Early focus and big ambitions
I arrived at Flatiron Health’s small Tribeca office for my first day of work about four and a half years and 450 employees ago; Flatiron was just 16 employees in October of 2013. Like many early startup employees, my role was pretty loosely defined: in my case I would be working on something having to do with data analytics and data integration. At the time, Flatiron’s focus was on developing an analytics product for small cancer centers and most of my focus in those early days would be on organizing and integrating our data. In my first few weeks it became clear that there were a number of more urgent challenges that we needed to address.
Our cofounder Zach did most of my onboarding, which consisted of a brain dump of the oncology data ecosystem — whiteboard scribbles of all of the systems in which patient data was trapped: a map of where the most valuable insights into patients’ disease lived. It was a mess. There were electronic health records systems, billing systems, practice management systems, radiology systems, lab information systems, document management systems, the list went on. Data flowed between these systems in various structured and unstructured formats, arrows indicated the direction of information exchange.
As the first hire in the newly created data role, this messy diagram became my roadmap. Navigating the arrows in the diagram with a laser focus on finding where the most relevant information for clinicians and patients lived would define the team’s journey over the next couple of years.
Healthcare clearly had different challenges than anything I had done before (robotics and supply chain at Amazon): the domain was critically important and the data was complex and messy. To address these challenges, we built a unique data team: our focus would be on digging in with customers and taking the time to understand the nuances of how stakeholders in healthcare defined value. With humility and empathy for the difficulty of the problems that we were trying to solve, our mandate would be to deliver on the promise of generating value from health record data, where other startups and IT vendors had failed before.
Getting deep with users and products
In the fall of 2013, we only had a couple of customers and a very early product that had not yet offered value back to our providers. It was clear that, while there were data integration issues, my time would be better spent helping the team find product-market-fit.
I quickly embedded with our product and engineering teams and joined their effort to define the initial product offering. One counterintuitive finding in our early research was that if we wanted to help cancer centers provide better care to patients, one of the best opportunities was to focus our product on alleviating the burden of medical billing. Billing insurance companies for oncology drugs and services is incredibly complicated and mistakes can lead to delays in treating patients or patients getting unexpected bills. Doctors at small cancer centers felt relatively comfortable with the tools they had for treating patients, but did not feel the same way about their practices’ billing and administration. Practices that could not stay on top of their billing could go out of business; not a good outcome for the practice or the patients that may have to switch oncologists in the middle of their treatment.
One specific process that we honed in on was the process of auditing bills for oncology drugs as they were being sent to insurers. We saw several practices manually validating bills across their various systems to make sure that the accounting had been done correctly: drug dispensing systems, billing systems and medical administration records.
This manual process was time-intensive and we suspected that it was error-prone. I quickly prototyped an algorithm that could match drugs in different coding systems and different units of measure to automatically audit bills that were going to insurers. The algorithm seemed to work — the automated process was catching errors that practices weren’t finding with their manual processes. The early prototype got productionized with engineering into a beta product and rolled out at handful of sites. At these early sites we found thousands of errors, and saved both practices and patients hundreds of thousands of dollars after rolling out the tool — then called “Billing Insights”.
Many of our earliest product insights, including this one, came from understanding both the technical complexity of medical data as well as how care is delivered in an oncology practice. Our team would focus on the intersection of the healthcare system and deep knowledge of our data. We realized that we could make a real difference for our customers, it would just take patience and humility to collaborate with the different stakeholders across the healthcare system.
A Data Insights team
We used this early successful product launch as a template for how our team would operate moving forward: engineers who have the skills to dig in and understand nuanced oncology data could work closely with cancer centers and our product teams to drive forward valuable products.
We wanted candidates who were adept at digging into complex systems problems; the titles “Data Scientist” or “Data Engineer” just didn’t capture how much we prioritized this skillset. The candidates who came in through the “Data Scientist” funnel were not always interested in spending a significant amount of time digging into the complexities of the healthcare system and healthcare data. At the time we decided it would be helpful to differentiate our data role both publicly and internally at Flatiron.
After we spent time figuring out how we would operate, we defined our team’s mission: to accelerate our understanding of our customers’ problems and oncology data, and apply technology to those problems where appropriate. And with that came our team name: Data Insights Engineering.
Four years later
Since the early customer insight which helped us launch into billing and practice management analytics four years ago, the team has contributed to the launch of dozens of data products. The Data Insights team has grown as the company’s product surface area has grown; often times this product growth has actually been driven by the DI team.
Data Insights Engineers are now embedded into almost every product team at Flatiron. We are working on problems like clinical trial patient matching, defining how quality of care is measured in oncology, supporting our research offerings by analyzing the use of drugs in the real world, and helping support our internal data platforms and data warehousing teams.
The same inclination to untangle that early messy diagram on the whiteboard has driven the team deeper into different areas of oncology with the hope of uncovering insights behind the deeply human processes and the technical processes in healthcare. We dig with the hope that we can positively impact all stakeholders in oncology — from the researchers pushing our knowledge of oncology forward, to the providers on the front lines of care, and most importantly to the patients who have not always been best served by the IT systems in which their data has been trapped.
If you are a strong coder and fluent with data, but are mostly interested in digging in with our customers and making a huge impact in our healthcare system, give us a shout — we’re hiring!