OP129 The Use Of A Text-Mining Screening Tool For Systematic Review Of Treatments For Relapsed/Refractory Diffuse Large B-Cell Lymphoma

Introduction Human screening of title and abstracts in a systematic literature review (SLR) is labor intensive time-consuming. In many instances, thousands citations may be retrieved; the vast majority excluded upon screening. Text-mining semi-automates accelerates by identifying patterns relevant irrelevant citations, as labelled screener. One such text-mining tool, Abstrackr, uses an algorithm within active-learning framework to predict likelihood being relevant. The objective this study was assesses performance Abstrackr for abstract SLR treatments relapsed/refractory diffuse large B-cell lymphoma. Methods Citations identified from searches electronic databases were imported Abstrackr. An investigator-selected database terms indicating relevance research question uploaded. These partly informed inclusion/exclusion criteria. deemed most screened first (screening prioritization). Screening carried out until maximum prediction score 0.4 or less, based on previous experience literature, reached. Remaining unlikely did not undergo truncation). Separately, single-human screener all using Covidence. Results A total 7,723 154 initial uploaded Of these 2,572 (33 percent) before 0.39 Compared (conducted citations), workload saving associated with 5 days. 451 (6 proceeded full-text screening; ten (0.1 included final evidence base. No predicted Conclusions tools have potential reduce screening, without missing citations.