Dataset Information

Semi-automated screening of biomedical citations for systematic reviews.

ABSTRACT:

Background

Systematic reviews address a specific clinical question by unbiasedly assessing and analyzing the pertinent literature. Citation screening is a time-consuming and critical step in systematic reviews. Typically, reviewers must evaluate thousands of citations to identify articles eligible for a given review. We explore the application of machine learning techniques to semi-automate citation screening, thereby reducing the reviewers' workload.

Results

We present a novel online classification strategy for citation screening to automatically discriminate "relevant" from "irrelevant" citations. We use an ensemble of Support Vector Machines (SVMs) built over different feature-spaces (e.g., abstract and title text), and trained interactively by the reviewer(s). Semi-automating the citation screening process is difficult because any such strategy must identify all citations eligible for the systematic review. This requirement is made harder still due to class imbalance; there are far fewer "relevant" than "irrelevant" citations for any given systematic review. To address these challenges we employ a custom active-learning strategy developed specifically for imbalanced datasets. Further, we introduce a novel undersampling technique. We provide experimental results over three real-world systematic review datasets, and demonstrate that our algorithm is able to reduce the number of citations that must be screened manually by nearly half in two of these, and by around 40% in the third, without excluding any of the citations eligible for the systematic review.

Conclusions

We have developed a semi-automated citation screening algorithm for systematic reviews that has the potential to substantially reduce the number of citations reviewers have to manually screen, without compromising the quality and comprehensiveness of the review.

SUBMITTER: Wallace BC

PROVIDER: S-EPMC2824679 | biostudies-literature | 2010 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Semi-automated screening of biomedical citations for systematic reviews.

Wallace Byron C BC Trikalinos Thomas A TA Lau Joseph J Brodley Carla C Schmid Christopher H CH

BMC bioinformatics 20100126

<h4>Background</h4>Systematic reviews address a specific clinical question by unbiasedly assessing and analyzing the pertinent literature. Citation screening is a time-consuming and critical step in systematic reviews. Typically, reviewers must evaluate thousands of citations to identify articles eligible for a given review. We explore the application of machine learning techniques to semi-automate citation screening, thereby reducing the reviewers' workload.<h4>Results</h4>We present a novel on ...[more]

PMID: 20102628

Similar Datasets

Project description:BackgroundWe evaluated the benefits and risks of using the Abstrackr machine learning (ML) tool to semi-automate title-abstract screening and explored whether Abstrackr's predictions varied by review or study-level characteristics.MethodsFor a convenience sample of 16 reviews for which adequate data were available to address our objectives (11 systematic reviews and 5 rapid reviews), we screened a 200-record training set in Abstrackr and downloaded the relevance (relevant or irrelevant) of the remaining records, as predicted by the tool. We retrospectively simulated the liberal-accelerated screening approach. We estimated the time savings and proportion missed compared with dual independent screening. For reviews with pairwise meta-analyses, we evaluated changes to the pooled effects after removing the missed studies. We explored whether the tool's predictions varied by review and study-level characteristics.ResultsUsing the ML-assisted liberal-accelerated approach, we wrongly excluded 0 to 3 (0 to 14%) records that were included in the final reports, but saved a median (IQR) 26 (9, 42) h of screening time. One missed study was included in eight pairwise meta-analyses in one systematic review. The pooled effect for just one of those meta-analyses changed considerably (from MD (95% CI) - 1.53 (- 2.92, - 0.15) to - 1.17 (- 2.70, 0.36)). Of 802 records in the final reports, 87% were correctly predicted as relevant. The correctness of the predictions did not differ by review (systematic or rapid, P = 0.37) or intervention type (simple or complex, P = 0.47). The predictions were more often correct in reviews with multiple (89%) vs. single (83%) research questions (P = 0.01), or that included only trials (95%) vs. multiple designs (86%) (P = 0.003). At the study level, trials (91%), mixed methods (100%), and qualitative (93%) studies were more often correctly predicted as relevant compared with observational studies (79%) or reviews (83%) (P = 0.0006). Studies at high or unclear (88%) vs. low risk of bias (80%) (P = 0.039), and those published more recently (mean (SD) 2008 (7) vs. 2006 (10), P = 0.02) were more often correctly predicted as relevant.ConclusionOur screening approach saved time and may be suitable in conditions where the limited risk of missing relevant records is acceptable. Several of our findings are paradoxical and require further study to fully understand the tasks to which ML-assisted screening is best suited. The findings should be interpreted in light of the fact that the protocol was prepared for the funder, but not published a priori. Because we used a convenience sample, the findings may be prone to selection bias. The results may not be generalizable to other samples of reviews, ML tools, or screening approaches. The small number of missed studies across reviews with pairwise meta-analyses hindered strong conclusions about the effect of missed studies on the results and conclusions of systematic reviews.

Project description:The vast collection of biomedical literature and its continued expansion has presented a number of challenges to researchers who require structured findings to stay abreast of and analyze molecular mechanisms relevant to their domain of interest. By structuring literature content into topic-specific machine-readable databases, the aggregate data from multiple articles can be used to infer trends that can be compared and contrasted with similar findings from topic-independent resources. Our study presents a generalized procedure for semi-automatically creating a custom topic-specific molecular interaction database through the use of text mining to assist manual curation. We apply the procedure to capture molecular events that underlie 'pain', a complex phenomenon with a large societal burden and unmet medical need. We describe how existing text mining solutions are used to build a pain-specific corpus, extract molecular events from it, add context to the extracted events and assess their relevance. The pain-specific corpus contains 765 692 documents from Medline and PubMed Central, from which we extracted 356 499 unique normalized molecular events, with 261 438 single protein events and 93 271 molecular interactions supplied by BioContext. Event chains are annotated with negation, speculation, anatomy, Gene Ontology terms, mutations, pain and disease relevance, which collectively provide detailed insight into how that event chain is associated with pain. The extracted relations are visualized in a wiki platform (wiki-pain.org) that enables efficient manual curation and exploration of the molecular mechanisms that underlie pain. Curation of 1500 grouped event chains ranked by pain relevance revealed 613 accurately extracted unique molecular interactions that in the future can be used to study the underlying mechanisms involved in pain. Our approach demonstrates that combining existing text mining tools with domain-specific terms and wiki-based visualization can facilitate rapid curation of molecular interactions to create a custom database. Database URL: •••

Project description:BACKGROUND:The usefulness of Google Scholar (GS) as a bibliographic database for biomedical systematic review (SR) searching is a subject of current interest and debate in research circles. Recent research has suggested GS might even be used alone in SR searching. This assertion is challenged here by testing whether GS can locate all studies included in 21 previously published SRs. Second, it examines the recall of GS, taking into account the maximum number of items that can be viewed, and tests whether more complete searches created by an information specialist will improve recall compared to the searches used in the 21 published SRs. METHODS:The authors identified 21 biomedical SRs that had used GS and PubMed as information sources and reported their use of identical, reproducible search strategies in both databases. These search strategies were rerun in GS and PubMed, and analyzed as to their coverage and recall. Efforts were made to improve searches that underperformed in each database. RESULTS:GS' overall coverage was higher than PubMed (98% versus 91%) and overall recall is higher in GS: 80% of the references included in the 21 SRs were returned by the original searches in GS versus 68% in PubMed. Only 72% of the included references could be used as they were listed among the first 1,000 hits (the maximum number shown). Practical precision (the number of included references retrieved in the first 1,000, divided by 1,000) was on average 1.9%, which is only slightly lower than in other published SRs. Improving searches with the lowest recall resulted in an increase in recall from 48% to 66% in GS and, in PubMed, from 60% to 85%. CONCLUSIONS:Although its coverage and precision are acceptable, GS, because of its incomplete recall, should not be used as a single source in SR searching. A specialized, curated medical database such as PubMed provides experienced searchers with tools and functionality that help improve recall, and numerous options in order to optimize precision. Searches for SRs should be performed by experienced searchers creating searches that maximize recall for as many databases as deemed necessary by the search expert.

Dataset Information

Semi-automated screening of biomedical citations for systematic reviews.

Background

Results

Conclusions

Publications

Semi-automated screening of biomedical citations for systematic reviews.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets