Digital quality management method and system, device, storage medium

Abstract

The application discloses a digital quality management method and system, equipment and a storage medium, and belongs to the technical field of digital quality management and quality performance analysis, and aims to solve problems of quality data multi-source heterogeneous dispersion, low efficiency of manual tracing of quality abnormalities and the like in consumable quality management, complete standardized management by collecting full-link multi-source target data, calculate the correlation value between various data by using a maximum information coefficient method, determine a target quality relationship, construct a medical consumable quality knowledge graph by taking the target data as entity nodes and the target quality relationship as a weighted directed edge, perform upstream path traversal and node contribution quantization calculation based on the graph, realize quality abnormality root cause positioning and output quality management decision results, and the application can be widely applied to consumable whole-process quality management, abnormality tracing and fine quality performance evaluation, and effectively improves quality management efficiency and decision scientificity.

Description

Technical Field

[0001] This invention relates to the field of digital quality management and quality performance analysis technology, specifically to a digital quality management method, system, electronic device and storage medium for the entire chain of medical consumables, which can be applied to the whole process quality assessment, anomaly tracing and refined quality control in the medical supply chain scenario. Background Technology

[0002] With the continuous growth in the usage and variety of high-value medical consumables in hospitals, end-to-end quality control of consumables has become a core component of hospital medical quality management. End-to-end quality management is a core component of supply chain operation and control, and quality performance analysis and anomaly tracing capabilities are key indicators for measuring the level of quality management. In the context of medical consumables management, consumable quality is directly related to clinical medical safety, placing high demands on the accuracy, timeliness, and traceability of end-to-end quality management.

[0003] Current practices in medical consumables quality management suffer from three main shortcomings: First, the fragmentation and heterogeneity of quality data are prominent. Quality data from procurement and supply, clinical use, quality inspection, and logistics warehousing belong to different business systems, resulting in inconsistent data standards and fragmented business processes. This makes it difficult to form a complete end-to-end quality data system and supports overall quality performance analysis. Second, the efficiency of tracing quality anomalies is low. When adverse events related to consumables occur, manual cross-system data retrieval and step-by-step verification are often used, leading to long processing cycles, high labor costs, and reliance on personnel experience for location results, resulting in insufficient quality risk response capabilities. Third, the accuracy of quality correlation analysis is limited. Traditional analysis methods are mostly based on simple linear statistical logic, making it difficult to effectively capture the nonlinear coupling relationships between quality-influencing factors at multiple stages. This makes it impossible to accurately quantify the contribution of each factor to the quality outcome and provide quantitative performance evaluation basis for quality management decisions. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention discloses a digital quality management method, system, device, and storage medium to solve the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a digital quality management method, comprising the following steps:

[0006] S1. Acquire multiple types of target data throughout the entire medical consumables supply chain, including quality-related data from procurement and supply, clinical use, quality inspection, and logistics and warehousing.

[0007] S2. Based on the distribution characteristics of each target data, calculate the correlation value between variables corresponding to any two types of target data; in response to the correlation value being greater than a preset correlation threshold, determine the target quality relationship between the corresponding two types of target data.

[0008] S3. Based on all target data and the target quality relationships between each target data, construct a knowledge graph of medical consumable quality;

[0009] S4. Based on the medical consumables quality knowledge graph, perform quality analysis and root cause localization of target medical consumables, and output quality management decision results.

[0010] Preferably, step S1, which involves acquiring multiple types of target data across the entire supply chain, specifically includes:

[0011] Heterogeneous raw data were collected from the procurement management system, clinical information system, quality inspection and follow-up system, and logistics traceability system, respectively.

[0012] The heterogeneous raw data is processed by missing value imputation, outlier removal, standardized encoding, and field mapping to obtain structured target data;

[0013] The target data of each stage are linked and aligned according to the batch identifier of medical consumables to form a multi-dimensional quality dataset with consumable batches as the core.

[0014] Preferably, in step S2, the step of calculating the correlation value between variables corresponding to any two types of target data based on the distribution characteristics of each target data is performed using the maximum information coefficient method. This method corresponds to the processing of quantifying the correlation degree of heterogeneous supply chain data based on probability distribution. The output correlation value serves as the quantitative input basis for subsequent target quality relationship determination. Specifically, it includes: dividing variables X and Y corresponding to any two types of target data into grids in a two-dimensional space and calculating the joint probability distribution within each grid cell; calculating the mutual information value based on the joint probability distribution and marginal probability distribution, and traversing all feasible grid division methods to obtain the maximum mutual information value; obtaining the maximum information coefficient MIC(X,Y) as the correlation strength value through normalization processing. The maximum information coefficient has a value range of [0,1] and is used to quantify the correlation degree of the two types of target data; determining the correlation direction by combining the monotonic change trend of the variables corresponding to the two types of target data, and using the correlation strength value with directional attributes as the correlation value. The absolute value of the correlation value is equal to the value of the corresponding maximum information coefficient. The calculation formula for the maximum information coefficient is:

[0015] ,

[0016] In the formula, This represents the mutual information between variables X and Y, where 'a' represents the number of grid divisions along the X-axis, 'b' represents the number of grid divisions along the Y-axis, and 'n' represents the total number of sample data. This indicates taking the smaller value between a and b; the formula for calculating the mutual information is:

[0017] ,

[0018] In the formula, This represents the joint probability when variable X takes the value x and variable Y takes the value y. This represents the marginal probability that variable X takes the value x. This represents the marginal probability that variable Y takes the value y.

[0019] Preferably, in step S2, the step of determining the target quality relationship between the two types of target data in response to a correlation value greater than a preset correlation threshold specifically includes:

[0020] The preset correlation threshold is set to a value range of 0.3 to 0.8; when the absolute value of the correlation value corresponding to two types of target data is greater than the preset correlation threshold, it is determined that there is a quality influence relationship between the two types of target data; the correlation boundary value is set to a value range of 0.5 to 0.7, and based on the positive or negative attribute of the correlation value and the size relationship between the absolute value of the correlation value and the correlation boundary value, the quality influence relationship is divided into four types: positive high correlation, positive low correlation, negative high correlation, and negative low correlation, which are used as the target quality relationship.

[0021] Preferably, in step S3, a medical consumables quality knowledge graph is constructed based on all target data and the target quality relationships between each target data point, specifically including:

[0022] Define the entity types of the knowledge graph, including consumable batch entities, supplier entities, acceptance indicator entities, surgical operation entities, logistics and warehousing node entities, and prognosis entities;

[0023] Map each target data to a corresponding type of entity node;

[0024] The target quality relationship is mapped as directed attribute edges between entity nodes. Each directed attribute edge is assigned a correlation value, influence direction, associated link, and confidence label to form a weighted directed medical consumables quality knowledge graph.

[0025] Preferably, in step S4, the step of performing quality analysis and root cause localization of the target medical consumable based on the medical consumable quality knowledge graph specifically includes:

[0026] When a quality abnormality is detected in the target medical consumable, the entity node corresponding to the abnormal outcome is taken as the starting node;

[0027] In the medical consumables quality knowledge graph, a breadth-first traversal and path reasoning are performed along the directed attribute edges to trace upstream associated acceptance indicator nodes, consumable batch nodes, surgical operation nodes, and logistics and warehousing nodes.

[0028] The contribution of each upstream node to the quality anomaly is calculated. This contribution corresponds to the process of quantifying the influence of nodes after path reasoning. The output contribution value provides a quantitative basis for the ranking and determination of quality root cause nodes. A predetermined number of nodes with the highest contribution ranking are determined as quality root cause nodes. The output includes the root cause nodes, related paths, and the degree of influence. The formula for calculating the contribution is:

[0029] ,

[0030] In the formula, This represents the contribution of the i-th upstream node. This represents the set of all paths from the starting node to the i-th upstream node. This represents the product of the correlation values ​​of each directed attribute edge on the k-th path. This represents the path length of the k-th path.

[0031] The present invention also provides a digital quality management system, including a multi-source data governance module, a quality correlation mining module, a knowledge graph construction module, and a quality traceability analysis module;

[0032] The multi-source data governance module is used to access heterogeneous raw data from various stages of the entire medical consumables supply chain, including procurement and supply, clinical use, quality inspection, and logistics and warehousing. It sequentially performs data cleaning, field standardization coding, and consumables batch-level correlation and alignment processing to output structured multi-category target data.

[0033] The quality correlation mining module is used to receive the target data output by the multi-source data governance module, calculate the maximum information coefficient between any two types of target data based on the probability distribution characteristics of each target data as the correlation value, compare the correlation value with a preset correlation threshold, and obtain the target quality relationship between the corresponding target data according to the comparison result and the positive and negative attributes of the correlation value.

[0034] The knowledge graph construction module is used to construct a medical consumables quality knowledge graph with multi-dimensional attribute labels, using various target data as corresponding entity nodes and the target quality relationship as weighted directed edges, and to complete structured storage.

[0035] The quality traceability analysis module is used to, when a quality abnormality of a target medical consumable is detected, take the entity node corresponding to the abnormal outcome as the starting point, perform upstream path traversal and node contribution quantification calculation in the medical consumable quality knowledge graph, locate the root cause node of the quality, and output the quality management decision results containing the associated path and the degree of impact.

[0036] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the digital quality management method described above.

[0037] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the digital quality management method described above.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] 1. This invention improves the ability to fuse multi-source quality data and the accuracy of correlation analysis. By standardizing the data governance process, it breaks down the barriers between quality data from multiple systems and forms a unified quality dataset with consumable batches as the core. The maximum information coefficient method can simultaneously capture the linear and nonlinear correlations between quality influencing factors, effectively improving the comprehensiveness and accuracy of quality correlation mining and providing a reliable data foundation for quality performance evaluation.

[0040] 2. This invention shortens the quality anomaly tracing cycle, improves the accuracy of root cause localization, constructs a weighted directed medical consumables quality knowledge graph, and combines breadth-first path traversal and node contribution quantification algorithms to achieve automated reverse tracing of quality anomalies. Compared with the traditional manual investigation mode, it significantly reduces the time for quality root cause localization, improves the objectivity and accuracy of localization results, and strengthens the ability to quickly respond to and handle quality risks.

[0041] 3. This invention constructs a closed-loop digital quality control system across the entire chain, supporting refined quality decision-making and forming a complete digital quality management chain of quality data collection, correlation mining, knowledge graph construction, and anomaly tracing analysis, providing quality control departments with clear rectification directions and quantitative performance basis. Attached Figure Description

[0042] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0043] In the attached diagram:

[0044] Figure 1 This is a flowchart of a digital quality management method according to the present invention;

[0045] Figure 2 This is a simplified diagram of the module structure of a digital quality management system according to the present invention. Detailed Implementation

[0046] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0047] like Figure 1 As shown, this embodiment provides a digital quality management method. This embodiment is applied to the whole-chain quality management scenario of high-value medical consumables in a tertiary-level Class A hospital. The whole-chain data of procurement, acceptance, clinical use and postoperative follow-up of 987 orthopedic implant consumables are selected as the calculation sample. It aims to solve the problem that the hospital consumable supply chain data is scattered in multiple business systems and it is difficult to quickly locate the root cause after the occurrence of adverse quality events.

[0048] S1. Collection and management of multi-source medical quality target data: This step is used to access and standardize heterogeneous medical quality data from various aspects of the hospital, eliminate the differences in data format and dimensionality between different information systems, and provide a unified dataset for subsequent correlation calculations.

[0049] In the specific implementation process, raw data from four stages—procurement, warehousing and acceptance, clinical use, and postoperative follow-up—were accessed through the Hospital Resource Planning (HRP) system, the Medical Consumables Management System, the Hospital Information System (HIS), and the Adverse Event Reporting and Follow-up System. The procurement data included consumable batch numbers, supplier qualification levels, sterilization parameters, and factory inspection indicators; the warehousing and acceptance data included visual inspection results, sterility test values, packaging integrity, and storage environment parameters; the clinical use data included patient basic information, surgical operation parameters, surgeon qualification levels, and consumable usage duration; and the postoperative follow-up data included incision healing level, infection incidence, rejection indicators, and prognostic scores.

[0050] For the collected heterogeneous raw data, data cleaning and standardization processes are performed sequentially:

[0051] 1) Missing value filling: Linear interpolation is used to fill continuous numerical fields. The filling rule is to take the average of the two valid records before and after the field as the filling value for the missing position. This method is suitable for continuous fields such as warehouse temperature and humidity, and operation time.

[0052] 2) Outlier Removal: The 3σ criterion is used for identification and removal. That is, when the difference between a value and the mean of the field is greater than three times the standard deviation, it is determined to be an outlier and removed. Taking the "Consumables Storage Temperature" field in this embodiment as an example, the mean of all samples is 22.1℃, the standard deviation is 0.8℃, and the three-standard-deviation threshold is 2.4℃. A certain record shows a measured storage temperature of 25.3℃, which differs from the mean by 3.2℃, greater than 2.4℃, and is therefore determined to be an outlier and removed.

[0053] 3) Field Mapping and Unified Encoding: Mapping and aligning of fields with the same name and batch identifier across different systems is performed. For example, the procurement batch number format in the HRP system is “CG-20240615-0032”, and the warehousing batch number format in the consumables management system is “RK240615089”. Using the mapping rules between production date and serial number, these are uniformly converted to an 18-digit standard medical consumables batch number in the format “YL2024061500000032”.

[0054] 4) Numerical Normalization: A minimum-maximum normalization method is used to map all numerical fields to the [0,1] interval, eliminating dimensional differences between different fields and improving the accuracy of subsequent correlation calculations. The normalization formula is: ,

[0055] In the formula, This represents the target value after normalization, where x represents the value of the original field. This indicates the minimum value of this field among all samples. This indicates the maximum value of this field across all samples.

[0056] After data cleaning and standardization, the data from the four stages of procurement, acceptance, use, and follow-up were correlated and aligned using the standard consumable batch number as the primary key. The entire data chain corresponding to the same batch of consumables was matched and integrated to form a multi-dimensional medical quality dataset centered on the consumable batch, which serves as the target data for subsequent steps. In this embodiment, 987 sets of valid batch data were obtained after processing, each set containing 24 quality-related fields from different stages.

[0057] S2. Correlation Calculation of Medical Quality Data Based on Maximum Information Coefficient: This step quantifies the degree of correlation between any two types of medical quality data based on the probability distribution characteristics of the target data. It can simultaneously capture linear and nonlinear correlations, adapting to the complex correlation characteristics of multi-factor coupling in medical scenarios. The output correlation value serves as the basis for determining quality relationships. This embodiment uses the Maximum Information Coefficient (MIC) to calculate correlations. This can be implemented using the minepy library in Python, or it can be implemented independently based on the following formula. The maximum information coefficient is used to quantify the correlation strength, and the direction of influence between variables is determined by monotonic trends. Combining the MIC strength and the direction of influence, the subsequent quality relationship classification is completed.

[0058] The specific calculation process is as follows:

[0059] 1) Select any two types of target data corresponding to variables X and Y. In this embodiment, “humidity of consumable storage environment” is used as variable X and “incidence of postoperative incision infection” is used as variable Y. “remaining days of sterilization validity” is used as variable X and “incidence of implant rejection” is used as variable Y. A two-dimensional scatter space is constructed based on 987 sets of sample data.

[0060] 2) Divide the two-dimensional space into a grid, dividing it into 'a' intervals along the X-axis and 'b' intervals along the Y-axis, forming... The grid cells are defined by the following conditions: where a and b are integers not less than 2, and satisfy the constraints. Iterate through all combinations of values ​​for a and b that satisfy the constraints, count the number of samples in each grid cell, and calculate the joint probability distribution of each grid cell, as well as the marginal probability distributions of the two variables. The total number of samples in this embodiment... , Therefore, the total number of grid divisions does not exceed 61.

[0061] 3) Calculate the mutual information of the two variables based on the joint probability and marginal probability. The formula for calculating mutual information is: ,

[0062] In the formula, This represents the joint probability when variable X takes the value x and variable Y takes the value y. This represents the marginal probability that variable X takes the value x. This represents the marginal probability that variable Y takes the value y.

[0063] 4) Take the maximum mutual information value under all grid partitioning methods, and obtain the maximum information coefficient through normalization as the correlation value of the two types of target data. The formula for calculating the maximum information coefficient is: ,

[0064] In the formula, This indicates taking the smaller value between a and b.

[0065] Following the above method, the maximum information coefficient between all pairwise combinations of target data is calculated to obtain a matrix of correlation values ​​between all pairwise target data. In this embodiment, the maximum information coefficient between the humidity of the consumable storage environment and the incidence of postoperative incision infection is 0.68, showing a positive trend, and the corresponding correlation value is recorded as 0.68; the maximum information coefficient between the remaining days of sterilization validity and the incidence of implant rejection is 0.62, showing a negative trend, and the corresponding correlation value is recorded as -0.62.

[0066] Furthermore, the determination of medical quality relationships based on threshold judgment involves identifying the quality impact relationships between different medical quality data based on the comparison results of the correlation value and the preset threshold, providing a classification basis for the edge construction of the knowledge graph. The preset correlation threshold is calibrated based on historical hospital quality data, with a value range of 0.3 to 0.8, preferably set to 0.5 in this embodiment; the association boundary value is used to distinguish the association strength, with a value range of 0.5 to 0.7, preferably set to 0.6 in this embodiment.

[0067] In the specific implementation process, when the absolute value of the correlation value between the two types of target data is greater than the preset correlation threshold, it is determined that there is a quality influence relationship between the two types of target data; when the absolute value of the correlation value is less than or equal to the preset correlation threshold, it is determined that there is no significant quality influence relationship between the two types of target data and they are not included in the subsequent map construction scope.

[0068] For data pairs that have a quality impact relationship, based on the positive or negative attribute of the correlation value and the relationship between the absolute value and the correlation threshold, four types of target quality relationships are identified: when the correlation value is greater than 0 and the absolute value is greater than the correlation threshold, it is determined to be a positive high correlation; when the correlation value is greater than 0 and the absolute value is less than or equal to the correlation threshold, it is determined to be a positive low correlation; when the correlation value is less than 0 and the absolute value is greater than the correlation threshold, it is determined to be a negative high correlation; and when the correlation value is less than 0 and the absolute value is less than or equal to the correlation threshold, it is determined to be a negative low correlation.

[0069] In this embodiment, the correlation between the humidity of the consumable storage environment and the incidence of postoperative incision infection is 0.68. Its absolute value is greater than the preset correlation threshold of 0.5, and its absolute value is greater than the correlation boundary value of 0.6 and the sign is positive. Therefore, the two are determined to be a positively correlated target quality relationship, that is, when the humidity of the storage environment increases, the incidence of postoperative incision infection increases synchronously. The correlation between the remaining days of sterilization validity period and the incidence of implant rejection is -0.62. Its absolute value is greater than the preset correlation threshold of 0.5, and its absolute value is greater than the correlation boundary value of 0.6 and the sign is negative. Therefore, the two are determined to be a negatively correlated target quality relationship, that is, the longer the remaining days of sterilization validity period, the lower the incidence of rejection.

[0070] S3. Construction of Medical Quality Knowledge Graph: This step transforms medical quality target data and quality relationships into a structured weighted directed knowledge graph, providing a graph structure foundation for subsequent quality traceability. In this embodiment, Neo4j graph database is used to complete the storage and query of the graph.

[0071] In the specific implementation process, six entity types are first defined for the knowledge graph: consumable batch entity, acceptance indicator entity, surgical operation entity, prognosis entity, supplier entity, and warehousing node entity. Entity mapping is performed according to the business domain to which the field belongs: procurement data is mapped to consumable batch entity and supplier entity nodes; acceptance data is mapped to acceptance indicator entity nodes; clinical usage data is mapped to surgical operation entity nodes; follow-up data is mapped to prognosis entity nodes; and warehousing environment data is mapped to warehousing node entity nodes. Each entity node contains its own attribute information. For example, the consumable batch entity includes consumable code, sterilization parameters, and supplier attributes; the surgical operation entity includes operation duration, surgeon information, and operation difficulty attributes.

[0072] The target quality relationships identified in the preceding steps are mapped to directed attribute edges between entity nodes. The direction of the edges points from the source node to the affected node; for example, a warehousing node entity points to an acceptance indicator entity, and an acceptance indicator entity points to a prognostic outcome entity. Four attribute labels are attached to each directed attribute edge: the first is the correlation value, i.e., the correlation calculation result of the corresponding target data pair; the second is the direction of influence, which can be positive or negative; the third is the link in the chain, marking the medical supply chain link that the relationship crosses, such as procurement-acceptance, acceptance-clinical; and the fourth is the confidence label, which is the correlation value itself, representing the credibility of the quality relationship. This is used to filter low-confidence paths during subsequent root cause localization, avoiding invalid associations from interfering with the judgment results.

[0073] After constructing the nodes and edges, a weighted directed medical quality knowledge graph is formed. In this embodiment, a total of 126 entity nodes and 312 directed attribute edges are constructed, covering the quality association paths of the entire consumable supply chain. The graph supports incremental updates. When a new batch of complete end-to-end data is added, only the corresponding entity nodes and associated edges are added; for existing edges, the relevance values ​​are recalculated based on the new data and the edge weights are updated, without the need to rebuild the entire graph.

[0074] S4. Knowledge Graph-Based Medical Quality Analysis and Root Cause Locator: This step, based on the constructed quality knowledge graph, performs reverse tracing and contribution quantification when adverse medical quality events occur, accurately locates the root causes of quality issues, and outputs actionable medical quality management decisions.

[0075] The specific implementation process is as follows:

[0076] 1) Abnormal Trigger: Real-time monitoring of postoperative follow-up and adverse event reporting data. When the incidence of adverse events corresponding to a batch of consumables exceeds a preset warning threshold, the root cause localization process is triggered, using the entity node corresponding to the abnormal outcome as the traversal starting node. In this embodiment, the warning threshold for postoperative incision infection of orthopedic implants is 3%. The postoperative infection rate corresponding to a batch of consumables is 4.8%, exceeding the warning threshold. Therefore, the entity node for the prognostic outcome "postoperative incision infection" is used as the starting node.

[0077] 2) Path Traversal: Perform a breadth-first traversal in reverse along the directed attribute edges in the knowledge graph, tracing back to all upstream related nodes. The traversal depth is set to 3 levels. This value is based on the length of the medical consumables supply chain. Tracing from prognosis to raw material procurement typically does not exceed 3 levels of transmission links. A 3-level traversal can cover more than 95% of the quality impact paths while avoiding traversal redundancy. During the traversal, only edges with a confidence level greater than or equal to the preset relevance threshold are retained. All reachable paths from the starting node to each upstream node are recorded, along with the edges and corresponding weights contained in each path.

[0078] 3) Contribution Calculation: For each upstream node, its contribution to the quality anomaly is calculated based on all reachable paths. The shorter the path and the higher the edge weight, the stronger the influence of the corresponding node. The contribution calculation formula is: ,

[0079] In the formula, This represents the contribution of the i-th upstream node; This represents the set of all paths from the starting node to the i-th upstream node; It represents the product of the relevance values ​​of all directed attribute edges on the k-th path; This represents the path length of the k-th path, i.e., the number of edges contained in the path. In this embodiment, the "warehouse environment humidity" node has two reachable paths: the first path has a length of 2, and the correlation values ​​of the two edges on the path are 0.68 and 0.72, respectively, with a corresponding weight product of 0.4896; the second path has a length of 3, and the correlation values ​​of the three edges on the path are 0.55, 0.61, and 0.70, respectively, with a corresponding weight product of 0.2349. Substituting these values ​​into the formula, the contribution of this node is calculated to be 0.4896 / 2 + 0.2349 / 3 = 0.3231.

[0080] 4) Root Cause Determination and Result Output: All upstream nodes are sorted from highest to lowest contribution, and a predetermined number of nodes at the top of the ranking are identified as root cause nodes. The number of root causes is set based on hospital quality management experience; the core root causes of a single batch of quality anomalies are usually 1-3. In this embodiment, the top two nodes are selected as root causes, balancing accuracy in location and efficiency in rectification implementation. In this embodiment, the "warehouse environment humidity" node ranks first in contribution and is identified as the main root cause of the excessive postoperative infection rate. The final output quality management decision results include the root cause node name, details of the associated path, numerical values ​​of the impact of each node, and corresponding rectification suggestions.

[0081] like Figure 2 As shown, this embodiment also provides a digital quality management system for executing the above-mentioned digital quality management method. The system includes four functional units: a multi-source data governance module, a quality correlation mining module, a knowledge graph construction module, and a quality traceability analysis module. Each module operates collaboratively in sequence according to the data flow direction.

[0082] The multi-source data governance module comprises a data access unit, a data cleaning unit, and a batch alignment unit. The data access unit connects to the hospital's internal HRP, HIS, and consumables management systems via standardized interfaces, enabling scheduled batch access of heterogeneous raw data from procurement, acceptance, clinical, and follow-up stages. The data cleaning unit incorporates algorithms for missing value imputation, outlier removal, and field standardization to process the raw data. The batch alignment unit uses the standard consumable batch number as the primary key to perform data correlation and matching across multiple stages, outputting structured target data and pushing it to the quality correlation mining module.

[0083] The quality correlation mining module includes a distribution statistics unit, a coefficient calculation unit, and a relationship determination unit. The distribution statistics unit is used to statistically analyze the probability distribution characteristics of each medical quality target data; the coefficient calculation unit has built-in maximum information coefficient calculation logic and outputs the correlation value between each pair of target data; the relationship determination unit classifies and determines the quality relationship of data pairs based on preset thresholds and boundary values, outputs the target quality relationship set, and transmits it to the knowledge graph construction module.

[0084] The knowledge graph construction module comprises an entity definition unit, a graph generation unit, and a graph storage unit. The entity definition unit pre-configures attribute rules for six types of medical quality entities, enabling the mapping of target data to entity nodes. The graph generation unit transforms target quality relationships into weighted directed edges, completing the generation and incremental updating of the graph structure. The graph storage unit interfaces with the Neo4j graph database to achieve persistent storage of the medical quality knowledge graph and provides a query interface to the quality traceability analysis module.

[0085] The quality traceability analysis module includes an anomaly detection unit, a path traversal unit, and a root cause output unit. The anomaly detection unit monitors postoperative follow-up and adverse event data in real time, identifies quality anomalies, and triggers the traceability process; the path traversal unit executes a breadth-first graph traversal algorithm to trace the upstream related paths of the anomaly nodes; the root cause output unit calculates and sorts the contribution of each node, and outputs the root cause localization results and quality management decisions.

[0086] This embodiment also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements all the steps of the aforementioned digital quality management method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, RAM, plug-in hard disk, smart memory card, secure digital card, flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage device. The device can be deployed in the hospital's information technology department computer room, adapting to the hospital's existing information architecture. The processor may be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0087] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements all the steps of the aforementioned digital quality management method. The storage medium can be any available medium accessible to a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.

[0088] The method disclosed in this embodiment can significantly improve the efficiency and accuracy of tracing quality issues in the hospital supply chain, and effectively solve the problem of difficulty in unified correlation analysis of multi-source heterogeneous medical data, providing quantitative decision support for hospital quality control departments.

[0089] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital quality management method, characterized in that, Includes the following steps: S1. Acquire multiple types of target data throughout the entire medical consumables supply chain, including quality-related data from procurement and supply, clinical use, quality inspection, and logistics and warehousing. S2. Based on the distribution characteristics of each target data, calculate the correlation value between variables corresponding to any two types of target data; in response to the correlation value being greater than a preset correlation threshold, determine the target quality relationship between the corresponding two types of target data. S3. Based on all target data and the target quality relationships between each target data, construct a knowledge graph of medical consumable quality; S4. Based on the medical consumables quality knowledge graph, perform quality analysis and root cause localization of target medical consumables, and output quality management decision results.

2. The digital quality management method according to claim 1, characterized in that: Step S1, which involves acquiring multiple types of target data across the entire supply chain, specifically includes: Heterogeneous raw data were collected from the procurement management system, clinical information system, quality inspection and follow-up system, and logistics traceability system, respectively. The heterogeneous raw data is processed by missing value imputation, outlier removal, standardized encoding, and field mapping to obtain structured target data; The target data of each stage are linked and aligned according to the batch identifier of medical consumables to form a multi-dimensional quality dataset with consumable batches as the core.

3. The digital quality management method according to claim 1, characterized in that: In step S2, the step of calculating the correlation value between variables corresponding to any two types of target data based on the distribution characteristics of each target data is performed using the maximum information coefficient method. This method corresponds to the processing of quantifying the correlation degree of heterogeneous supply chain data based on probability distribution. The output correlation value serves as the quantitative input basis for subsequent determination of target quality relationships, specifically including: For any two types of target data corresponding to variables X and Y, a grid is divided in two-dimensional space, and the joint probability distribution within each grid cell is calculated. The mutual information value is calculated based on the joint probability distribution and the marginal probability distribution, and all feasible grid division methods are traversed to obtain the maximum mutual information value. The maximum information coefficient MIC(X,Y) is obtained through normalization and used as the association strength value. The formula for calculating the maximum information coefficient is as follows: ， In the formula, This represents the mutual information between variables X and Y, where 'a' represents the number of grid divisions along the X-axis, 'b' represents the number of grid divisions along the Y-axis, and 'n' represents the total number of sample data. This indicates taking the smaller value between a and b; the formula for calculating the mutual information is: ， In the formula, This represents the joint probability when variable X takes the value x and variable Y takes the value y. This represents the marginal probability that variable X takes the value x. This represents the marginal probability that variable Y takes the value y.

4. The digital quality management method according to claim 3, characterized in that: In step S2, the step of determining the target quality relationship between the two types of target data in response to a correlation value greater than a preset correlation threshold specifically includes: The preset correlation threshold is set to a value range of 0.3 to 0.8; when the correlation value corresponding to two types of target data is greater than the preset correlation threshold, it is determined that there is a quality influence relationship between the two types of target data; the correlation boundary value is set to a value range of 0.5 to 0.7, and based on the positive or negative attribute of the correlation value and the relationship between the absolute value of the correlation value and the correlation boundary value, the quality influence relationship is divided into four types: positive high correlation, positive low correlation, negative high correlation, and negative low correlation, which are used as the target quality relationship.

5. The digital quality management method according to claim 1, characterized in that: In step S3, based on all target data and the target quality relationships between each target data point, a medical consumables quality knowledge graph is constructed, specifically including: Define the entity types of the knowledge graph, including consumable batch entities, supplier entities, acceptance indicator entities, surgical operation entities, logistics and warehousing node entities, and prognosis entities; Map each target data to a corresponding type of entity node; The target quality relationship is mapped as directed attribute edges between entity nodes. Each directed attribute edge is assigned a correlation value, influence direction, associated link, and confidence label to form a weighted directed medical consumables quality knowledge graph.

6. The digital quality management method according to claim 1, characterized in that: Step S4 involves performing quality analysis and root cause localization of the target medical consumable based on a medical consumable quality knowledge graph. Specifically, this includes: When a quality abnormality is detected in the target medical consumable, the entity node corresponding to the abnormal outcome is taken as the starting node; In the medical consumables quality knowledge graph, a breadth-first traversal and path reasoning are performed along the directed attribute edges to trace upstream associated acceptance indicator nodes, consumable batch nodes, surgical operation nodes, and logistics and warehousing nodes. The contribution of each upstream node to the quality anomaly is calculated. This contribution corresponds to the process of quantifying the influence of nodes after path reasoning. The output contribution value provides a quantitative basis for the ranking and determination of quality root cause nodes. A predetermined number of nodes with the highest contribution ranking are determined as quality root cause nodes. The output includes the root cause nodes, related paths, and the degree of influence. The formula for calculating the contribution is: ， In the formula, This represents the contribution of the i-th upstream node. This represents the set of all paths from the starting node to the i-th upstream node. This represents the product of the correlation values ​​of each directed attribute edge on the k-th path. This represents the path length of the k-th path.

7. A digital quality management system, applied to the digital quality management method according to any one of claims 1-6, characterized in that: It includes a multi-source data governance module, a quality correlation mining module, a knowledge graph construction module, and a quality traceability analysis module; The multi-source data governance module is used to access heterogeneous raw data from all links of the entire medical consumables supply chain, and output structured target data after data cleaning, standardized coding and batch alignment. The quality correlation mining module is used to calculate the maximum information coefficient as the correlation value based on the distribution characteristics of the target data, and determine the target quality relationship between the target data through threshold comparison and positive and negative attribute classification. The knowledge graph construction module is used to construct and store a medical consumables quality knowledge graph with multi-dimensional attribute labels, using target data as entity nodes and target quality relationships as weighted directed edges. When a quality anomaly is detected, the quality traceability analysis module performs upstream path traversal and contribution quantification calculation based on the knowledge graph to locate the root cause of the quality issue and output the quality management decision results.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the digital quality management method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the digital quality management method as described in any one of claims 1 to 6.

Images