Establishment Of ADR Prediction System For Traditional Chinese Medicine Injection Containing Panax Notoginseng Saponins

Objective: Taking traditional Chinese medicine injection containing panax notoginseng saponins as an example,this project explored and established a highly sensitive and highly specific pre-drug adverse reaction prediction system that could be used in clinical practice based on individual characteristics of patients,so as to provide relevant medication suggestions for clinical workers and reduce the risk of adverse reactions.Methods: By combining nested case-control study with prescription sequence analysis,the data of patients with adverse reactions after using Xuesaitong injection and Xueshuantong injection was collected from 5 hospitals in Sichuan Province,including personal information,disease,combined medication and adverse reaction prognosis.Patients without adverse reactions were selected in a matching and random manner according to 1:4,and two data sets with a sample size of 530 were finally formed.Through the traditional mathematical statistics research method,the influence factors related to its adverse reactions were explored.Then,four machine learning algorithms,K nearest neighbor,support vector machine,random forests and gradient boosting decision tree,were used to construct the model and select the classifier with the best performance.Finally,the network prediction platform of adverse drug reactions of xuesaitong and xueshuantong injection was established.Results: 1.The population of adverse reactions was mainly among the middle-aged people aged 45-59 years old with junior high school education and below.The incidence of female was slightly higher than that of male.Among them,the most common diseases in the population were orthopedic diseases,cardiovascular and cerebrovascular diseases and digestive system diseases.Dexamethasone sodium phosphate injection,pantoprazole and amlodipine were frequently used in drug combination.Most of the adverse reactions occurred within 1 day,and the number of people with moderate severity of adverse reactions was the largest.The most adverse reactions involved organs were skin and accessory damage,followed by systemic and respiratory system damage.2.The traditional mathematical statistics were used to study the risk factors affecting adverse drug reactions of panax notoginseng saponins.The study found that the occurrence of adverse reactions may be related to the individual situation of the patient,the disease,the combination of drugs,biochemical examination indicators,etc.3.The Python language was used to build the model in Pycharm.In the random group,the classifier with gradient boosting decision tree combined with the SMOTE oversampling algorithm showed the best performance,with an accuracy rate of 83.65% and an AUC value greater than 90%.Finally,it was applied to Django,a Web application framework,to establish a network prediction platform for adverse drug reactions of traditional Chinese medicine injection containing panax notoginseng saponins.Conclusion: The prediction model established by machine learning can be used to classify the adverse reactions of the patients using panax notoginseng saponins injection.The application of the network platform can easily and quickly provide the prediction results and relevant suggestions for clinical workers,which has a high practical value for identifying and screening high-risk groups of adverse drug reactions.