Literature Screening

When completing a comprehensive literature search, it's common to have retrieved articles that are outside of the project's scope. The purpose of screening is to eliminate material that does not meet your inclusion criteria. The research topic, search strategy, as well as other considerations such as whether or not to include gray literature, will all determine the size of the screening corpus. Screening is performed by a subject matter expert, often by the person who has requested the project. However, multiple screeners are encouraged in an effort to speed up the screening process. Multiple, independent screeners may be warranted depending on the project; this can also help to reduce bias during screening. The screening phase often requires an initial title & abstract screening followed by a full-text review to assess eligibility, however, projects intended to gather all relevant literature on a particular topic may only require a single screening phase followed by coding/organization of the literature.

Pre-screening

Prior to the screening phase, the research librarian will work to ensure that the metadata for each record is accurate and complete. The initial screening phase relies heavily on titles and abstracts. Depending on the size of the screening corpus, this process may take weeks to complete. During this time, our team will ensure that you have a SWIFT Active Screener account (IF applicable to your project - not every project will warrant the use of SWIFT). We encourage project screeners to review the SWIFT video tutorials during this time. When the screening team is ready, the research librarian and screeners will meet to review and customize inclusion/exclusion questions and answers, spotlight terms, project reports, and metadata extraction.

Machine Learning Assisted Screening

Screening can be extremely time-consuming. In an effort to make the screening more efficient, Research Services provides access to SWIFT Active Screener. SWIFT is a web-based application, which integrates machine learning algorithms to save time. Through the use of seed articles and the screener's actions to include or exclude individual articles, SWIFT is able to prioritize articles during the screening process, pushing the most relevant articles to the top of the list. Through the use of the algorithms, SWIFT is also able to provide the screener with a real-time estimate of document recall so that a screener knows when they can stop screening. The visual dashboard provides screeners with information on the progress of the screening.