How Literature Review Automation and Efficient Processes Streamline RWE HEOR Studies

In this webinar roundtable, experts from Evidence Partners, IQVIA, Johnson & Johnson, and Yoh for Sanofi reflect on how literature review automation helps solve the daily challenges facing research professionals.

• How DistillerSR streamlines research
• Applications of real world evidence and data
• Literature review challenges in different contexts
• Managing non-published literature
• Importance of system efficiency and performance
• Specific benefits for rare disease research

Patti Peeples, PhD, RPh, kicks off this panel by remarking that more than 3 million scientific articles are published in English every year, a number that is growing by almost 10% annually. Researchers from all fields face the common dilemma of keeping up with and making sense of this high volume of information. Disciplines that deal with the evidentiary basis of pharmaceuticals, diagnostics, and medical devices like health economics and outcomes research (HEOR) are particularly challenged by this, and it is exacerbated by the increasing emphasis on real world evidence (RWE) and real world data (RWD).

“Dealing with this volume of literature, which can be incredibly heterogeneous, is like rowing a leaky boat across a big data sea, and the primary source of that leak is the traditional literature review.”

Manual literature reviews that utilize spreadsheets are time-consuming, difficult, and prone to errors. This panel discussion moderated by Patti features Shirely Sylvestor, MD, MPH, Richa Goyal Rai, MPharm, MBA, Vardhini Ganesh, MS, and Lee-Anne Bourke, BSc, MLIS, who each represent different roles within literature review and evidence management. The goal of this panel is to identify common challenges and offer solutions to conduct better, faster, and more accurate literature reviews.

Lee-Anne first provides background information on DistillerSR, which manages the entire literature review lifecycle from searching to reporting. This system automates the management of literature collection, triage, and assessment using intelligent workflows and artificial intelligence (AI). The workflow is based on the user’s own literature review protocols and methods, and is flexible and modifiable by the user. By removing administrative aspects, the reviewer is able to focus only on critical tasks. Literature review automation manages the growing volume of published literature and reduces the human effort to capture complex data from the literature. DistillerSR is used by a range of industries from pharma to academia, further highlighting its flexibility.

“In every use case, including HEOR, the more layers of automation you build into DistillerSR, the more time you can save while being assured that integrity of your protocol is maintained.”

Shirley, Richa, and Vardhini next provide insights into how they use RWE and RWD in their respective roles. Shirley works to drive policy changes in clinical practice to transform patient lives, and therefore uses RWE not necessarily to comply with regulatory requirements, but to inform decision making as it relates to policies for seeking and receiving care. Richa instead works with secondary data in the form of systematic or targeted reviews that feed into RWE for reimbursement purposes. For Vardhini, a data scientist, RWE means avoiding redundancies in efforts to come up with cutting-edge technologies and drugs for the public as soon as possible.

“COVID-19 has been the best example. Imagine if we had all sat in our nutshells, sitting in our laboratories, doing our own thing and not exchanging information, not tapping into RWD or RWE: we may not have accomplished having the emergency-authorized vaccine.”

Since their RWE use cases are different, major challenges and pain points differ as well. For Richa, a major challenge has been managing the increasing number of articles within the same time frame and at the same level of quality. The biggest challenge for Vardhini has been having clear-cut inclusion and exclusion criteria; focusing on major outcomes and not just throwing a wide net is a particular pain point when done manually, especially when there are hundreds of reviews to read and understand. For both of these challenges, automation is the right approach.

Lee-Anne next comments on the ability to properly capture RWE studies given the limitations in appropriate indexing. It is critical to index and use keywords when searching to balance precision and recall. With DistillerSR, users can create labels and filters and use a hierarchical data extraction process to triage and move data to different buckets in the workflow. At the end of the process, thanks to a strong reporting tool, users will be able to see relationships that might have been otherwise overlooked.

Vardhini, Richa, and Shirley find that a great deal of RWE is located in non-published or gray literature. Vardhini deals extensively with clinical trials, so these types of registries as well as conference abstracts are important; exploring all relevant databases based on the nature of an application is well-formatted in DistillerSR. Richa points out that social media (e.g., patient groups on Facebook) is a vast network for RWE. For Shirley, reading and having access to specific research interventions is important for improving maternal health outcomes. When dealing with policy changes for maternal health equity, it is critical to elevate the voices of women and emphasize that women are not just numbers, but people with stories and experiences.

“The social listening part is actually very critical and we have undertaken various … research projects in which we are using social listening as a way to elevate the woman’s voice as well.”

Shirley also shares an example wherein looking at the differential impact of COVID-19 by sex turned out to not be as simple as initially thought. Early in the pandemic, there was evidence that men had higher mortality, which later evened out. Shirley notes that answering this question involved deep-diving into the literature to find data in obscure supplemental information or footnotes. From the vast amount of initial information they had, Shirley’s team was able to narrow this down to only 35 articles that could help them answer the question.

“We were really looking for that needle in the haystack, and we found that needle in the haystack usually in the supplemental information because … journals usually … limit the amount of information that you can publish, and it’s not common for people to publish the data in a disaggregated way.”

Next, Vardhini comments on options for displaying and reporting data at the end of the extraction process. While it is okay to report data in spreadsheets, a major obstacle is reshaping the data into the outcome that should be visualized. DistillerSR helps with data extraction and offers insights to come up with birds-eye view reporting. Lee-Anne also stresses the importance of thinking about what data points are needed at the end. DistillerSR provides an iterative process that accommodates workflow changes so that nothing is lost or missed, and is also flexible in reporting (i.e., can be granular or broad).

The panelists next speak on system efficiency and performance. Lee-Anne shares that there are a number of case studies available online that attest to the time saved through using DistillerSR, which is accomplished through removing manual administrative work and reducing the chance of error. Richa, for example, started out using DistillerSR’s keyword highlighting feature and has since graduated to the AI functionality. Vardhini also notes that DistillerSR has the capability of using existing forms in different projects, which further speeds up the review process. Shirley stresses that performance and efficiency are key for helping streamline research in a meaningful way, and that features like visualization are an added value.

Lastly, Lee-Anne and Shirley touch on how DistillerSR has specifically benefited the area of rare diseases. Lee-Anne explains that a key component to aiding rare disease research is designing a protocol and workflow for casting a wider net, since researchers do not always have all the information they would like and data are not always clearly reported. Shirley also notes that there may be a lot of failed products or compounds in the rare disease space, so expanding the search to these areas and learning from the failures of others could be important. Furthermore, listening to what patients with rare diseases are saying about their symptoms and how they are receiving care is critical, and looking outside of the box will give tremendous insights.