Data Mining Analysis Of Adverse Reactions Of Shenmai Injection Based On Nested Case Control

With the frequent use of Shenmai injection,adverse drug reaction(ADR)caused by it has increased year by year.Generally,the incidence of ADR is relatively small,and its ratio is very different.Moreover,the composition of proprietary Chinese medicines is complex,causing the safety problems of Shenmai injection to be complex and difficult to analyze.In addition,there are many variables such as disease factors and drug factors,and factors may interact with each other,thus traditional methods based on probability theory and mathematical statistics are difficult to identify key factors.Most existing ADR studies lack control groups,and a single database analysis does not fully explain the law of ADR.Therefore,this study combines the ADR data of the spontaneous reporting system with the patient-rich data of the electronic medical records system to integrate medical expertise knowledge.Incidence of ADR and non-ADR cases was matched 1:2 according to the nested case-control method,and the data mining technology combined with mathematical statistics was used to explore the relationship between diseases of patients,drugs and ADR,and to establish a predictive model of ADR occurrence.The four parts of this study are as follows:(1)Preliminary analysis of the basic situation of the occurrence and prognosis of ADR was carried out through mathematical statistics,and the factors that may affect the occurrence and prognosis of ADR are preliminarily detected.(2)Combined with pharmaceutical expertise,data preprocessing such as transform,fill,and reduce,and feature screening was performed to exclude abnormal data.(3)The important factors affecting the occurrence of ADR were explored by data mining and mathematical statistics,attention was paid to analyzing the drug factors and disease factors that affect the occurrence of ADR,and a predictive model of ADR was established.(4)The important influencing factors of ADR prognosis were explored through data mining combined with mathematical statistics.Results: The overall accuracy of the five models is good(test set 85.71%);AUC(test set 88.4%)is larger;Gini coefficient(test set 0.769)is larger,and the difference between the groups is smaller,and the performance of the ROC evaluation model is reliable.The major factors affecting the occurrence of ADR in Shenmai injection are: total number of drugs,number of diseases,intravenous drip rate,single dose,total days of drug use,amount of vehicle,level of liver function,type of vehicle,heart failure grade and age.The drug factors from large to small are: the total number of drugs,potassiumions affecting drugs,intravenous drip rate,brain-waking drugs,β lactam antibiotics,a single dose,Chinese medicine ingredients,antiemetic drugs,liver protection drugs,total days of medication and solvent.And the most important disease factors are tumors and chest diseases.The influence factors affecting the prognosis of Shenmai ADR from large to small are: number of adverse reactions,change of infusion device,comorbidity index,drug treatment,ADR intervention time,ADR termination time and non-drug treatment.The predictive model predicts that the probability of ADR occurring is a minimum of 35% and a maximum of 87%,the maximum detectable ADR is 66.7% and the correct rate is 76.64%.The above study provided evidence for monitoring of Shenmai injection ADR and rational drug use,and provided new ideas for research patterns about influencing factors and predictive analysis for ADR of traditional Chinese medicine injections.