Health care startup Apervita has a compelling vision for the health care industry: democratize information by creating a self-service marketplace where institutions and researchers can exchange information - even algorithms - that were previously not accessible.
But there's a catch: that's a heck of a lot of data. One health care institution alone can have millions of records spread across multiple data sources. Since they launched two years ago, Apervita has been building their platform on the MongoDB database.
Seven months ago, the marketplace went live. It's still in the invite-only phase, but with clients the likes of the Cleveland Clinic and Mayo Clinic already on board, Apervita is off to a strong start. That means hiring: the company has doubled in size in six months, with thirty employees and counting.
At MongoDB World, I had a chance to talk with two Apervita executives, Michael Oltman and Aaron Symanski, about their ambitions, and the challenges ahead.
During the keynote customer panel, Oltman spoke to the "gap" in the health care industry that Apervita addresses:
There are thousands of researchers, doctors and surgeons who spend a lot of time trying to improve the quality of health care. But they come up against this hurdle: the data they built their application on only works against their database. In order to take it from one institution to the next, they have to reboil the ocean. All that data is different across different institutions... There's ton of data out there, but there's no knowledge.
During our chat, Oltman elaborated on why the divide between researchers and institutions is a problem:
We're trying to bring two sides together have a lot of trouble of meeting in the middle. Both sides will benefit the health care industry as a whole if they can start to work together. The researchers, which we sometimes call the informatics experts, they're on one side, and all the data is in the institutions, the clinics, or the pharmacies. It's really hard for them to actually use each others' data.
There's a ton of information out there that's not being used for knowledge in the health care space. It's not coming back to help the quality of care, or improve the practice of health care. We want to help take the researchers, and take all their efforts, which is sometimes a 5 to 10 year cycle, and allow them to become business owners, entrepreneurs, and monetize that.
How Apervita conquered its launch obstacles
Two years from the company launch, Apervita's marketplace is open, albeit in an invite-only phase, and is bridging that data gap. But to build Apervita, Oltman and Symanski had to overcome a number of blockers in their path, including data privacy, compliance, and disparate data sources with high volumes.
On the privacy front, Apervita addresses this point by providing clear guidelines on data ownership. Oltman:
We don't own the data. The government may change that in the future and say, "No, a patient owns their own data no matter what," but today in the U.S., the institution owns the data. We're trying to provide tools that say "You own the data, you control the data, you assign and provision the data to the appropriate people. We can't see it. We can't resell it. We can't access it. We're not saying we're going to take your data. We're saying you own your data. We're just a platform for you.
The problem of massive/diverse data sources was central to Apervita's early decision to run on MongoDB. Oltman had some exposure to MongoDB from his prior work in financial services, where he saw firsthand the issues with "legacy" database technology and the "bottlenecks" those systems caused. As Oltman put it on the keynote stage:
All of you are individuals, and your data is unique at every health care institution you go to. The uniqueness of the data led us to a place where we can't deal with a schema, we can't deal with a database where we have to try to pigeonhole pieces of data into well know areas and well known descriptors.
That led Oltman's team to an evaluation of NoSQL databases:
My experience evaluating MongoDB in its early days, then seeing where the roadmap had taken it and looking at the scale [was the key]. We have a big scaling problem: everyone has a ton of data, and we need to evaluate as much of it as possible. MongoDB was a very quick decision for us, and we're very happy with it. Our entire platform runs on MongoDB, everything from user information to analyticsl results to patient information. It's all one database. We want to keep it a simple, scaleable stack.
After go-live: how do you price an algorithm?
Even with the best of plans, go-live can provoke a caffeine binge. I asked the guys about the feedback they've received from their initial customers. So far, the input has been very positive, though with the expected learning curve on all sides. The authors haven't had an equivalent experience before, and aren't always ready to think like a business. Example: they may not know how to price an algorithm on the marketplace.
Oltman summed up their clients' response to the go-live: "Wow! This really seems to be working, but here's what I really need." That leads Apervita into new development cycles, as well as rigorous Q/A and testing. Symanski:
We had to get more rigorous with our release process. The next thing you know, we need a Q/A team, and we need a deployment and a DevOps team. It used to just be six developers and we all did everything. It's great, because that's what we're aiming for. But so far, those who have been involved with us - the early pilots - are all very enthusiastic and really excited about the potential.
As they onboard more authors and institutions, I had one burning question: how do you price an algorithm?
The answer: you either price it low for wide distribution, or higher for bigger margins. Or you experiment until you find the right equilibrium (Apervita's suthors are free to set their own prices). It also depends on the value of the algorithm and where the author thinks that particular market is going. Some algorithms have high price points due to their material returns. Example: an algo that saves one percent of a Medicare reimbursement for a hospital is a huge financial asset.
Growing pains: how will MongoDB scale?
During his talks, Oltman was open about their own growing pains and what is looking for with MongoDB. Apervita is eager to get onto 3.0 (released in February), and improve their write performance using the Wired Tiger storage engine (first shipped with 2.8). With constant compliance chores pertaining to the moving targets of regulations like HIPAA, Oltman is also looking ahead to release 4.0 and additional encryption and security features, including enhanced tracking and auditing.
Oltman and Symanski are confident MongoDB will scale with them. Some of that means tapping into a growing community of experienced MongoDB practitioners. I get the sense that their confidence in MongoDB is bolstered by their description of MongoDB engineers working alongside them, pushing the boundaries and solving issues. Oltman also loves the ease of use:
Every developer in our company has an entire environment sitting on their laptop. There's no central server they have to connect to. Just grab the latest package, put it on the machine, and we're off and running. It really is a quick adoption, you just have to model and think about your architecture and use case.
Looking ahead from the keynote stage, Oltman envisioned a day when visual medical documents could one day be indexed:
If you want to throw some challenges out there, to be able to process radiology indexes and imagery would totally transform the health care industry. EKG readings, heart rate readings... An x-ray is a giant document with a lot of info on it. To be able to index that!
But for now, Apervita has their eyes on the prize, which is building up the marketplace until they open the gates to the entire medical community. They aren't there yet, but based on their early adopters, Oltman thinks Apervita is on the right track:
To be able to bring all this data into one place without worrying about its structure, without worrying about what the data source was, and then giving researchers the tools to deal with it later, really liberates [our clients]. It allows them to stop worrying that "I've just spent six months boiling the ocean to enable one application."
Well, now you can enable 100 applications by just loading your application into our system and doing it in a different way. It's just a huge variety of information - and now we're turning that information into knowledge.
A worthy pursuit indeed - and a story to watch.