal and child health quality rules inherently rely on a holistic or global view. For example, they may require assessments of preconception, multiple prenatal checkups, delivery records, postpartum visits, and child follow-up records to determine the completeness of the entire life cycle path. Alternatively, they may require comprehensive analysis across the province to identify issues such as duplicate registrations across institutions, centralized data entry, and abnormally low high-risk labeling rates. The effective execution of these rules objectively depends on a data warehouse or data lake environment with full data and wide table / fact table structures. Furthermore, considering the differences in storage organization and index optimization goals between online and offline databases, the performance cost of executing the same rule on different engines can differ by orders of magnitude. Therefore, without a real-time traffic distribution mechanism for rules, a structural conflict inevitably arises between the system's performance tolerance and the depth of data quality control.

[0013] To address the aforementioned issues, the embodiments provided in this application employ the underlying technical principles of combining rule-based metadata modeling, real-time performance scoring, and triage decision-making in the health information quality control system. First, define rule metadata for each health data quality rule, including parameters such as scope, data dependency, time window, computational complexity, business real-time performance, and risk level. By quantifying and encoding the above metadata dimensions, establish a real-time performance scoring function to numerically evaluate the real-time execution capability of each rule.

[0014] Based on this, online and offline thresholds are set, and the rules are automatically divided into three types according to the real-time performance score: online execution only, offline execution only, and mixed online and offline execution, thus forming the rule distribution result.

[0015] For complex rules classified as hybrid online and offline quality control rules, they are further broken down into online sub-rules that rely only on the current record and lightweight historical cache and meet online latency constraints, and offline sub-rules that rely on individual full historical records, cross-institutional records or statistical aggregated data, through rule semantic parsing. These sub-rules are then sent to the memory rule execution engine and the offline batch processing engine for execution, respectively.

[0016] Meanwhile, a lightweight historical cache is built on the online side using a key-value index structure and a preset time window, supporting small-scale historical dependency verification such as the most recent K records and the institution's current day records with O(1) or O(logN) time complexity. On the offline side, based on the full or incremental data in the data warehouse or data lake, high-complexity rules are executed according to a preset scheduling plan to generate record-level, individual-level, and institution-level quality tags. This achieves the synergy between online real-time quality control and offline re-evaluation, eliminating the structural contradiction that performance and quality are difficult to balance under a single execution mode, and effectively solving the problem that data quality rules of the maternal and child health information system cannot be efficiently and reasonably scheduled under an online / offline hybrid architecture.

[0017] It is worth noting that the embodiments provided in this application are applied to a health information quality control system with an online / offline hybrid architecture, including an online quality control engine and an offline quality control engine. Specifically, the health information quality control system includes: The online quality control engine is used to perform online data quality verification on the target health data records when health data is accessed through an online interface to the health data platform, and outputs the online verification results as real-time quality control results.

[0018] The offline quality control engine is used to perform offline data quality assessment based on historical health data in the health data storage system within a preset offline scheduling period, and output the offline assessment results as offline quality control results.

[0019] The interface access layer is used to adapt access protocols, convert data formats, and perform basic legality verification on health data reported from maternal and child health care institutions, midwifery institutions, and community health service institutions. It then sends the processed target health data records to the online quality control engine and stores the target health data records and subsequent historical health data in the health data storage system.

[0020] The rule management module is used to maintain the pre-configured set of health data quality rules and the rule metadata corresponding to each data quality rule. It supports the addition, modification, publication and version management of data quality rules.

[0021] The rule-based routing module is used to score data quality rules in real time based on rule metadata and a preset real-time scoring model. It also classifies each data quality rule into at least one of the following categories: online quality control rules, offline quality control rules, and hybrid online-offline quality control rules. The module generates rule-based routing results and provides these results to the online and offline quality control engines.

[0022] The quality tag library is used to associate and store real-time quality control results with offline quality control results, and organizes quality tags according to record level, individual level, organization level and data domain dimensions to support subsequent quality analysis and rule optimization.

[0023] Furthermore, the aforementioned online quality control engine specifically includes: The online access module is used to receive target health data records from the interface access layer, and match the set of online quality control rules and the set of online sub-rules associated with the target health data records based on the rule-based traffic distribution results and the data domain to which the target health data records belong.

[0024] The memory rule execution engine is used to execute online quality control rule sets and online sub-rule sets on target health data records in memory. It verifies field integrity, value range, enumeration value validity, and logical consistency within the same record. Based on a lightweight historical cache, it executes data quality rules that depend on the individual's most recent K records or the organization's records for the day, and obtains real-time quality control results.

[0025] The lightweight history cache module is used to store health data window records within a preset time window, and constructs a key-value index structure based on at least one of the individual identifier, organization identifier, and date identifier, providing the memory rule execution engine with low-latency access capabilities for querying the most recent K records of an individual and the records of an organization for the current day.

[0026] Furthermore, the aforementioned offline quality control engine includes: The offline rule configuration module is used to configure the corresponding offline execution frequency, applicable data domain, and scheduling priority for each offline quality control rule and offline sub-rule based on the rule routing results.

[0027] The offline scheduling module is used to trigger offline quality control tasks corresponding to each offline quality control rule and offline sub-rule according to the offline execution frequency within a preset offline scheduling period, and to send the data range and data domain information to be evaluated to the offline rule execution module.

[0028] The offline rule execution module is used to execute offline quality control rules and offline sub-rules based on full or incremental historical health data in a data warehouse or data lake. It performs offline data quality re-evaluation on record-level quality issues, individual full-history consistency issues, cross-organization duplicate record issues, and organization-level statistical anomaly issues, and generates offline quality control results including at least one of record-level quality tags, individual-level quality tags, and organization-level quality tags.

[0029] Figure 1 This is a flowchart illustrating a health information quality control method according to an example embodiment of this application. Figure 1 As shown, the health information quality control method provided in this embodiment includes: S110. Based on the rule-based traffic distribution results, call the online sub-rules of the online quality control rules and / or the online-offline hybrid quality control rules to perform real-time data quality verification on the target health data that has been accessed.

[0030] In this step, when health data is accessed to the health data platform through an online interface, based on the rule-based traffic distribution results, the online quality control rules and / or online sub-rules of the online-offline hybrid quality control rules are invoked on the accessed target health data to perform real-time data quality verification and obtain real-time quality control results.

[0031] In the existing IT infrastructure of maternal and child health care, midwifery institutions, and community health service institutions, health data is typically reported in real-time to regional or provincial health data platforms through various business systems, such as maternal and infant management systems, child health care systems, and immunization systems. Health data quality control is mostly deployed at the data warehouse layer or report statistics layer, using batch processing frameworks (such as offline SQL scripts and ETL tasks) for full or periodic checks. There is no online rule matching and execution mechanism for individual real-time reported records, resulting in significant quality control delays. Furthermore, in the online processing chain, most platforms have not introduced dedicated in-memory rule execution engines and lightweight historical cache structures, lacking the ability to quickly retrieve the most recent K records or the institution's current day's records based on key-value indexes such as individual identifiers, institution identifiers, and date identifiers. This prevents rules relying on recent historical data from being executed online within acceptable latency.

[0032] To address this, the aforementioned real-time data quality verification can be achieved by receiving target health data records from reporting systems of maternal and child health institutions, midwifery institutions, and community health service institutions via an online access module. Then, based on the rule routing results and the data domain to which the target health data record belongs, a set of online quality control rules and a set of online sub-rules associated with the target health data record are matched. Finally, the online quality control rules and the set of online sub-rules are executed on the target health data record in the in-memory rule execution engine to obtain the real-time quality control results for the target health data record.

[0033] Furthermore, regarding the specific implementation of the online quality control rule set and online sub-rule set for executing target health data records in the memory rule execution engine, it can be achieved by executing single-record verification rules on the target health data records. These single-record verification rules include at least field integrity checks, field value range checks, enumeration value validity checks, and logical consistency checks within the same record. Then, historical records related to the target health data records are retrieved from the lightweight historical cache based on at least one of the individual identifier, institution identifier, and date identifier to support the execution of rules with limited historical dependencies. Finally, based on the historical records, data quality rules are executed to detect duplicate registrations within a short period and duplicate vaccinations of the same child within the same institution.

[0034] Optionally, the aforementioned lightweight historical cache is used to store health data window records within a preset time window. The lightweight historical cache is based on a key-value index structure and provides time-complexity query capabilities for querying the most recent K records of an individual and the records of an organization for the current day.

[0035] By pre-establishing rule-based traffic distribution results, data quality rules are divided into online quality control rules and online-offline hybrid quality control rules based on their real-time scoring capability. Each rule is labeled with its applicable data domain and triggering conditions. This enables the online access module to quickly match the corresponding set of online quality control rules and set of online sub-rules based on the data domain to which the record belongs and the traffic distribution results when it receives target health data records from maternal and child health institutions, midwifery institutions, and community health service institutions.

[0036] Secondly, in the memory rule execution engine, rule compilation or rule interpretation execution technology is used to transform online rules into an intermediate representation suitable for memory computing. Combined with a lightweight historical cache based on key-value index structure, using individual identifier, organization identifier, date identifier, etc. as primary keys, the system performs O(1) or near-O(1) time complexity queries on health data window records within a preset time window, providing low-latency data access for verification rules that rely on the individual's most recent K records or the organization's current day records.

[0037] Furthermore, during rule execution, single-record verification rules such as field integrity checks, field value range checks, enumeration value validity checks, and intra-record logical consistency checks are configured as operators that can be executed in parallel or in a pipelined manner. Optimization techniques such as sharing intermediate results in memory, short-circuiting conditions, and batch processing are used to reduce redundant calculations and ensure high concurrency processing capabilities under the condition that the single-record verification latency is controlled.

[0038] Through the above mechanism, the online quality control engine achieves accurate matching and efficient execution of rules for recording target health data without relying on large-scale offline computing resources, ensuring the implementation of real-time quality control technology from the perspectives of system architecture and data processing.

[0039] S120. Based on the rule-based traffic distribution results, the offline data quality is reassessed by calling offline sub-rules of offline quality control rules and / or online-offline hybrid quality control rules on historical health data stored in the health data platform.

[0040] In this step, within a preset offline scheduling period, based on the rule-based traffic distribution results, the offline quality control rules and / or offline sub-rules of the online-offline hybrid quality control rules are invoked on the historical health data stored in the health data platform to perform offline data quality re-evaluation and obtain offline quality control results.

[0041] Specifically, the offline data quality reassessment mentioned above can be achieved by configuring offline execution frequency and data domain dependency information for each offline quality control rule and offline sub-rule based on the rule-based traffic distribution results. Then, the corresponding offline quality control task is triggered by the offline scheduling module at a preset scheduling time. Next, the offline quality control rules and offline sub-rules are executed in the data warehouse or data lake based on full or incremental historical health data to obtain the offline quality control results. These offline quality control results include at least one of the following: record-level quality markers, individual-level quality markers, and organization-level quality markers.

[0042] S130. Conduct quality control on health data based on real-time and offline quality control results.

[0043] Specifically, real-time quality control results can be fed back to the front-end business system to provide real-time alerts for data errors or risks during data entry or patient visits. Alternatively, offline quality control results can be pushed to relevant institutions via tasks or messages to indicate objects and fields that need to be supplemented, revisited, or corrected. Alternatively, real-time and offline quality control results can be linked and stored in a quality tag library for subsequent quality analysis and rule optimization.

[0044] In this embodiment, when health data is accessed to the health data platform via an online interface, based on the rule-based traffic distribution results, online sub-rules of online quality control rules and / or online-offline hybrid quality control rules are invoked on the accessed target health data to perform real-time data quality verification and obtain real-time quality control results. Then, within a preset offline scheduling period, based on the rule-based traffic distribution results, offline sub-rules of offline quality control rules and / or online-offline hybrid quality control rules are invoked on the historical health data stored in the health data platform to perform offline data quality re-evaluation and obtain offline quality control results. Finally, based on the real-time quality control results and offline quality control results, quality control is performed on the health data, thereby achieving the synergy between real-time quality verification and periodic offline quality re-evaluation of health data without increasing unacceptable response latency in the front-end business system.

[0045] In other words, the above embodiment sets up both an online quality control engine and an offline quality control engine in its system architecture. In terms of rule management, it uses rule routing to break down each health data quality rule into parts suitable for online execution and parts suitable for offline execution. When a medical institution or maternal and child health care institution reports new health records through an interface, the system automatically calls the online quality control rules and the online sub-rules in the online-offline hybrid rules based on the rule routing results to quickly verify the newly accessed data. This is equivalent to setting up a real-time gatekeeper at the entry point, intercepting format errors, obvious logical conflicts, and other problems as much as possible without significantly increasing the front-end waiting time, and generating real-time quality control results. At the same time, the system starts the offline quality control engine according to a preset offline scheduling cycle, performs batch scanning and re-evaluation of the large amount of historical data accumulated in the platform, calls the offline quality control rules and the offline sub-rules in the online-offline hybrid rules, and conducts in-depth health checks from multiple dimensions such as the entire individual process, records from multiple institutions for one person, and statistical anomalies over a period of time. This identifies global, trend, and combined quality problems that are difficult to detect through online checks, forming offline quality control results.

[0046] exist Figure 1 Based on the illustrated embodiment, it is also worth noting that in existing health information quality control practices, especially in large-scale, cross-institutional data environments such as provincial maternal and child health information platforms, the configuration and execution of data quality rules typically rely on manual experience to determine their operation mode and location: some rules are directly configured for execution in online interface stages, while others are scheduled for execution in end-of-day or monthly offline batch processing tasks. Generally, only the check logic is described in the rule text; standardized rule metadata is not established for each data quality rule, such as the data scope affected by the rule, the data types and volumes it depends on, the time span involved, computational complexity, real-time business requirements, and data quality risk level. Without the representation of these structured technical attributes, the system cannot objectively assess whether a rule is suitable for online execution or must be executed offline. This leads to rule deployment strategies varying based on the experience of project and maintenance personnel, resulting in unpredictable online interface performance, failure to promptly intercept critical errors, and high rule maintenance costs.

[0047] In response, Figure 2 This is a flowchart illustrating a health information quality control method according to another example embodiment of this application. Figure 2 As shown, in this embodiment... Figure 1 Before S110 in the embodiment, the following is also included: S210. Obtain the pre-configured set of health data quality rules.

[0048] In this step, a pre-configured set of health data quality rules can be obtained. Each data quality rule in the set has corresponding rule metadata.

[0049] Specifically, a rule configuration module can be used to model the health data quality inspection logic for various business scenarios, generating a health data quality rule set. Each data quality rule is stored as a structured rule object. Then, a rule metadata record is generated for each data quality rule in the health data quality rule set, and this record is associated with and stored as a corresponding data quality rule. The rule metadata characterizes the technical attributes related to the corresponding data quality rule, and includes at least one of the following: rule identification information, applicable data domain information, rule target hierarchy information, data scope information the rule depends on, time window information involved in rule execution, rule computational complexity information, rule real-time business requirements information, and rule risk level information.

[0050] S220. Based on the rule metadata, determine the real-time score for the data quality rule.

[0051] In this step, the real-time performance score of the data quality rules can be determined based on the rule metadata, and the real-time performance score result corresponding to each data quality rule can be obtained.

[0052] Specifically, at least one of the following can be quantitatively scored in the rule metadata of each data quality rule: The scope parameter is used to characterize the data range in which the rule applies, including at least one of record level, individual level, institutional level, and provincial level. Data dependency parameters are used to characterize the data type and data range of rule dependencies, including at least one of the following: current record, individual's most recent K records, individual's full historical record, cross-institutional historical record, and statistical aggregated data; The time window parameter is used to characterize the time span of the data on which the rule depends, including at least one of the following: single event, most recent preset duration, the entire pregnancy, and the entire life cycle; The computational complexity parameter is used to characterize the computational complexity required to execute the rule; Business real-time parameters are used to characterize the real-time requirements of rules in business scenarios; The risk level parameter is used to characterize the risk level of the data quality problem corresponding to the rule. Based on the quantitative scoring results of parameters such as scope of application, data dependency, time window, computational complexity, business real-time performance, and risk level, the real-time performance score of the corresponding data quality rules is determined according to the preset weighted calculation model.

[0053] It is worth noting that key indicators such as the execution time of a single data quality rule, the number of data records involved in each execution, the false hit rate, the frequency of concurrent execution, and the execution failure rate depend not only on the complexity of the rule logic itself and the number of fields it depends on, but also on a combination of factors such as changes in the number of access institutions, differences in reporting behavior in different regions, growth in the size of data table records, changes in data distribution structure, and adjustments to database indexing strategies and underlying architecture. The execution cost and hit rate of the same rule may vary by orders of magnitude in different time windows and different institutional scenarios.

[0054] These execution characteristics are essentially random processes that change over time. If the system only manually estimates and fills in a few static metadata fields when a rule goes online, and neither collects rule execution characteristic information in the online and offline quality control engines during the rule's operation, nor performs windowed aggregation and statistical updates of the execution characteristic information over time, then fields such as execution time, amount of data checked, and hit rate in the rule metadata will quickly become disconnected from the actual running state. Real-time performance scoring will rely on outdated static configurations rather than real-time data for a long time. As a result, even after a sudden increase in data scale, changes in the institutional access structure, or adjustments to the system architecture, rules with high time consumption, high concurrency, and high hit rate will still be incorrectly classified as online quality control rules, or rules that could be executed online in a lightweight manner will be kept in the offline quality control engine for a long time. This directly leads to problems such as slow online interface response, missed real-time alarms, and delayed offline quality control results.

[0055] To address this, in the online and / or offline quality control engines, rule execution characteristic information can be collected and written to the rule execution log library each time a data quality rule is executed. The rule execution characteristic information includes rule identifier, engine type identifier, data domain identifier, organization identifier, execution start time, execution end time, number of check records per execution, number of error records per execution, and execution success identifier.

[0056] Based on the rule execution log library, the rule execution feature information within the preset time window is aggregated and statistically analyzed in a rolling manner. Runtime rule statistics are generated according to the combination dimension of rule identifier, data domain identifier and / or organization level identifier. Runtime rule statistics include at least one of the following: average execution time, quantile execution time, average number of check records, error hit rate, concurrent execution frequency and execution failure rate.

[0057] Then, the runtime rule statistics are updated to the rule metadata corresponding to the data quality rules, so that the rule metadata can dynamically reflect the execution characteristics of the data quality rules in the current runtime environment.

[0058] Furthermore, the various indicators representing execution time, data access scale, false hit rate, concurrent execution frequency, and execution failure rate in the runtime rule statistics can be normalized to obtain normalized indicator values ​​within a preset numerical range. These normalized indicator values ​​are then weighted using preset weighting coefficients to obtain the real-time performance score corresponding to the data quality rules. Specifically, the runtime rule statistics at different times are smoothed using an exponentially weighted moving average algorithm. The real-time performance score is calculated based on the smoothed runtime rule statistics to reduce the impact of short-term abnormal fluctuations on the real-time performance score.

[0059] Specifically, a unified execution interception point is set in the rule execution framework of both the online and offline quality control engines. Each time a data quality rule is executed, the interception point automatically collects rule execution characteristic information such as rule identifier, engine type identifier, data domain identifier, organization identifier, execution start time, execution end time, number of check records in a single execution, number of error records in a single execution, and execution success identifier. The collected execution characteristic information is then written to the rule execution log library in batches through non-blocking methods such as asynchronous buffer queues. This ensures that the business thread overhead is controllable while completing fine-grained logging of the rule execution process.

[0060] Subsequently, based on the rule execution log library, the system performs rolling aggregation statistics on the stored rule execution characteristic information according to preset time windows, such as the most recent 5 minutes, the most recent 1 hour, the most recent 24 hours, etc., and preset aggregation dimensions, such as the combination of rule identifier and data domain identifier and / or organizational level identifier. It uses statistical methods such as average, quantile, ratio, and count to calculate runtime rule statistics for each data quality rule within the current time window, such as average execution time, quantile execution time, average number of check records, error hit rate, concurrent execution frequency, and execution failure rate. Then, it uses overwrite update or incremental update methods to synchronize the above runtime rule statistics to the rule metadata record corresponding to each rule, so that the rule metadata always reflects the real execution status within the most recent running cycle.

[0061] Based on this, when determining the real-time performance rating of data quality rules according to rule metadata, the dynamically maintained runtime rule statistics can be directly used as input. Combined with normalization and weighted calculation methods, the real-time performance rating results can be generated. This enables the real-time performance rating to be instantly aware of changes in the execution environment and to balance and control statistical noise. As a result, the dynamic adaptive division and gradual adjustment of rule categories (online / offline / online-offline hybrid) are realized, ensuring the long-term stable operation of health information quality control methods under the online / offline hybrid architecture and the continuous optimization of quality control effects.

[0062] S230. Determine the rule distribution result based on the real-time performance score results corresponding to each data quality rule.

[0063] In this step, based on the real-time performance score of each data quality rule, the data quality rules are divided into at least one of the following categories: online quality control rules, offline quality control rules, and hybrid online-offline quality control rules, thus obtaining the rule distribution result.

[0064] Specifically, data quality rules with a real-time performance score greater than or equal to the online threshold can be classified as online quality control rules. Data quality rules with a real-time performance score less than or equal to the offline threshold can be classified as offline quality control rules. Data quality rules with a real-time performance score between the online and offline thresholds can be classified as hybrid online and offline quality control rules.

[0065] It's worth noting that, regarding the above steps, after real-time performance scoring and rule routing, only the entire rule can be labeled as online, offline, or hybrid. However, hybrid rules remain at the label level, meaning that so-called online / offline hybrid quality control rules are still executed as a whole during actual runtime, without fine-grained division of labor between the online and offline engines. Without a structured semantic description of the rules, the system cannot automatically identify which logic can be quickly completed using only the current record and a small number of cached records, and which logic must wait for the data to settle in a data warehouse or data lake before execution.

[0066] To address this, the classification of hybrid online / offline quality control rules can be further refined by performing rule semantic parsing on the target data quality rules, which are then split into online and offline sub-rules. The online sub-rules rely solely on the current health data record and cached historical data within a preset range, and their computational complexity meets the latency constraints for online execution. The offline sub-rules rely on at least one of the following: individual full historical records, cross-institutional records, or statistically aggregated data.

[0067] Specifically, the decomposition of the aforementioned target data quality rules can be achieved by parsing the rule expressions corresponding to the target data quality rules to construct a rule syntax tree representing the logical structure of the rule expressions. Based on the rule syntax tree, the rule expressions are then decomposed into multiple basic inspection units. Each basic inspection unit includes at least one or more of the following: field-level inspection units, record-level relationship inspection units, short-term historical comparison inspection units, and long-term historical or cross-institutional statistical inspection units.

[0068] Then, for each basic inspection unit, based on the data dependency range information, time window information and computational complexity information corresponding to the basic inspection unit in the rule metadata, it is determined whether the basic inspection unit only depends on the current health data record and cached historical data within the preset range, and whether its computational complexity meets the latency constraints of online execution.

[0069] Next, the basic check units that satisfy the requirement of relying only on the current health data record and cached historical data within a preset range, and whose computational complexity meets the latency constraints for online execution, are reorganized into online sub-rules according to the logical connection relationships in the rule syntax tree.

[0070] In addition, basic check units that do not meet the requirements of relying solely on current health data records and cached historical data within a preset range, or that do not meet the latency constraints of online execution, are reorganized into offline sub-rules according to the logical connection relationships in the rule syntax tree.

[0071] It is understandable that after completing the real-time scoring of the rules and dividing the data quality rules into online quality control rules, offline quality control rules, and online-offline hybrid quality control rules, for each target data quality rule marked as an online-offline hybrid quality control rule, the rule semantics are parsed using a preset rule syntax tree or intermediate representation, and the complex expression in the rule composed of logical AND / OR, conditional branches, function calls, subqueries, etc., is decomposed into multiple basic inspection units.

[0072] Secondly, for each basic inspection unit, based on its corresponding rule metadata and the data access pattern extracted during the parsing process, it is determined whether the unit only depends on the target health data record itself and the historical records that can be provided by the lightweight cache within the preset time window, and whether its estimated computational complexity meets the latency threshold preset by the online interface. If it does, the inspection unit is classified into the online sub-rule; otherwise, it is classified into the offline sub-rule.

[0073] Next, the inspection units that will be classified into online sub-rules are recombined according to the logical relationships in the original rules to generate rule expressions for the online sub-rules, and the data access paths for the online sub-rules are marked so that they only need to call the cache interface in the online quality control engine without accessing the data warehouse.

[0074] At the same time, the inspection units that will be classified into offline sub-rules will be reorganized into offline sub-rule expressions according to their original dependencies, and the data domains and execution plans required for cross-organization, cross-time range, or statistical aggregation queries will be configured for the offline sub-rules.

[0075] Ultimately, the system registers online and offline sub-rules in the online and offline rule bases respectively, and establishes a parent-child relationship mapping for the original target data quality rules in the rule distribution results. This enables the online quality control engine to automatically execute online sub-rules when data is accessed, and the offline quality control engine to automatically execute offline sub-rules within the scheduling cycle.

[0076] It is worth noting that, given that S220 has already introduced real-time performance scoring, further solutions are needed to address how to implement a gradual adjustment strategy within each preset traffic distribution adjustment cycle. This strategy should involve controlled quantity / proportion, prioritized order, and version traceability for adjustments to online quality control rule categories, offline quality control rule categories, and mixed online / offline quality control rule categories. Specifically, each time the traffic distribution results are recalculated, the traffic distribution results under the same metadata version are versioned using the rule distribution version number. Furthermore, an upper limit is set on the number or proportion of data quality rules that can be adjusted in a single instance. The data quality rules are then ranked according to their contribution to system resource consumption, prioritizing adjustments only for target data quality rules with a high contribution to system resource consumption. This avoids a concentrated switch of a large number of rule execution categories in a short period, reducing the impact of rule distribution strategy changes on the online / offline engine load and overall system performance. This allows for gradual adjustment and gradual effectiveness of rule distribution results, ensuring that the provincial maternal and child health information platform maintains a stable and controllable operating state even with continuous rule optimization and dynamic adaptive traffic distribution.

[0077] To address this, the rule-based traffic allocation results can be recalculated based on the latest real-time performance score within a preset traffic allocation adjustment period. A rule allocation version number can be assigned to each calculated result, and results calculated based on the same rule metadata version can be associated with the same rule allocation version number. Within each traffic allocation adjustment period, the number or proportion of data quality rules whose execution categories can be adjusted is limited to a preset threshold. Data quality rules are then ranked according to their contribution to system resource consumption, prioritizing adjustments to target data quality rules with higher contributions to system resource consumption across online, offline, and hybrid online / offline quality control rule categories. This allows for a gradual adjustment and effective implementation of the rule-based traffic allocation results.

[0078] Specifically, when each preset traffic splitting adjustment period arrives, the system first generates a candidate rule splitting scheme based on the latest real-time performance score in the rule metadata. This scheme represents the ideal online / offline / mixed category of each data quality rule under the current score. The system then performs a differential comparison with the currently effective rule splitting scheme to obtain a set of candidate rules for which all categories have changed during the current period.

[0079] Subsequently, the system calculates the contribution index of each candidate rule to system resource consumption based on runtime rule statistics such as average execution time, average number of check records, and concurrent execution frequency. The candidate rules are sorted from high to low according to their contribution. Based on the preset adjustment quantity limit or proportion limit, the top N (or top X%) rules are selected from the sorted results as the target rule set for the actual execution category adjustment in this cycle. The remaining candidate rules are not adjusted for the time being and are postponed to subsequent diversion adjustment cycles.

[0080] Based on this, the system only performs changes to the online quality control rule category, offline quality control rule category, and online-offline hybrid quality control rule category for the target rule set. A new effective rule distribution scheme is generated accordingly, and a new rule distribution version number is assigned to this scheme and stored in the rule distribution version repository. This version number is then associated with the corresponding rule metadata version. During runtime, both the online and offline quality control engines consistently determine the rule category and schedule execution based on the rule distribution scheme corresponding to the currently effective version number.

[0081] When the system detects that the current traffic distribution adjustment is detrimental to performance or causes anomalies, it can quickly roll back to the previous version of the rule-based traffic distribution scheme using the version number, further improving system controllability. Through the aforementioned periodic recalculation and differential selection mechanism, combined with hard limits on single-period adjustment quotas and priority control based on contribution ranking, the update of rule-based traffic distribution results is ensured to be a small-scale, high-impact, and reversible gradual process. This avoids the instantaneous load impact caused by a full replacement update strategy, thus ensuring the long-term stable operation and continuous adaptive optimization of the health information quality control method in an online / offline hybrid architecture.

[0082] Figure 3 This is a schematic diagram of the structure of a health information quality control system according to an example embodiment of this application. For example... Figure 3 As shown, the health information quality control system 300 provided in this embodiment includes: The quality verification module 310 is used to perform real-time data quality verification on the target health data when health data is accessed through the online interface, based on the rule-based diversion results, by calling the online quality control rules and / or the online sub-rules of the online-offline hybrid quality control rules to obtain real-time quality control results. The quality assessment module 320 is used to, within a preset offline scheduling period, based on the rule diversion result, call the offline quality control rules and / or the offline sub-rules of the online-offline hybrid quality control rules to reassess the offline data quality of historical health data stored in the health data platform, and obtain the offline quality control result. The quality control system 330 is used to perform quality control on the health data based on the real-time quality control results and the offline quality control results.

[0083] Figure 4 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. For example... Figure 4 As shown, the electronic device 400 provided in this embodiment includes: a processor 401 and a memory 402; wherein: Memory 402 is used to store computer programs, and the memory may also be flash memory.

[0084] Processor 401 is used to execute the execution instructions stored in the memory to implement the various steps in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0085] Alternatively, the memory 402 can be either standalone or integrated with the processor 401.

[0086] When the memory 402 is a device independent of the processor 401, the electronic device 400 may further include: Bus 403 is used to connect the memory 402 and the processor 401.

[0087] This embodiment also provides a readable storage medium storing a computer program, which, when executed by at least one processor of an electronic device, enables the electronic device to perform the methods provided in the various embodiments described above.

[0088] This embodiment also provides a program product including a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the methods provided in the various embodiments described above.

[0089] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0090] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for quality control of health information, characterized in that, It is applied to health information quality control systems with an online / offline hybrid architecture that includes both online and offline quality control engines; The method includes: When health data is accessed to the health data platform through an online interface, based on the rule-based traffic distribution results, the online sub-rules of the online quality control rules and / or the online-offline hybrid quality control rules are invoked on the accessed target health data to perform real-time data quality verification and obtain real-time quality control results. Within a preset offline scheduling period, based on the rule-based traffic distribution results, the offline quality control rules and / or the offline sub-rules of the online-offline hybrid quality control rules are invoked on the historical health data stored in the health data platform to perform offline data quality re-evaluation and obtain offline quality control results. The health data is subject to quality control based on the real-time quality control results and the offline quality control results.

2. The health information quality control method according to claim 1, characterized in that, Before the health data is accessed through the online interface to the health data platform, the following is also included: Obtain a pre-configured set of health data quality rules, wherein each data quality rule in the set has corresponding rule metadata; Based on the rule metadata, the real-time performance score of the data quality rule is determined, and the real-time performance score result corresponding to each data quality rule is obtained; Based on the real-time performance scoring results corresponding to each of the data quality rules, the data quality rules are divided into at least one of the following categories: online quality control rules, offline quality control rules, and hybrid online and offline quality control rules, to obtain the rule distribution results.

3. The health information quality control method according to claim 2, characterized in that, The data quality rules are classified into at least one of the following categories: online quality control rules, offline quality control rules, and hybrid online / offline quality control rules, including: The data quality rules with a real-time performance score greater than or equal to the online threshold are classified as online quality control rules. The data quality rules whose real-time performance score is less than or equal to the offline threshold are classified as offline quality control rules. Data quality rules whose real-time performance score falls between the online threshold and the offline threshold are classified as hybrid online and offline quality control rules.

4. The health information quality control method according to claim 3, characterized in that, The division of the online / offline hybrid quality control rules also includes: The target data quality rules, which are classified into online and offline hybrid quality control rules, are subjected to rule semantic parsing, and the target data quality rules are split into online sub-rules and offline sub-rules. The online sub-rules rely only on the current health data records and cached historical data within a preset range, and their computational complexity meets the latency constraints of online execution. The offline sub-rules rely on at least one of the following: individual full historical records, cross-institutional records, or statistical aggregated data.

5. The health information quality control method according to claim 1, characterized in that, The real-time data quality verification includes: The system receives target health data records from reporting systems of maternal and child health care institutions, midwifery institutions, and community health service institutions through an online access module. Based on the rule-based traffic splitting results and the data domain to which the target health data record belongs, match the online quality control rule set and online sub-rule set associated with the target health data record; The online quality control rule set and online sub-rule set are executed on the target health data record in the memory rule execution engine to obtain the real-time quality control result for the target health data record.

6. The health information quality control method according to claim 5, characterized in that, The execution of the online quality control rule set and online sub-rule set on the target health data record in the memory rule execution engine includes: The target health data record is subjected to a single record verification rule, which includes at least field integrity check, field value range check, enumeration value validity check, and logical consistency check within the same record. Retrieve historical records related to the target health data record from the lightweight history cache based on at least one of the following: individual identifier, organization identifier, and date identifier; Based on the historical records, data quality rules are implemented to detect duplicate registrations within a short period of time and duplicate vaccination records for the same child within the same institution.

7. The health information quality control method according to claim 6, characterized in that, The lightweight history cache is used to store health data window records within a preset time window, and the lightweight history cache is based on a key-value index structure.

8. The health information quality control method according to claim 1, characterized in that, The offline data quality reassessment includes: Based on the rule-based traffic distribution results, configure the offline execution frequency and data domain dependency information for each of the offline quality control rules and offline sub-rules; The offline scheduling module triggers the corresponding offline quality control task at a preset scheduling time. Based on full or incremental historical health data in a data warehouse or data lake, the offline quality control rules and offline sub-rules are executed to obtain offline quality control results. The offline quality control results include at least one of record-level quality markers, individual-level quality markers, and institutional-level quality markers.

9. The health information quality control method according to claim 1, characterized in that, The quality control of the health data based on the real-time quality control results and the offline quality control results includes: The real-time quality control results are fed back to the front-end business system to provide real-time alerts for data errors or risks during data entry or medical treatment. The offline quality control results are pushed to relevant institutions in the form of tasks or messages to indicate the objects and fields that need to be supplemented, revisited, or corrected. The real-time quality control results are associated with the offline quality control results and stored in a quality tag library for subsequent quality analysis and rule optimization.

10. A health information quality control system, characterized in that, include: Online / offline hybrid architecture of online and offline quality control engines; The hybrid architecture includes: The quality verification module is used to perform real-time data quality verification on the target health data when health data is accessed through the online interface, based on the rule-based traffic distribution results, by calling the online quality control rules and / or the online sub-rules of the online-offline hybrid quality control rules, and obtaining real-time quality control results. The quality assessment module is used to, within a preset offline scheduling period, based on the rule-based traffic distribution results, call the offline quality control rules and / or the offline sub-rules of the online-offline hybrid quality control rules to reassess the offline data quality of historical health data stored in the health data platform, and obtain the offline quality control results. The quality control system is used to perform quality control on the health data based on the real-time quality control results and the offline quality control results.

Images