Applications Of Big Data Technology In The Rational Use Of Pediatric Medicines

Rational use of medicines requires that patients receive medications appropriate to their clinical needs,in doses that meet their own individual requirements,for an adequate period of time,and at the lowest cost to them and their community.As a special group whose organs and physiological functions in all aspects of the body have not yet being matured,children's capability of drug metabolism is relatively fragile,and they are prone to adverse drug reactions(ADR).Therefore,special attention should be paid to the rational use of children's medicines.However,currently,the problems of irrational use of medicines in pediatric clinical medication are still prominent.On one hand,the proper varieties,dosage forms,and specifications for children's medicines are still insufficient compared to adults' medicines.On the other hand,non-standard medication approaches,such as unlicensed and off-label use of medicines,are still widespread in clinical medication practice.These phenomena have brought potential risks to the pediatric medication.With the rapid development and deployment of hospital information systems(HIS),some hospitals have started to use clinical medication decision support softwares(CMDS)to help review the pediatric medication to reduce medication errors.These softwares are usually built upon massive clinical medication information collected from medicine instructions and guides,and they use complicated rules for medication review and risk warning.However,due to the insufficiency of pediatric clinical medication resources,the accuracy,reliability,and comprehension of the collected pediatric medication information are difficult to guarantee.Furthermore,due to the lack of intelligent and flexible review rules,these systems usually produce a large number of invalid risk warnings that are ignored by medical staffs during clinical practice,lowering the effectiveness of these softwares with regard to medication decision support.Objective:This research aims to find the problems of existing CMDS softwares in pediatric clinical medication practice,by conducting a case study on one of the CMDS softwares in a children's hospital.Upon this basis,this research aims to explore the applications of big data technologies on the rational use of pediatric medicines,and to propose a prescription-big-data-driven evaluation methodology for the rational use of pediatric medicines,providing the decision support for pediatric medication.Methods:A prescription-big-data-driven evaluation platform for the rational use of pediatric medicines was proposed.By analyzing the massive prescription data leveraging big data technologies,a pediatric medication knowledge base was built automatically,and the clinical medication patterns were learned by the platform.A pediatric medicine dosage modeling and risk evaluation application and a pediatric medicine concomitant modeling and risk evaluation application were proposed to provide medication decision supports for both doctors and pharmacists.The pediatric medicine dosage modeling and risk evaluation application was built upon the massive pediatric prescription data collected from two tertiary hospitals.By leveraging the Apache Spark distributed computing framework,the medicine dosage records were extracted,as well as the corresponding impacting factors including the patients' physiological information,the types of indications,and the concomitant medicines.By leveraging neural networks and ensemble learning techniques,a hybrid prediction model of the medicine dosage against the above-mentioned factors was built using the machine learning library MLlib.Based upon the model,a risk evaluation method for the medicine dosages in new prescriptions was proposed to provide decision support for doctors and pharmacists.The pediatric concomitant medicine modeling and risk evaluation application was also built upon the massive pediatric prescription data collected from two tertiary hospitals.The medicine frequency and concomitant medicine records were extracted by leveraging the Apache Spark distributed computing framework.By leveraging graph modeling techniques,a concomitant medicine graph model was built using the graph library GraphX.Based upon the model,a risk evaluation method for the concomitant medicines in new prescriptions using sub-graph searching algorithm was proposed to provide decision support for doctors and pharmacists.Results:After data cleansing for the massive pediatric prescription data collected from two tertiary hospitals,254 824 medication records and 4 508 331 medication records were obtained,respectively.For pediatric medicine dosage modeling and risk evaluation,the proposed model achieved a prediction error of 12.5%ad 21.3%,respectively,which demonstrated the model's ability to predict medication dosage for children of different age and weight groups,different indications,and different concomitant medicines.The risk evaluation on problematic prescriptions achieved a recall of 96.3%,indicating the model's sensibility to identify dosage risks.For pediatric concomitant medicine modeling and risk evaluation,two concomitant medicine graphs containing 376 nodes and 460 nodes were constructed,respectively.The evaluation results on prescriptions with concomitant medicines achieved an accuracy of 99.8%,demonstrating the model's ability to accurately evaluate the risks of pediatric concomitant medicines.Conclusion:The prescription big data cover the detailed records of patients'diagnosis and treatment process in hospital,containing the precious clinical medication experiences of medical staffs.These data are of great importance for understanding,predicting,and evaluating the clinical medication patterns and risks in children's hospitals.By leveraging big data and artificial intelligence algorithms,various impacting factors related to rational use of medicines were extracted,the pediatric medication knowledge bases were built,and the patterns of clinical medication dosages and concomitant medicines were learned automatically,enabling an intelligent,flexible,and personalized platform for evaluating the rational use of medicines.Two applications based upon the platform,including pediatric medication dosage modeling and risk evaluation,and pediatric concomitant medicine modeling and risk evaluation,were proposed and evaluated,providing decision support for doctors and pharmacists.