HIStalk Interviews Dave Hodgson, CEO, Project Ronin

Dave Hodgson is co-founder and CEO of Project Ronin of San Mateo, CA.

Tell me about yourself and the company.

I’m a molecular biologist by training. I started my first job on the genome project in Cambridge in the UK. From there, I moved to the United States and worked for a small biotech that was selling genome data to pharmaceutical companies. Then I was part of bioinformatics  for several years for Pfizer, and then knowledge management for Pfizer. Then to  Roche, where I ran scientific computing  for a few years before I decided to completely pivot the career out of pharmaceutical executive land and became part of startups in the Bay area. I lived near Palo Alto and I worked as chief technology officer for a diagnostics company and for a telehealth company. Then I was the first chief technology officer at One Medical, which is now part of Amazon.

After all of that — having been in pharma, the genome project, telehealth, and primary care — it was time to become a consultant. I did a lot of healthcare consulting for a while, and during that was introduced to Dr. David Agus and to Larry Ellison, who were enthusiastic about how we might apply data science and thoughtful clinical interface design to something that could be embedded inside the medical record system to assist decision-making in complicated diseases such as cancer. Project Ronin was the founding of that idea. That was several years ago, and we’ve been working on that challenge ever since.

What drives Larry Elllison’s interest in healthcare and how does he see healthcare and technology merging?

A lot of us have seen the devastation that terrible diseases like cancer can do to the individual, to loved ones, to families, and to friends. It’s a truly horrible situation to be in. So many of us have seen that, and him, too. Then you think about how might you really improve the quality of care —  not just in the United States, but everywhere — and the quality of decision making. 

Cancer is complicated. It’s actually multiple different diseases. Patients have different goals. They have different desires. I want to live as long as possible. I want to stay as healthy as possible, I want to optimize my wellness for now, or I want to do whatever it takes to survive. The diversity continues to their genotype, their other clinical situations, and other things that are going on with them.

Patients are very, very individual. When you are thinking about that quality of decision-making, you have to take into all of that into account. But ideally, you could leverage the entire corpus of knowledge that we have about the disease, from everything published to everything that has happened in the past, and also everything that has happened to a patient like this, from their social situation to their clinical history to their genotype. If you were in the horrible situation of having some kind of tumor diagnosis, you would really love it if your doctor had at their fingertips every single piece of information possible. 

Given that desire and that need, you can start to look at the process of clinical decision-making — diagnosis, selection of treatment, management of treatment, management of survivorship — as a data problem. How might we bring all the world’s clinical knowledge into the space between patient and provider for their most optimized decision-making? Larry, the other founders, all of the team at Project Ronin, many of the clinicians that we work with, and the patients that are on the platform are all quite aligned on that desire to have all the world’s data be in a place where clinicians and patients can use it to make the best decisions for them possible.

What psychology is involved with trying to put all of that together using technology?

It’s a complicated problem. You have many different types of variables to consider — the desires and the situation of the patient, the experience of the oncologist, and the clinical biology of the particular disease that the patient has. Treating late-stage lung cancer is extremely different from early-stage prostate cancer, and very different from mid-stage breast cancer. Very, very different situations. You have all these very different variables. 

The assembly of the data and the processing of those data is extremely complex. That is why this has been a personal mission. Let’s do something very difficult that will have such a broad benefit. There is not only a data assembly complexity, but then there is a psychology that you need to present those data to the clinical team in a way that they can use it, digest it, and take action on it.

That means that you have to present it inside the medical record system. No clinician ever, anywhere, wants to step out of their patient record, charting what happened in the encounter with a patient and then logging into another portal to go look up information or anything else. Although they have to do that in certain cases, they don’t want to do it, because it is clumsy and inefficient. They are already seeing many patients, their day is very busy, and it’s complicated. 

The first source of psychology is to serve up data that is relevant to this particular patient in this particular situation inside the medical record system, so that the clinician has an efficient access to very rich information that they can use to assist or validate or even qualify some of the decisions that they are making.

Similarly, the patient is always wanting to know, what can I expect to happen next? Am I on the right treatment for my situation? If I have questions, can I reach my clinical team? We think very carefully about how we surface the answers to those questions, again under the direction of the clinical team and doing things the right way. But there is some human-centric design very much involved in this that is paired up with all of that data assembly and data rendering that we do. 

You’re probably getting a sense that it’s a pretty hard problem. We don’t believe that anyone has really solved it yet. That’s why we’re working very, very hard to show that it can be done.

A cancer patient with means and knowledge will often seek out the best available expert at an organization such as Sloan Kettering or MD Anderson. Can the scale of technology democratize that for for patients who lack connections, the ability to travel, or insurance coverage to seek out a super-specialist?

Very much, and that was one of our founding desires, to take the expertise that is known by the very best, highly specialized oncologists in the very best academic medical centers and make that knowledge available to practicing oncologists. That’s essentially what we are doing.

One of our best and biggest partners is one of the largest academic medical centers. We have published with them and we are developing the platform with that very goal. How do we package the expertise and the data so that a community oncologist can take advantage of it? We are working with a community practice where when they pull up the patient chart, they see the reference data and the data insights that we add to that. It is supercharging their knowledge in a particular specialty.

Typically in academic medical centers or some of the larger cancer centers, you have practice oncologists that specialize in a particular tumor type, kidney cancer or whatever. Then in the community, you tend to have a little bit more generalist oncologists who are seeing a breast cancer case in one appointment and then a prostate cancer in another. You want to be able to equip them to know, what are all my choices in the right way of treating this particular situation? What has historically been done? What do the reference guidelines suggest? What does the literature suggest? That can take a long time if you do it by hand. We automate that and then present that in an integrated way.

Are providers and pharma connecting in new ways around real-world evidence, clinical trials enrollment, and post-marketing surveillance?

We are seeing that, too. There’s definitely a desire for those worlds to be less separate than they were.

There are a few dimensions where that makes a lot of sense in the priorities of both parties, and that is to enroll the right patients for all clinical trials. There’s a lot of new medicines in the pipelines of pharmaceutical companies that are oncology drugs. There is, and has been for several years, a desire to find the right qualified patients to be enrolled in a trial and the patient’s attributes that would qualify them in or out of any particular trial. A lot of that data is in the patient chart, sitting in the medical record system. There’s an obvious place there to look for eligibility and enrollment by integrating those two systems. 

Then the other part that you mentioned of real-world evidence. There’s clearly a desire to have some kind of companion for the patient through parts of their journey, such as managing their wellness and managing their general interaction with their primary care doc. If they are in a situation where they are diagnosed with cancer, that there would be some companion that would take them through that, including if they found themselves as part of a clinical trial. You would want that companion app, let’s call it, to be with them through that, where it’s collecting the appropriate data of how they are experiencing treatment. 

Then not only have that be an input into how the treatment is performing, but also help the patient manage their side effects and symptoms, which is part of the Ronin platform as well. We do that symptom monitoring and capture of patient-reported outcomes and patient experience.

How much data is needed to make the “patients like me” concept clinically useful, especially for uncommon conditions?

We have done a lot of work in how to acquire the right patient records and then structure them, because clinical data is very much dominated by clinical notes, encounter notes that are all text. They are written in a certain clinical language that is a little bit difficult to manage. There’s a lot of work to do with the data, cleaning up and mapping to a central model. We do a lot of that.

The good news is that over time, we have become pretty good at not requiring an enormous data set or enormously high quality data set. Over our experience in the last few years, we are getting better at doing not only the cleanup, but also requiring less voluminous amounts of data. With a few hundred records, we can do quite a lot of trending and analytics on those data sets to be in a position to serve up insights in a qualified, thoughtful, and high-integrity way. We have a lot of standards around data quality. Our QA and QC processes are robust and strict, so that anything that we put before a clinician or a patient is rigorously tested and validated first.

The early days of precision medicine had limited applicability since few correlations existed between genomic data and condition management options. Will advances make precision or personalized medicine more of a standard?

Very much so. We see that certainly every day. In that data view that we have built with our collaborators that we serve up inside the medical record in the patient’s chart, we show all of the known genomic biomarkers that the patient has been tested for, and then the literature that shows any particular consideration of those. If the patient has lung cancer and is their EGFR is positive, there’s good literature around which treatments may or may not be effective because of the presence or absence of that biomarker. 

In oncology, we spend most of our time in those correlations between biomarker presence or absence and which treatments that information suggests that you should use. Those are becoming quite well published, and therefore, we want to be able to have those reference data be available to the clinician. We are seeing that progress as the science progresses and as the the clinical evidence progresses. We are serving that up when it has been published and validated in all the right ways.

What will the company’s next few years look like?

I want two things to happen, and I would have to speculate greatly on whether they will happen. Obviously there’s a lot of hullabaloo about what will AI in medicine will really look like. We are obviously very invested in that, and we have developed some pretty effective large language models for in the generative AI space. We have some really exciting prototypes that we are taking through some validation, some research processes. In the next couple of years, we are going to begin to identify the most appropriate, safe, and effective uses of AI algorithms, machine learning, and deep learning, including large language models, to the practice of medicine. I hope that we will see the safe and effective demonstrations of those, and the company is heavily invested in that.

The second, which is my dream, honestly, is a greater ubiquity of value-based care reimbursement, where the incentives for practicing medicine in the US are driven by getting paid when the quality outcome for the patient is met rather than getting paid on the volume of procedures that are performed. Value-based care has been conceptually around for decades and has made slow progress, but my dream is that that progress would go faster and that there would be more and more reimbursement using value-based care structures. Technology has a role to play in enabling that. That brings the aligned incentives that we really crave that will really drive a lot better outcomes and a lot better economics.