ML model saves up to 85% of time for the clinical research team
ML model boosts efficiency of systematic literature review by 85%, so the clinical research team of an F100 biopharma to focus on high-quality research
MQuest
The client’s clinical research team performed the crucial literature reviews related to vaccines research manually.
- The time spent on literature reviews took up to a year from query to manuscript submission and typically over 2000 documents are reviewed
- Abstract/full-text reviews took anywhere from 1-3 months and typically only 0.05-1.0% of reviewed abstracts are relevant to the research question.
- Once a systematic review is completed, the probability of new relevant published articles being added is very high.
The manual process was tedious and took up time that could have been used for higher-quality research work. The new solution would need to address this and also provide an interface to access recently-published articles post the systematic review for up-to-date research.
MSolve
The MResult team formulated an algorithm to expedite the speed of literature reviews.
- Subject Matter Experts (SMEs) provided historical labeled data (i.e. extracts + IDs of relevant abstracts) and other parameters such as population, interventions, comparators, outcomes, study design (PICOS), and inclusion/exclusion criteria
- Machine learning algorithms used this information to rank abstracts and reduce the time spent manually reviewing them
- Features were extracted using tf-idf and cosine similarity
- An interface allowed users to upload and rank new search results on demand so that they could answer research questions more efficiently.
MPact
Automating the systematic literature review process:
- Saved 60-85% of the time taken, increasing efficiency
- Provided the latest research at the team’s fingertips at any point in time
- Freed up time for the team to focus on high-quality research.
