The Colorado Center for Personalized Medicine (CCPM), a partnership among the University of Colorado Denver, UCHealth, Children’s Hospital Colorado, and CU Medicine, and located in the Denver area, is driving forward on precision medicine research with a focus on analyzing patient records and genetic data to predict disease risk and develop targeted treatments based on an individual’s health history. To do this research work, CCPM, which is a part of UCHealth, requires the ability to examine the health history and genetic makeup of thousands of patients.
Health Data Compass serves as the enterprise health data warehouse for CCPM and plays a foundational role in integrating patient genomic data from CCPM and electronic health records from UCHealth, Children’s Hospital Colorado and CU Medicine, including external records such as insurance claims, public health records and environmental data.
Michael Ames, associate director for Health Data Compass and director of enterprise architecture for CCPM, of which Health Data Compass comprises one part, says Compass is able to link patient records from those institutions to create a longitudinal record that does not exist within the EHRs alone. “Our data warehouse combines records from across those organizations, about six million total lives, and the intent is to enable new kinds of biomedical discovery and open up new opportunities for operational excellence within the hospitals by doing things with the data that we can’t do when we operate in the silos.”
Further, Ames says, “By bringing the children’s records and the adult records together, we can get a more complete picture of patient care. Especially for patients with chronic disease that may begin treatment in a pediatric setting and then move into an adult setting. By combining records, we can uncover better and more complete descriptions of patient care and procedures so we pull all those things together and make it available for a wide-range of doctors, clinicians and administrators from across those partners for a variety of purposes.”
However, Health Data Compass’s first attempt to develop a traditional, on-premises data warehouse to support data analytics, which went live in early 2015, ran into a number of problems. “It was not long before we started confronting challenges and difficulties with this on-premises enterprise infrastructure, which really stemmed from the overhead, costs, time, effort and risk associated with having to manage our own internal on-premises infrastructure,” Ames says. “Even partnering very closely with our highly skilled health information technology team, we found that it was taking a substantial amount of time and energy and money just to keep the lights on, keeping up with the patching on these systems, keeping up with the security concerns, and all the things you have to do when you take on the responsibility for managing the infrastructure.” The on-premises system also didn’t scale for the center’s analytics needs.
In order to compile the vast amounts of data from these hospitals, UCHealth needed a way to streamline information on patient plans of care and medical research under one roof so that each organization would have the data readily available for collaboration. Working with over six million patients and eight major data sources, each organization had its own on-premise data, which became difficult to efficiently share.
With this mind, Health Data Compass began a project to migrate to cloud-based health data management utilizing the Google Cloud Platform. The overall aim of the project was to bring data from silos up to the cloud to make the data more useable, comparable, flexible and available.
Cloud-based data management has grown significantly in healthcare in the past several years in response to the need for operational efficiency and as a strategic decision to better support data analytics capabilities and to innovate more efficiently, according to many health IT industry leaders.
Over a six-month pilot project, which was completed in October 2016, Health Data Compass migrated to a new enterprise data warehouse running on a Google Cloud Platform, working with Tableau, a visual analytics solutions provider. One substantial hurdle to getting the pilot project up and running, according to Ames, was getting all the C-suite executives on board with migrating to the cloud. “We set about six months of work, doing little proof of concepts, and then we went to our steering committee and CIOs across the organizations and said, ‘We want to make this real and do a focused six-month pilot.’ But, the kick is, we have done this with dummy data and synthetic data and we need do it with real data, and see if everything we need to do in our warehouse with real data, which means protected health information (PHI) and HIPAA (Health Insurance Portability and Accountability Act) concerns, and that means a heavy focus on security,” he notes.
While Ames and other project leaders were confident that the cloud-based platform was secure, there were concerns about configuring and managing data securely in the cloud. “Google’s own security on the platform and software level was Fort Knox, but they gave us keys to the doors, and if we aren’t careful, we can leave the wrong doors and windows open,” he says. Project leaders needed expertise on proper security configuration on the cloud, “to give compliance and security officers the confidence that the data would be safe and sound.”
Health Data Compass partnered with Tectonic, a cloud consulting firm, to develop security design documentation. Tectonic, which provided advisory services, also helped develop training programs to work on the Google platform as well as best practices, which feed into the organization’s policies and procedures.
While there were a number of technical hurdles with the project, Ames notes that the collaborative nature of Health Data Compass’s analytics initiatives, as it involves four different healthcare organizations, poses organizational political challenges. “Getting the data out of the silos and getting people to share data and use it for different purposes, those are challenges, but I think we largely overcame those,” he says.
And with this particular pilot project to migrate to the cloud, Ames says “winning the battle for hearts and minds” and gaining support from security and compliance officers, as well as legal representation from the healthcare organizations, was a significant challenge. “I literally have a meeting once a month with those three roles from all four of those organizations, and getting those twelve people in alignment, to support our deployment of this sensitive data to the cloud, is a substantial task. Part of how we accomplished that was by bringing our own skill set to the table and bringing the experience and credibility of our external partners to the table as well,” he says.
The goal of the six-month project was to show that moving to cloud-based data management would be “functional, faster, easier to use and at least equally secure relative to what we were doing on-premise,” Ames says, adding, “By the end of that six-month period, we have proven that in a big way.”
Moving from the legacy data warehouse to the cloud has reduced infrastructure operating costs by 60 percent, according to Health Data Compass, and has improved performance by 50 percent. Ames explains, “In terms of performance, we measure end-to-end extract transfer load (ETL), or how we get data out of source systems, transferred into our systems, data models changed and harmonized into the central mode that we use for reporting analysis in the warehouse. So, the moment we pick up that byte of data to the moment that we have it ready to be analyzed in the data warehouse.” What’s more, operating the master patient index to identify duplicate records and linking patient names and birth dates across the different organizations used to be an eight-hour process. That process now takes about 15 minutes, Ames says.
According to Ames, cloud-based data management supports the advanced analytics capabilities of Health Data Compass because the platform can handle integrating and analyzing massive amounts of data in a high-performance manner.
“Looking at Google BigQuery, which is the beating heart of our new data warehousing architecture, we are able to deploy our many terabytes of data to integrate from many different sources to do the data transformation and analytical queries at record speed with zero effort put into management of virtual machines or networking,” he says. “Our movement to the cloud brought us a platform where the storage is, essentially, infinitely scalable and the performance scales according to our needs automatically, so we can incorporate whatever data we can get our hands on and make that available for analytics purposes.”
Looking ahead, Health Data Compass, as part of the Colorado Center for Personalized Medicine, has a foundational focus on integrating genomic data, and, to this end, the organization is currently working on a research study to collect DNA samples from thousands of patients over the next several years to develop a DNA bio banking study protocol, and to link that data to other data the organization is collecting. “This is the very definition of a big data project, given the volumes that we’re talking about and the number of ways we want it to be analyzed. It’s going to enable Health Data Compass to play a key role in the Center’s objectives toward improving patient healthcare by giving doctors the tools to understand patient health at the molecular level.”
He continues, “So somebody comes in with a common disease, a standard of care, a prescription of course of treatment, we can analyze that person at the molecular level and determine if that treatment is actually optimal for them. And that is a huge amount of science and research behind that, but it starts with the collection of the data, the linking of the data with the rest of the medical record and making that available to our researchers and clinicians. And, Health Data Compass is right in the center of that process,” Ames says.