Research On Pediatric Rational Drug Use Leveraging Prescription Big Data

According to the statistics of the World Health Organization(WHO),improper medication has become an important factor affecting the safety of rational medication.With the development of medical informatization process,some medical institutions have been equipped with rational drug decision support software for drug review to effectively improve children’s rational drug use level.However,this kind of software are mostly based on established rules to regulate wrong prescriptions,so it is not suitable for pediatrics clinical environments that require intelligent review.Due to the children’s special characteristic,there are few special medicines for children,and there is still a lack of pediatric medication guidance.Children are more prone to unreasonable medication than adults.In this paper,we integrate pediatric historical medical history big data and external medical field knowledge to mine potential children’s medication experience and rules for children’s rational medication research in a data-driven way.The research contents are elaborated as follows:Firstly,mine medications contained in the pediatric historical medical records to accurately evaluate the risk of pediatric drug interactions.We integrate prescription data from pediatric medical record data with external contraindications information to construct a drug interaction map.Then we apply community detection algorithm Louvain to extract higher level medication knowledge.Finally,we provide the prescription risk warning based on the proposed indicator of drug interaction risk.In this paper,we evaluate our framework on a real-world anonymized prescription dataset collected from a tertiary hospital in Fujian province.The results show that the accuracy of our proposed method in the test prescription set reaches 84.2%,and outperforms other baselines.Secondly,model the complex relationship between medicines and diagnosis based on pediatric historical medical record big data to make reasonable drug recommendations for children.The main steps are as follows:we extract the patient’s physiological feature and diagnose feature from the historical medical record big data,and then concatenate external drug interaction feature as model input.Finally,we use the sequence generation model Transformer to make personalized pediatric rational medication recommendations.In this paper,we leverage the MIMIC_Ⅲ pediatric dataset and the pediatric dataset of a tertiary hospital in Fujian Province to verify the validity of the model.The results show that our method is more effective than other baseline methods in the accuracy of drug recommendation.In this paper,we use big data technology and deep learning framework to mine medication rules and knowledge in pediatrics historical data to make up for the lack of existing rational medication review system.It can provide decision-making support for clinical pharmacy management and patients’ daily drug use,which can help improve the safety of children’s rational drug use while saving manpower and material resources.