Evidence pollution chain traceability system and method based on data analysis

Abstract

The invention discloses an evidence pollution chain traceability system and method based on data analysis, and relates to the technical field of evidence-based medicine, and the system comprises an evidence preprocessing and quality grading module, a problem test recognition module, a pollution chain tracking and positioning module, an evidence correction and reanalysis module, a clinical guide updating module and a long-term management and monitoring module. The method comprises the following steps: performing quality grading on original research, identifying problem test, constructing a pollution chain, removing problem data, performing re-analysis to generate a correction conclusion, updating a clinical guide according to the correction conclusion, and establishing a long-acting monitoring mechanism. According to the method, full-chain processing from pollution identification and source tracing to evidence correction and guide updating is realized, the reliability of the evidence chain of evidence-based medicine is effectively guaranteed, and scientific support is provided for clinical decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of evidence-based medicine, and in particular to a system and method for tracing the source of an evidence contamination chain based on data analysis. Background Art

[0002] In the field of evidence-based medicine, the scientific nature of clinical decision-making is highly dependent on the reliability of evidence from original studies, systematic reviews, meta-analyses, and clinical guidelines. However, there is a core problem in existing technologies: some original studies become "problematic trials" due to incomplete trial design documents, defects in basic data, or contradictory conclusions. However, they may still be included in subsequent systematic reviews and meta-analyses, and then cited by clinical guidelines as the basis for recommendations, forming a "contamination chain" from problematic trials to clinical practice. This problem causes contaminated evidence to continue to influence clinical decisions, bringing potential risks to medical practice. Existing technologies lack a full-chain solution that can accurately identify problematic trials, systematically trace the contamination chain, dynamically correct evidence, and update clinical guidelines, and are unable to effectively prevent contaminated evidence from misleading clinical practice.

[0003] In response to the above problems, the present invention proposes an evidence contamination chain tracing system and method based on data analysis. Summary of the Invention

[0004] The purpose of the present invention is to provide a system and method for tracing the source of evidence pollution chain based on data analysis to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for tracing the source of an evidence pollution chain based on data analysis, comprising the following steps: S1. Pre-process and quality grade the original research, and grade the original research using direct judgment indicators such as experimental design, basic data quality, and consistency of conclusions to provide initial evidence for subsequent processing; S2. Identify problem trials. Compare the non-compliant items in the evidence with known problem characteristics to form a list of problem trials. S3. Track and locate contaminated literature and construct a contamination chain. Based on the identified problematic trials, track the cited literature records, locate contaminated systematic reviews, meta-analyses, and clinical guidelines, and sort out the contamination transmission path. S4. Correct and reanalyze the contaminated evidence, remove the problematic test data, and then re-analyze the data to generate a revised conclusion. S5. Update clinical guidelines based on the revised conclusions, locate related clinical guidelines and specific reference locations, adjust recommendations, and create an update record; S6. Carry out long-term management and monitoring of evidence, integrate all processing results to improve the dynamic management mechanism, and support the continuous operation of the system.

[0006] Said S1 further comprises the following steps: S1.1: Extract information from original research literature, including study design type, sample selection criteria, data collection process, intervention details, outcome measure definitions, statistical analysis methods, and original result presentation format. Convert unstructured information into standardized data entries using pre-defined structured tables. S1.2: Construct a judgment-based assessment system based on the need for evidence contamination risk screening to judge the trial design, basic data quality, and conclusion consistency, and form a list of yes / no status judgment results for each sub-item; among them, the sub-items included in the trial design are the complete documentation of random sequence generation, the execution record of allocation concealment, and the existence of a filing agreement for blind implementation; the sub-items of basic data quality include whether the original data of the main outcome indicators are complete, and whether the code used for statistical analysis and the original data set are directly accessible; the sub-items of conclusion consistency include whether the conclusion direction preset in the research plan is consistent with the actual results, and whether the original data calculation results match the effect size described in the conclusion; S1.3: Based on the results of the quality assessment sub-items, the evidence level is divided into the following levels: Baseline evidence corresponds to all 7 sub-items meeting the standards, available evidence corresponds to 5 sub-items meeting the standards, restricted evidence corresponds to 3 sub-items meeting the standards, and excluded evidence corresponds to 1 sub-item meeting the standards; The baseline evidence and available evidence were then screened as initial evidence.

[0007] Said S2 further comprises the following steps: S2.1: Based on the seven quality assessment sub-items in S1.2, connect to the database that stores information on known problematic tests and simultaneously obtain specific records of these marked problematic tests failing to meet the standards in the seven sub-items. These records include the test title, research team, publication journal, conclusions, and the reason for being identified as a problematic test. At the same time, search for supplementary literature on these problematic tests through professional literature databases to extract specific descriptions of their failures in the S1.2 sub-items. Finally, based on the unique test identifier, associate and integrate this information with the failure criteria for the seven sub-items in S1.2 to form a multi-dimensional dataset consisting of problematic tests, failed sub-items, and specific defect descriptions. This provides standardized material for S2.2 to extract problem features that match the sub-items. S2.2: Based on the aggregated multi-dimensional data set and the seven sub-items of quality assessment, the common characteristics of the problem tests are extracted and categorized. Clear judgment criteria are established for each characteristic, with the judgment criteria being whether it directly corresponds to a sub-item or not, thus forming a structured problem feature library. S2.3: Receive the judgment results of each sub-item of the initial evidence, extract all sub-items in the negative state in the test to be identified, match them with the features in the problem feature library, mark the tests with one or more matching features as suspected problem tests, and compile them into a list of problem tests.

[0008] Said S3 further comprises the following steps: S3.1: Based on the trial identifiers in the problematic trial list, connect to the academic literature database to search and obtain all literature records cited by the problematic trial. Extract the cited literature title, publication journal, publication date, author team, and original text of the cited paragraph. Separately mark the literature type and the position of the included evidence list for the three types of literature: systematic review, meta-analysis, and clinical guidelines. Structured storage of data based on the correspondence between the problematic trial and the cited literature. S3.2: Analyze the content association between the problematic trial and the cited literature one by one. For systematic reviews and meta-analyses, check whether their inclusion criteria include the problematic trial and whether the data are recorded in the data extraction form. For clinical guidelines, check whether the evidence sources clearly cite the problematic trial. At the same time, confirm whether the problematic trial is located in the conclusion section, data consolidation section, and designated section of the evidence list in the cited literature. Mark the cited literature that meets any of the above conditions as contaminated literature and compile a list. S3.3: Based on the list of contaminated literature, organize the dissemination levels and paths from the problematic trials to the contaminated literature, divide them into initial dissemination nodes, secondary dissemination nodes, and terminal dissemination nodes, record the citation relationship, citation purpose, and evidence transmission direction between each node, and present the complete contamination chain in a structured chart; among them, the initial dissemination node represents the systematic review and meta-analysis that directly cites the problematic trials; the secondary dissemination node represents the clinical guidelines that directly cite the initial dissemination node; and the terminal dissemination node represents the clinical guidelines that directly cite the secondary dissemination node.

[0009] The S4 further comprises the following steps: S4.1: Structured extraction of trial data from contaminated systematic reviews and meta-analyses. Then, match the unique trial identifiers in the problem trial list to locate the problematic data items, remove them from the original dataset, and integrate the validated data to form a clean dataset. S4.2: Retrieve the statistical analysis method records of the original study, including the effect size pooling model, heterogeneity test method, confidence interval calculation method, and significance judgment criteria. Recalculate the new effect size, pooled statistics, and confidence interval based on the clean data set according to the original method. S4.3: Extract the original conclusion and revised conclusion of the contaminated evidence, compare the differences in effect direction, numerical magnitude, and significance between the two, and analyze the influence weight of the problematic experimental data in the original analysis by calculating the sample size proportion and the effect size proportion. The calculation formula for the influence weight W is: W = α × W1 + (1-α) × W2; where the influence weight W is determined by the combined contribution of the sample size W1 and the weighted contribution of the effect size W2, and α is the weight coefficient; when the weight W exceeds the preset threshold, the original conclusion is discarded and the revised conclusion is retained; In addition, the sample size contribution W1 reflects the proportion of the problem test sample size in the total analysis, and its calculation formula is: W1=n i / N; where n i is the sample size of the problematic trial, which represents the number of subjects included in the trial, derived from the sample selection criteria and original data extracted in S1.1; N is the total sample size of all included trials in the original analysis, which represents the sum of the sample sizes of all included trials in the contaminated systematic review and meta-analysis; The weighted contribution of effect size W2 reflects the weighted proportion of the problem experiment in the combined effect size. The calculation formula is: W2=w i / Ʃw j ; Among them, w i represents the effect size weight of the problem experiment, and the value range of i is [1, n]. Its calculation formula is wᵢ=1 / Var(Eᵢ), where Var(Eᵢ) is the variance of the effect size Eᵢ of the problem experiment, which comes from the statistical analysis method extracted in S1.1 and the standard deviation of the effect size in the original results; Ʃw j is the sum of the effect size weights of all included trials in the original analysis; Then, based on the research field, characteristics of the target population, and clinical practice scenarios, it is determined whether the recommendations of the associated clinical guidelines need to be adjusted, and a standardized report is formed that includes a summary of the original conclusions, revised implementation steps, details of changes in key indicators, and the final revised conclusions.

[0010] The S5 further comprises the following steps: S5.1: Based on the standardized report of the revised conclusion, match the documents marked as clinical guidelines in the contaminated literature list to determine the target clinical guidelines that need to be updated. Based on the citation location information recorded in the contamination chain, locate the specific chapters, paragraphs, and clauses in the target clinical guidelines that cite the contaminated evidence, and form an association map between clinical guidelines, citation location, and revised conclusion; S5.2: Based on correlation mapping and standardized reporting, assess the original recommendation's reliance on the contaminated evidence and the degree to which the revised conclusion differs from the original evidence to adjust the recommendation, producing an analytical report that includes the impact level and adjustment suggestions. S5.3: Based on the association mapping, identify the content that needs to be adjusted in the target clinical guidelines, modify the recommendations based on the adjustment suggestions in the analysis report, record the comparison of the original recommendations before and after the update, the basis for the adjustment, the operation time and the executor, associate the contamination chain structure information to confirm the contaminated transmission path, and generate a standardized update report.

[0011] The S6 further comprises the following steps: S6.1: Real-time connection to academic literature databases and clinical guideline publishing platforms to obtain newly published research and guideline data, retrieve problem feature libraries and contamination transmission patterns, and perform double verification on new data. The verification method is as follows: first, compare the problem feature library to identify potential problem trials, then check whether known problem trials and contaminated literature are cited to identify potential contaminated literature, and finally, based on the risk level, provide a graded warning and trigger the pre-verification process; S6.2: Integrate clinical guideline update records and all previous processing results, establish evidence dissemination records covering the entire chain from original research to clinical guidelines, synchronize new processing results in real time, and automatically update the evidence dissemination status and status changes of related nodes.

[0012] An evidence contamination chain traceability system based on data analysis, comprising the following contents: the system comprises an evidence preprocessing and quality grading module, a problem trial identification module, a contamination chain tracking and positioning module, an evidence correction and reanalysis module, a clinical guideline update module and a long-term management and monitoring module; the evidence preprocessing and quality grading module extracts information from original research literature and converts it into standardized data entries, makes a yes / no judgment based on seven sub-items of trial design, basic data quality and conclusion consistency, divides the evidence level according to the number of sub-items that meet the standards, and screens benchmark evidence and available evidence as initial evidence; the problem trial identification module, based on the above seven sub-items, connects to the database to obtain non-standard records of known problem trials and integrates them into a multi-dimensional data set, extracts and forms a problem feature library corresponding to the non-standard sub-items, compares the no status sub-items in the initial evidence to identify the problem trials and form a list; the contamination chain tracking and positioning module is based on the problem trials The list retrieves the cited literature records, marks the contaminated systematic reviews, meta-analyses and clinical guidelines, divides the initial, secondary and terminal transmission nodes according to the transmission level, and constructs a contamination chain including citation relationships and transmission directions; the evidence correction and reanalysis module: removes the problematic test data from the contaminated evidence to form a clean data set, reanalyzes it using the original statistical method, and forms a standardized report containing the revised conclusion by calculating the impact weight of the problematic test; the clinical guideline update module matches the clinical guidelines to be updated based on the revised conclusions, locates the specific location of the contaminated evidence cited in the guidelines, evaluates and adjusts the recommendations, records the update details and generates a standardized report; the long-term management and monitoring module is used to monitor newly published studies and guidelines in real time, identifies potential problematic tests and contaminated literature through double verification and issues graded warnings, integrates the results of the entire process to form a record of the entire chain of evidence dissemination and dynamically updates the status of related nodes.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. Accurately identify problematic experiments: Based on seven sub-item evaluation criteria, including experimental design, basic data quality, and conclusion consistency, combined with comparison of problem feature libraries, we can accurately identify problematic experiments from initial evidence and avoid missing the source of contamination.

[0014] 2. Build a complete contamination transmission chain: Through citation tracking, clarify the contamination path of problematic trials to systematic reviews, meta-analyses, and clinical guidelines, clearly locate the level and scope of dissemination of contaminated evidence, and provide precise targets for targeted corrections.

[0015] 3. Form a standardized evidence revision plan: After eliminating problematic data from contaminated evidence, use the original statistical method to reanalyze and quantify the impact weight of the problematic trials, and generate revised conclusions that can be directly used for guideline updates to ensure the reliability of the evidence. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a method for tracing the source of an evidence contamination chain based on data analysis according to the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] Example: Figure 1 As shown, the present invention provides a technical solution. A method for tracing the source of an evidence pollution chain based on data analysis, comprising the following steps: S1. Pre-process and quality grade the original research, and grade the original research using direct judgment indicators such as experimental design, basic data quality, and consistency of conclusions to provide initial evidence for subsequent processing; S2. Identify problem trials. Compare the non-compliant items in the evidence with known problem characteristics to form a list of problem trials. S3. Track and locate contaminated literature and construct a contamination chain. Based on the identified problematic trials, track the cited literature records, locate contaminated systematic reviews, meta-analyses, and clinical guidelines, and sort out the contamination transmission path. S4. Correct and reanalyze the contaminated evidence, remove the problematic test data, and then re-analyze the data to generate a revised conclusion. S5. Update clinical guidelines based on the revised conclusions, locate related clinical guidelines and specific reference locations, adjust recommendations, and create an update record; S6. Carry out long-term management and monitoring of evidence, integrate all processing results to improve the dynamic management mechanism, and support the continuous operation of the system.

[0019] Said S1 further comprises the following steps: S1.1: Extract information from original research literature, including study design type, sample selection criteria, data collection process, intervention details, outcome measure definitions, statistical analysis methods, and original result presentation format. Convert unstructured information into standardized data entries using pre-defined structured tables. S1.2: Construct a judgment-based assessment system based on the need for evidence contamination risk screening to judge the trial design, basic data quality, and conclusion consistency, and form a list of yes / no status judgment results for each sub-item; among them, the sub-items included in the trial design are the complete documentation of random sequence generation, the execution record of allocation concealment, and the existence of a filing agreement for blind implementation; the sub-items of basic data quality include whether the original data of the main outcome indicators are complete, and whether the code used for statistical analysis and the original data set are directly accessible; the sub-items of conclusion consistency include whether the conclusion direction preset in the research plan is consistent with the actual results, and whether the original data calculation results match the effect size described in the conclusion; S1.3: Based on the results of the quality assessment sub-items, the evidence level is divided into the following levels: Baseline evidence corresponds to all 7 sub-items meeting the standards, available evidence corresponds to 5 sub-items meeting the standards, restricted evidence corresponds to 3 sub-items meeting the standards, and excluded evidence corresponds to 1 sub-item meeting the standards; The baseline evidence and available evidence were then screened as initial evidence.

[0020] Said S2 further comprises the following steps: S2.1: Based on the seven quality assessment sub-items in S1.2, connect to the database that stores information on known problematic tests and simultaneously obtain specific records of these marked problematic tests failing to meet the standards in the seven sub-items. These records include the test title, research team, publication journal, conclusions, and the reason for being identified as a problematic test. At the same time, search for supplementary literature on these problematic tests through professional literature databases to extract specific descriptions of their failures in the S1.2 sub-items. Finally, based on the unique test identifier, associate and integrate this information with the failure criteria for the seven sub-items in S1.2 to form a multi-dimensional dataset consisting of problematic tests, failed sub-items, and specific defect descriptions. This provides standardized material for S2.2 to extract problem features that match the sub-items. S2.2: Based on the aggregated multi-dimensional data set and the seven sub-items of quality assessment, the common characteristics of the problem tests are extracted and categorized. Clear judgment criteria are established for each characteristic, with the judgment criteria being whether it directly corresponds to a sub-item or not, thus forming a structured problem feature library. S2.3: Receive the judgment results of each sub-item of the initial evidence, extract all sub-items in the negative state in the test to be identified, match them with the features in the problem feature library, mark the tests with one or more matching features as suspected problem tests, and compile them into a list of problem tests.

[0021] Said S3 further comprises the following steps: S3.1: Based on the trial identifiers in the problematic trial list, connect to the academic literature database to search and obtain all literature records cited by the problematic trial. Extract the cited literature title, publication journal, publication date, author team, and original text of the cited paragraph. Separately mark the literature type and the position of the included evidence list for the three types of literature: systematic review, meta-analysis, and clinical guidelines. Structured storage of data based on the correspondence between the problematic trial and the cited literature. S3.2: Analyze the content association between the problematic trial and the cited literature one by one. For systematic reviews and meta-analyses, check whether their inclusion criteria include the problematic trial and whether the data are recorded in the data extraction form. For clinical guidelines, check whether the evidence sources clearly cite the problematic trial. At the same time, confirm whether the problematic trial is located in the conclusion section, data consolidation section, and designated section of the evidence list in the cited literature. Mark the cited literature that meets any of the above conditions as contaminated literature and compile a list. S3.3: Based on the list of contaminated literature, organize the dissemination levels and paths from the problematic trials to the contaminated literature, divide them into initial dissemination nodes, secondary dissemination nodes, and terminal dissemination nodes, record the citation relationship, citation purpose, and evidence transmission direction between each node, and present the complete contamination chain in a structured chart; among them, the initial dissemination node represents the systematic review and meta-analysis that directly cites the problematic trials; the secondary dissemination node represents the clinical guidelines that directly cite the initial dissemination node; and the terminal dissemination node represents the clinical guidelines that directly cite the secondary dissemination node.

[0022] The S4 further comprises the following steps: S4.1: Structured extraction of trial data from contaminated systematic reviews and meta-analyses. Then, match the unique trial identifiers in the problem trial list to locate the problematic data items, remove them from the original dataset, and integrate the validated data to form a clean dataset. S4.2: Retrieve the statistical analysis method records of the original study, including the effect size pooling model, heterogeneity test method, confidence interval calculation method, and significance judgment criteria. Recalculate the new effect size, pooled statistics, and confidence interval based on the clean data set according to the original method. S4.3: Extract the original conclusion and revised conclusion of the contaminated evidence, compare the differences in effect direction, numerical magnitude, and significance between the two, and analyze the influence weight of the problematic experimental data in the original analysis by calculating the sample size proportion and the effect size proportion. The calculation formula for the influence weight W is: W = α × W1 + (1-α) × W2; where the influence weight W is determined by the combined contribution of the sample size W1 and the weighted contribution of the effect size W2, and α is the weight coefficient; when the weight W exceeds the preset threshold, the original conclusion is discarded and the revised conclusion is retained; In addition, the sample size contribution W1 reflects the proportion of the problem test sample size in the total analysis, and its calculation formula is: W1=n i / N; where n i is the sample size of the problematic trial, which represents the number of subjects included in the trial, derived from the sample selection criteria and original data extracted in S1.1; N is the total sample size of all included trials in the original analysis, which represents the sum of the sample sizes of all included trials in the contaminated systematic review and meta-analysis; The weighted contribution of effect size W2 reflects the weighted proportion of the problem experiment in the combined effect size. The calculation formula is: W2=w i / Ʃw j ; Among them, w i represents the effect size weight of the problem experiment, and the value range of i is [1, n]. Its calculation formula is wᵢ=1 / Var(Eᵢ), where Var(Eᵢ) is the variance of the effect size Eᵢ of the problem experiment, which comes from the statistical analysis method extracted in S1.1 and the standard deviation of the effect size in the original results; Ʃw j is the sum of the effect size weights of all included trials in the original analysis; Then, based on the research field, characteristics of the target population, and clinical practice scenarios, it is determined whether the recommendations of the associated clinical guidelines need to be adjusted, and a standardized report is formed that includes a summary of the original conclusions, revised implementation steps, details of changes in key indicators, and the final revised conclusions.

[0023] The S5 further comprises the following steps: S5.1: Based on the standardized report of the revised conclusion, match the documents marked as clinical guidelines in the contaminated literature list to determine the target clinical guidelines that need to be updated. Based on the citation location information recorded in the contamination chain, locate the specific chapters, paragraphs, and clauses in the target clinical guidelines that cite the contaminated evidence, and form an association map between clinical guidelines, citation location, and revised conclusion; S5.2: Based on correlation mapping and standardized reporting, assess the original recommendation's reliance on the contaminated evidence and the degree to which the revised conclusion differs from the original evidence to adjust the recommendation, producing an analytical report that includes the impact level and adjustment suggestions. S5.3: Based on the association mapping, identify the content that needs to be adjusted in the target clinical guidelines, modify the recommendations based on the adjustment suggestions in the analysis report, record the comparison of the original recommendations before and after the update, the basis for the adjustment, the operation time and the executor, associate the contamination chain structure information to confirm the contaminated transmission path, and generate a standardized update report.

[0024] The S6 further comprises the following steps: S6.1: Real-time connection to academic literature databases and clinical guideline publishing platforms to obtain newly published research and guideline data, retrieve problem feature libraries and contamination transmission patterns, and perform double verification on new data. The verification method is as follows: first, compare the problem feature library to identify potential problem trials, then check whether known problem trials and contaminated literature are cited to identify potential contaminated literature, and finally, based on the risk level, provide a graded warning and trigger the pre-verification process; S6.2: Integrate clinical guideline update records and all previous processing results, establish evidence dissemination records covering the entire chain from original research to clinical guidelines, synchronize new processing results in real time, and automatically update the evidence dissemination status and status changes of related nodes.

[0025] An evidence contamination chain traceability system based on data analysis, comprising the following contents: the system comprises an evidence preprocessing and quality grading module, a problem trial identification module, a contamination chain tracking and positioning module, an evidence correction and reanalysis module, a clinical guideline update module and a long-term management and monitoring module; the evidence preprocessing and quality grading module extracts information from original research literature and converts it into standardized data entries, makes a yes / no judgment based on seven sub-items of trial design, basic data quality and conclusion consistency, divides the evidence level according to the number of sub-items that meet the standards, and screens benchmark evidence and available evidence as initial evidence; the problem trial identification module, based on the above seven sub-items, connects to the database to obtain non-standard records of known problem trials and integrates them into a multi-dimensional data set, extracts and forms a problem feature library corresponding to the non-standard sub-items, compares the no status sub-items in the initial evidence to identify the problem trials and form a list; the contamination chain tracking and positioning module is based on the problem trials The list retrieves the cited literature records, marks the contaminated systematic reviews, meta-analyses and clinical guidelines, divides the initial, secondary and terminal transmission nodes according to the transmission level, and constructs a contamination chain including citation relationships and transmission directions; the evidence correction and reanalysis module: removes the problematic test data from the contaminated evidence to form a clean data set, reanalyzes it using the original statistical method, and forms a standardized report containing the revised conclusion by calculating the impact weight of the problematic test; the clinical guideline update module matches the clinical guidelines to be updated based on the revised conclusions, locates the specific location of the contaminated evidence cited in the guidelines, evaluates and adjusts the recommendations, records the update details and generates a standardized report; the long-term management and monitoring module is used to monitor newly published studies and guidelines in real time, identifies potential problematic tests and contaminated literature through double verification and issues graded warnings, integrates the results of the entire process to form a record of the entire chain of evidence dissemination and dynamically updates the status of related nodes.

[0026] This example uses the "Efficacy Study of a New Antihypertensive Drug A" in the field of evidence-based medicine as a scenario, simulates the entire process from original research quality assessment to clinical guideline update, and specifically illustrates the practical application of the evidence contamination chain tracing method based on data analysis.

[0027] An original research article entitled "A randomized controlled trial of antihypertensive drug A versus conventional drugs for the treatment of hypertension" was selected, and information was extracted according to a pre-set structured table: the study design type was a randomized controlled trial (RCT); the sample selection criteria were patients with essential hypertension (systolic blood pressure 140-160 mmHg), excluding those with severe heart and kidney insufficiency; the data collection process was baseline data: blood pressure before medication and blood pressure measurements 4 and 8 weeks after intervention, recorded by uniformly trained nurses; the intervention measures were oral antihypertensive drug A (5 mg / day) in the experimental group and oral 4 mg / day in the control group. Conventional drug B (10 mg / day) was used for an 8-week course of treatment. The outcome measures were the primary outcome, the decrease in systolic blood pressure after 8 weeks, and the secondary outcome, the incidence of adverse reactions. The statistical analysis method was the independent sample t-test to compare differences between the groups, and the effect size was expressed as "mean difference (MD) ± standard deviation (SD)." The original result was a decrease in systolic blood pressure of (15.2±3.5) mmHg in the experimental group and (10.1±4.2) mmHg in the control group. The conclusion was described as "the efficacy of antihypertensive drug A was significantly better than that of drug B (MD=5.1 mmHg, P<0.05)."

[0028] The above information was then converted into standardized data entries, such as "Original data of the main outcome: experimental group n=100, control group n=100, complete record of 8-week systolic blood pressure reduction value" and "Statistical code: not provided in the appendix".

[0029] Based on the need for evidence contamination risk screening, a "yes / no" judgment was made for seven sub-items: Regarding the completeness of the trial design documents, the random sequence generation document lacked a clear algorithm and generation process record (no); the allocation concealment execution record did not mention confidentiality measures and unblinded registration (no); and the blinding implementation record agreement did not specify the blinding operation process for patients and assessors (no); Regarding the integrity of the basic data, the primary outcome raw data were not missing (yes); the code and original data set used for statistical analysis were not directly accessible (no); Regarding the consistency of the conclusions, the direction of the conclusions pre-specified in the study protocol was consistent with the actual results (yes), and the calculated results of the raw data were consistent with the effect size described in the conclusions (yes). The judgment results were 3 "yes" and 4 "no".

[0030] According to the rule of "yes = 1 point, no = 0 point", the total score is 3 points, corresponding to "limited evidence". Because this study was subsequently included in a systematic review, it is temporarily classified as initial evidence.

[0031] Based on the above 7 sub-items, we connected to the "Known Problem Trial Database" to obtain a related record: "A Short-term Efficacy Study of Antihypertensive Drug A" published in 2020 was marked as a problem trial due to "missing random sequence generation documents" and "missing allocation concealment records"; an additional search of its supplementary literature revealed that its "statistical code is unavailable", which is highly consistent with the sub-item that did not meet the standards of this study.

[0032] The data is integrated to form a multi-dimensional data set, which includes problem test identification, sub-items that do not meet the standards, and specific defect descriptions, such as "Problem test identification T001, sub-items that do not meet the standards are random sequence generation documents, allocation hidden records, and statistical codes. The specific defect description is that the random sequence generation algorithm is not recorded, allocation hiding is not implemented, and the statistical analysis code is not disclosed."

[0033] Based on the above dataset, we extracted the problem features corresponding to seven sub-items, including missing random sequence generation documents, corresponding to "Does the complete document generated by the random sequence exist = no?", missing allocation concealment records, corresponding to "Does the execution record of allocation concealment exist = no?", and unavailable statistical code, corresponding to "Can the code used for statistical analysis and the original dataset be directly obtained = no?"

[0034] The “No” status sub-items in this study were extracted: the random sequence generation document was missing, the allocation concealment record was missing, and the statistical code was unavailable. They were completely matched with the problem feature library (3 items matched), so they were marked as “problem trials” and included in the problem trial list with the number T002.

[0035] Based on the title and DOI of the problematic trial T002, we searched databases such as PubMed and Cochrane Library, obtaining a total of eight citations. These included two systematic reviews and meta-analyses (e.g., "A Systematic Review and Meta-Analysis of Antihypertensive Drug A for the Treatment of Hypertension," published in 2022) and one clinical guideline ("Guidelines for the Prevention and Treatment of Hypertension in Adults (2023 Edition)"). Citation relationships were stored in a structured manner, for example, "Systematic review S001 cites T002, and the data extraction table records its 'systolic blood pressure decrease MD = 5.1 mmHg'" and "Guideline G001 cites S001 as part of the evidence for 'recommending antihypertensive drug A for mild to moderate hypertension'."

[0036] Citation analysis: The inclusion criteria for the systematic review S001 explicitly included T002, and the data extraction form recorded its outcome data (meeting the "contaminated literature" criteria). The guideline G001 explicitly cited S001 in its evidence sources section, and the "Recommendations" section mentioned "Based on the results of the meta-analysis, antihypertensive drug A is more effective" (meeting the "contaminated literature" criteria). This resulted in a list of contaminated literature: S001 (systematic review) and G001 (clinical guideline).

[0037] Nodes are divided according to the level of dissemination: the initial dissemination node is S001 (which directly cites T002 and incorporates its data); the secondary dissemination node is G001 (which directly cites S001 and serves as the basis for the recommendation); the dissemination path is T002 → S001 → G001, and the direction of evidence transfer is "problem trial → systematic review → clinical guideline." The document type and citation location of each node are marked (e.g., the section citing S001 in G001 is "3.2 Antihypertensive Drug Selection").

[0038] Data from the five included trials (including T002) were extracted from the systematic review S001: the sample size of the experimental group of trial number T002 (problem trial) was 100, the sample size of the control group was 100, the systolic blood pressure reduction MD (mmHg) was 5.1±3.8, and the effect size variance Var(E) was 0.038 (SD=0.195); the sample size of the experimental group of trial number T003 was 150, the sample size of the control group was 150, the systolic blood pressure reduction MD (mmHg) was 4.2±4.1, and the effect size variance Var(E) was 0.041; The sample size for the experimental group was 80, the sample size for the control group was 80, the MD reduction in systolic blood pressure (mmHg) was 3.8±3.6, and the effect size variance (Var(E)) was 0.036. The sample size for the experimental group was 120, the sample size for the control group was 120, the MD reduction in systolic blood pressure (mmHg) was 4.5±3.9, and the effect size variance (Var(E)) was 0.039. The sample size for the experimental group was 90, the sample size for the control group was 90, the MD reduction in systolic blood pressure (mmHg) was 3.9±4.0, and the effect size variance (Var(E)) was 0.040. The data from T002 were matched with its unique identifier and removed, forming a clean dataset for the remaining four trials.

[0039] The original statistical methods for S001 were retrieved: a random-effects model was used to pool effect sizes, and heterogeneity was tested using the I² statistic with a 95% confidence interval (CI). Recalculation based on the clean data set showed a pooled MD of 4.1 mmHg (95% CI: 3.2-5.0), with an I² of 25% (low heterogeneity). The pooled MD for the original analysis (including T002) was 4.6 mmHg (95% CI: 3.8-5.4), and the effect size decreased by 0.5 mmHg after excluding T002.

[0040] The original conclusions and revised conclusions of the contaminated evidence were extracted, the differences were compared, and the influence weights were analyzed: the sample size contribution W1 was calculated as the total sample size of the problem experiment T002 nᵢ = 200 (100 + 100), the total sample size of the original analysis N = (100 + 100) + (150 + 150) + (80 + 80) + (120 + 120) + (90 + 90) = 1340, W1 = 200 / 1340 ≈ 0.149 (14.9%); The weighted contribution of effect size W2 is calculated as the effect size weight of the problem experiment wᵢ=1 / Var(Eᵢ)=1 / 0.038≈26.3, the total weight of the original analysis Σwⱼ=26.3+(1 / 0.041)+(1 / 0.036)+(1 / 0.039)+(1 / 0.040)≈129.1, and W2=26.3 / 129.1≈0.204 (20.4%). The comprehensive influence weight W takes α=0.5 (default value), W=0.5×0.149+0.5×0.204≈0.176 (17.6%), and the preset threshold is 15%. Because W>15%, it is judged that the problem experiment has a significant impact on the original conclusion.

[0041] The original conclusion, "Antihypertensive Drug A significantly reduced systolic blood pressure compared to conventional drugs (MD = 4.6 mmHg, P < 0.01)," was then revised to "Antihypertensive Drug A reduced systolic blood pressure compared to conventional drugs (MD = 4.1 mmHg, P < 0.05)," with a reduced effect size and weakened significance. A standardized report was then created, clarifying that "after eliminating the problematic trial, the superiority of Antihypertensive Drug A decreased."

[0042] Based on the revised conclusion report, match G001 (Guidelines for the Prevention and Treatment of Hypertension in Adults (2023 Edition)) in the contaminated literature list and locate the paragraph in its "3.2 Selection of Antihypertensive Drugs" chapter that quotes S001: "Antihypertensive drug A is recommended as the first-line medication for mild to moderate hypertension (evidence level A)".

[0043] Based on association mapping and standardized reporting, the original recommendation's reliance on contaminated evidence was assessed as "high" (70% of the evidence came from S001). The effect size difference between the revised conclusion and the original evidence was 0.5 mmHg (10.9%). It was determined that the recommendation level needed to be lowered, and an analysis report including the impact level and adjustment suggestions was formed.

[0044] Based on the association mapping, the content that needs to be adjusted in the target clinical guidelines is locked, and the corresponding paragraph in G001 is modified to: "Antihypertensive drug A can be used as a second-line drug for mild to moderate hypertension (evidence level B)", and the update details are recorded: before the update, it was "Antihypertensive drug A is recommended as a first-line drug (evidence level A)", and after the update, it is "Can be used as a second-line drug (evidence level B)". The basis for the adjustment is that the effect size is reduced and the strength of evidence is weakened after the problem trial in S001 is eliminated. The executor is the guideline revision group, and the time is March 15, 2024; generate a standardized update report and synchronize it to the long-term management module.

[0045] Real-time monitoring of the newly published "Study on the Efficacy of Antihypertensive Drug A in Diabetes and Hypertension" in April 2024, compared with the problem feature library, found that its "random sequence generation document is missing" and "statistical code is unavailable", matching two problem features, and was determined to be a "high-risk potential problem trial", triggering the pre-verification process.

[0046] Integrate the entire process data, including the problem characteristics of T002, the correction records of S001, the update records of G001, etc., establish a transmission chain record: "T002→S001 (corrected)→G001 (updated)", and mark the associated status of the new warning test, such as "pending verification, not included in any systematic evaluation."

[0047] Through this method, the problematic trial T002 was successfully identified, and its contamination path to the systematic review S001 and clinical guideline G001 was traced. The exaggerated efficacy conclusions were corrected, and the guideline recommendation level was lowered. Ultimately, new contamination risks were prevented through long-term monitoring, effectively ensuring the reliability of the evidence chain of evidence-based medicine.

[0048] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for tracing the source of evidence pollution chains based on data analysis, characterized by: The following steps are involved: S1. Pre-process and quality grade the original research, and grade the original research using direct judgment indicators such as experimental design, basic data quality, and consistency of conclusions to provide initial evidence for subsequent processing; S2. Identify problem trials. Compare the non-compliant items in the evidence with known problem characteristics to form a list of problem trials. S3. Track and locate contaminated literature and construct a contamination chain. Based on the identified problematic trials, track the cited literature records, locate contaminated systematic reviews, meta-analyses, and clinical guidelines, and sort out the contamination transmission path. S4. Correct and reanalyze the contaminated evidence, remove the problematic test data, and then re-analyze the data to generate a revised conclusion. S5. Update clinical guidelines based on the revised conclusions, locate related clinical guidelines and specific reference locations, adjust recommendations, and create an update record; S6. Carry out long-term management and monitoring of evidence, integrate all processing results to improve the dynamic management mechanism, and support the continuous operation of the system.

2. The method for tracing the evidence contamination chain based on data analysis according to claim 1 is characterized by: Said S1 further comprises the following steps: S1.1: Extract information from original research literature, including study design type, sample selection criteria, data collection process, intervention details, outcome measure definitions, statistical analysis methods, and original result presentation format. Convert unstructured information into standardized data entries using pre-defined structured tables. S1.2: Construct a judgment-based assessment system based on the need for evidence contamination risk screening to judge the trial design, basic data quality, and conclusion consistency, and form a list of yes / no status judgment results for each sub-item; among them, the sub-items included in the trial design are the complete documentation of random sequence generation, the execution record of allocation concealment, and the existence of a filing agreement for blind implementation; the sub-items of basic data quality include whether the original data of the main outcome indicators are complete, and whether the code used for statistical analysis and the original data set are directly accessible; the sub-items of conclusion consistency include whether the conclusion direction preset in the research plan is consistent with the actual results, and whether the original data calculation results match the effect size described in the conclusion; S1.3: Based on the results of the quality assessment sub-items, the evidence level is divided into the following levels: Baseline evidence corresponds to all 7 sub-items meeting the standards, available evidence corresponds to 5 sub-items meeting the standards, restricted evidence corresponds to 3 sub-items meeting the standards, and excluded evidence corresponds to 1 sub-item meeting the standards; The baseline evidence and available evidence were then screened as initial evidence.

3. The method for tracing the evidence contamination chain based on data analysis according to claim 2 is characterized by: Said S2 further comprises the following steps: S2.1: Based on the seven quality assessment sub-items in S1.2, connect to the database that stores information on known problematic tests and simultaneously obtain specific records of these marked problematic tests failing to meet the standards in the seven sub-items. These records include the test title, research team, publication journal, conclusions, and the reason for being identified as a problematic test. At the same time, search for supplementary literature on these problematic tests through professional literature databases to extract specific descriptions of their failures in the S1.2 sub-items. Finally, based on the unique test identifier, associate and integrate this information with the failure criteria for the seven sub-items in S1.2 to form a multi-dimensional dataset consisting of problematic tests, failed sub-items, and specific defect descriptions. This provides standardized material for S2.2 to extract problem features that match the sub-items. S2.2: Based on the aggregated multi-dimensional data set and the seven sub-items of quality assessment, the common characteristics of the problem tests are extracted and categorized. Clear judgment criteria are established for each characteristic, with the judgment criteria being whether it directly corresponds to a sub-item or not, thus forming a structured problem feature library. S2.3: Receive the judgment results of each sub-item of the initial evidence, extract all sub-items in the negative state in the test to be identified, match them with the features in the problem feature library, mark the tests with one or more matching features as suspected problem tests, and compile them into a list of problem tests.

4. The method for tracing the evidence contamination chain based on data analysis according to claim 1 is characterized by: Said S3 further comprises the following steps: S3.1: Based on the trial identifiers in the problematic trial list, connect to the academic literature database to search and obtain all literature records cited by the problematic trial. Extract the cited literature title, publication journal, publication date, author team, and original text of the cited paragraph. Separately mark the literature type and the position of the included evidence list for the three types of literature: systematic review, meta-analysis, and clinical guidelines. Structured storage of data based on the correspondence between the problematic trial and the cited literature. S3.2: Analyze the content association between the problematic trial and the cited literature one by one. For systematic reviews and meta-analyses, check whether their inclusion criteria include the problematic trial and whether the data are recorded in the data extraction form. For clinical guidelines, check whether the evidence sources clearly cite the problematic trial. At the same time, confirm whether the problematic trial is located in the conclusion section, data consolidation section, and designated section of the evidence list in the cited literature. Mark the cited literature that meets any of the above conditions as contaminated literature and compile a list. S3.3: Based on the list of contaminated literature, organize the dissemination levels and paths from the problematic trials to the contaminated literature, divide them into initial dissemination nodes, secondary dissemination nodes, and terminal dissemination nodes, record the citation relationship, citation purpose, and evidence transmission direction between each node, and present the complete contamination chain in a structured chart; among them, the initial dissemination node represents the systematic review and meta-analysis that directly cites the problematic trials; the secondary dissemination node represents the clinical guidelines that directly cite the initial dissemination node; and the terminal dissemination node represents the clinical guidelines that directly cite the secondary dissemination node.

5. The method for tracing the evidence contamination chain based on data analysis according to claim 1 is characterized by: The S4 further comprises the following steps: S4.1: Structured extraction of trial data from contaminated systematic reviews and meta-analyses. Then, match the unique trial identifiers in the problem trial list to locate the problematic data items, remove them from the original dataset, and integrate the validated data to form a clean dataset. S4.2: Retrieve the statistical analysis method records of the original study, including the effect size pooling model, heterogeneity test method, confidence interval calculation method, and significance judgment criteria. Recalculate the new effect size, pooled statistics, and confidence interval based on the clean data set according to the original method. S4.3: Extract the original conclusion and revised conclusion of the contaminated evidence, compare the differences in effect direction, numerical magnitude, and significance between the two, and analyze the influence weight of the problematic experimental data in the original analysis by calculating the sample size proportion and the effect size proportion. The calculation formula for the influence weight W is: W = α × W1 + (1-α) × W2; where the influence weight W is determined by the combined contribution of the sample size W1 and the weighted contribution of the effect size W2, and α is the weight coefficient; when the weight W exceeds the preset threshold, the original conclusion is discarded and the revised conclusion is retained; In addition, the sample size contribution W1 reflects the proportion of the problem test sample size in the total analysis, and its calculation formula is: W1=n i / N; where n i is the sample size of the problematic trial, which represents the number of subjects included in the trial, derived from the sample selection criteria and original data extracted in S1.1; N is the total sample size of all included trials in the original analysis, which represents the sum of the sample sizes of all included trials in the contaminated systematic review and meta-analysis; The weighted contribution of effect size W2 reflects the weighted proportion of the problem experiment in the combined effect size. The calculation formula is: W2=w i / Ʃw j ; Among them, w i represents the effect size weight of the problem experiment, and the value range of i is [1, n]. Its calculation formula is wᵢ=1 / Var(Eᵢ), where Var(Eᵢ) is the variance of the effect size Eᵢ of the problem experiment, which comes from the statistical analysis method extracted in S1.1 and the standard deviation of the effect size in the original results; Ʃw j is the sum of the effect size weights of all included trials in the original analysis; Then, based on the research field, characteristics of the target population, and clinical practice scenarios, it is determined whether the recommendations of the associated clinical guidelines need to be adjusted, and a standardized report is formed that includes a summary of the original conclusions, revised implementation steps, details of changes in key indicators, and the final revised conclusions.

6. The method for tracing the evidence contamination chain based on data analysis according to claim 1 is characterized by: The S5 further comprises the following steps: S5.1: Based on the standardized report of the revised conclusion, match the documents marked as clinical guidelines in the contaminated literature list to determine the target clinical guidelines that need to be updated. Based on the citation location information recorded in the contamination chain, locate the specific chapters, paragraphs, and clauses in the target clinical guidelines that cite the contaminated evidence, and form an association map between clinical guidelines, citation location, and revised conclusion; S5.2: Based on correlation mapping and standardized reporting, assess the original recommendation's reliance on the contaminated evidence and the degree to which the revised conclusion differs from the original evidence to adjust the recommendation, producing an analytical report that includes the impact level and adjustment suggestions. S5.3: Based on the association mapping, identify the content that needs to be adjusted in the target clinical guidelines, modify the recommendations based on the adjustment suggestions in the analysis report, record the comparison of the original recommendations before and after the update, the basis for the adjustment, the operation time and the executor, associate the contamination chain structure information to confirm the contaminated transmission path, and generate a standardized update report.

7. The method for tracing the evidence contamination chain based on data analysis according to claim 1 is characterized by: The S6 further comprises the following steps: S6.1: Real-time connection to academic literature databases and clinical guideline publishing platforms to obtain newly published research and guideline data, retrieve problem feature libraries and contamination transmission patterns, and perform double verification on new data. The verification method is as follows: first, compare the problem feature library to identify potential problem trials, then check whether known problem trials and contaminated literature are cited to identify potential contaminated literature, and finally, based on the risk level, provide a graded warning and trigger the pre-verification process; S6.2: Integrate clinical guideline update records and all previous processing results, establish evidence dissemination records covering the entire chain from original research to clinical guidelines, synchronize new processing results in real time, and automatically update the evidence dissemination status and status changes of related nodes.

8. A system for tracing the source of an evidence contamination chain based on data analysis, applied to a method for tracing the source of an evidence contamination chain based on data analysis according to any one of claims 1 to 7, characterized in that: Includes the following: The system includes an evidence preprocessing and quality grading module, a problem trial identification module, a contamination chain tracking and location module, an evidence revision and reanalysis module, a clinical guideline update module, and a long-term management and monitoring module. The evidence preprocessing and quality grading module extracts information from original research literature and converts it into standardized data entries. It makes yes / no judgments based on seven sub-items: trial design, basic data quality, and conclusion consistency. It then divides the evidence level by the number of sub-items that meet the standards and selects baseline evidence and available evidence as initial evidence. The problem trial identification module is based on the above 7 sub-items, connects to the database to obtain the non-compliant records of known problem trials and integrates them into a multi-dimensional data set, extracts and forms a problem feature library corresponding to the non-compliant sub-items, compares the negative status sub-items in the initial evidence to identify the problem trials and form a list; the contamination chain tracking and positioning module retrieves the cited literature records based on the problem trial list, marks the contaminated systematic reviews, meta-analyses and clinical guidelines, divides the initial, secondary and terminal transmission nodes according to the transmission level, and constructs a contamination chain including citation relationships and transmission directions; the evidence correction and reanalysis module: removes the contaminated evidence from the contaminated evidence The problem trial data forms a clean data set, which is re-analyzed using the original statistical method. By calculating the impact weight of the problem trial, a standardized report containing the revised conclusion is formed; the clinical guideline update module matches the clinical guidelines to be updated based on the revised conclusion, locates the specific location of the contaminated evidence cited in the guideline, evaluates and adjusts the recommendations, records the update details and generates a standardized report; the long-term management and monitoring module is used to monitor newly published research and guidelines in real time, identify potential problem trials and contaminated literature through double verification and issue graded warnings, integrate the results of the entire process to form a full-chain record of evidence dissemination and dynamically update the status of related nodes.

Images