Trial Banks Affiliates Talk 3398：银行分支机构的审判秀.pptVIP

Outline Background Trial Bank Project DBP Goals Domain Problem Want to advance clinical science and practice evidence-based medicine based on results from scientific experiments, e.g., from randomized controlled trials (RCTs) randomization equalizes known and unknown confounders between intervention arms any observed differences in outcome should be due only to the intervention HIV Trials Two controversial areas of HIV with conflicting RCT evidence Prevention of mother-to-child-transmission (MTCT) Structured Treatment Interruption Very complex heterogenous trials need computational support to visualize, analyze, apply these trials Need Computational KBs of RCTs RCT journal articles not ideal key descriptors of experimental method not always reported not easily computable KB of information on RCT design execution results would allow for better computer support for advancing clinical science and practice Outline Background Trial Bank Project DBP Goals Trial Bank Software Bank-a-Trial web-based program for entering trials into RCT Bank (annotation) clinical descriptions of trial features (slot values) are in UMLS terms RCT Presenter web-based browser of RCT Bank entries SecurePresenter for browsing restricted-access entries Ontology: RCT Schema Models clinical trials instantiated using UMLS terms to describe clinical concepts Frame-based model (in Ocelot) 7 levels deep 188 frames, 601 unique slots average 9.8 slots/frame 13 frames (7%) have multiple parents 187/601 slots (31%) take other frames as values Available at / RCT Presenter Demo RCT Presenter is open access /Presenter/ Bank-a-Trial Demo /BaT/ Outline Background Trial Bank Project DBP Goals DBP Goals Trial Bank instances visualizing collections of related but heterogenous clinical trial instances using sample HIV medical record to select “applicable” MTCT and STI trials in RCT Bank meta-analysis RCT Schema towards a reference ontology for clinical trials Goal 1: Visualization Visualization (co