Edtior's Note: Each month, the editors at SearchHealthIT recognize an innovative software, service or technology approach. Ayasdi is our September 2015 selection.
Release date: Ayasdi Care released in 2014, Ayasdi Core released in 2013, Ayasdi Iris released in 2013
Ayasdi, based in Menlo Park California, was created when a mathematics Ph.D student, Gurjeet Singh, and his advisor, Gunnar Carlsson decided to productize the work they'd been doing at Stanford with Topological Data Analysis (TDA) in 2008.
"Gunnar Carlsson came up with the idea that you can drive understanding from very complex data sets by using the principles of topology or principle of shape," said Patrick Rogers, chief marketing officer at Ayasdi.
In 2005 Carlsson received a $10 million grant from DARPA (Defense Advanced Research Projects Agency) and the National Science Foundation and in 2012 Ayasdi received venture funding.
What Ayasdi does
Ayasdi, a software offered in either SaaS or on-premises form, uses topological data analysis to enable researchers to find patterns and anomalies in complex data sets with the help of algorithms, Rogers said. "We make complex data useful whether it be people in the healthcare profession or financial services or other industries."
Vance MooreMercy Health System
When using TDA it is believed that every data set has a shape and once the shape of the data is known, it is much easier to select the appropriate algorithm in order to discover insights.
"[The software] can look across these available algorithms, help select one of the ones that are most optimal for understanding that data and then be able to represent the results back to a user," Rogers said.
Ayasdi's software -- which can be installed on-premises or hosted by Ayasdi -- does this automatically, making gathering insights from big data much more manageable.
Why Ayasdi matters in healthcare
Ayasdi's software proves valuable in healthcare because it solves a common problem many physicians and healthcare organizations struggle with.
"One of [healthcare's] top challenges, and it's true for all hospitals, is reducing variation in clinical care. When they have a patient come in for a particular procedure, it can be wildly variable in terms of patient outcomes, readmittance rates, cost of the procedure and so forth," Rogers said. "One of their goals is to provide more consistency to their patients and they're trying to manage that consistency across 2,000 physicians all of whom have different best practices for treating a patient."
For example, Mercy -- a health system based in St. Louis that has 46 facilities in 24 communities -- found it had significant variation around certain procedures and wanted to develop best practices for these procedures based on their data.
"They looked at all the medical records associated with these procedures, and this includes things like lab tests and doctors' orders and medications, and they gave us effectively all that data and we were able to sort through that and determine what were optimal pathways that led to the best outcomes," Rogers said. "We're able to find groupings within their data of successful procedures and some very surprising and impactful findings came out of this."
Sorting through and deriving meaning from such large, complex and varied data sets is no easy feat.
"The EMR records for patients are very complex, they're not particularly complete and they're generally very messy so extracting patterns from them is just hard period," said Johan Grahnen, principal data scientist at Ayasdi. "Using our topological data analysis technology we can take this really messy data set that has patterns in it but patterns that are extremely difficult to discern with traditional methods."
With prescriptions, and lab tests and X-rays and so on, it can be very difficult to not only keep track of everything but also find a coherent pattern, Grahnen said. Ayasdi's software, with the help of TDA, allows hospitals to take treatment data, organize it and extract patterns. This can be done for certain portions of the data and for certain groups of patients.
"We can look at every specific group of patients that are similar to each other and extract the patterns for that group and then … you can go find which group had the best outcomes, which group had the best results in terms of cost or in terms of ambulation or whatever it might be and then roll that out as a best practice downstream," Grahnen said.
What a user says
About a year and a half ago Mercy first began using Ayasdi's software.
"A lot of what we do and how we do it is dependent on an individual physicians experience, where they were trained, what they like and don't do … that's basically the way it works," said Vance Moore, senior vice president of operations at Mercy. "What we know certainly from other industries [is that] uncontrolled variation is the enemy of quality. If we understand what drives the best quality, the best outcomes, why wouldn't we all want to follow that general pattern?"
This was the impetus for using Ayasdi, Moore said. Before this technology, doctors and physicians would gather together to try to build out care paths. But in doing it this way everyone brought their own individual biases to the table and the care paths that resulted from gatherings like this were "bloated," Moore said. This was because one physician might say one test should be done and another physician would say a different test should also be done and so on.
One example within Mercy is with hip replacement surgery. There are ten different physicians using ten different products, Moore said. He questioned whether one product is truly better than the other and, therefore, truly worth the cost. Because, "why should we pay the higher price for the product that's not delivering the higher result?" Moore asked.
Moore said they also realized there are multiple applications within healthcare for Ayasdi's software including submitting claims.
"It's providing us more granularity of information on why we get denied and who denies us and we're seeing patterns within that as well," Moore said. "We're determining for one reason or another, is it because we have poor documentation via an individual? Or do we have a particular payer that's pushing back disproportionally to somebody else? We can have a much more refined conversation around … what's going on here because we can't afford to be delayed in our payment or denied."
Moore said that based on their pilot testing period of Ayasdi, topological data analysis technology could save Mercy between 5% to 15% in direct variable costs by identifying the most effective and efficient care paths.
If this initial study is correct, Moore estimates there is an opportunity to save about $100 million.
More on topological data analysis
Explore Gunnar Carlsson's paper "Topology And Data" to learn more.
When a hospital uses Ayasdi, all the relevant information is pulled into the tool, then the tool crunches the data and delivers results in the form of a list, Moore said. The preferred care path appears at the top of the screen with the reference points in the development of that care path appearing below. The tool will also show a physician how they compare to the standard concerning treatment they've delivered to patients.
Ayasdi offers its services by annual subscriptions to use their software via a SaaS or on-premises offering. Ayasdi declined to provide pricing information.
Editor's note: This story has been updated to correct Mercy's name and its number of facilities.
Learn more about data analysis:
Data analytics will be included in clinical decision support tools
The potential of pairing data analytics and IoT
The key to productive healthcare analytics? High-quality data