How a rare disease company improved clinical trial recruitment with natural language processing
Beghou Consulting helps company develop a more data-driven clinical trial strategy for an oncology drug.
Pharmaceutical companies are increasingly leveraging artificial intelligence and machine learning to boost their commercial impact. However, many companies still struggle to design, pilot, and operationalize AI/ML models efficiently. Given the inherent power of AI/ML models, the emergence of digital-first competitors, and, most importantly, the entry into an era of micro launches, companies need to find ways to expedite their deployment of AI/ML.
In this era of micro launches, companies will need to commercialize products quicker and with fewer resources per launch. Deploying AI/ML models will help companies improve their full scope of launch preparedness efforts – from clinical trial planning and execution to commercial strategy. And the stakes are high for emerging pharma companies that bet nearly everything on a successful launch.
Challenge: Overcoming information overload and clutter
One rare disease company sought to improve its recruitment and retention of participants for a clinical trial for an oncology drug. The company needed to recruit and retain the right patients (e.g., those affected by the disease the therapy treated and ideally ones who would remain in the trial through completion).
A field team (medical science liaisons) interfaces with a contract research organization and health care professionals to gain insights that help the company recruit and retain patients for the trial. These various interactions generate notes – some digital, some handwritten, some thorough, some sparce – that the headquarters-based clinical team needs to comb through and blend with data and insights from other sources to generate insights that drive effective action in the clinical trial recruitment process.
The problem was that this information collation and analysis work was manual and took months to complete. And, as with any manual process, the risk of human error was not negligible. Further, the slow process of generating insights limited the agility and effectiveness of the company’s clinical trial recruitment efforts.
How could this company blend data from a variety of sources quickly and accurately and generate meaningful insights that would help it improve the quality of its clinical trials?
Approach: Unlocking insights with natural language processing
Natural language processing is an AI/ML technique that allows companies to sort, analyze and generate insights from qualitative information in a variety of formats. Teams can then use the NLP solution to rigorously analyze data and generate meaningful insights. That is what we did in this case.
We reviewed the company’s array of field notes and identified, smart tagged them using NLP and then categorized them, again using NLP, for further use into:
How could this company blend data from a variety of sources quickly and accurately?
This NLP model comprehended notes in various formats and languages, intelligently interpreting typos, acronyms and indirect references. The NLP model allowed the company to route the right insights to the right stakeholders and trigger timely action related to patient recruitment and retention.
Results: A more precise and effective clinical trial strategy
Within just six weeks, we successfully built and deployed this impactful NLP model, achieving an impressive accuracy rate of 95% compared to the prior manual process. With ongoing use, the model’s accuracy has continued to improve.
The NLP model triggered downstream notifications that stakeholders could act on. For instance, it:
- Provided the company’s CRO with valuable insights into the broader competitive landscape.
- Highlighted cross-trial patient challenges (themes) to principal investigators at different sites.
- Offered recruitment and logistical insights to CRO leads.
These timely and curated insights helped these groups take effective steps to enhance clinical trial recruitment and retention.
This deployment serves as proof that AI/ML projects need not be cumbersome endeavors with extensive timelines. Using a custom NLP model, we helped this rare disease company transform a previously laborious and time-consuming process that took months into a streamlined system capable of generating actionable insights within minutes.