How Biotech and Pharma Use AI Today
When you ask who is “doing” AI the best, the answer is rarely a Fortune 500 Pharma company, and there are a few reasons for this. Most people think of genomics for healthcare AI applications. Another common one is drug discovery, which is very linear.
According to Dr. Adam Jenkins, the linear nature of these applications can hide some of the most interesting ways that pharmaceuticals are using AI after commercialization. Post-launch, the opportunities aren’t nearly as direct, but that means there could be some interesting applications. In his ODSC East talk, “Integrating Data Science into Commercial Pharma: The Good, The Bad, and The Validated,” he outlines how biotech and pharma companies can better leverage AI post-launch.
Post Launch Opportunities
There are three main customers companies can target using data science and machine learning.
- Consumers — How do we make customer experiences pleasant? How do we encourage customers to use our drug/therapy?
- Healthcare providers — How do we measure receptiveness? How do we know when to approach?
- Payers — What are the difficulties with payments? How do we identify when you’ll be lenient with payments?
The post-launch field is an area where data science can thrive. There are three main ways that data science could revolutionize and provide insight into the post-launch field of commercial pharmaceuticals.
Data Integrity and Understanding
There’s a trade-off between the cost and depth of data. Pharma deals with three levels of data:
- Pharma outside the US — rolled up at geographic level. Least expensive
- US In-house — Lots of info about the single disease. — mid-range expensive
- Third-Party claims — everything about the patient. Most expensive but also the most useful.
If you’re a smaller biotech firm, you may focus more on outside the US data, whereas if you’re a Fortune 500, you may have better access to more extensive and more expensive data. And the type of data will determine the types of ML you can perform. Limited granularity is a far different beast than wide unknown granularity.
Another difficulty is that pharma has its own lingo. If you’ve never worked with the lingo before, it can be challenging to bring in expertise to the data. Doctors sometimes do strange things to accomplish different things for patients, and this must be accounted for when examining the data.
Pharma is also one of the last industries relying so heavily on salespeople, instead of fancy virtual marketing. This is a unique way to receive data. This field force can utilize ML to its best — if unorthodox — advantage. It provides competitive intelligence, business optimization, and heuristic creation.
AI and BI Intersection
One thing that’s often overlooked — the person utilizing your AI isn’t the one creating it. AI is like quicksand. There’s so much opportunity, but you must stop and think realistically. With all the data pharma has, it’s easy for AI to fail.
Simple goes a long way. Instead of struggling against something that isn’t working, think simply and rationally. Long-term knowledge from the field force is ripe for leveraging in this case.
Commercial pharma returns aren’t immediate. It can take a long time to determine if an initiative works, three, six, even 12 months later. This length also creates some noise in the data.
Pharma also can’t ever know what patients and doctors look like. The nature of the field creates a blind field, coming to grips with not knowing everything is critical. 70% accuracy is a good number for pharma, unlike other fields.
More information isn’t better. Using a dashboard to tell a story from the BI standpoint is crucial to AI adoption.
Validation of Commercial
A pharma’s data is limited in four main ways:
- pharma must be careful how many touchpoints exist
- once a customer is lost, they’re gone forever
- there are minimal time frames to leverage
- pharma must adhere strictly to government regulation
Algorithms must exist around these limitations to perform validation. The timing and framing of the actual questions matter. These are set in stone guardrails.
This is where the field force comes back in; they can provide vital feedback and insight for the validation. Plus, understanding the optimization process is key. Strong statistical knowledge must be present before launching a full marketing campaign. You only get limited chances.
Key Takeaways on AI in Biotech and Pharma
Dr. Jenkins has four key takeaways for implementing AI strategies in the commercial biotech and pharma space:
- Understand your data
- Understand your customers
- “Don’t get pretty.”
- Don’t get compound mistakes
Once Pharmaceutical companies understand their limitations plus the unique advantages of a physical sales force, implementing AI post-launch should come a lot easier.
Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform.
