Epidemiological Analysis On Adverse Events Following Immunization Of Guangdong 2005-2016 Based On The Application Of Data Mining

ObjectsImmunization play an important role in preventing infectious diseases.identifying rare and delayed responses is essential to reduce adverse reactions of vaccination,build parental confidence,and maintain a high level of vaccination.Since May 2005,Guangdong Province has established a surveillance system in accordance with the WHO guidelines,system detected Adverse Events Following immunization(AEFI).This study explored data mining technology to study adverse reactions of Gutangdong province from 2005 to 2016,to understand the epidemiological characteristics of adverse reactions to immunization in Guangdong Province,and improving the immunization system with scientific evidence.MethodsFour Date Manning methods are used to detected suspected signals of AEFI of Gutangdong province from 2005 to 2016,including the proportional reporting ratio(PRR),the reporting odds ratio(ROR),the empirical Bayes Gamma-Poisson Shrinker(GPS),and the Bayesian confidence propagation neural network(BCPNN),and feedback Data Manning methods used to AEFI surveillance.Results1.The rates and distribution of AEFI of Guangdong province from 2005 to 2016.In 12 years,a total of 54 092 AEFI cases were reported in Guangdong,with an annual increase of 26.82%,All 126 counties(100%)reported AEFIs;95.64%of cases were reported within 48h,and 98.80%were investigated within 48h.The ratios of cases for the Pearl Delta Region(PDR)and NonPDR were 1.95:1,and 1.48:1 for males and females.Cases?1 year old accounted for 52.05%;85.49%of cases occurred Id after vaccination.The AEFI incidence rates in 2005-2016 ranged from 15.77 to 193.10 per million doses;the incidence rates for rare vaccine reactions ranged from 5.40 to 42.20 per million doses.2.Characteristics of AEFI.The most common vaccine reactions were fever,local swelling and induration;the most common rare vaccine reactions were anaphylactic rash and sterile abscess.The incidence of serous rare vaccine reactions was very low,including Febrile Convulsion,Thrombocytopenic Purpura,Henoch-SchOnlein Purpura and Anaphylactic Shock.The reported incidence of non-severe AEFI was 10.73/1 million doses(Rubella Attenuated Live Vaccine)-856.18/1 million doses(7-valent Pneumococcal Polysaccharide Conjugate Vaccine);the reported incidence of severe AEFI was 0.00/1 million doses(Hepatitis A and B Combined Vaccine)-13.27/1 million doses(23-valent Pneumococcal Polysaccharide Vaccine);and the Common Vaccine Reaction response rate was 2.93/1 million doses(Rubella Attenuated Live Vaccine)-588.47/1 million doses(Diphtheria,Tetanus and Acellular Pertussis,Inactivated Poliomyelitis and Haemophilus Influenzae Type b Conjugate Combined Vaccine);the incidence of rare vaccine reactions was 5.85/1 million doses(Rubella Attenuated Live Vaccine)to 299.15/1 million doses(7-valent Pneumococcal Polysaccharide Conjugate Vaccine).3.Four DPAs,ROR,PRR,BCPNN and GPS,were used to analyze the rare vaccine reactions in 13089 cases.43,43,29 and 20 signals were found,and 27,27,27 and 11 signals were found with serious rare vaccine reactions.Hepatitis A Attenuated Live Vaccine(HepA-L)detected Anaphylactic Shock and Laryngeal Edema with high value,the number of cases were 14,15;Varicella Attenuated Live Vaccine(VarV)were found Anaphylactic Shock and Laryngeal Edema with high value,the number of cases were 14,12.Abnormal safety signals of HepA-L and VarV varicella vaccine were proven by other studies.4.In this study,We used the specific vaccine rare reaction more than 3 times baseline as golden standard,which the incidence of all vaccine rare reactions as the baseline,analysis of Receive Operator Characteristic curve(ROC)Curves of four Data Mining Technologies.The sample size of PRR were 43 cases,and AUC(Area under the ROC,AUC)=0.88(95%CI:0.75-0.96,P<0.001),ROC curve had statistical significance;The maximum value of Youden index was 0.71,the best cut-off value was>1.88(sensitivity=88.46%,specificity=82.35%).ROR sample size were 43 cases,AUC=0.88(95%CI=0.74-0.96,P<0.001),ROC curve had statistical significance;The maximum value of Youden index was 0.71,the optimal cut-off value was>1.89(sensitivity=88.46%,specificity=:82.35%).BCPNN,sample size were 29 cases,and AUC=0.80(95%CI:0.61-0.92,P=0.002).ROC curve had statistical significance.The maximum value of Youden index was 0.69,the optimal cut-off value was>1.57(sensitivity=89.47%,specificity=80.0%).GPS,sample size were 20 cases,and AUC=0.750(95%CI:0.51-0.91,P=0.062),ROC curve had no statistical significance.The maximum value of Youden index was 0.56,and the optimal cut-off value was>3.01(sensitivity = 56.25%,specificity = 100%).ConclusionsData mining technology can be used to detect abnormal vaccine safety signals in surveillance system,which PRR,ROR,BCPNN could detect signals of AEFI in high validity.Guangdong AEFI surveillance was sensitivity.The incidence rates of rare vaccine reactions were not higher than the expected,AEFI mainly occurs in the younger age group,incidence rates of multivalent vaccines.