Imagine 25 years ago. You are working on a systematic review. You are making hard copies of reports and studies, manually highlighting the relevant information, filing included studies in a cabinet and moving the excluded studies elsewhere. Your eyes are strained, you’ve been working on this for months at this point, and you don’t see an end in sight.
Fast-forward to 15 years ago. You are using a spreadsheet on your computer to manually input information. It’s still time-consuming, but at least you have freed up some space and eliminated most of the towering paper stacks in your office. But there’s still a problem. Your team isn’t communicating the way you wish they would, the spreadsheet isn’t fool-proof, and there are still many possibilities for errors.
Jump ahead to just five years ago. You left the spreadsheet behind and are now using reference management software or specialized systematic review software to complete your reviews. It’s faster, more efficient, with numerous features to help mitigate errors and facilitate a credible review. It’s still not perfect. You’re facing a huge demand for new evidence in your field, and the process could still stand to be faster and more efficient.
But the next wave is coming. What if I told you that the next innovation in systematic review is closer than 30, 10, or even 5 years away? It’s happening right now.
Technology is advancing at speeds we have never seen before. The next generation of systematic reviews are using features like automation and futuristic AI machine-learning to facilitate faster and more accurate reviews.
Driving Factors for Change
Automation in a systematic review is a product of necessity. Without the ability to produce evidence-based research quickly, many products and innovations would be slow to market, and we would struggle to keep up with the demands of up-to-date evidence. We need solutions to make systematic reviews faster and more efficient. Automation and artificial intelligence (AI) hold many of the answers. Some of the main driving factors for the development of systematic review automation technology include:
- The extremely cumbersome, manual process of current SR’s.
- The need to produce evidence faster.
- Insufficient reviewers to keep up with the demand of up-to-date evidence and the onslaught of new publications.
- Manual data management produces errors that are often difficult to detect.
- Project oversight on multi-participant reviews that is inefficient, difficult, and error-prone.
- The need to produce transparent, auditable, and reproducible results.
How is Automation Applied to SR Software Today?
Today’s systematic review software evolved from the need to automate manual, time-intensive tasks. Previously, the tools available to support efficient and credible reviews were seriously lacking. Unsophisticated instruments that are inefficient and prone to error generation are no longer acceptable for today’s research team. They need systematic literature review tools that will positively influence their workflow, resulting in improved efficiency and lower costs for their systematic review.
However, we aren’t yet at the place where reviews can be completed by a robot or artificial intelligence without human help. Features such as automatic inclusion/exclusion tracking as well as flagging and removing duplicate studies are just a couple of examples of how automation speeds up time-intensive tasks. Although today’s systematic reviews are supported by automation and AI, they are still very much completed by real people.
The Future: Advanced AI
AI is making huge technological strides today. The systematic review community will soon reap the benefits of using machine-learning features to further facilitate and improve the systematic review process.
AI review solutions use “classifiers” to answer questions and to find similarities or differences between references. Classifiers are statistical models that rely on natural language processing (NLP) to answer closed-ended (multiple choice or check all that apply) questions about references automatically and then group, or classify results accordingly.
So how can AI improve the efficiency of your systematic reviews? Here are a few ways that artificial intelligence can help:
- Enable users to create and train their own custom AI classifiers using screening data collected by humans during the review process.
- Use AI classifiers to automatically answer closed-ended form questions using the same process a human would use, with a complete audit trail.
- Automatically test precision and recall of your custom AI classifiers so you can confirm accuracy.
- Act as a second screener or QC reviewer to compare against human results, further improving efficiency and making it easier to do more with fewer resources.
So where do you stand when it comes to using technology in your systematic reviews? Take a look at the timeline below:
If your organization is not leveraging automation and machine learning in your systematic reviews, you risk falling behind. It’s time to upgrade your processes to stay ahead of evolving best practices and new regulations. Discover how DistillerSR can help you complete better, faster systematic reviews. Request a demo today.
- AI and Systematic Reviews: What's Real and What's Not
- Here's Why You Should Stop Relying on Spreadsheets for Your Literature Review Process
- No Second Screener? There's a Robot for That