Some of the most compelling SAP HANA use cases pertain to healthcare; this year's Sapphire Now event was no exception. At the end of Steve Singh's day two keynote, he announced three partners for SAP Connected Health - CancerLinQ, Castlight Health and Dharma.
The day prior, we had CancerLinQ on our Sapphire Now video couch. We also got a story from two data scientists from the Stanford Center for Inherited Cardiovascular Disease. Both stories bring into focus a new era of medicine, where unified data sets and database breakthroughs change approaches to fighting disease - and personalizing treatment.
First up,Kevin Fitzpatrick, CEO of CancerLinQ LLC:
ASCO represents 40,000 oncologists and their care teams. As Fitzpatrick tells it, they partnered with SAP to build a rapid learning system, now called CancderLinQ, which is a not-for-profit, wholly owned subsidiary of ASCO. The goal of CancerlinQ? To compile data on every consecutive cancer patient that comes through ASCO's practices, assemble that in a database, and provide feedback back to the care teams.
The result? "Feedback like they've never had before."
Turns out SAP was serious about getting involved. After issuing an RFP, Fitzpatrick's team got a personal visit from Hasso Plattner and Bill McDermott:
We got the usual RFP responses, but in addition,with SAP we got a personal visit from the leadership saying, "This is important to us. We want to do this with you."
Fitzpatrick on CancerlinQ's goals:
- To assemble real-world data on every consecutive patient, across all age groups.
- To turn that data into an actionable database that allows the physician to see her own patients in near real-time.
- To understand how the patient is doing relative to the accepted domestic and international standards of care.
CancerlinQ has been live on HANA for six months:
We've now assembled this data into a research and analytic database that has over a million patients in it today, and will grow rapidly over time. [This will] allow for subtle signals to emerge from the daily noise of patient care.
One nifty aspect of this project: the doctors who have contributed these one million records are volunteering their time to input. There is no financial incentive for them to input this data:
In the age in which we live, we have physicians, busy physicians, signing up in droves for this and there's no economic incentive. They're not paid to do this, it's actually extra work to look at the reports that come back to utilize this platform. They do it because they care about quality... This shows, I think, the hunger that the practicing oncologist has a for a feedback system, a quality assurance system from their peers.
There are two goals driving this project: improving patient care, and sticking it to cancer:
We live in an era now where there's tremendous strides being made with lots of new therapies, lots of insights, especially in the genomic space. Good things are happening. What we want to do is continue this migration of moving cancer from an acute and rapidly fatal illness, in many cases, to a controlled chronic disease like you might control hypertension.
In situations like this, I want to know if the vendor partnership is window dressing, or if the technology makes a fundamental difference. Fitzpatrick told me that HANA matters here because of the need for data scale:
Jon Reed: Kevin, you also made the point that the volume of data that is supported by HANA is pretty important. You were saying that a doctor might only see one patient in the course of their lifetime that managed to survive cancer for a particular reason. They would just write that off as an anomaly... but when you put that into a bigger database, maybe there's ten other use cases like that. You can start to figure out why.
Kevin Fitzpatrick: Exactly. There's a lot of noise in the real world of clinical evidence. As you assemble these data sets - especially having the HANA in-memory system where you can rapidly analyze large numbers of patients - there are important signals that emerge through the noise about toxicities, about side effects, about people who have wonderful responses to certain therapies, about sub-populations that may not be responding to a therapy in a way that you might have expected, allowing us then to target new approaches to those groups.
The research benefits of CancerlinQ are clear. But how does this improve patient care? Fitzpatrick:
Only 3 percent of patients get enrolled in clinical trials, which is where the data is most carefully curated. 97 percent of patients do not qualify into a trial. The general population of older patients, of sicker patients, of ethnically diverse patients, are not represented in the current data sets that we have. CancerLinQ seeks to serve that population, and to make the idiosyncrasies that might occur there more visible.
Fitzpatrick's reference to the genome angle was timely, given our shoot with two data scientists from the Stanford Center for Inherited Cardiovascular Disease, Daryl Waggot, Genetics Data Scientist at Stanford University, and Dr. Euan Ashley, Associate Professor of Cardiovascular Medicine and Genetics:
Their mission: creating the "next generation medical record," based on the human genome. They've been working on this project, dubbed with the fancy name of "The SAP Project," for two years. Ashley told me this project wouldn't have been possible even three years ago:
We've been thinking about the onslaught of big data and its impact on medicine for many years. Sequencing a patient's genome used to be 3 billion dollars just a few years ago. Now, for 1,000 dollars, we can do that.
Medecine is being impacted by an influx of new data streams:
We can put watches on people and we get data for their heart rate. All these data streams are coming into medicine, and being integrated with the data that's already there, x-rays, electrocardiograms, patient records, notes that have been dictated by physicians.
The problem is ensuring that data has an impact on patients:
What we want to do is think of a way of integrating that in a way that made it easy for the patient and easy for the doctor, so that they can improve diagnosis and ultimately improve treatment.
So the next-gen patient record must pull together diverse data sources: wearable data, medical records, and the human genome. Their project pulls all this together.
But in order to that, they had to figure out how to scale the data. Enter SAP HANA:
One of the challenges we initially had was scaling genome analysis. It's not trivial... to scale the number of patients that you see within a clinic. That's a challenge. When we ported the system that we currently have to SAP HANA, we were able to do literally thousands of patients within a day through that system in real time. That opens the doors to scaling across a whole medical system.
So why did HANA stand out from other options? Waggot:
The idea of in-memory databases was something that we've always been really aware of, and HANA seemed like it was best positioned to do it. The other thing was that we had access to an innovation team in Palo Alto at Stanford, which was incredibly skilled at design planning and implementation. They managed to take an idea that we had, and implemented it within literally three months. It's pretty incredible.
The system is now running as a prototype with 50 genome records. But ambitious scale up plans are in the works - they'd like to get to ten million records. Their two year goal? To have this system serving numerous health care facilities, including Stanford's. Ashley is motivated by the first patients in the system:
One of my patients that I see in clinic had been diagnosed with a heart muscle problem, cardiomyopathy. The system went into the medical record. It did not show language parsing of all the information in there. It pulled out the diagnosis of a heart muscle disease. It then automatically looks up the patient's genome to see which genes and which variants for that patient are associated with cardiomyopathy, and then it connects that with the wearable data and surfaces it all on a dashboard.
Both projects raise the typical questions about patient privacy, a topic Fitzpatrick and I discussed. Those issues are solvable, in my view, IF there is a rigorous approach to data security and opt-in awareness for patients. The American healthcare system looms as a bureaucratic beast that must be tamed.
Still, the potentials here are enticing. Give me the principle of precision medicine over the mass prescriptions of one-size-fits-all pharmaceuticals. The technology is finally accessible enough to make these ideals feasible. As Ashley says, we now have a chance to move from generic exams to a continuous care model:
What I really want to see is wearable sensors doing continuous monitoring to change the paradigm of healthcare from something that's episodic, where you go in every six months, or two years, to something where there's a continuous monitoring... The medical record becomes intelligent.