Research And Implementation Of Personal Health Records Fusion Method Based On Clinical Data

With the development and improvement of hospital information system,people are no longer satisfied with the operations to the clinical data such as storing and inquiring in the hospital information system.It is hoped that clinical data can not only describe the business logic of the treatment process,but also can analyze the association between them.For example,the same patient goes to the same or different hospitals for treatment at different times,then his clinical data will be distributed in different clinical systems,this will cause patient's history clinical records incomplete and personal health records unable to establish.These problems can not be solved by hospital information systems now.Aiming to the real requirements of hospital,we proposed a personal health records fusion method based on clinical data.Two fusion methods were designed for homogeneous and heterogeneous databases.The purposes of the two fusion methods are both to find different clinical records of the same patient and merge these clinical records to form the personal health records of the patient.Personal health records can help patients to get on disease warning and prevention,they can also help doctors to give personalized treatment to patients and provide clinical assistance.The main contribution of this paper is as followed:1.We designed a personal health records fusion framework based on clinical data.This framework includes two parts,one is the personal health records fusion method of homogeneous databases based on clustering.In this method,we extract the key attributes of patients in the database,and use clustering and instance matching method to identify different clinical records of the same patient,then merge these records to form the health records of the patient.The other is the personal health records fusion method of heterogeneous databases based on linked data.In this method,we converse the clinical data in the relational database into linked data with the characteristics of linked data andfind the different resources representing the same patient through similarity calculation.Based on this,we merge different clinical records of the same patient to form the personal health records.2.We proposed a clinical data clustering algorithm.In this paper,we cluster the clinical records in the homogeneous databases through improved K-Means clustering algorithm.We extract the attributes such as name and age of the patient and establish multi-dimensional coordinates to represent clinical records.At the same time,we change the selection method of the initial cluster centers in traditional K-Means algorithm to improve the accuracy of clustering and design the object distance formula to divide clusters.3.We proposed a recognition algorithm of clinical data based on linked data.In this paper,we calculate the similarities of diseases and drugs with external linked data sources and give the method of attribute weight adjusting dynamically.We also design the similarity calculation formula based on linked data to determine whether the two resources are representing the same patient.4.We implemented a prototype system to validate the methods of this paper.We establish a prototype system to accomplish the personal health records fusion method based on the clinical data in different hospitals with Openlink Virtuoso database.At the same time,we compare our system with other clinical analysis systems.It is proofed that the proposed personal health records fusion method of this paper has high feasibility and usability.