10 min read

Using ML/AI to Find the Right Patients at the Right Time: A Conversation with AVEO Oncology

Featured Image

By: Dan Fisher, Principal, IPM.ai

Ganesh Rajaratnam, Senior Director of Commercial Operations, Insights & Analytics at AVEO Oncology spoke with IPM.ai’s Senior Vice President, Andrea Lacianca and Associate Principal Ivan Kouchlev, to discuss patient finding at key intervals of the treatment journey.  

AVEO, a commercial-stage oncology-focused biopharma company, partnered with IPM.ai to uncover patients for its new treatment targeting adults with relapsed or refractory renal cell carcinoma (RCC). Low incidence rates complicated patient finding, while targeting those on the cusp of progressing to their third or later systemic therapy presented an even greater challenge — especially considering the aggressive nature of the disease.

To ensure a successful commercial launch, Ganesh not only wanted to find the right patients but to predict their progression prior to the next line of therapy being initiated. Armed with this information, the AVEO sales team could intercept and educate HCPs about its drug at exactly the right time. 

Such highly targeted predictive analysis is only possible through machine learning/artificial intelligence (ML/AI) powered by real world data (RWD). IPM.ai’s extensive RWD universe consists of the diagnosis, medical procedure, and prescription claims of over 300 million de-identified US patients from the last decade, enabling patient finding for rare and specialty diseases.  

To address AVEO’s challenge, a model was created from a robust, qualified cohort. Patterns in the claims data were analyzed 30-60 days prior to the endpoint of a patient being prescribed a third-line therapy. The results were impressive – the ML generated meaningfully different variables compared to patients that did not progress. Through these insights, AVEO’s team uncovered future patients. 

IPM.ai’s data is refreshed weekly and delivered directly to AVEO’s CRM, just like a news feed. When an alert is triggered, AVEO deploys a medical rep to contact the specific physician (linked by NPI number) treating an appropriate patient at an opportune moment of their health journey. By partnering with IPM.ai, AVEO has elevated script lift and above all, made a positive impact on patient lives. 

Throughout the process, Ganesh and Ivan have worked together to refine the model based on what’s been most effective. The more fine-tuned and precise it becomes, the better the outcomes for AVEO and RCC patients alike. 

For more information on how we help organizations like AVEO Oncology accelerate the development and commercialization of precision medicine for rare and specialty diseases reach out at hello@ipm.ai.  

Read on for the full conversation between AVEO Oncology and IPM.ai or watch it here.

Andrea Lacianca (AL): Today we are discussing how IPM.ai partnered with AVEO to find the right patients who could benefit from their new treatment at the right time, by applying machine learning and artificial intelligence to real world data. Ganesh, can you provide an overview of AVEO? What is its mission and what your role is within the organization?

Ganesh Rajaratnam (GR): AVEO is a commercial-stage oncology focused biopharmaceutical company, committed to delivering medicines that provide a better life for patients with cancer. AVEO currently markets an FDA approved medicine in the US for the treatment of adult patients with relapsed or refractory cell carcinoma (RCC). AVEO continues to develop its portfolio of medicines and immuno-oncology combinations in RCC and other indications and has several programs in clinical.

I'm a Senior Director of Commercial Operations, Insights & Analytics. As a member of the commercial leadership team, one of my primary areas of focus is working closely with the leadership team in developing and driving strategy on continued launch execution to help drive innovation, prioritization and commercialization.

Additionally, I oversee all aspects of commercial operations, which includes sales, operations, the commercial infrastructure, data strategy, key analytics and reporting, and US market research to inform both our short-term and long-term business.

AL: Can you tell us a little bit more about the launch as well as some of the objectives you were trying to achieve and the challenges that you were looking to overcome?

GR: A critical component of my role initially in this launch period is really helping our sales team focus their time on the right set of health care providers at crucial times. We're looking for a way to not only find the right types of patients, but also look for signs of progression before the next line of therapy is initiated, that way we can educate HCPs about AVEO’s medicine at the time the treatment decision is being made. As you may know, RCC is a rare disease and finding patients who have progressed past two or most systemic therapies makes customer targeting even more complex. RCC is aggressive and decisions about new lines of therapy are made very quickly. For us, it's even more vital to ensure that we can have our sales team make the right call with the health care provider at the right time.

AL: What were you looking for in a partner to achieve those goals and overcome those challenges? In other words, what brought you to IPM.ai?

GR: While we were initially exploring other options to achieve the objectives of targeting —and targeting itself is a complex area — finding the right HCP at the right time, we realized that the question of the right timing was especially difficult to address. When it comes to finding the right HCP, IPM.ai has created an extremely robust database as the foundation for all their work. This allowed us to see patient volume that was not captured in some of our other data sources and complemented our data set nicely. This gave us a lot of confidence that we were already starting from a very solid foundation. 

Where IPM.ai truly differed was in the ability to predict the progression beforehand. Rather than just relying on what we thought were signals of progression, we wanted to learn what those signals actually are in the real world, and then use them to predict future patients. 

AL: How did you learn about IPM.ai initially? 

GR: I had the good fortune of working with IPM.ai in one of my prior roles leveraging similar predictive AI algorithms to help find the appropriate patients at the right time.

AL: Ivan, how did IPM.ai help AVEO, particularly in terms of connecting to the right physicians at the right point in time?

Ivan Kouchlev (IK): To find the right physicians, we need to find the right patients. The first question becomes, how do we find the patients that might benefit from AVEO? IPM.ai has licensed a de-identified patient database that spans over the past decade. We see over 300 million patients in our data whose diagnosis, medical procedure, and prescription claims are updated weekly. For AVEO, we looked for patients with confirmed RCC diagnosis and evidence of multiple systemic therapies. As a starting point, we ensure that any warning or predictions can only be made from this qualified, very robust cohort. Then we created a machine learning model to study claims patterns in the 30 to 60 days before patients reached the endpoint suited to AVEO’s therapy. The model ingested the known patient claims and identified the variables that are meaningfully different compared to those patients who did not progress.

Some of these variables were novel and unexpected. Other variables turned out to be more mundane but happened to occur in a sequence or frequency that turned out to be meaningful. The key thing for anyone listening is that there usually is no simple or seemingly magical solution to a complicated problem, which is precisely why we use machine learning to identify these patterns of progression.

Finally, we turned on the AI. AVEO now receives weekly alerts that predict which health care providers are seeing patients who look like they're getting close to progressing to a new therapy. The alerts allow them to provide just-in-time education to a health care provider who is likely to benefit from this information.

AL: Ivan, was there anything unique about this project for IPM.ai? 

IK: Yes, predicting and finding rare patient leads and rare diseases in oncology is a daily practice for us at IPM.ai, however, AVEO’s problem statement was significant in that we predicted patient progression within a 30-to-60-day window. We are increasingly doing this kind of work to help our clients have timely interactions across aggressive, as well as indolent oncology indications — but that prediction window was unique to this situation. 

AL: What's the process and timeline like for getting a project like this off the ground and operational for an organization like AVEO?

IK: What we generally are speaking about happens in three steps. We first need to define the outcome that we're measuring. In AVEO’s case, it was looking for those patients who had been diagnosed with RCC and had multiple systemic therapies. We need to identify those patients that progressed and then also define the pool of eligible patients from which predictions can be made. Step two is to study those patients. We built a machine learning model to study the signals in patients that look like they would have been suited to AVEO’s therapeutic option. The third step is generating the predictions and deploying that on an ongoing basis so that we can share these leads in a timely way. From start to finish, this kind of project can take anywhere between six to nine weeks — with AVEO we were able to accomplish this in just about two months.

AL: Ganesh, what positive outcomes has AVEO seen so far from working with IPM.ai on this particular solution? What, if anything, have you heard from your sales team or anyone else in your organization about the project?

GR: As you can imagine, as we are going through the launch phase, we do a lot of reporting on what's working and what's not working to really understand where we can better spend our time — not just to salesforce, but all of us, and make sure we are doing it diligently. With the IPM.ai project alerts, I've actually frequently heard from our sales leadership and our sales teams on how an alert has resulted in a meaningful call with a health care provider. And sometimes it's a new target that we come across because we're getting a complementary set of targets from this work. More recently, I've heard of these calls turning into a success where the right patient has been found and put on the product. That's great news for us because we’re getting our product out to patients in need. 

The other piece of this is Covid, which is the big elephant in the room. It's been very hard to get access to a lot of institutions because they've been shut down across the country. With IPM.ai, we're getting alerts that are meaningful. When our sales teams are making these calls with the health care providers, they're very meaningful, impactful calls. Generally, the call is at the right time, because we have a sense of when that patient might be progressing in their treatment paradigm. What I really want to stress above all else is knowing that we are making a meaningful and a positive impact on the lives of RCC patients. 

AL: Given that this is a fairly new approach to uncovering patients with a current need, as well as those approaching a potential line of therapy change, what was the process that you took to educate and gain buy-in from others within your organization? 

GR: It’s pretty new, pioneering this type of approach. Getting that cross-functional engagement across all the leaders of the company was vital initially, and really educating them about the benefits and about how the program works.

I laid out the approach and took them through the steps involved. It was rolled out in stages. We first took it to our executive team to make sure that we had covered it from a strategic perspective and had all the right parameters in place. Then it was sales leadership, as they're going to be the ones who are essentially owning this, and then management with their team. We then went to our individual sales reps, our oncology account managers, to make sure that they understood the program, the parameters, and what these signals were going to be effectively telling them as we receive them.

The other piece is to set expectations that this is a program that's working with a machine algorithm in the background that’s going to try and predict patient alerts. So, to help explain that process and make them understand that it's not going to be perfect all the time, but that's fine as we’re going in the right direction.

Then the biggest piece is getting the implementation done right. You have to make sure it gets executed all the way through. We focused on the implementation and made sure we had the right resources and tools, where we’ll receive these alerts on a weekly basis. We also continued training and reinforced the process that was continued to create the leads. I share the success stories that I receive across the nation so folks understand that they should really pay attention to the alerts that we're getting. 

Before we used to have an email and a static report that we present to our team. What we’ve taken now as the next step forward is essentially putting it through our CRM and using some language to help drive this to the CRM at the forefront.

If you have an iPad, you get a notification that comes out through a news feed. This is almost like an IPM.ai news feed that says, and you have these possible alerts you should act on. Reaching out to the providers is the next best action associated with these alerts.

AL: Was there anything surprising about the implementation journey that you learned?

GR: Honestly, I think the success of the program has exceeded my expectations and that of my sales leadership and their teams as well. Some of them might have seen something like this in a different flavor in their prior lives, but seeing the success that we've gotten out of this has been very fruitful. This is also because we provide feedback to IPM.ai. I've worked in close partnership with Ivan on the alert types that have worked or not.

Ivan and his team have fine-tuned the alerts that we get further. It's always the continuous feedback loop about what works and what doesn't; IPM.ai has made it better for us as we go forward. 

AL: Ivan, would you say that's typical for organizations first undertaking this type of approach to patient finding? What should pharmaceutical and biotech leaders keep in mind when engaging with IPM.ai? 

IK: Talking about the results first, it's always a good thing when the insights we generated go better than expected. We do typically hear from our clients that this approach uncovers new opportunities, or it helps them pinpoint exactly where they should spend their time and energy compared to traditional targeting approaches. I think it all culminates in the implementation as well, which we've just heard from Ganesh is quite exemplary too. I think that the use and adoption of the insights that we're sharing here have really taken it the extra mile. 

There are a few things that I think would be useful for biopharma leaders to know when working with us. The first is that biopharma leaders don't have to make the choice about where they purchase claims data from and how they're going to generate insight from it. We have the required tools, people and technology under one roof at IPM.ai, which allows our clients to be more efficient with their budgets and how they get these kinds of insights.

Second, while artificial intelligence and machine learning is a big part of what we do, we do offer complimentary analytics like business rules-based targeting, market assessments and patient profiling analyses, to name a few. Again, we have the team and processes in place to help our clients understand their business more broadly.

The third thing is that I'd like to touch on how we can specifically bolster targeting efforts. First, we can build managed email campaigns that provide support, in addition to deploying any ongoing alerts. This reinforces the work that we've done and takes our clients one step closer to an omnichannel approach to interacting with their customers.

And on top of that, we can layer on additional information like health care providers, account affiliations, where they practice, and even their patient's payer mix for more enhanced targeting. I think these things would go a long way for some of our clients or other biopharma leaders to know when working with us.

AL: Thanks, is there anything else you’d like to add?

GR: No, thank you for the time today.

IK: We look forward to building on the partnership. 

About the Author

Dan Fisher

Principal, IPM.ai 

IMG_6644 copy


As Principal for IPM.ai, Dan leads a team that utilizes machine learning, artificial intelligence and advanced analytics to deliver valuable insights that guide and accelerate the clinical and commercial decisions of life sciences companies. With a focus on specialty markets, Dan’s deep expertise in rare disease and oncology disease states helps biopharma clients better understand and more effectively uncover ideal patients and their health care providers. Prior to joining IPM.ai, Dan led commercial operations and clinical analytics projects for ZS Associates. He holds a Master of Business Administration (MBA) from Vanderbilt University.

1 min read

Using Predictive Intelligence to Uncover the Rare Disease Patient Journey

In this podcast interview, Dan Fisher, managing director and practice lead for IPM.ai, explains the use of artificial intelligence for stitching...

Read More

2023 Survey: Automation Is Ubiquitous for Orphan and Specialty Drug Commercialization

87% of Biopharma Respondents Use AI/ML for Post-Market Activities, But Key Missed Opportunities Remain

Read More