Introduction
As a technology company with deep roots in the machine learning and AI space, we’ve seen how AI applied correctly can transform the business process of an industry. Over the last few years, we’ve started applying machine learning to many of the facets of Pharmaceutical Medical Affairs & Commercial functions with the goal of improving efficiency, insights, and patient outcomes. In short, AI for Medical Affairs involves collecting and digitally analyzing relevant content, looking for the patterns and trends that drive the business.
Our experiences across the different facets of Medical Affairs has led to the creation of this white paper on how to apply AI to Med Affairs without over-reaching or looking for a magic bullet that will solve every problem in Medical Affairs.
In this paper we’ll look at a number of specific examples of problems and solutions; some of them targeted at improving the competitiveness of a drug, while others focus on enabling the department to function more efficiently and achieve improvements in performance. The point of examining these cases is to look for the overarching patterns that can be reused throughout these projects to improve the consistency of the results and save money in implementation.
The Role & Responsibility of Medical Affairs
It’s likely that most people reading this primer are aware of the role of Medical Affairs within a pharmaceutical company, but for any that don’t: Medical Affairs is basically the bridge between the drug development / research functions and the marketing, sales and medical professionals that utilize the drugs.
One of the best descriptions we’ve come across is from TriNet Pharma: “They synthesize information and translate findings about therapeutics /drugs into language that can be better understood by company staff whose primary expertise is not scientific. Additionally, Medical Affairs typically operates an information center to respond
to unsolicited product inquiries from healthcare professionals. The information center is especially busy around the time of a market launch.”
AI-driven tools are helping Medical Affairs teams excel in:
- Fielding Medical Information Requests from Medical Professionals
- Medical Congress Insights Analysis
- Scientific Congress & Publication Strategy
- Advisory Board Insights
- Key Opinion Leader Engagement
- And more…
All of these areas of responsibility as well as others can be aided by the judicious use of AI to help a MedAffairs team in understanding how to message their drugs, monitor the competitors, and meet regulatory requirements. Let’s examine four specific examples to see exactly how NLP and AI can be leveraged for Medical Affairs problems.
Text Analytics & NLP
Text Analytics forms the foundation of numerous Natural Language Processing (NLP) features, including named entity recognition, categorization, and sentiment analysis.
In broad terms, these NLP features aim to answer four questions:
- Who is talking?
- What are they talking about?
- What are they saying about those subjects?
- How do they feel?
Medical Information AI in Japanese
All Medical Affairs teams will be familiar with the difficulties of managing the information used to respond to medical information inquiries. Information changes, it grows week by week, and needs to be pulled from multiple sources. But by combining machine learning and keyword search, it is possible to make the process of managing medical information more efficient. This is demonstrated by an example from Japan, where most MedInfo requests are made over the phone instead of electronically, and while the goal is still to return the appropriate answer from the FAQs or published literature the business problem is different.
If a question takes longer than 1 minute to answer then it is elevated to a Medical Professional (MP), thereby dramatically increasing the cost of the
MedInfo request.
By combining machine learning and keyword search in a unified platform, we were able to build a system that reduces average call time without sacrificing quality of responses. Imagine a search interface with an AI model behind it that processes the question against the FAQs and other literature to provide the best answers at the top of the results list.
The system recognizes the absolute need for humans in the loop due to regulatory requirements,
but makes these humans faster and more accurate which reduces the need to elevate requests to an
MP which significantly reduces cost.
- Lexalytics designed a medical information system that reduced average call time without sacrificing quality of responses.
Compendia Analysis in Oncology
NCCN is a US-based non-profit organization that distributes Compendia, the standard practices of which drugs to use to treat which cancers, in what combinations and in what circumstances (e.g. only use this one if this other class isn’t effective anymore, or use this one in conjunction with that one).
Physicians and HCPs use Compendia for treatment decisions, insurers use it for approving prescriptions, so it has a major impact on how often a drug is used. Because it’s so important, many companies have a small team that manually compares each version of the Compendia as it comes out, but you can imagine how difficult a process a manual review of these updates is.
We worked with a large pharma company to build a tool that shows what has changed from version to version, highlighting the relevant text, and allowing them to annotate key findings from the Compendia. The technology behind this project is a relatively simple differencing model, but the business value is immense.
Content Aggregation
Another common problem that occurs in most larger pharmaceutical companies is the siloing of data and inconsistent tagging of content for a given medical condition. This means fewer people benefit from collected insights and the organization can’t fully leverage the value of those insights. This is less an AI problem than it is a business process problem.
We work with some of the leading MedAffairs consultants in designing and building disease- and condition-specific taxonomies that can be applied across all of the content in a given organization. This is typically done by combining the tags that already exist along with tags suggested by our models into a single taxonomy that can be applied on top of existing content to allow for use across different parts of the organization.
This is a more time-consuming project due to the need for human oversight in the creation of the merged domain specific taxonomy, but the business value is obvious. Being able to make an apples to apples comparison of Congress, field insights, research content, and all the rest means that you can perform detailed data analytics across all of your content for use in marketing, compliance and all the other areas that Medical Affairs concentrates on.
- Lexalytics helped leading MedAffairs consultants streamline the process for designing and building disease- and condition-specific taxonomies.
Field Insights & Engagement Tracking
Often field MSL teams record brief free text interactions following their recent HCP discussions. Prior to the broad use of NLP technologies, these interactions were stored and often underutilized with company CRMs such as Veeva or Salesforce.com.
However, with the use of NLP tools combined with customized taxonomies and advanced visualizations, Medical Affairs teams can now gauge, in one place, what’s going on during those high-cost, high value, MSL customer interactions.
Key themes can objectively be tracked longitudinally as well as sentiment/perceptions, data gaps, and other areas. This significantly increases the value of the MSL and ensures HCP feedback is better incorporated into medical strategies.
- Lexalytics helped leading pharma companies track, predict, and analyze MSL insights from text interactions following HCP discussions.
Conclusion
The four examples presented here are just a snapshot of AI-driven projects delivered for pharmaceutical company clients. Increasingly, Medical Affairs broadly spans both the internal organization and interactions with external stakeholders. This leads to increased expectations of what Medical Affairs can deliver. In today’s data-driven world, it is no longer possible for Medical Affairs to meet those expectations without adopting new tools and technologies.
Some organizations have kept pace with developments in technology and have been using AI-driven tools to exploit the potential commercial benefits of big data and analytics. But many pharmaceutical companies are not as far along on this journey and face a widening gulf between them and the competition.
For Medical Affairs, AI-driven tools are unlocking all manner of possibilities to improve or support commercial benefits and patient outcomes. The challenge is to be aware of the technologies that are available, and to adopt the ones that will best achieve the organization’s goals. You can find out which AI-driven tools could help you get your jobs done by reaching out to us for an informal conversation.
/