Health medical data quality evaluation method and system, storage medium and electronic device

Abstract

The invention discloses a health medical data quality evaluation method and system, a storage medium and electronic equipment, and the method comprises the following steps: 1, building a health medical data quality index system, each quality index comprising one or more core themes; step 2, identifying key quality indexes in the health medical data quality index system by using DEMATEL, calculating the influence degree, the affected degree, the centrality and the cause degree of each quality index, constructing an influence relation graph based on the centrality and the cause degree, and analyzing the quadrant of each quality index in the influence relation graph so as to determine the quality index of the health medical data; determining the importance degree of each quality index in the framework; constructing an influence relationship among visual indexes of a multi-level confrontation topology hierarchical structure model in combination with AISM, orderly placing each quality index in different levels by the confrontation topology hierarchical structure model, and connecting each node by using a directed line segment to represent the influence relationship; and 3, calculating the weight of each quality index by using ANP, and evaluating the data quality in combination with FCE.

Description

Technical Field

[0001] This invention belongs to the field of data quality assessment technology, specifically relating to a method, system, storage medium, and electronic device for assessing the quality of health and medical data. Background Technology

[0002] With the rise of the "data as a product" concept, data is increasingly being regarded as a crucial strategic resource. Health and medical data, characterized by its multi-source and heterogeneous nature, also carries patient privacy and clinical research value, thus being widely considered a valuable asset. Despite the widespread recognition of the value of health and medical data, data quality issues severely restrict its application in clinical research and healthcare management decisions. Health and medical data may face quality problems such as missing data, inconsistencies, and insufficient timeliness at each stage of its lifecycle. A review of existing literature reveals that scholars generally believe health and medical data quality assessment mainly includes two aspects: data quality indicators and indicator measurement standards. First, traditional data quality indicators such as completeness and timeliness are used to characterize data quality, and then qualitative or quantitative measurement standards for each indicator are defined. However, a unified consensus has not yet been reached on the definitions and measurement standards of these quality indicators, making it difficult to objectively describe the unique attributes of data quality in the context of health and medical care, thus keeping health and medical data quality assessment work at the theoretical stage.

[0003] The existing data quality framework has laid the theoretical foundation for health and medical data quality assessment, but it still has the following shortcomings: First, it lacks an indicator system for assessing health and medical data quality; second, it fails to fully consider the influence relationship between various data quality indicators, making it difficult to clarify the degree of influence of each indicator on the overall data quality; and third, it lacks assessment methods for health and medical data quality. Summary of the Invention

[0004] To address three major problems in existing health and medical data quality assessments—namely, the lack of a systematic characterization of data quality indicators, unclear relationships between various data quality indicators, and a lack of data quality assessment methods—this invention provides a systematic health and medical data quality assessment method, system, storage medium, and electronic device. First, this invention comprehensively characterizes data quality through a health and medical data quality indicator system. Second, it uses Decision Laboratory Analysis (DEMATEL) and Adversarial Explanatory Structure Model (AISM) methods to identify key quality indicators within the health and medical data quality indicator system and reveal the relationships between these indicators. Finally, it utilizes Analytic Network Analysis (ANP) and Fuzzy Comprehensive Evaluation (FCE) methods from multi-criteria decision-making approaches to assess the quality of health and medical data. Based on the performance of each quality indicator, it provides decision-making support for health and medical institutions to improve and manage data quality.

[0005] The present invention adopts the following technical solution: A method for assessing the quality of health and medical data includes three main steps: qualitative analysis, structural modeling, and comprehensive evaluation. The specific steps are as follows: Step 1: Qualitative Analysis Establish a quality indicator system for health and medical data; each quality indicator includes several core themes to further explain the meaning of the indicator; Step 2, Structural Modeling: This study utilizes DEMATEL to identify key quality indicators within the health and medical data quality indicator system, calculates the influence, affectedness, centrality, and causation of each quality indicator, and constructs an influence relationship diagram based on centrality and causation. By analyzing the quadrant of the influence relationship diagram, the importance of each quality indicator within the framework is clarified. Furthermore, it combines AISM to construct a multi-level adversarial topology hierarchical model to visualize the influence relationships between indicators. The adversarial topology hierarchical model systematically places each quality indicator at different levels, using directed line segments to connect the nodes to represent the influence relationships. Step 3: Comprehensive Evaluation The ANP is used to calculate the weights of each quality indicator and combined with FCE to assess data quality.

[0006] Preferably, in step one, the quality indicators include: metadata standardization, interoperability, comprehensiveness, numerical integrity, sufficiency, accessibility, compliance, comparability, reliability, durability, timeliness, uniqueness, traceability, relevance, interpretability, and scarcity.

[0007] Preferably, step two, structural modeling, includes the following implementation steps: S2.1: Using triangular fuzzy numbers (l,m,u) to describe the pairwise influence relationships between quality indicators, a structural self-interaction matrix is ​​constructed. ; matrix elements This indicates the k-th evaluator's opinion on indicator A. i For A j The assessment results of the degree of impact; S2.2: Defuzzify all structural self-interaction matrices using the CFCS method to obtain the direct influence matrix S=(s ij ) n ;s ij This represents the evaluation index A by all evaluators 1, 2, ..., k using triangular fuzzy numbers. i For indicator A j The real value obtained after defuzzifying the result of the degree of influence using the CFCS method is used as the real value s. ij Indicator A i For indicator A j The degree of direct impact.

[0008] The CFCS method is a defuzzification method that converts triangular fuzzy numbers into real values, which quantify index A. i For A j The degree of influence is expressed as: (1) Given k structural self-interaction matrices X 1 ,X 2 ,…,X k Indicator A i For A j Results of impact assessment Calculate all evaluation results Regular triangular fuzzy numbers The expression is: (2) Calculate the left-hand regularization value and right-hand regularization value The expression is: (3) Calculate the overall regularization value f k The expression is: (4) For all The regularization value f k Summing yields the deblurred result, expressed as: S2.3: Calculate the comprehensive influence matrix T; the comprehensive influence matrix quantifies index A. i For A j The magnitude of the impact, including direct impact and indirect impact transmitted through other indicators, is expressed as: Where M is the matrix that directly affects the normalized matrix S, and t ij This indicates that in the comprehensive influence matrix T=(t ij ) n Indicator A i For A j The overall degree of impact; S2.4: Based on the comprehensive influence matrix, calculate the influence degree, affected degree, centrality, and causal degree of each quality indicator. The expression is as follows: S2.5: Construct an influence relationship diagram; the influence relationship diagram is a two-dimensional coordinate graph with centrality as the horizontal axis and causality as the vertical axis; quality indicators are represented by coordinate points in the graph, and the coordinate positions are determined according to centrality and causality; based on the average centrality and causality of all quality indicators, the influence relationship diagram is divided into four quadrants, corresponding to Q1 - Contingent indicators, Q2 - Influential indicators, Q3 - Autonomous indicators, and Q4 - Dependent indicators, respectively, to clarify the status and importance of each indicator in the indicator system; Steps S2.1-S2.5 above describe in detail the process of using DEMATEL to identify key quality indicators within the health and medical data quality indicator system.

[0009] S2.6: Calculate the fuzzy reachability matrix FR=(l ij ) n , l ij Indicates quality index A i For A j The cumulative impact intensity includes both direct and indirect effects; the fuzzy reachability matrix refers to the fuzzy relation matrix obtained by performing a fuzzy relation transitive closure operation on the comprehensive impact matrix T. The expression for obtaining the fuzzy reachability matrix is: Among them, the transitive closure operation of fuzzy relations uses the Zade operator. By the intersection operator Union operator Composition, where M is the representation of the Zade operator in the fuzzy operators; T k This refers to the result of k intersection operations between the comprehensive influence matrix T and itself.

[0010] S2.7: Calculate the reachability matrix by setting a threshold. Filter out influence relationships with an influence level below a threshold in the fuzzy reachability matrix and obtain the truncated reachability matrix. : S2.8: Hierarchical partitioning; calculation of the reachability set of each indicator. , advance set and intersection Two types of stratified results were iteratively extracted according to the following two categories of rules; Rule 1: Draw sequentially those that satisfy the rules. Indicator A i ; Rule 2: Draw sequentially those that satisfy the rules. Indicator A i ; S2.9: Calculate the reachability matrix L of the contracted points r The condensed reachability matrix refers to the set of quality indicators with loop relationships identified based on the truncation of the reachability matrix, and the condensation operation is used to merge multiple indicators that influence each other in the loop relationship into a new indicator to simplify the complex dependencies between indicators. S2.10: Directed line segments are used to connect the nodes to represent the influence relationships, specifically as follows: Calculate the skeleton matrix K; the skeleton matrix is ​​used to eliminate non-critical influence relationships between indicators, retaining the critical influence relationships between indicators, and is the basic skeleton for constructing the adversarial topology hierarchical structure model, expressed as: In this step, the skeleton matrix K is a matrix with {0,1} as its elements. If any element in the skeleton matrix K is 1, it indicates that a certain index A... i For indicator A j In adversarial topological hierarchies, there are directed line segment links, denoted by index A. i The starting arrow points to A. j .

[0011] Steps S2.6-S2.10 above describe in detail the specific process of visualizing the influence relationships between indicators in constructing a multi-level adversarial topology hierarchical model using AISM.

[0012] Preferably, in step three, the comprehensive evaluation, the weights of each quality indicator are calculated using ANP (see steps S3.1 and S3.2) and combined with FCE assessment (see steps S3.3 and S3.4) to evaluate data quality; specifically, this includes the following steps: S3.1: Calculate the hypermatrix W; the hypermatrix is ​​the result of column normalization of the comprehensive influence matrix T in step S3; S3.2: Calculate the limiting hypermatrix The aforementioned limiting hypermatrix refers to the result of performing an infinite power operation on the hypermatrix. When all columns of the hypermatrix converge to stable values, the final weights of each quality index are obtained, expressed as: S3.3: Construct the membership matrix R, expressed as: in, It refers to the set of factors consisting of n quality indicators being evaluated; It is a set of comments consisting of m possible evaluation choices for quality indicators; r ij It is a quality indicator u i Evaluation of level v jThe degree of membership.

[0013] S3.4: The comprehensive evaluation result G is expressed as: Among them, g j This indicates the degree of membership of the data quality assessment results to the evaluation decision level; based on the principle of maximum membership and the principle of weighted average, the qualitative evaluation level and quantitative evaluation results of data quality are obtained respectively.

[0014] This invention also discloses a health and medical data quality assessment system for performing the above-described method, comprising the following modules: Qualitative analysis module: used to establish a quality indicator system for health and medical data, with each quality indicator containing several core themes; Structural Modeling Module: Utilizing DEMATEL, this module identifies key quality indicators within the health and medical data quality indicator system, calculates the influence, affectedness, centrality, and causation of each indicator, and constructs an influence relationship diagram based on centrality and causation. By analyzing the quadrant of the influence relationship diagram, the module clarifies the importance of each quality indicator within the framework. It also combines AISM to construct a multi-level adversarial topology hierarchical model to visualize the influence relationships between indicators. This adversarial topology hierarchical model systematically places each quality indicator at different levels, using directed line segments to connect the nodes to represent the influence relationships. Comprehensive evaluation module: Uses ANP to calculate the weight of each quality indicator and combines it with FCE to evaluate data quality.

[0015] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method or system.

[0016] The present invention also discloses 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 above-described method or system.

[0017] Compared with the prior art, the present invention has the following significant technical effects: 1. This invention constructs a health and medical data quality indicator system. This indicator system describes the quality characteristics of health and medical data from multiple dimensions. It further refines general data quality indicators such as completeness, consistency, and timeliness through different core themes, and also incorporates indicators such as compliance and interoperability that are considered due to the usage scenarios and security requirements of health and medical data. It has strong adaptability and scalability and can support the quality assessment of health and medical data under different data sources and application scenarios.

[0018] 2. This invention analyzes the importance of each indicator in the health and medical data quality indicator system and the influence relationship between the indicators; the influence relationship diagram output by DEMATEL clarifies the status and importance of each indicator with four quadrants, and the adversarial topology hierarchical structure model output by AISM reveals the mutual influence relationship between the indicators. The two systematically present the internal structural characteristics of the health and medical data quality indicator system.

[0019] 3. This invention establishes a health and medical data quality assessment model; it uses the ANP and FCE methods to calculate the weights of each indicator and comprehensively evaluate the quality of health and medical data, thereby achieving a comprehensive assessment of health and medical data quality; and it combines the influence relationship diagram, the adversarial topology hierarchical structure model, and the comprehensive evaluation results to provide management decision-making basis for health and medical industry related institutions to improve and manage data quality. Attached Figure Description

[0020] Figure 1 This is a flowchart of a preferred embodiment of the present invention for a method of assessing the quality of health and medical data.

[0021] Figure 2 This is the influence relationship diagram involved in step two of the preferred embodiment of the present invention.

[0022] Figure 3 This is a diagram of the adversarial topological hierarchy model involved in step three of the preferred embodiment of the present invention.

[0023] Figure 4 This is a block diagram of a health and medical data quality assessment system according to a preferred embodiment of the present invention. Detailed Implementation

[0024] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] like Figure 1 This embodiment discloses a method for assessing the quality of health and medical data, which includes three main steps: qualitative analysis, structural modeling, and comprehensive evaluation, as detailed below: Step 1: Establish a health and medical data quality indicator system, with each quality indicator containing several core themes; Step 2: Use DEMATEL to identify key quality indicators within the health and medical data quality indicator system, calculate the influence, affectedness, centrality, and causation of each quality indicator, and construct an influence relationship diagram based on centrality and causation. By analyzing the quadrant of each quality indicator in the influence relationship diagram, clarify the importance of each quality indicator in the framework; combine AISM to construct a multi-level adversarial topology hierarchical structure model to visualize the influence relationships between indicators. The adversarial topology hierarchical structure model orderly places each quality indicator at different levels, and uses directed line segments to connect each node to represent the influence relationship. Step 3: Use ANP to calculate the weights of each quality indicator and combine it with FCE to evaluate data quality.

[0026] The following is a more detailed description of each step.

[0027] The first step in this embodiment: Qualitative analysis This embodiment provides a health and medical data quality assessment index system, including the following 16 quality indicators: metadata standardization, interoperability, comprehensiveness, numerical integrity, sufficiency, accessibility, compliance, comparability, reliability, durability, timeliness, uniqueness, traceability, relevance, interpretability, and scarcity.

[0028] The primary method for constructing the aforementioned health and medical data quality assessment index system is literature analysis. An initial set of indicators for constructing the health and medical data quality index system was extracted from existing literature. Literature searches were conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, using PubMed and Web of Science databases as data sources. Keywords such as "indicator," "dimension," and "framework" were employed to ensure the comprehensiveness of the search results. After initial literature acquisition, duplicate literature was removed, and then literature was screened according to pre-defined inclusion and exclusion criteria. Full-text review was conducted to determine the research literature that met the requirements.

[0029] Document coding primarily employs grounded theory to perform three levels of coding on target documents: primary coding, intermediate coding, and advanced coding. Table 1 uses the formation process of the indicator "metadata normalization" as an example, illustrating the coding objects of the three levels and providing coding examples and results. First, primary coding extracts data quality terms and their definitions from the target document text, generating initial tags in the format "data quality related terms: definition or description." Intermediate coding categorizes and integrates all initial tags according to their similarity, forming several core themes. Finally, advanced coding further summarizes the core themes into an initial set of health and medical data quality indicators. The health and medical data quality indicator system proposed in this invention includes 16 indicators and 29 core themes, as shown in Table 2.

[0030] Table 1. Examples of Document Coding Table 2 Health and Medical Data Quality Indicator System Step 2, Structural Modeling: Use DEMATEL-AISM to analyze the internal structural characteristics of the health and medical data quality indicator system.

[0031] For the 16 health and medical data quality indicators determined in Step 1, the influence relationships between each pair of the 16 indicators are described using multi-level linguistic variables as shown in Table 3, including "absolute influence," "strong influence," "relatively strong influence," "moderate influence," "relatively weak influence," "weak influence," and "no influence," thereby constructing a structured self-interaction matrix. In this embodiment, the influence relationships between the 16 indicators are evaluated using the above-mentioned multi-level linguistic variables, ultimately forming 12 structured self-interaction matrices.

[0032] A direct influence matrix is ​​constructed. Based on the CFCS method, the 12 structural self-interaction matrices are defuzzified, resulting in a 16×16 direct influence matrix S. Each quality index is divided by the maximum sum of the values ​​in each row of S to normalize the direct influence matrix, further yielding the comprehensive influence matrix T, expressed as: Table 3. Multilevel linguistic variables and their triangular fuzzy numbers In this embodiment, the comprehensive influence matrix As shown in the matrix below; the i-th row or j-th column of the matrix corresponds to index A in Table 2. i Or A j .

[0033] Based on the comprehensive influence matrix, calculate the influence degree D of each quality indicator. i Influence level C i Centrality M i , Reason R i The influence refers to a certain indicator A. i For other indicators A j The sum of the direct and indirect effects is expressed as: The degree of influence refers to a certain indicator A i From other indicators A within the system j The combined expression for the direct and indirect effects is: Centrality is the sum of influence and affectedness, used to measure the importance of a particular indicator within the entire system. The higher the centrality, the more central the indicator is in the system.

[0034] Causality is the difference between influence and being influenced, and is used to determine the type of role a certain indicator plays in a system. A positive causality indicates that the indicator mainly affects other indicators as a dependent variable, while a negative causality indicates that the indicator is mainly affected by other indicators as an outcome variable.

[0035] Table 4 presents the influence, affectedness, centrality, and causation of 16 indicators. As shown in Table 3, the top 5 indicators in terms of centrality are comprehensiveness (A3), sufficiency (A5), interpretability (A15), relevance (A4), and traceability (A13), while the top 5 indicators in terms of causation are scarcity (A16), compliance (A7), metadata standardization (A2), timeliness (A11), and reliability (A9). These indicators hold a more important position in the indicator system.

[0036] Table 4. Centrality, Causality, and Ranking of Each Indicator To gain a more intuitive understanding of the importance of each indicator within the system, Figure 2 The influence relationship diagram visualizes this importance. Based on the mean centrality and mean causality of all quality indicators, Figure 2 It is divided into four quadrants: the first quadrant is for contingency indicators, the second quadrant is for influence indicators, the third quadrant is for independent indicators, and the fourth quadrant is for dependent indicators.

[0037] The contingency indicators in the first quadrant include timeliness (A11) and accessibility (A6). Contingency indicators have strong centrality and positive causality, meaning they both influence and are influenced by other indicators. This indicates that these indicators must be optimized in a targeted manner, tailored to the specific application scenarios, during the process of health and medical data quality management.

[0038] The impact indicators in the second quadrant include metadata standardization (A1), interoperability (A2), numerical integrity (A4), compliance (A7), comparability (A8), reliability (A9), and scarcity (A16). Impact indicators have low centrality but positive causality, indicating that these indicators primarily influence other indicators while being less affected themselves. Among them, scarcity (A16) has the highest centrality, indicating its high importance in the health and medical data quality indicator system.

[0039] The independent metrics in the third quadrant include persistence (A10) and uniqueness (A12). Both have centrality and causality below the average, have limited impact on other metrics, and are not significantly affected by them. Although not central to the overall data quality framework, independent metrics are still an important component of data quality and remain a fundamental guarantee of data quality in specific scenarios.

[0040] The dependency indicators in the fourth quadrant include comprehensiveness (A3), sufficiency (A5), traceability (A13), relevance (A14), and interpretability (A15). Dependency indicators have strong centrality but negative causality, which means they are highly dependent on other indicators in the system.

[0041] To further understand the influence relationships among the 16 quality indicators, this embodiment uses the DEMATEL method to visualize these relationships using a topological hierarchical model. The Zadeh fuzzy operator is applied to the comprehensive influence matrix T. Performing fuzzy relation transitive closure operation yields the fuzzy reachability matrix FR, expressed as: Among them, the Zade operator The intersection operator refers to The union operator refers to ;T k This refers to the result of k intersection operations between the comprehensive influence matrix T and itself.

[0042] The fuzzy reachability matrix in this embodiment As shown below; the i-th row or j-th column of the matrix corresponds to index A in Table 2. i Or A j .

[0043] Set threshold The fuzzy relations of the fuzzy reachability matrix are converted into 0-1 binary relations, and the cut-off reachability matrix is ​​constructed. The expression is: This embodiment considers the following three topological features of the topological hierarchy model when selecting the threshold: hierarchy, connected components, and loops. First, a finite set of thresholds is obtained based on the fuzzy reachability matrix FR. This set contains 88 different possible thresholds arranged in ascending order, specifically: Secondly, threshold intervals are constructed using any two adjacent thresholds. Any threshold within a threshold interval can construct the same topological hierarchy model. For example, a topological hierarchy model constructed using any threshold within the threshold interval (0.10592, 0.10937) has 4 levels, 1 connected component, and 2 loops. Finally, the topological features of the topological hierarchy models for all 88 threshold intervals are derived by traversing the data, and the changes in each topological feature are analyzed.

[0044] To ensure that the selected threshold clearly reflects the influence relationship between indicators, this invention proposes the following threshold selection principle: prioritize thresholds with a large number of levels, few connected components, and containing certain loops. Based on this principle, the final threshold λ = 0.14065 is determined to obtain the cut-off reachability matrix L. λ Specifically: Based on the truncated reachability matrix, the reachability set R(A) of each indicator is calculated. i ), look-ahead set Q(A i ) and intersection T(A i Two types of stratified results are extracted iteratively according to the following two types of rules; Rule 1: Draw sequentially those satisfying T(A) i )=R(A i Indicator A) i The indicators extracted step by step by this rule are located at different levels from top to bottom, and the corresponding topological hierarchy is called a UP-type hierarchy, such as... Figure 3 As shown in (a) of the diagram.

[0045] Rule 2: Draw sequentially those satisfying T(A) i )=Q(A i Indicator A) i As shown in Table 4, the indicators extracted step by step by this rule are located at different levels from bottom to top. The corresponding topological hierarchy is called a DOWN-type hierarchy, such as... Figure 3 As shown in (b) of the diagram.

[0046] The two stratification results indicate that the indicator system contains an independent element, uniqueness (A12), and a loop consisting of comprehensiveness (A3) and sufficiency (A5), with indicators within this loop influencing each other. Further point reduction operations can be performed on all indicators within the loop to obtain a point-reduced reachable matrix L. r The point reduction operation refers to merging the loop composed of comprehensiveness (A3) and sufficiency (A5) into a new index and denoting it as "comprehensiveness + sufficiency" (B1).

[0047] Based on the reachability matrix L rCalculate the skeleton matrix K, which represents the core influence relationship framework of the health and medical data quality indicator system. It extracts the most representative and decisive influence paths from the complex influence relationships between indicators, and its expression is: The skeleton matrix in this embodiment is: Based on the skeleton matrix and the hierarchical results of the two types of rules, a structure is constructed as follows: Figure 3 The adversarial topology hierarchy model shown has Rule 1 and Rule 2 forming a UP-type hierarchy model and a DOWN-type hierarchy model, respectively. The 16 indicators are distributed across different levels according to the two types of hierarchical results. Directed edges between levels represent the influence relationships between indicators, while bidirectional edges represent loops.

[0048] Depend on Figure 3 It can be seen that the 16 quality indicators are organized into a 9-layer topological hierarchical structure model. The UP-type and DOWN-type hierarchical structure models are basically consistent in overall structure. This model indicates that there are three active elements in the health and medical data quality indicator system: metadata standardization (A1), interoperability (A2), and comparability (A8). These active elements refer to the elements distributed at different levels in the adversarial topological hierarchical structure model. There is a loop between comprehensiveness (A3) and sufficiency (A5), reflecting their mutual influence. When formulating health data quality improvement strategies, they should be considered as a whole. At the threshold λ = 0.14065, uniqueness (A12), as an independent element in the system, is not connected to other indicators and neither affects nor is affected by other indicators.

[0049] After removing independent elements from the system, the adversarial topology hierarchy diagram is further divided into three layers: the causal layer, the result layer, and the intermediate layer. The causal layer refers to compliance (A7) and scarcity (A16) at layer 9. These indicators serve only as the starting points of directed edges, exerting a fundamental and dominant influence on the system. The result layer refers to traceability (A13) and relevance (A14) at layer 1, primarily influenced by interpretability (A15). The intermediate layer comprises the remaining 11 indicators distributed across layers 2 to 8. These indicators act as intermediaries between the causal and result layers. The interrelationships among the intermediate layer indicators reveal the overall characteristics of the health and medical data quality system: the formation of health and medical data quality depends on the synergy and interaction between multiple indicators, and cannot be dominated by a single indicator.

[0050] Step 3: Comprehensive Evaluation: Use ANP to calculate the weights of each quality indicator, combine it with FCE to evaluate data quality, and use a specific health and medical dataset as a case study.

[0051] Normalize the columns of the comprehensive influence matrix T from step one and calculate the hypermatrix W: By exponentiation of the hypermatrix to the power of infinite powers, the limiting hypermatrix is ​​obtained. Each column of the limiting hypermatrix converges to the same value, and this column vector represents the weight of each indicator. Table 5 shows the weights and ranking results of the health and medical data quality indicators.

[0052] Table 5. Indicator Weights and Ranking Results This embodiment uses the structured datasets ZA1-D, ZA2-D, and ZH-D of type 2 diabetes from a company surveyed in Step 1 as the data quality assessment objects. This company is a healthcare enterprise providing digital healthcare technologies such as internet hospitals, patient follow-up platforms, and remote multidisciplinary collaboration systems to medical institutions. The three datasets were scored sequentially on 16 quality indicators from 0 to 100, and a fuzzy comprehensive evaluation membership matrix was established. A five-level linguistic variable set of "excellent, good, moderate, poor, and bad" was used as the evaluation criteria. Table 6 shows the scoring results and their mean values ​​for each indicator.

[0053] Table 6. Scores of ZA1-D, ZA2-D, and ZH-D metrics for datasets. Based on the indicator weights in Table 5, the comprehensive evaluation results of datasets ZA1-D, ZA2-D, and ZH-D are calculated as follows: According to the maximum membership principle of fuzzy comprehensive evaluation, the maximum membership degrees of the three datasets ZA1-D, ZA2-D, and ZH-D are 0.5557, 0.5024, and 0.4940, respectively, corresponding to evaluation results of "Good", "Excellent", and "Good". As shown in Table 6, ZA2-D scores higher than ZA1-D and ZH-D on 10 indicators, including relevance (A14), reliability (A9), and interpretability (A15), indicating that the data quality of this dataset meets the needs of type 2 diabetes-related research to a certain extent. The advantages of ZA2-D are mainly reflected in its comprehensiveness (A3) and sufficiency (A5), with a data field consisting of 411 records and 32 clinical indicators, including fasting blood glucose and glycated hemoglobin. However, ZA2-D scores lower on persistence (A10) and timeliness (A11), indicating a certain trade-off between data depth and timeliness.

[0054] Compared to ZA2-D, the overall data quality of ZA1-D and ZH-D is lower. ZA1-D scored lowest in 12 out of 16 indicators, only performing well in timeliness (A11), compliance (A7), and durability (A10). Although ZA1-D includes recently collected data, its limited data volume (only 262 records) results in lower comprehensiveness (A3) and sufficiency (A5) compared to ZA2-D and ZH-D. ZH-D's membership between "good" and "excellent" is relatively close, and its performance in comprehensiveness (A3), numerical completeness (A4), compliance (A7), reliability (A9), and scarcity (A16) is similar to ZA2-D. Although ZH-D's overall quality is still lower than ZA2-D, it still has some usability.

[0055] Based on the comprehensive evaluation results above and the analysis results in step two, this embodiment proposes the following three management insights for health and medical institutions to improve the quality of health and medical data.

[0056] (1) Emphasizing the important role of health and medical data quality evaluation. In this embodiment, the fuzzy comprehensive evaluation method in the criterion decision-making method was combined with three datasets of a certain company, ZA1-D, ZA2-D, and ZH-D, to achieve a comprehensive assessment of health and medical data quality. Although the health and medical data quality indicator system qualitatively represents multiple dimensions of data quality with 16 indicators, describing the various aspects of data quality through indicators is only the initial step in data quality evaluation. How to select appropriate methods to qualitatively or quantitatively measure each indicator becomes another challenge. Each indicator should adopt a data quality assessment method that fits the actual situation, and the indicators should be further refined according to specific task requirements to support the monitoring and management of health and medical data quality.

[0057] (2) Establish a data compliance inspection mechanism to ensure that the data processing process complies with relevant laws and regulations and medical industry standards. Health and medical data are highly sensitive, and data security issues such as privacy protection undoubtedly make data compliance an inevitable requirement in the development of health and medical data. Figure 2 Influence relationship diagram and Figure 3 The anti-topological hierarchy models all indicate that compliance (A7) has a significant impact on the indicator system, and there is an urgent need for a data compliance mechanism to help relevant institutions proactively address potential health and medical data security risks.

[0058] (3) Coordinate the needs of different stakeholders. The "data as a product" concept emphasizes close collaboration among stakeholders throughout the data lifecycle to ensure that the data product meets business objectives in terms of quality. As a byproduct of medical research, the development and management of health and medical data involves close collaboration among stakeholders throughout the medical research process. At any stage of the data lifecycle, a lack of effective collaboration may lead to a gap between the actual data quality and expectations, affecting the application value of the data in supporting clinical decision-making.

[0059] like Figure 4 As shown, this embodiment discloses a health and medical data quality assessment system for performing the above-described method, including the following modules: Qualitative analysis module: used to establish a quality indicator system for health and medical data, with each quality indicator containing several core themes; Structural Modeling Module: Utilizing DEMATEL, this module identifies key quality indicators within the health and medical data quality indicator system, calculates the influence, affectedness, centrality, and causation of each indicator, and constructs an influence relationship diagram based on centrality and causation. By analyzing the quadrant of the influence relationship diagram, the module clarifies the importance of each quality indicator within the framework. It also combines AISM to construct a multi-level adversarial topology hierarchical model to visualize the influence relationships between indicators. This adversarial topology hierarchical model systematically places each quality indicator at different levels, using directed line segments to connect the nodes to represent the influence relationships. Comprehensive evaluation module: Uses ANP to calculate the weight of each quality indicator and combines it with FCE to evaluate data quality.

[0060] Other aspects of this embodiment can be found in the above method embodiments.

[0061] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method or system.

[0062] The present invention also discloses 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 above-described method or system.

[0063] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A method for assessing the quality of health and medical data, characterized in that, Includes the following steps: Step 1: Establish a health and medical data quality indicator system, with each quality indicator containing one or more core themes; Step 2: Use DEMATEL to identify key quality indicators within the health and medical data quality indicator system; calculate the influence, affectedness, centrality, and causation of each quality indicator, and construct an influence relationship diagram based on centrality and causation. By analyzing the quadrant of each quality indicator in the influence relationship diagram, clarify the importance of each quality indicator in the framework; combine AISM to construct a multi-level adversarial topology hierarchical structure model to visualize the influence relationships between indicators. The adversarial topology hierarchical structure model orderly places each quality indicator at different levels, and uses directed line segments to connect each node to represent the influence relationship. Step 3: Use ANP to calculate the weights of each quality indicator and combine it with FCE to evaluate data quality.

2. The health and medical data quality assessment method as described in claim 1, characterized in that, In step one, the quality indicators include: metadata standardization, interoperability, comprehensiveness, numerical integrity, sufficiency, accessibility, compliance, comparability, reliability, durability, timeliness, uniqueness, traceability, relevance, interpretability, and scarcity.

3. The health and medical data quality assessment method as described in claim 1, characterized in that, Step two specifically includes the following implementation steps: S2.1: Using triangular fuzzy numbers (l,m,u) to describe the pairwise influence relationships between quality indicators, a structural self-interaction matrix is ​​constructed. Where n represents the number of quality indicators, i, j = 1,2,…,n; matrix elements This indicates that the k-th evaluator's opinion on indicator A is... i For indicator A j The assessment results of the degree of influence; triangular fuzzy number Let k represent the evaluator's attitude towards index A. i For indicator A j The lower bound, peak bound, and upper bound of the assessed impact level; S2.2: Defuzzify all structural self-interaction matrices using the CFCS method to obtain the direct influence matrix. S ij Indicator A i For indicator A j The degree of direct impact; S2.3: Calculate the comprehensive impact matrix T; the comprehensive impact matrix quantifies indicator A. i For indicator A j The magnitude of the impact, including direct impact and indirect impact transmitted through other indicators, is expressed as: Where I is the identity matrix, M is the matrix that directly affects the normalized matrix S, and t ij In the comprehensive influence matrix Indicator A i For indicator A j The overall degree of impact; S2.4: Based on the comprehensive influence matrix, calculate the influence degree, affected degree, centrality, and causal degree of each quality indicator, with the following expressions: S2.5: Construct an influence relationship diagram; the influence relationship diagram refers to a two-dimensional coordinate graph with centrality as the horizontal axis and causality as the vertical axis; quality indicators are represented by coordinate points in the graph, and the coordinate positions are determined according to centrality and causality; based on the average centrality and causality of all quality indicators, the influence relationship diagram is divided into four quadrants, corresponding to Q1-contingency indicators, Q2-influence indicators, Q3-independent indicators, and Q4-dependent indicators, respectively, to clarify the importance of each indicator in the indicator system; S2.6: Calculate the fuzzy reachability matrix The fuzzy reachability matrix refers to the fuzzy relation matrix obtained by performing fuzzy relation transitive closure operation on the comprehensive influence matrix T. ij Indicates quality index A i For A j The expression for obtaining the fuzzy reachability matrix by calculating the cumulative influence intensity is as follows: Among them, the transitive closure operation of fuzzy relations uses the Zade operator. ;T k This refers to the result of k intersection operations between the comprehensive influence matrix T and itself. S2.7: Calculate the reachability matrix by setting a threshold. Filter out influence relationships with an influence level below a threshold in the fuzzy reachability matrix and obtain the truncated reachability matrix. : S2.8: Hierarchical partitioning; calculation of the reachability set of each indicator. , advance set and intersection Two types of stratified results were iteratively extracted according to the following two categories of rules; Rule 1: Draw sequentially those that satisfy the rules. Indicators ; Rule 2: Draw sequentially those that satisfy the rules. Indicators ; S2.9: Calculate the reachable matrix L at the contracted point. r The aforementioned condensed reachability matrix refers to the set of quality indicators with loop relationships identified based on the truncation of the reachability matrix, and the condensation operation is used to merge multiple mutually influential indicators in the loop relationships into a new indicator. S2.10: Calculate the skeleton matrix K; the skeleton matrix is ​​used to eliminate non-critical influence relationships between indicators and retain the critical influence relationships between indicators. It is the basic skeleton for constructing the adversarial topology hierarchical structure model, and its expression is: 。 4. The health and medical data quality assessment method as described in claim 3, characterized in that, In step S2.2, the CFCS method is a defuzzification method that converts triangular fuzzy numbers into real values, where the real values ​​quantify index A. i For indicator A j The degree of influence is expressed as: (1) Given k structural self-interaction matrices Medium Indicator A i For indicator A j Results of impact assessment Calculate all evaluation results Regular triangular fuzzy numbers The expression is: (2) Calculate the left-hand regularization value and right-hand regularization value The expression is: (3) Calculate the overall regularization value The expression is: (4) For all regularization value Summing yields the deblurred result, expressed as: 。 5. The health and medical data quality assessment method as described in claim 4, characterized in that, Step three is as follows: S3.1: Calculate the hypermatrix W; the hypermatrix is ​​the result of column normalization of the comprehensive influence matrix T in step S2.3; 3.2: Calculating the Limiting Hypermatrix The aforementioned limiting hypermatrix refers to the result of exponentiation of the hypermatrix. When all columns of the hypermatrix converge to stable values, the final weights of each quality index are obtained, expressed as: S3.3: Construct the membership matrix R, expressed as: in, It refers to the set of factors consisting of n quality indicators being evaluated; It is a set of comments consisting of m possible evaluation choices for quality indicators; r ij It is a quality indicator u i Evaluation of level v j Membership degree; S3.4: The comprehensive evaluation result G is expressed as: in, It is an indicator weight vector. Indicates quality indicators A1 to A n The weight value, g j This indicates the degree of membership of the data quality assessment results to the evaluation decision level; based on the principle of maximum membership and the principle of weighted average, the qualitative evaluation level and quantitative evaluation results of data quality are obtained respectively.

6. A health and medical data quality assessment system for performing the method as described in any one of claims 1-5, characterized in that, Includes the following modules: Qualitative analysis module: used to establish a quality indicator system for health and medical data, with each quality indicator containing several core themes; Structural Modeling Module: Utilizing DEMATEL, this module identifies key quality indicators within the health and medical data quality indicator system, calculates the influence, affectedness, centrality, and causation of each indicator, and constructs an influence relationship diagram based on centrality and causation. By analyzing the quadrant of the influence relationship diagram, the module clarifies the importance of each quality indicator within the framework. It also combines AISM to construct a multi-level adversarial topology hierarchical model to visualize the influence relationships between indicators. This adversarial topology hierarchical model systematically places each quality indicator at different levels, using directed line segments to connect the nodes to represent the influence relationships. Comprehensive evaluation module: Uses ANP to calculate the weight of each quality indicator and combines it with FCE to evaluate data quality.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-5 or the system of claim 6.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1-5 or the system of claim 6.

Images