Getting to know Pomelo’s Data Science team

At Pomelo, we are scaling a virtual value-based maternity care model to improve access, reduce disparities, and eliminate healthcare gaps for families nationwide. We constantly use data to better understand our patients' needs and behaviors, make our clinicians more efficient, and measure our impact on outcomes.

We sat down with Sherry, a data scientist, to learn more about what makes Pomelo's data team unique. 

Why did you decide to join Pomelo?

In college, I learned about data science through my interest in public policy and civic technology. I discovered a community tackling problems to improve people's lives, from replacing lead pipes to improving access to government benefits and using data to design more efficient and scalable solutions.  Over the years, I sought out roles where I could work on challenging, high-impact problems, wear many different hats, and be on teams with smart, thoughtful people.

When I learned that Pomelo was building a scalable way to improve maternal health outcomes, my initial reaction was, "How does this not already exist?"  I was excited to join a company with an ambitious, impactful mission and so many opportunities to use data to help drive its success. 

How does the data team at Pomelo operate?

Every company says data is core to its work, but data is genuinely core to how every function at Pomelo operates. We use data to identify eligible patients, understand patient risk factors and prior medical history, and measure the success of our interventions in engaging patients and improving outcomes.  

Our Insights team comprises data scientists, software engineers, data engineers, and product managers and supports the business's data needs. As our company and priorities have changed over time, the problems we focus on and the tools we use have evolved, too. As we navigate this change, our team follows a few important, constant principles – 

1. We invest in centralizing our data and building core data models. Before I joined Pomelo, the team invested in storing all patient health information in a centralized FHIR store, a data standard that we use as the source of truth for every Pomelo patient's health history. We also set up a centralized data warehouse for storing data on everything from clinical interactions and outcomes to marketing and patient engagement. Finally, we continuously build dbt models that standardize and transform data on key entities that we then use as the starting point for more complex analyses. 
2. We make analytics-ready data accessible across the whole company.  With a small data team and a very data-driven culture, there are more daily data requests than we could handle if we completed each one ourselves. So, we use a BI tool that enables anyone at Pomelo to easily modify dashboards and charts (or build their own!). Stakeholders can view key metrics, answer ad-hoc questions, and form hypotheses.  Making analytics self-service enables teams to get answers to their data questions faster and allows the data team to focus on building scalable, self-service data products and tackling new, more complex challenges. 
3. We work directly with other functions to understand challenges and design solutions. Every data scientist knows that working cross-functionally is part of the job. At Pomelo, we collaborate with stakeholders across teams every day. The data team joins early discussions in new problem areas to gather context before building a solution. We shadow and ask our nurses, providers, and specialists questions when solving patient care problems. Data science at Pomelo goes beyond building dashboards and reports – we scope pilots and implement new solutions to address challenges from patient outreach to engagement to outcomes measurement.  
   

How do you think about measuring success at Pomelo?

As Pomelo quickly grows to serve more and more patients every month, we need to simultaneously be able to measure our impact on patient outcomes and continuously identify where and how we need to improve. To give a few examples, we need to understand:  

• Which interventions impact patient outcomes most, such as reduced avoidable Emergency Department visits and Neonatal Intensive Care Unit (NICU) admissions? 
• Which patients engage the most in our programs? Who are the patients we aren't currently engaging but should be, and how do we engage them? 
• How do improved outcomes translate to healthcare cost savings for our payer partners? 
  

Answering these questions is not a month- or quarter-long project; defining and measuring patient outcomes is perpetual, given how often we experiment and evolve our care model.  The work we've done to characterize patients' interactions, risk factors, and outcomes is the foundation for all of our in-flight and upcoming work, aiming to answer questions like the ones above and inform the decisions that enable us to scale   

We are just getting started and looking for more impact-driven, collaborative data people to join our team. To learn more, check out our open roles.

‍