Hospital scientific research paper influence comprehensive evaluation method

Abstract

This application provides a comprehensive evaluation method for the impact of hospital research papers, including: identifying whether a citation belongs to the core of the recommended text or the peripheral of the background discussion based on the evidence level label and the chapter level of each citation, and determining its recommendation strength level to obtain the core and peripheral attributes and recommendation strength level of each citation; assessing the actual support of each citation for clinical decision-making by combining the core and peripheral attributes and recommendation strength level of each citation, classifying citations into high decision support and low decision support categories to obtain citation decision support classification labels; summarizing the adjusted individual citation conversion scores by paper dimension, and merging the scores of other types of citations that are not cited in the guidelines to obtain the paper's comprehensive clinical conversion score.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a comprehensive evaluation method for the impact of hospital research papers. Background Technology

[0002] In the comprehensive evaluation of the impact of hospital research papers, whether clinical research findings can truly guide doctors' daily diagnostic and treatment decisions is one of the core indicators for measuring a hospital's research level and ability to improve medical quality. Papers cited in clinical practice guidelines are generally considered important evidence with high clinical translational value, as guidelines represent the current medical community's consensus and recommendations on best practices. This citation mechanism makes guidelines a bridge connecting basic research and clinical application, and is crucial for assessing the structure of research output in hospital departments. However, existing evaluation methods have significant shortcomings in handling guideline citations. For example, a method for analyzing and ranking the academic influence of subject literature in a citation database (publication number CN103729432B) discloses a technique for calculating and ranking academic influence values ​​based on citation networks. This method neglects the identification of the specific position, level of evidence, and strength of recommendation of papers cited in clinical practice guidelines. These methods often only count whether a paper is included in the guideline, ignoring the significant differences in the actual role the same paper plays within the guideline. Some papers, although appearing in the guideline's reference list, are only used as background information, supporting explanations, or secondary evidence, rather than key evidence supporting core recommendations. This phenomenon leads evaluation systems to easily confuse marginal citations with core citations, resulting in a systematic overestimation of the clinical decision-making influence of numerous papers and thus distorting the true structure of hospital clinical research output. The core technical challenge lies in the fact that guidelines clearly indicate the level of evidence and strength of recommendation for cited literature, but this annotation information is not effectively analyzed and utilized in existing evaluation systems. Guidelines typically categorize recommendations into different levels such as strong, moderate, and weak recommendations based on the quality of evidence, while also indicating the specific literature supporting the recommendation and its level of evidence. For example, a randomized controlled trial might be listed as low-quality evidence in a guideline, used only to support a weak recommendation of "considerable"; while another paper on a similar topic might be considered high-quality evidence, directly supporting a core treatment pathway of "strong recommendation." It is precisely because of the lack of fine-grained identification of the relationship between citation location, level of evidence annotation, and strength of recommendation that the system cannot distinguish the true differences in the weight of these citations in the clinical decision-making chain, thus causing a systematic bias of overestimating marginal citations and underestimating core citations. Therefore, accurately capturing the specific location and level of a paper's citation in clinical practice guidelines, the level of evidence it is cited in, and the strength of the supporting recommendations, and dynamically judging its actual supporting role in clinical decision-making based on these factors, has become a key issue in scientifically evaluating the translational value of hospital clinical research papers. Summary of the Invention

[0003] This invention provides a method for comprehensively evaluating the impact of hospital research papers, including: All citation records of the target paper were collected from the clinical practice guidelines. The literature identifier, chapter level, evidence level label and original influence score of each citation were extracted to obtain the original citation dataset containing chapter level and evidence level. Based on the evidence level label and chapter level of each citation, identify whether the relevant citation belongs to the core citation of the recommended main text or the peripheral citation of the background discussion, and determine its recommendation strength level, thus obtaining the core and peripheral attributes and recommendation strength level of each citation; By combining the core marginal attributes and recommendation strength level of each citation, we assess its actual support for clinical decision-making and classify citations into high decision support and low decision support categories, thus obtaining citation decision support classification labels. Based on the citation decision support classification label, the original influence scores of each citation in the original citation dataset are adjusted in layers. An enhancement coefficient is applied to citations with high decision support, and a decay coefficient is applied to citations with low decision support, to obtain the adjusted single citation conversion score. The adjusted individual citation conversion scores are summarized by paper dimension, and the scores of other types of citations that are not cited in the guidelines are merged to obtain the paper's comprehensive clinical conversion score.

[0004] Furthermore, the method also includes: Based on the threshold range of the overall clinical translation score in the clinical application category, the paper is determined to belong to the basic research, transitional, or clinical application category. The category and decision support classification label are then stored in the evaluation database to obtain the final clinical research output category of the paper.

[0005] Furthermore, the process involves collecting all citation records of the target paper from the clinical practice guidelines, extracting the document identifier, chapter level, evidence level label, and original influence score for each citation, and obtaining the original citation dataset containing chapter level and evidence level, including: The distribution area of ​​citation records of target papers is located in the clinical practice guidelines text. The document identification information of each citation record is identified by the string pattern matching method. The document identification information includes a unique document code and a bibliographic number. The document identification information is compared with a pre-established target paper metadata database to determine the set of valid citation records belonging to the target paper. For each citation record in the set of valid citation records, its chapter level in the guide text is obtained. The chapter level is divided into first-level chapter citations, second-level chapter citations and appendix citations according to the guide directory structure. At the same time, the evidence level information marked in the text paragraph where the citation record is located is extracted. The evidence level information is parsed into high-quality evidence, medium-quality evidence and low-quality evidence. Based on the chapter level attribution and the evidence level information, the original influence score corresponding to the document identification information in the target paper meta-database is retrieved. The original influence score, chapter level attribution, evidence level information and document identification information are associated with fields to obtain the original citation dataset containing chapter level, evidence level and original influence score.

[0006] Furthermore, the identification of whether a citation belongs to the core of the recommended text or the peripheral of the background discussion, based on the evidence level label and the chapter level of each citation, includes: Extract the chapter level and evidence level information of each citation record from the original citation dataset. Determine the text functional area of ​​the citation location based on the chapter level. When the chapter level is a first-level chapter citation, it is determined to be a recommended text area. When the chapter level is a second-level chapter citation or an appendix citation, it is determined to be a background discussion area. The text functional area label of each citation record is obtained. For the text functional area label, core edge attribute identification is performed in combination with the evidence level information corresponding to the citation record. When the text functional area label is the recommended text area and the evidence level is high-quality evidence or medium-quality evidence, it is determined to be a core citation. When the text functional area label is the background discussion area or the evidence level is low-quality evidence, it is determined to be an edge citation, thus obtaining the core edge attribute label of each citation record.

[0007] Furthermore, obtaining the core edge attributes and recommendation strength level for each reference includes: Based on the core edge attribute labels, the records marked as core references are classified into recommendation strength levels. The recommended sentence text of the paragraph where the record marked as a core reference is located is obtained, and the recommendation expression feature words are identified. The recommendation expression feature words are divided into strong recommendation indicating mandatory recommendation, medium recommendation indicating suggestion, and weak recommendation indicating optional recommendation. The recommendation strength level corresponding to the core reference is determined according to the matching results. The core edge attribute labels are associated with the recommendation strength level. For records marked as edge references, their recommendation strength level is uniformly assigned as "no recommendation support", thus obtaining the core edge attribute and recommendation strength level of each reference.

[0008] Furthermore, by comprehensively considering the core peripheral attributes and recommendation strength level of each citation, the actual support level for clinical decision-making is assessed, and citations are divided into high decision support and low decision support categories, resulting in citation decision support classification labels, including: The core edge attribute labels and recommendation strength levels are obtained from each reference record. For records marked as core references, corresponding support weight values ​​are assigned according to their recommendation strength levels. Strong recommendations correspond to high support weight values, medium recommendations correspond to medium support weight values, and weak recommendations correspond to low support weight values, thus obtaining the support weight value for each core reference. For the support weight value, the core edge attribute label of the reference record is combined to perform decision support classification. When the core edge attribute label is a core reference and the support weight value is a high support weight value or a medium support weight value, it is classified into the high decision support class. When the core edge attribute label is a core reference but the support weight value is a low support weight value, or the core edge attribute label is an edge reference, it is classified into the low decision support class, thus obtaining the reference decision support classification label for each reference record.

[0009] Furthermore, based on the citation decision support classification labels, the original influence scores of each citation in the original citation dataset are adjusted hierarchically, with a decay coefficient applied to low decision support classes and an enhancement coefficient applied to high decision support classes, to obtain the adjusted single citation conversion score, including: The original influence score and citation decision support classification label of each citation record are obtained from the original citation dataset. Based on the citation decision support classification label, the high decision support class and the low decision support class are identified. The enhancement coefficient is retrieved for the high decision support class citations and the attenuation coefficient is retrieved for the low decision support class citations to obtain the adjustment coefficient corresponding to each citation record. The adjusted citation score is obtained by multiplying the adjustment coefficient with the original influence score. The adjusted citation score is stored as a single citation conversion score, and the single citation conversion score is associated with the document identification information of the citation record.

[0010] Furthermore, it also includes consistency verification of the citation decision support classification labels with the original influence scores, evidence level labels, and chapter level information in the original citation dataset, resulting in an adjusted original influence score for each citation, carrying an adjustment direction marker, evidence level, and chapter level. Specifically, this includes: The original influence score, evidence level label, and chapter level information of each citation record are retrieved from the original citation dataset. At the same time, the citation decision support classification label corresponding to the citation record is obtained. The classification label is associated with the evidence level label and chapter level information to obtain the attribute information set of each citation record. For each reference record in the attribute information set, a consistency check is performed between the classification label and the attribution determination basis. Specifically, for high decision support, the check is performed to see if the chapter level is a first-level chapter reference and whether the evidence level label is high-quality evidence or medium-quality evidence. For low decision support, the check is performed to see if the chapter level is a second-level chapter reference or appendix reference, or whether the evidence level label is low-quality evidence. The consistency check result for each reference record is obtained. Based on the consistency verification results, the reference records that have passed the verification are assigned adjustment direction tags. The adjustment direction tags of high decision support references are assigned enhancement adjustment, and the adjustment direction tags of low decision support references are assigned decay adjustment, thus obtaining reference records carrying adjustment direction tags. The reference record carrying the adjustment direction mark is associated with the original influence score, evidence level label, and chapter level information and stored in a field association to obtain the original influence score to be adjusted for each reference carrying the adjustment direction mark, evidence level, and chapter level.

[0011] Furthermore, the adjusted individual citation conversion scores are aggregated by paper dimension, and other citation scores not cited in the guidelines are merged to obtain the paper's comprehensive clinical translation score, including: Based on the document identification information, all individual citation conversion scores are grouped by paper dimension. Multiple guideline citation conversion scores belonging to the same paper are accumulated and summarized to obtain the guideline citation summary score of the paper. The non-guideline citation score of the paper is retrieved from the target paper metadata database. The total score of guideline citations is then combined with the non-guideline citation score to obtain the paper's overall clinical translation score.

[0012] Furthermore, based on the threshold range of the clinical application category within the overall clinical translation score, the paper is determined to belong to the basic research, transitional, or clinical application category. The category and decision support classification label are then stored in the evaluation database to obtain the final clinical research output category of the paper, including: The total score of the paper in clinical translation is compared with a pre-set threshold interval, which is divided into three continuous segments: basic research interval, transitional interval, and clinical application interval. The paper is classified into basic research, transitional, or clinical application categories based on the interval in which the total score falls, thus obtaining the paper's category determination result. The classification result is associated with the citation decision support classification label corresponding to the paper, and stored in the evaluation database to obtain the final clinical research output classification of the paper.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a comprehensive evaluation method for the impact of hospital research papers. It systematically collects all citation records of target papers from clinical practice guidelines, extracting the document identifier, chapter level, evidence level label, and original impact score for each citation to construct an original citation dataset containing chapter level and evidence level. Then, based on the evidence level and chapter level characteristics, it intelligently identifies whether each citation belongs to the core of the recommended text or the peripheral of the background discussion, and determines its recommendation strength level, thereby assigning core / peripheral attributes and recommendation strength levels to the citations. Based on this, it comprehensively evaluates the core / peripheral attributes and recommendation strength, etc. The evaluation method assesses the actual support of citations for clinical decision-making, precisely categorizing citations into high and low decision-support categories. Subsequently, a differentiated stratified adjustment is applied to the original impact scores: an enhancement coefficient is applied to high-support citations, and a decay coefficient to low-support citations, resulting in an adjusted individual citation conversion score that better reflects the true clinical value. Finally, the adjusted scores are aggregated by paper dimension and integrated with other citation scores not cited in guidelines to form a comprehensive clinical translation score for the paper. Based on this score, the paper is classified as basic research, transitional research, or clinical application, and this score, along with the decision-support category label, is stored in the evaluation database. The core technical effect of this invention lies in breaking through the limitations of traditional methods that rely solely on citation counts or journal impact factors. It achieves a scientific evaluation of paper impact based on the actual application support of clinical guidelines, significantly improving the accuracy and guidance of the clinical translation value assessment of hospital research outputs. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method for comprehensively evaluating the impact of hospital research papers according to the present invention.

[0015] Figure 2 This is a schematic diagram of a comprehensive evaluation method for the impact of hospital research papers according to the present invention.

[0016] Figure 3 This is another schematic diagram of a comprehensive evaluation method for the influence of hospital research papers according to the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] like Figures 1-3 This embodiment of a comprehensive evaluation method for the impact of hospital research papers may specifically include: S101. Collect all citation records of the target paper from the clinical practice guidelines text, extract the literature identifier, chapter level, evidence level label and original influence score of each citation, and obtain the original citation dataset containing chapter level and evidence level.

[0019] The distribution area of ​​citation records for target papers is located in the clinical practice guidelines. A string pattern matching method is used to identify the document identification information of each citation record. This document identification information includes a unique document code and a bibliographic number. The document identification information is compared with a pre-established target paper metadata database to determine the set of valid citation records belonging to the target paper. For each citation record in the valid citation record set, its chapter level in the guideline text is obtained. The chapter level is divided into three categories based on the guideline's table of contents structure: first-level chapter citations, second-level chapter citations, and appendix citations. Simultaneously, the evidence level information marked in the text paragraph containing the citation record is extracted. This evidence level information is parsed according to the original guideline annotations as high-quality evidence, medium-quality evidence, and low-quality evidence. Based on the chapter level and the evidence level information, the original influence score corresponding to the document identification information in the target paper metadata database is retrieved. The original influence score, chapter level, evidence level information, and document identification information are correlated to obtain an original citation dataset containing chapter level, evidence level, and original influence score, which is used for subsequent influence calculation and analysis.

[0020] When locating citation records of target papers in clinical practice guidelines, a string pattern matching method is used to scan the entire text of the guidelines and identify text fragments that conform to the characteristics of literature citation format.

[0021] For example, when a number enclosed in square brackets appears in the main text of the guide, that number points to the corresponding reference entry in the bibliography list. By comparing the identified bibliographic information with a pre-established target paper metadata database one by one, citation records that match the target papers are filtered out, forming a set of valid citation records.

[0022] In one possible implementation, the chapter hierarchy is determined based on the hierarchical numbering characteristics within the guideline's table of contents. First-level chapter citations are located in the main body of the guideline's recommendations, typically identified by a single-level Arabic numeral number, and carry key evidence supporting the core treatment pathways. Second-level chapter citations are located in the sub-discussion paragraphs under each topic, presented in a double-level numbering format, and are used to supplement explanations or provide literature on alternative solutions. Appendix citations appear in the supplementary materials section at the end of the guideline, and are mostly references cited in background information or methodological explanations.

[0023] Specifically, extracting evidence level information requires analyzing the level markers immediately following citations in the guideline text. Clinical guidelines typically indicate the quality of evidence assessment of cited literature in parentheses. High-quality evidence corresponds to rigorously designed randomized controlled trials or systematic reviews, medium-quality evidence corresponds to clinical studies with some risk of bias, and low-quality evidence corresponds to observational studies or expert consensus. By identifying the correlation between these markers and the location of citations, the structured extraction of evidence level information is achieved.

[0024] It should be noted that the target paper metadata database pre-stores the original impact score of each paper, which is calculated based on the citation frequency after publication and the journal's impact factor. Based on the aforementioned chapter hierarchy and evidence level information, the original impact score of the corresponding paper is retrieved. The document identification information, chapter hierarchy, evidence level information, and original impact score are then linked and stored to obtain an original citation dataset containing chapter hierarchy and evidence level, providing a structured data foundation for subsequent fine-grained evaluation of citation value.

[0025] S102. Based on the evidence level label and chapter level of each citation, identify whether the citation belongs to the core citation of the recommended main text or the peripheral citation of the background discussion, and determine its recommendation strength level to obtain the core and peripheral attributes and recommendation strength level of each citation.

[0026] The chapter level and evidence level information of each citation record are extracted from the original citation dataset. Based on the chapter level, the text functional region of the citation is determined. If the chapter level is a first-level chapter citation, the citation is located in the recommended main text area; if it is a second-level chapter citation or appendix citation, the citation is located in the background discussion area. This yields a text functional region label for each citation record. For each text functional region label, core edge attribute identification is performed in conjunction with the corresponding evidence level information. When the text functional region label is in the recommended main text area and the evidence level is high-quality or medium-quality evidence, the citation is considered a core citation. When the text functional region label is in the background discussion area or the evidence level is low-quality evidence, the citation is considered a edge citation. This yields a core edge attribute label for each citation record. Based on the core edge attribute labels, the records marked as core citations are classified into recommendation strength levels. The recommended statement text of the paragraph containing the cited record is obtained, and the recommended expression feature words are identified. These feature words include words indicating mandatory recommendations (strong recommendation), words indicating suggestion (medium recommendation), and words indicating optional recommendations (weak recommendation). The recommendation strength level supported by the core citation is determined based on the matching results. The core edge attribute labels are associated with the recommendation strength level. For records marked as edge citations, their recommendation strength level is uniformly assigned as no recommendation support, thus obtaining the core edge attributes and recommendation strength level of each citation.

[0027] In the structure of clinical practice guidelines, different chapter levels serve different clinical decision-making functions. First-level chapters typically correspond to the core recommendations of the guideline, directly outlining treatment suggestions for specific clinical problems; the references cited in this section constitute direct evidence supporting the recommendations. Second-level chapters and appendices primarily provide background information, methodological explanations, or supplementary discussions, with the cited references mostly serving as supplementary materials. Based on this structural characteristic, determining the textual functional area where a citation is located by its chapter level allows for the differentiation of the citation's role within the guideline.

[0028] Specifically, when a citation is labeled as a first-level chapter citation, it indicates that the citation appears in the main body of the guideline's recommendations, and its text functional area label is assigned to the recommendation text area. When the chapter level is labeled as a second-level chapter citation or an appendix citation, the text functional area label is assigned to the background discussion area. This classification reflects the differences in the positioning of different chapter content in the guideline's writing specifications.

[0029] In one possible implementation, the identification of core and peripheral attributes employs a joint judgment rule combining text functional area labels and evidence level information. Citations within the recommendation text area that are simultaneously labeled as high-quality or medium-quality evidence indicate that the citation is both in a core decision-making position and has strong evidentiary support, meeting the dual conditions of a core citation. Citations located in the background discussion area, regardless of their evidence level, do not directly support the formation of the recommendation and are therefore classified as peripheral citations. Similarly, even if located within the recommendation text area, if the evidence level is labeled as low-quality evidence, its supporting role in the recommendation is limited, and it is also classified as a peripheral citation.

[0030] It should be noted that the strength of recommendation is determined based on the strength of the wording in the recommendation statements within the guidelines. The determination of the strength of recommendation is independent of the level of evidence. The level of evidence reflects the quality of the research itself, while the strength of recommendation reflects the attitude of the guideline writing expert group towards the treatment measure. There is a correlation between the two, but they are not entirely corresponding: high-quality evidence usually supports a strong recommendation, but it may also be marked as a moderate or weak recommendation due to factors such as benefit-risk ratio and cost-effectiveness; low-quality evidence, in the absence of higher-quality evidence, may also be marked as a strong recommendation due to clinical urgency. Therefore, the strength of recommendation is determined solely based on the characteristic words in the recommendation statements and is not directly constrained by the level of evidence. Clinical guidelines typically use wording with clear differences in strength when expressing recommendations. Strong recommendations are characterized by mandatory words such as "should," "must," and "recommended," indicating that the treatment measure should be routinely implemented in clinical practice. Moderate recommendations are characterized by suggestive words such as "recommend," "preferably," and "can be considered for priority," indicating that the measure is applicable in most cases but clinicians are allowed to adjust it according to specific situations. The characteristic words for weak recommendations include optional words such as "may", "or may", and "may be considered under specific circumstances", indicating that the measure is only provided as an alternative for reference.

[0031] For example, in cardiovascular disease diagnosis and treatment guidelines, if a recommendation states, "For patients with acute myocardial infarction, percutaneous coronary intervention should be performed as early as possible after the onset of the disease," the word "should" is a mandatory term, and the cited literature supporting this statement is marked as a strong recommendation. If the statement states, "It is recommended that hypertensive patients undergo regular blood pressure monitoring," the word "recommend" is a suggestive term, and the corresponding citation is marked as a moderate recommendation. If the statement states, "Combination therapy may be considered for some patients," the word "may be considered" is an optional term, and the corresponding citation is marked as a weak recommendation.

[0032] In one embodiment, the identification of recommendation expression feature words is achieved by obtaining the complete recommendation sentence text of the paragraph containing the cited record. The predicate verb and its modifiers are extracted from this sentence text and matched against a pre-established recommendation expression feature word library. This feature word library is constructed based on a clinical guideline corpus, formed through expert annotation and machine learning training, and stores corresponding word sets according to three categories: mandatory, advisory, and optional. For example, the mandatory category includes "must," "should," etc.; the advisory category includes "recommended," "appropriate," etc.; and the optional category includes "can," "optional," etc. Upon successful matching, the recommendation strength level supported by the citation is returned. Furthermore, for records marked as marginal citations, since they do not directly support any recommendation opinion, there is no need for a recommendation strength level identification process; their recommendation strength level field is uniformly assigned the value of "no recommendation support." This processing method clearly distinguishes the differences in evaluation dimensions between core citations that support clinical decision-making and marginal citations that only serve as background references. By associating core and marginal attribute tags with recommendation strength levels, a complete attribute annotation for each citation record is formed. Core citations carry a recommendation strength level label of strong, moderate, or weak recommendation, while marginal citations carry a label of no recommendation support. The two types of citations have a clear distinguishing attribute basis in the subsequent assessment of their support for clinical decision-making.

[0033] S103. Based on the core marginal attributes and recommendation strength level of each citation, assess its actual support for clinical decision-making, and classify citations into high decision support and low decision support categories to obtain citation decision support classification labels.

[0034] The core and peripheral attribute tags and recommendation strength levels are obtained from each citation record. For records marked as core citations, corresponding support weight values ​​are assigned based on their recommendation strength levels. These support weight values ​​are determined according to the correlation between the recommendation strength level and clinical decision-making: strong recommendations correspond to high support weight values, moderate recommendations to median support weight values, and weak recommendations to low support weight values. This yields the support weight value for each core citation. Based on these support weight values, decision support classification is performed using the core and peripheral attribute tags of the citation record. If the core and peripheral attribute tag is a core citation and the support weight value is a high or median support weight value, the citation is classified into the high decision support category. If the core and peripheral attribute tag is a core citation but the support weight value is a low support weight value, or if the core and peripheral attribute tag is a peripheral citation, the citation is classified into the low decision support category. This yields the citation decision support classification tag for each citation record.

[0035] In the evaluation of clinical practice guideline citations, the supporting role of citations at different recommendation strength levels differed significantly in clinical decision-making. The weighting of support values ​​reflects the correspondence between recommendation strength levels and the degree of relevance to clinical decision-making. A strong recommendation means the treatment should be routinely implemented in clinical practice, and its supporting literature directly guides physicians' decision-making behavior, thus it is assigned a high support weighting. A moderate recommendation means the measure is applicable in most cases but allows clinicians to adjust according to specific situations; its supporting literature's decision-making guidance is less than that of a strong recommendation, and it is assigned a median support weighting. A weak recommendation is only provided as an alternative for reference; its supporting literature has a weaker binding force on decision-making, and it is assigned a low support weighting.

[0036] Specifically, a high support weight value indicates that the recommendation supported by the citation has a mandatory attribute in clinical practice, and clinicians typically follow this recommended path in diagnosis and treatment when faced with relevant conditions. A median support weight value indicates that the recommendation supported by the citation has a reference attribute, and doctors decide whether to adopt it after comprehensively considering individual patient differences. A low support weight value indicates that the recommendation supported by the citation only has an optional attribute, and can only be considered under specific conditions.

[0037] In one possible implementation, the assessment of decision support level employs a joint judgment rule combining core and peripheral attribute labels and support weight values. Records marked as core citations have already been confirmed through prior steps to be located within the recommendation text and possess a certain level of evidence quality. Based on this, the actual degree of support for clinical decision-making is further differentiated according to the support weight value. When the support weight value is high, it indicates that the citation is both in a core decision-making position and supports a mandatory recommendation, thus its decision support level is determined to be high. When the support weight value is median or low, although the citation still falls within the scope of core citations, the strength of its supported recommendation is relatively weak, thus its decision support level is determined to be medium.

[0038] It should be noted that a simplified approach is used to determine the decision support level of marginal citations. Since marginal citations have already been identified in the background discussion area or as low-quality evidence in the preceding stages, their recommendation strength level is marked as "no support," indicating that the citation does not directly support any recommendation. Regardless of the citation frequency in the guideline references, it does not directly constrain clinicians' diagnostic and treatment decisions; therefore, it is uniformly classified as low support.

[0039] For example, in a diabetes treatment guideline, a randomized controlled trial paper on metformin as a first-line drug is listed as high-quality evidence and appears in the core recommendation paragraph of a first-level chapter. The recommendation states that "metformin should be the first-line hypoglycemic drug for patients with type 2 diabetes." This citation is labeled as a core citation, its recommendation strength is strong, its support weight is high, and its decision support level is high. However, in the same guideline, another observational study on diabetes epidemiology only appears in the background section of the appendix. Its core citation is labeled as a marginal citation, and its decision support level is directly determined to be low. After labeling the decision support level of all citations, a binary classification process is performed according to a pre-defined classification and merging rule. Citations at both the high and medium support levels provide varying degrees of positive support for clinical decision-making and are classified as high decision support. Citations at the low support level provide no direct support for clinical decision-making or only provide background information and are classified as low decision support.

[0040] Understandably, the purpose of this binary classification design is to differentiate the weighting of citations in the assessment of clinical translational value. High decision support citations indicate that the research findings have been incorporated into the core evidence framework of clinical decision-making pathways, having a substantial impact on medical practice. Low decision support citations, although appearing in the guideline's literature list, are limited to background information or secondary references and should be treated differently when assessing the clinical translational value of the paper. Through this classification process, each citation record acquires a clear citation decision support classification label. This label will serve as the core basis for subsequent impact score adjustments, resulting in a differentiated allocation of evaluation weights between high and low decision support citations.

[0041] S104. Based on the citation decision support classification label, the original influence scores of each citation in the original citation dataset are adjusted in layers. A decay coefficient is applied to the low decision support class, and an enhancement coefficient is applied to the high decision support class to obtain the adjusted single citation conversion score.

[0042] The original influence score and citation decision support classification label of each citation record are obtained from the original citation dataset. Based on the citation decision support classification label, the citation is identified as belonging to the high or low decision support category. For high decision support citations, a pre-configured enhancement coefficient is retrieved; for low decision support citations, a pre-configured attenuation coefficient is retrieved. The enhancement coefficient is a positive number greater than one, and the attenuation coefficient is a positive number less than one, resulting in an adjustment coefficient for each citation record. The adjustment coefficient is multiplied by the original influence score. Multiplying the original influence score of a high decision support citation by the enhancement coefficient amplifies the value, while multiplying the original influence score of a low decision support citation by the attenuation coefficient reduces the value, resulting in an adjusted citation score. This adjusted citation score is stored as the single citation conversion score for that citation record. The single citation conversion score is then associated with the document identification information of that citation record to obtain the adjusted single citation conversion score.

[0043] When adjusting the original impact scores in a stratified manner, the adjustment coefficients reflect the differentiated weights of different decision support categories of citations in the clinical translational value assessment. The enhancement coefficient is set to a positive number greater than one, amplifying the original impact score of high decision support citations after calculation, reflecting the substantial contribution of this type of citation to clinical diagnosis and treatment decisions. The attenuation coefficient is set to a positive number less than one, reducing the original impact score of low decision support citations after calculation, reflecting the marginal status of this type of citation in the clinical decision-making chain.

[0044] Specifically, the values ​​of the enhancement and attenuation coefficients are pre-configured according to industry standards for evaluating citations in clinical practice guidelines. High-support citations include core evidence supporting strong and moderate recommendations, which directly guide physicians' treatment decisions; the recommended range for the enhancement coefficient is 1.2 to 1.5. Low-support citations serve only as background references or secondary evidence; the recommended range for the attenuation coefficient is 0.5 to 0.8.

[0045] In one possible implementation, the score calculation uses a simple multiplication method. The original influence score is used as the multiplicand, and the adjustment coefficient is used as the multiplier; the two are multiplied together to obtain the adjusted citation score. This calculation method maintains the relative order of scores while simultaneously differentiating the numerical values ​​between different categories of citations.

[0046] For example, a paper is cited twice in a cardiovascular disease diagnosis and treatment guideline. The first citation is located in the core recommendation paragraph of a first-level chapter and supports a strong recommendation. Its citation decision support category is high decision support, with an original influence score of 80. Multiplying this by an enhancement coefficient of 1.3 yields a single citation conversion score of 104. The second citation is located in the appendix background introduction section. Its citation decision support category is low decision support, with the same original influence score of 80. Multiplying this by a decay coefficient of 0.6 yields a single citation conversion score of 48.

[0047] It should be noted that the storage of individual citation conversion scores adopts a field association method with document identification information. Each citation record establishes an index relationship through a unique document code and bibliographical number, and the adjusted score is recorded as a quantitative indicator of the conversion value of that citation record. Through the above-mentioned hierarchical adjustment process, high-decision-support citations and low-decision-support citations are clearly distinguished numerically, so that the citation value of the same paper in different guidelines and in different positions is reflected differently, avoiding the crude treatment of treating all guideline citations the same in traditional evaluation methods.

[0048] Retrieve the decision support classification label for each citation, as well as the original influence score, evidence level label, and chapter level information corresponding to that citation in the original citation dataset. Identify the classification label for each citation and verify whether its evidence level label and chapter level are consistent with the classification label's attribution. Determine that the score adjustment direction for high decision support citations is enhancement adjustment and the score adjustment direction for low decision support citations is decay adjustment. Obtain the original influence score to be adjusted for each citation, carrying the adjustment direction label, evidence level, and chapter level.

[0049] The original influence score, evidence level label, and chapter level information of each citation record are retrieved from the original citation dataset. Simultaneously, the citation decision support classification label corresponding to that citation record is obtained. The classification label is then associated with the evidence level label and chapter level information to obtain a set of attribute information for each citation record. For each citation record in the attribute information set, a consistency check is performed between the classification label and the attribution determination criteria. If the classification label is high decision support, it is checked whether its chapter level is a first-level chapter citation and whether the evidence level label is high-quality or medium-quality evidence. If the classification label is low decision support, it is checked whether its chapter level is a second-level chapter citation or appendix citation, or whether the evidence level label is low-quality evidence. This yields a consistency check result for each citation record. Based on the consistency check results, adjustment direction tags are assigned to the citation records that pass the check. High decision support citations are assigned an enhancement adjustment tag, while low decision support citations are assigned a decay adjustment tag, resulting in citation records carrying adjustment direction tags. The reference record carrying the adjustment direction mark is associated with the original influence score, evidence level label, and chapter level information and stored in a field association to obtain the original influence score to be adjusted for each reference carrying the adjustment direction mark, evidence level, and chapter level.

[0050] Before adjusting the scores of citation records, multidimensional attribute information for each citation record is retrieved from the original citation dataset. This attribute information includes four fields: original impact score, evidence level label, chapter / level information, and citation decision support classification label. The original impact score reflects the paper's basic academic performance in terms of citation frequency and journal impact factor; the evidence level label records the guideline's evaluation of the cited literature's quality; the chapter / level information identifies the citation's position within the guideline's text structure; and the classification label represents the comprehensive assessment result of the citation's clinical decision support capability from the preceding evaluation phase.

[0051] Specifically, the association reading of attribute information sets uses document identification information as the index key. Each citation record is located to the corresponding entry in the original citation dataset through its unique document code and bibliographic number, and the attribute values ​​stored in that entry are extracted to form a structured record containing complete attribute information.

[0052] In one possible implementation, the purpose of consistency verification is to verify whether the classification label determination result maintains logical consistency with its original attribution basis. Since the classification label is derived based on a comprehensive judgment of evidence level labeling and chapter hierarchy information, theoretically, there should be a definite correspondence between the classification label and these two original attributes. Consistency verification identifies any anomalies in the data transmission process by retrospectively checking this correspondence. For high-decision-support citations, the verification conditions are: the chapter hierarchy must be a first-level chapter citation, and the evidence level label must be high-quality or medium-quality evidence; both conditions must be met simultaneously. For low-decision-support citations, the verification conditions are: the chapter hierarchy must be a second-level chapter citation or appendix citation, or the evidence level label must be low-quality evidence; meeting either condition is sufficient to pass the verification.

[0053] It should be noted that the consistency verification is based on the following business logic: high decision support citations represent core evidence directly supporting clinical recommendations, and they must simultaneously possess both positional advantage and quality advantage. If a citation is marked as high decision support but its chapter level is not a first-level chapter citation, or its evidence level is low-quality evidence, it indicates a contradiction between the classification label and the original attributes, and the consistency verification result for that record fails. The verification conditions for low decision support citations are relatively lenient; as long as either the positional attribute or the quality attribute meets the characteristics of a marginal citation, the verification is considered successful.

[0054] For example, in a cancer treatment guideline, a clinical trial paper on the efficacy of targeted therapy drugs is marked as high decision support. Its chapter level is a first-level chapter citation, and its evidence level is labeled as high-quality evidence. Both attributes meet the verification criteria for high decision support, and the consistency verification result is passed. Another basic research paper on the pathogenesis of tumors is marked as low decision support. Its chapter level is an appendix citation. Although its evidence level is labeled as medium-quality evidence, its consistency verification result is also passed because its chapter level meets one of the criteria for low decision support. After completing the consistency verification, adjustment direction markers are assigned to the verified citation records. Adjustment direction markers indicate whether subsequent score adjustments will enhance or diminish the citation. A high decision support citation is assigned an enhancement adjustment marker, meaning the original influence score of the citation will be amplified in subsequent calculations. A low decision support citation is assigned a diminishing adjustment marker, meaning the original influence score of the citation will be reduced in subsequent calculations. Furthermore, the citation record carrying the adjustment direction marker is stored in a field-associative manner with the original influence score, evidence level label, and chapter level information. This multi-field association storage method preserves the complete attribute characteristics of the citation record, allowing the original evidence quality and location characteristics of the citation to be traced during subsequent score adjustment operations when performing enhancement or attenuation calculations. Through the above-mentioned processing flow of attribute retrieval, consistency verification, and adjustment direction markers, each citation record forms an original influence score to be adjusted, carrying the adjustment direction marker, evidence level, and chapter level, providing complete data input for subsequent stratification coefficient adjustment calculations.

[0055] It should be noted that the adjusted citation score may exceed the normalized range [0, 100] of the original influence score. When the original influence score of high decision support citations is close to 100 and the enhancement coefficient is 1.5, the adjusted score may reach 150. To maintain the consistency of the evaluation system, the adjusted citation score will not be normalized again, allowing it to exceed the original range, in order to truly reflect the numerical difference between high decision support citations and low decision support citations. In the subsequent summary stage, the upper limit of the overall clinical translation score will be naturally determined by the actual adjustment results, without setting an artificial upper limit constraint.

[0056] S105. Summarize the adjusted individual citation conversion scores by paper dimension, and merge the scores of other types of citations that are not cited in the guidelines to obtain the paper's comprehensive clinical conversion score.

[0057] Based on the document identification information, all individual citation conversion scores are grouped by paper dimension. Multiple guideline citation conversion scores belonging to the same paper are summed to obtain the paper's total guideline citation score. The non-guideline citation score of the paper is retrieved from the target paper metadata database. This non-guideline citation score represents the basic citation score generated by the paper's citations in academic journals and conference papers. The total guideline citation score and the non-guideline citation score are then combined and summed to obtain the paper's overall clinical translation score.

[0058] After adjusting the conversion scores for individual citations, all citation conversion scores are grouped and summarized by paper dimension. The grouping is based on the document identification information carried by each citation record. Multiple citation records of the same paper in different guidelines are identified and matched through their unique document codes, which are unique identifiers generated based on international standards such as DOI or PMID.

[0059] Specifically, the scores for all guideline citations belonging to the same paper are summed up to form the paper's overall score in the clinical practice guideline citation dimension. If a paper is cited in multiple practice guidelines from different specialties, the scores for each of those citations are included in the summation.

[0060] It should be noted that the academic impact of a paper is reflected not only in citations to clinical practice guidelines, but also in the basic citation score generated by citations in other academic journals and conference papers. The non-guideline citation score is directly retrieved from the target paper's metadata database, and this score is continuously updated after publication. The overall clinical translation score of the paper is formed by combining the total guideline citation score and the non-guideline citation score. This total score reflects the paper's comprehensive impact on both clinical decision support and academic communication.

[0061] S106. Based on the threshold range of the clinical application category in which the overall clinical translation score is located, determine whether the paper belongs to the basic research category, the transitional category, or the clinical application category. Store the category and decision support classification label in the evaluation database to obtain the final clinical research output category of the paper.

[0062] The overall clinical translation score of the paper is compared with a pre-defined threshold range, which is divided into three consecutive segments: basic research, transitional, and clinical application. If the total score falls into the basic research segment, the paper is classified as basic research; if it falls into the transitional segment, it is classified as transitional; and if it falls into the clinical application segment, it is classified as clinical application. This results in the paper's classification. The classification result is then correlated with the corresponding citation decision support classification label and stored in the evaluation database to obtain the final clinical research output classification of the paper.

[0063] After calculating the overall clinical translation score of the paper, the classification is determined according to a pre-set threshold range. The threshold range is divided into three continuous segments: basic research, transitional, and clinical application. For example, when S < 5, it is basic research; when 5 ≤ S < 8, it is transitional; and when S ≥ 8, it is clinical application. Here, S is the overall clinical translation score.

[0064] Specifically, the basic research category corresponds to papers with lower overall clinical translation scores, indicating that the research findings are still in the basic exploration stage and have not been widely adopted by clinical practice guidelines. The transitional category corresponds to papers with intermediate overall clinical translation scores, indicating that the research findings have begun to enter the clinical translation process but have not yet formed core decision support. The clinical application category corresponds to papers with higher overall clinical translation scores, indicating that the research findings have been adopted as core evidence by clinical practice guidelines.

[0065] In one embodiment, if the overall clinical translation score of a paper falls into the clinical application category, its category determination result is clinical application. This determination result is associated with the decision support classification label of the paper in various guidelines and then stored in the evaluation database to form the final clinical research output category record of the paper.

[0066] If the technical solution of this application involves the collection, storage, use, processing, transmission, provision, disclosure, or deletion of personal information, the products using this technical solution have clearly and understandably informed the users of the personal information processing rules before processing personal information, and have obtained the individuals' voluntary consent in accordance with the law. If the technical solution of this application involves sensitive personal information (such as biometrics, religious beliefs, specific identities, medical and health information, financial accounts, and location tracking), the products using this solution have obtained the individuals' separate consent before processing sensitive personal information, and have also met the requirement of "express consent," ensuring that individuals make authorization decisions voluntarily based on full knowledge.

[0067] Specific implementation methods include, but are not limited to, the following: setting up clear and prominent signs at personal information collection devices such as cameras and sensors to inform relevant personnel that they have entered the scope of personal information collection and that their personal information will be collected and processed. If an individual voluntarily enters the collection scope after being informed, it is deemed that they have agreed to the collection of their personal information; or using obvious icons, text descriptions, or other means on the terminal device or system interface for personal information processing to inform them of the rules for personal information processing, and obtaining the individual's explicit authorization through interactive methods such as pop-up prompts, check confirmation boxes, or asking the individual to upload their personal information themselves.

[0068] The aforementioned personal information processing rules should include, but are not limited to, the name and contact information of the personal information processor, the specific purpose of personal information processing, the processing method, the types of personal information processed, the retention period, and the methods and procedures for individuals to exercise their relevant rights.

[0069] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A comprehensive evaluation method for the impact of hospital research papers, characterized in that, include: All citation records of the target paper were collected from the clinical practice guidelines. The literature identifier, chapter level, evidence level label and original influence score of each citation were extracted to obtain the original citation dataset containing chapter level and evidence level. Based on the evidence level label and chapter level of each citation, identify whether the relevant citation belongs to the core citation of the recommended main text or the peripheral citation of the background discussion, and determine its recommendation strength level, thus obtaining the core and peripheral attributes and recommendation strength level of each citation; By combining the core marginal attributes and recommendation strength level of each citation, we assess its actual support for clinical decision-making and classify citations into high decision support and low decision support categories, thus obtaining citation decision support classification labels. Based on the citation decision support classification label, the original influence scores of each citation in the original citation dataset are adjusted in layers. An enhancement coefficient is applied to citations with high decision support, and a decay coefficient is applied to citations with low decision support, to obtain the adjusted single citation conversion score. The adjusted individual citation conversion scores are summarized by paper dimension, and the scores of other types of citations that are not cited in the guidelines are merged to obtain the paper's comprehensive clinical conversion score.

2. The method for comprehensively evaluating the impact of hospital research papers according to claim 1, characterized in that, The method further includes: Based on the threshold range of the overall clinical translation score in the clinical application category, the paper is determined to belong to the basic research, transitional, or clinical application category. The category and decision support classification label are then stored in the evaluation database to obtain the final clinical research output category of the paper.

3. The method for comprehensively evaluating the impact of hospital research papers according to claim 1, characterized in that, The process involves collecting all citation records of the target paper from the clinical practice guidelines, extracting the document identifier, chapter level, evidence level label, and original influence score for each citation, resulting in an original citation dataset containing chapter level and evidence level, including: The distribution area of ​​citation records of target papers is located in the clinical practice guidelines text. The document identification information of each citation record is identified by the string pattern matching method. The document identification information includes a unique document code and a bibliographic number. The document identification information is compared with a pre-established target paper metadata database to determine the set of valid citation records belonging to the target paper. For each citation record in the set of valid citation records, its chapter level in the guide text is obtained. The chapter level is divided into first-level chapter citations, second-level chapter citations and appendix citations according to the guide directory structure. At the same time, the evidence level information marked in the text paragraph where the citation record is located is extracted. The evidence level information is parsed into high-quality evidence, medium-quality evidence and low-quality evidence. Based on the chapter level attribution and the evidence level information, the original influence score corresponding to the document identification information in the target paper meta-database is retrieved. The original influence score, chapter level attribution, evidence level information and document identification information are associated with fields to obtain the original citation dataset containing chapter level, evidence level and original influence score.

4. The method for comprehensively evaluating the impact of hospital research papers according to claim 1, characterized in that, The process of identifying whether a citation belongs to the core of the recommended text or to the peripheral discussion based on the evidence level label and the chapter level of each citation includes: Extract the chapter level and evidence level information of each citation record from the original citation dataset. Determine the text functional area of ​​the citation location based on the chapter level. When the chapter level is a first-level chapter citation, it is determined to be a recommended text area. When the chapter level is a second-level chapter citation or an appendix citation, it is determined to be a background discussion area. The text functional area label of each citation record is obtained. For the text functional area label, core edge attribute identification is performed in combination with the evidence level information corresponding to the citation record. When the text functional area label is the recommended text area and the evidence level is high-quality evidence or medium-quality evidence, it is determined to be a core citation. When the text functional area label is the background discussion area or the evidence level is low-quality evidence, it is determined to be an edge citation, thus obtaining the core edge attribute label of each citation record.

5. The method for comprehensively evaluating the impact of hospital research papers according to claim 4, characterized in that, The core edge attributes and recommendation strength levels of each reference are obtained, including: Based on the core edge attribute labels, the records marked as core references are classified into recommendation strength levels. The recommended sentence text of the paragraph where the record marked as a core reference is located is obtained, and the recommendation expression feature words are identified. The recommendation expression feature words are divided into strong recommendation indicating mandatory recommendation, medium recommendation indicating suggestion, and weak recommendation indicating optional recommendation. The recommendation strength level corresponding to the core reference is determined according to the matching results. The core edge attribute labels are associated with the recommendation strength level. For records marked as edge references, their recommendation strength level is uniformly assigned as "no recommendation support", thus obtaining the core edge attribute and recommendation strength level of each reference.

6. The method for comprehensively evaluating the impact of hospital research papers according to claim 1, characterized in that, The core marginal attributes and recommendation strength level of each citation are combined to assess its actual support for clinical decision-making, classifying citations into high-decision-support and low-decision-support categories, resulting in citation decision-support classification labels, including: The core edge attribute labels and recommendation strength levels are obtained from each reference record. For records marked as core references, corresponding support weight values ​​are assigned according to their recommendation strength levels. Strong recommendations correspond to high support weight values, medium recommendations correspond to medium support weight values, and weak recommendations correspond to low support weight values, thus obtaining the support weight value for each core reference. For the support weight value, the core edge attribute label of the reference record is combined to perform decision support classification. When the core edge attribute label is a core reference and the support weight value is a high support weight value or a medium support weight value, it is classified into the high decision support class. When the core edge attribute label is a core reference but the support weight value is a low support weight value, or the core edge attribute label is an edge reference, it is classified into the low decision support class, thus obtaining the reference decision support classification label for each reference record.

7. The method for comprehensively evaluating the impact of hospital research papers according to claim 1, characterized in that, The process involves stratifying and adjusting the original influence scores of each citation in the original citation dataset based on the citation decision support classification labels. A decay coefficient is applied to low-decision support citations, and an enhancement coefficient is applied to high-decision support citations, resulting in adjusted individual citation conversion scores. This includes: The original influence score and citation decision support classification label of each citation record are obtained from the original citation dataset. Based on the citation decision support classification label, the high decision support class and the low decision support class are identified. The enhancement coefficient is retrieved for the high decision support class citations and the attenuation coefficient is retrieved for the low decision support class citations to obtain the adjustment coefficient corresponding to each citation record. The adjusted citation score is obtained by multiplying the adjustment coefficient with the original influence score. The adjusted citation score is stored as a single citation conversion score, and the single citation conversion score is associated with the document identification information of the citation record.

8. The method for comprehensively evaluating the impact of hospital research papers according to claim 7, characterized in that, This also includes verifying the consistency between the citation decision support classification label and the original influence score, evidence level label, and chapter level information in the original citation dataset. This yields an adjusted original influence score for each citation, carrying an adjustment direction marker, evidence level, and chapter level. Specifically, this includes: The original influence score, evidence level label, and chapter level information of each citation record are retrieved from the original citation dataset. At the same time, the citation decision support classification label corresponding to the citation record is obtained. The classification label is associated with the evidence level label and chapter level information to obtain the attribute information set of each citation record. For each reference record in the attribute information set, a consistency check is performed between the classification label and the attribution determination basis. Specifically, for high decision support, the check is performed to see if the chapter level is a first-level chapter reference and whether the evidence level label is high-quality evidence or medium-quality evidence. For low decision support, the check is performed to see if the chapter level is a second-level chapter reference or appendix reference, or whether the evidence level label is low-quality evidence. The consistency check result for each reference record is obtained. Based on the consistency verification results, the reference records that have passed the verification are assigned adjustment direction tags. The adjustment direction tags of high decision support references are assigned enhancement adjustment, and the adjustment direction tags of low decision support references are assigned decay adjustment, thus obtaining reference records carrying adjustment direction tags. The reference record carrying the adjustment direction mark is associated with the original influence score, evidence level label, and chapter level information and stored in a field association to obtain the original influence score to be adjusted for each reference carrying the adjustment direction mark, evidence level, and chapter level.

9. The method for comprehensively evaluating the impact of hospital research papers according to claim 1, characterized in that, The adjusted individual citation conversion scores are aggregated by paper dimension, and other citation scores not cited in the guidelines are merged to obtain the paper's comprehensive clinical translation score, including: Based on the document identification information, all individual citation conversion scores are grouped by paper dimension. Multiple guideline citation conversion scores belonging to the same paper are accumulated and summarized to obtain the guideline citation summary score of the paper. The non-guideline citation score of the paper is retrieved from the target paper metadata database. The total score of guideline citations is then combined with the non-guideline citation score to obtain the paper's overall clinical translation score.

10. The method for comprehensively evaluating the impact of hospital research papers according to claim 2, characterized in that, The process involves determining the paper's classification as basic research, transitional research, or clinical application based on the overall clinical translation score and its corresponding threshold range. The classification category and decision support label are then stored in the evaluation database to obtain the final clinical research output category of the paper, including: The total score of the paper in clinical translation is compared with a pre-set threshold interval, which is divided into three continuous segments: basic research interval, transitional interval, and clinical application interval. The paper is classified into basic research, transitional, or clinical application categories based on the interval in which the total score falls, thus obtaining the paper's category determination result. The classification result is associated with the citation decision support classification label corresponding to the paper, and stored in the evaluation database to obtain the final clinical research output classification of the paper.

Images