Construction Of Clinical Translation Potential Index Based On Multi-dimensional Characteristics

Whether in response to major public health emergencies or to improve people’s health,it is necessary to promote and accelerate the process of scientific research results to clinical application.However,the huge gap between basic research and clinical practice makes the investment of a large number of scientific research funds can not get significant results.Therefore,it is necessary to identify and predict scientific research achievements with clinical value and clinical translation potential in advance,and prioritize the allocation of scientific research resources,so as to promote and promote the progress of clinical translation of scientific research achievements.With the advent of Web 2.0,the academic communication environment has changed greatly,and various online media platforms have become the main stage for academic achievements to play an important role.Altmetrics indicators have emerged as a supplement to traditional citation indicators.If traditional citation indicators are still used to evaluate the clinical translation potential of scientific research achievements,there may be some drawbacks.After summarizing and combing existing relevant studies,this study combined with altmetrics indicators to construct a multidimensional characteristic index system for clinical citation prediction of papers based on the extraction of citations and clinical citation prediction of biomedical triangle related papers.Based on the multidimensional characteristic index system,Statistical analysis,difference analysis,mutual information score method,feature significance analysis and other methods are used to analyze and compare and verify the role of multidimensional characteristic indicators in the prediction of clinical citations in papers.Four groups of characteristic indicator system are constructed by controlling the two variables of numerical accumulation time and range of characteristic indicators by comparative experiment method.A total of 12 clinical reference prediction models are established by using three different machine learning prediction algorithms,Random Forest,Gradient Boosting Decision Tree and CatBoost.The accuracy,precision,recall and F1 values of each prediction model are calculated by ten fold cross-validation to evaluate and compare their prediction efficiency.Finally,the algorithm and characteristic index system with the best prediction efficiency are selected.Construct an index with more advantages in timeliness and accuracy to evaluate and reflect the clinical translation potential of the paper,so as to provide government departments and relevant institutions with more timeliness and reliability decision aid in scientific research funding,accelerate the clinical translation process of scientific research results,and provide references for the evaluation of clinical translation potential of the paper and the prediction of clinical citations of the paper.It also explores the application of altmetrics indicators in the field of clinical value of papers and prediction of clinical citations of papers.The main conclusions of this study are as follows:(1)The comparative experimental results of timeliness analysis,coverage analysis and paper clinical citation prediction validate the improvement effect of altmetrics indicators on the original feature index system of paper clinical citation prediction in terms of timeliness and accuracy.(2)Quantitative methods such as difference analysis,principal component analysis,correlation analysis,mutual information score and feature importance analysis,as well as the results of comparative experiments,verify that the multidimensional characteristic index system constructed in this study for the prediction of clinical paper citation has a certain scientificity and reliability.In addition,the papers that would receive clinical citations are higher in the number of citations,altmetrics indicators,and the numerical values of indicators characteristic of human or animal-related biomedical triangle models.(3)Based on the comparative experimental results,The Potential to Clinical Translation scores(PCT)is constructed based on the improved multidimensional characteristic index system(composed of citation indexes,altmetrics indicators and characteristic indexes of biomedical triangle model)and the best-performing CatBoost algorithm.The index has a certain accuracy in identifying the clinical citations of papers,and the value of the index can represent the clinical conversion potential of papers to a certain extent.