• Sustainability
  • DE&I
  • Pandemic
  • Finance
  • Legal
  • Technology
  • Regulatory
  • Global
  • Pricing
  • Strategy
  • R&D/Clinical Trials
  • Opinion
  • Executive Roundtable
  • Sales & Marketing
  • Executive Profiles
  • Leadership
  • Market Access
  • Patient Engagement
  • Supply Chain
  • Industry Trends

Harnessing the Power of Machine Learning


Jean Drouin M.D., CEO of Clarify Health Solutions, gives Pharm Exec his perspective on how biopharmaceutical companies are faring in the use of AI and machine learning

Clarify Health Solutions offers a technology and analytics platform to personalize and optimize patient care journeys. The San Francisco-based startup is focused on transforming healthcare processes and workflows, using the same technology that drives portfolio accounting for leading financial institutions such as J.P.  Morgan and Goldman Sachs. By seamlessly integrating large datasets, machine learning analytics, and real-time patient navigation capabilities on its platform, Clarify aims to illuminate a clear path to better patient care, empowering physicians, health systems and payers with the tools they need to improve outcomes, reduce costs and thrive in the emerging value-based world.

In October 2018, with the launch of its new life sciences division, Clarify made its platform and precise articulation of care journeys available to biopharmaceutical companies and CROs. The new industry-focused offerings are designed to help life sciences organizations better understand the clinical and economic elements of patient journeys, define cohorts for therapeutic intervention, and evaluate the potential or observed benefits of a therapy.

PharmExec caught up with Clarify’s CEO Jean Drouin M.D. to get his perspective on how biopharmaceutical companies are faring in the use of AI and machine learning, and how he sees this space evolving in the next few years. Prior to founding Clarify, Drouin established and led McKinsey Healthcare Analytics, helped set up England’s hospital regulator and served as the Head of Strategy for NHS London.

PharmExec: What changes are you seeing in the way companies are embracing AI and data analytics?

Jean Drouin

Jean Drouin: Across providers, payers and pharma, there's a growing sense that machine learning and AI are useful. But one of the challenges is that we're not yet at a stage where people have become expert assessors, evaluators, and buyers of what truly adds value – versus what is potentially academically or scientifically interesting but may not translate into insights that would be actionable or have a meaningful business impact.

Today, many people still use machine learning and AI interchangeably, as if the terms mean the same thing. However, there are profound differences in the applicability and sophistication of these techniques.

With machine learning, people say, "Oh, I must be able to tease out a new set of insights." And this is true, but what they forget is that machine learning, when it has the most impact, is often the application of well-known statistical methods through the computing power that's now available in cloud-based analytical platforms.

AI is useful where there are lots of manual tasks that a human being would normally have to do. A good example is using AI to spot credit card fraud. When it comes to health systems, AI can be used to more easily identify variations in patients’ response to therapy and clinicians’ performance. Essentially,  you can use a machine to rapidly detect the most varied events, and then go to those physicians or those patients and say: "Given the presence of one or more additional diseases or disorders (comorbidities), we ought to try a different therapy…" or “for someone in this condition, losing 20 pounds before surgery is critical to reducing the risk of serious complications and potential readmission to the hospital."

The question one needs to start with is what's the problem that you’re trying to solve? When is it helpful to be able to go through a statistical model faster and much more powerfully? Interestingly, when you help customers understand that those are the kinds of questions that they should ask, they then come back with more precise questions addressing what they feel they're missing.

In terms of using this technology to establish value, do you see that expertise evolving internally, or is it a case of relying increasingly on solution providers?

While we have seen expertise evolve internally in payer, provider and life sciences organizations, most of these organizations now recognize that, to be effective, they should bring in partners with top data science and engineering talent, who are 100 percent focused on developing and refining technology that can be scaled to drive meaningful impact.

From our conversations with large biopharmaceutical companies, it’s clear that they see the value in bringing in internal data and advanced analytics experts to oversee collaborative partnerships with external service providers.

Overall, we have witnessed a lot of fatigue in the industry as AI and machine learning have slowly started becoming meaningless buzz words. For that reason, it’s more important than ever that service providers offer a clear value proposition and transparent methodological approach. One of my tests when I hear a pitch is: Is the pitch primarily about the fancy methodology? Or is the pitch primarily about, "the business problem we're solving and the demonstrated impact we've had"? With the latter, there is more assurance that the AI or the machine learning is actually being done in a robust way and is pointed at a relevant business problem.

You've worked in lots of different countries. What kind of challenges or advances will we to continue to see from different countries with different data policies and different qualities of data?

Countries like the UK, Denmark, Sweden, and France tend to have very good clinical and social determinant data at the individual level. However, because, in single-payer systems, individual hospitals and physicians don't send out claims for everything that they do, it’s actually harder to get a view of the patient journey from a workflow aspect, relative to what we’re able to do in the US. So, there's going to be quite a lot of value in amalgamating the more holistic clinical and social datasets from Europe and Australia with the claims-based datasets in the US. And I see value in that.

On the data governance front, broadly speaking, we're headed to a world where companies like Apple will make it increasingly possible for individuals to be able to aggregate their data in one place and govern it themselves, that is, they will be able to give differential access to different stakeholders. My guess is we're 10 or 15 years away from that, but there's going to be more and more pressure from consumers who demand it.

How does Clarify Health Solutions differentiate itself from other companies in this space?

The engineering team that came to form Clarify were the Chief Technology Officer, the CIO, and several senior engineers at a financial services company called Advent Software. Advent had built the cloud-based analytics platform that powers the likes of J.P. Morgan and Goldman Sachs, and now 5,000 other customers. The platform reconciles large amounts of trading information to mark the books back to market, including automating the valuation of derivatives. We have imported the power, speed and security of financial services analytics into healthcare. It's the first time that healthcare has had a platform like that.

We’re able to union very large datasets, for example, the entire Medicare dataset and over 100 million lives of commercial claims data to data on social determinants of health, such as people's living status, and their education. Bringing all of that together and then using the power and speed of the platform to run machine learning and AI models allows us principally to do two things. One is to be much more precise about the relative performance of health systems, hospitals, and individual physicians. The second is we're then able to use that understanding of variation to put patients into groups, enabling much more precise views of the actual journeys of care they go through. So, with the size of dataset we have, we're able to more precisely predict the characteristics of a patient that lend that individual to proceed down a particular path of care.

The next thing for us, particularly in life sciences, is to take this superior understanding of care journeys and apply it to help make clinical trial design and management more effective and efficient. On the provider side, it's continuing to build out the real-time care guidance that we do across cardiac, maternity, and oncology.

Related Videos