At Evidence Partners, we are always striving to help our users work more efficiently. Today’s researcher is continuously bombarded with new data and publications. This torrent of information has reinforced the importance of automation in our review processes. There simply isn’t enough time or resources to continue doing systematic reviews without automating as much as possible.
Fortunately, automation is not just the future - it’s here now, with the newest DistillerSR updates taking a significant step towards building a more efficient, automated systematic review process.
Continuous AI Reprioritization
The goal when screening is to find all the relevant references to your review and move them to the next stage of your process. Continuous AI Reprioritization uses machine learning to learn from the references you are including and excluding and automatically reorder the ones you have left to screen, putting the most pertinent references in front of you first. This means that you find included references MUCH more quickly.
Check out this simulation of how many references these reviewers DIDN’T have to screen in order to find 95% of their included references, using Continuous AI Reprioritization.
The diagonal line denotes the traditional screening method.
AI Audit and AI Simulation
AI Audit uses machine learning to identify and highlight references that may have been excluded accidentally. While this functionality existed in the previous version of DistillerSR, the AI engine that powers it has been completely updated.
AI Audit is an extra set of eyes for your screening process and can greatly reduce false excludes, especially when only one human screener is used.
AI Simulation uses machine learning for two things. First, it can be used retrospectively on projects to identify how much time you would have saved during screening if you had used AI Reprioritization. The intent here is that you can run it against your existing project to determine how much AI Reprioritization would have helped.
AI Simulation also helps you find references that may have been accidentally included by highlighting the included references in which it is least confident.
Do you find yourself answering the same question, like “Is this a randomized controlled trial” or ”Is this an adverse event?”, many times in a review or across multiple reviews?
DAISY classifiers are AI classifiers that can be trained to answer these questions for you, simply by providing them with sets of references for which these questions have already been answered. Once trained, DAISY classifiers can be applied to your screening or assessment forms to answer their targeted questions automatically. This can completely automate some aspects of your screening or reference coding process.
Unlike many AI classifier solutions, DAISY classifiers can be created, tested for accuracy and then put to use answering questions quickly and easily, without extensive user training.
Reliable, fast deduplication is important in any review but it is essential in continuously updated or living reviews.
DistillerSR’s Natural Language Processing (NLP) powered Duplicate Detection Engine now has a Smart Quarantine feature, allowing users to specify thresholds and requirements for which references from duplicate pairs get quarantined automatically. Smart Quarantine enables users to bulk quarantine duplicate references with the confidence that they are in fact duplicates, and automatically quarantine the appropriate versions of the reference. This saves time having to manually go through and select which reference is the appropriate one to quarantine, and to visually determine if they are in fact duplicates.
Deduplication now also allows you to display Keyword Highlighting when comparing references to more easily determine if they are duplicates.
Automatic Reference Import
Keeping a literature review up-to-date is time-consuming, especially today when new research is continually published.
If you are an Ovid user, you know that Ovid Alerts can be set up to automatically email you with new references matching a saved search. DistillerSR takes it one step further by allowing those alerts to be sent directly to your DistillerSR project automatically. Additionally, when new references are added, project admins will receive a notification and reviewers will receive a notification that they have new references to review.
Are you starting to see a trend here?
Full-Text Retrieval and DOI Link
One of the easiest ways to save time and money in a literature review is to reduce friction in the process by connecting the tools in your arsenal. Our new Full-Text Retrieval features do just that by enabling users to connect their various reference procurement accounts to DistillerSR. If you have a Copyright Clearance Center (CCC) or Article Galaxy account, you can now access and procure your full text documents directly from DistillerSR. With these connections activated, users can quickly find, download, and purchase the full-text version of a reference from these third parties.
Even if you don’t have accounts with these services, DisitllerSR will provide your reviewers with a clickable link to the publisher’s reference page for each reference that has a DOI field in the reference.
Automation is the future of systematic reviews, and right now we’re in an exciting time where we have stopped asking “what if?” and starting asking “when?” The latest features included in this release are exciting stepping stones to furthering the use and acceptance of automation and AI in systematic reviews. We’ve already seen the proof that these features work with our COVID-19 initiatives, which makes it even more exciting to think about what the NEXT release will hold.
As always, our updates are inspired by our users and we will continue to create features designed to help researchers maximize their time and resources. What do you think? Which parts of your systematic review process are you most excited to automate?