Research And Implementation Of Adverse Drug Effect Signal Detection Technology Based On Clinical Case Base

The generation of massive data causes a big interest in secondary use of clinical data on different purposes, including medical research, education and management, which has become a hot biomedical informatics research. Meanwhile, as patient safety receives more and more concerns, Spontaneous Reporting Systems (SRS) are built by many countries and medical organizations, and computerized methods are developed for knowledge discovery in database, i.e. ADE signal detection. However, studies have shown that SRS suffers from underreporting, especially for those mild and common symptoms, which has bad influence on the efficiency of signal detection. Therefore, it is significant to detect ADE signal directly from clinical data. This thesis researched on the ADE signal detection technology based on clinical case base and developed a prototype accordingly.First, we constructed a case base for ADE signal detection study. The modeling solution for Clinical Data Warehouse (CDW) is adopted and a case base to support the ADE signal detection study is built. The case base is designed not only to support researches in ADE signal detection but also to achieve the capacity to other secondary use study applications by equipped with Clinical Data Warehouse features.23,898 cases with the information of medication orders and symptoms have been imported, and Natural Language Process (NLP) is used to extract symptoms from real medical documentations.Second, we conducted an ADE signal detection study based on the case base and developed configurable web tools to detect ADE signal. An algorithm is proposed to generate Drug-Event Combination (DEC) from clinical data. The Disproportional Analysis (DPA), which is simple but proved effective, is performed to detect signal from DECs. And a visualization tool with highly intensive information is provided. The mining platform allows configuration on several parameters, so the experiments on different interests and purposes can be conducted. Finally, the results of signal detection data analysis and evaluation are given.According to the status of secondary use of clinical data and ADE signal detection, the thesis proposed a set of ADE signal detection method from real clinical data, constructed a research-oriented case base and developed a pharmacovigilance tool for ADE signal detection. It is a useful practice for clinical data secondary use, which is significant for related research.