Artificial intelligence is poised to revolutionize the healthcare ecosystem. Providers are starting to apply AI in an effort to leverage patient data, therapy information and standards of care to make better treatment decisions. Payers are experimenting with AI to help with pre-authorization and claims adjudication. Whether or not the pharma industry is ready, AI is changing the way that pharma’s customers and stakeholders do business, thereby necessitating changes to pharma’s own commercial practices. And, according to my colleague John Piccone, those changes will be “dramatic.”
I recently spoke with John, a ZS principal and advanced data science expert who previously led IBM Watson Health’s life sciences offerings, about the status and influence of AI in healthcare. In my previous post, he and I discussed how AI is being successfully put to use in pharma R&D and what obstacles companies have to overcome to find success with AI on the commercial side. As we continued our conversation, we dove into what AI in healthcare will look like in the near future, and how pharma companies should start planning a commercial reinvention to keep pace—planning for a future in which, as John said, “pharma’s role is going to change from educating people to educating algorithms.”
Q: Let’s look forward a few years and imagine that a lot of pharma’s stakeholders—healthcare providers and payers—have embraced and adopted AI. That shift will have an adverse impact on pharma regardless of whether or not it’s using AI because in the future, it will be confronted with AI out in the real world affecting patients’ treatment decisions. If a physician’s treatment decisions are largely replaced by a pathway or an AI algorithm, how will pharma promote to an algorithm? That’s a bit extreme, but is pharma’s commercial model, or the way it goes to market, fundamentally under threat because of AI?
A: I think it is. I think that the commercial model that has developed in pharma over the 1970s, ’80s and ’90s, over the past 50 years, is going to change dramatically. If you think about the early days of pharmaceutical marketing—and the physical embodiment of this was Merck Manual or Physician’s Desk Reference—it was about educating physicians, providing them an educational channel, but allowing them to exercise their clinical judgment. Then, less so in the life-threatening illnesses but much more in the lifestyle illnesses, it became more of a promotional function than an educational function.
There are two specific examples that I can give of AI applications in provider and payer organizations that demonstrate how pharma marketing is going to go back to more of an educational function, and pharma’s role is going to change from educating people to educating algorithms. Pharma sales teams have said, “Well, if the decisions about therapies are being made by algorithms, how do I detail an algorithm?” We can predict how that’s going to look.
There are two applications of AI that I think are driving this and the first one is clinical decision support. There are multiple vendors with multiple technologies and pilots where they’ve been using AI to help physicians make better therapy recommendations for specific patients based on a comprehensive understanding of the patients’ health status from all of the data and knowledge available in their records, and a better understanding of all of the information related to the therapy options and the standards of care for treating a patient of that type.
When an AI system makes recommendations and presents the physician with evidence for the recommendations, the physician’s role becomes that of reviewer and approver, except in nuanced cases or when the physician disagrees with the AI recommendations. In this AI-driven future, the sales role largely goes away in the traditional sense because the physician is no longer the initiator of treatment recommendations. Instead, a set of algorithms are characterizing the patient and treatment options, and presenting a set of evidence-based, optimized choices.
The other example is pre-authorization and claims adjudication. Many health insurers are experimenting with AI, or in some cases have adopted wholesale systems, to perform the complicated workflow of pre-authorization and claims adjudication. AI can help insurers make decisions based on the policy conditions and riders for a particular patient’s healthcare policy, the standards of care and procedural guidelines within the payer and provider organizations, the formularies and the nuances of the patient’s health status and medical history.
Q: With AI positioned to augment or automate how both HCPs and insurers make decisions, how should pharma companies prepare?
A: In both cases, how does the pharmaceutical company now influence or more appropriately and properly inform those decisions? The pharmaceutical company needs to look at how these systems work and where they’re getting their information, and how they’re making decisions. First, most of these systems use a corpus of knowledge—medical literature, standards and guidelines, documented standards of practice and formulary documentation—to inform their decisions. And that knowledge is weighted based on the relevance, accuracy and popularity of those corpuses, so it’s going to be very important for pharma to understand what those corpuses are, how they’re ranked in terms of their application in these algorithms, and then how to publish into those corpuses in a credible way so that their products are accurately and sufficiently represented.
Second, it’s going to be important for pharma to learn how the algorithms that make the recommendations work so that they can play an active role in the training process by influencing the decision process and the learning process, both through the corpus and through collaborations with the organizations that are building or executing the algorithms.
AI has arrived in healthcare. It’s already starting to bring sweeping changes to both how healthcare is delivered and how players throughout the healthcare ecosystem conduct business. To keep up, pharma organizations are going to have to look beyond isolated AI applications—beyond thinking of AI and advanced data science as technological add-ons to their existing systems and processes—and rethink their commercial models more holistically. Stay tuned for my next post, in which John and I discuss a few ways to get started.