Q&A: Big Data Analytics Means Smarter Care at Seattle Children's Hospital
Patient queries have dropped from minutes to seconds as top-rated Seattle Children's Hospital -- with 350,000 patients annually -- turns to big data and analytics. Wendy Soethe, manager of enterprise data warehouse and business intelligence at the top-rated hospital, explains.
- By Linda L. Briggs
- August 13, 2013
Big Data's dramatic impact on healthcare offers a preview of how using analytics on vast amounts of data may play out in many fields. At Seattle Children's Hospital, new technology from IBM is improving patient treatment at the highly rated hospital. Patient queries are faster, staff members have a more complete view of trends in patient care, and spreadsheet-based queries that used to take weeks to compile manually are now done on-demand.
In this interview, Wendy Soethe, manager of the enterprise data warehouse at the hospital, explains how the hospital is better able to work with the thousands of data points associated with each child, thus improving the quality of care and treatment.
BI This Week: What are some of the ways in which Seattle Children's Hospital is using big data?
Wendy Soethe: Seattle Children's Hospital is using big data as part of our Clinical Standard Work (CSW) program, which defines patient populations and recommends an ideal protocol for each population, allowing us to ensure that every patient at the hospital receives the same standard of care. The enterprise data warehouse (EDW) currently integrates data from 10 sources across the hospital, including electronic medical records (EMRs) and billing systems. The CSW program is a large consumer of much of that data.
With insight into the thousands of patient data points, hospital doctors and nurses are now able to paint a more holistic picture of their patients, get answers to complex queries about potential treatments and procedures, and identify pathways of care for patients with particular needs, regardless of provider. Clinicians can also look back and evaluate treatment protocols in order to determine where the hospital needs to allocate more resources.
Can you talk about the numbers of patients or size of data stores you work with?
As one of the leading children's hospitals in the nation, Seattle Children's Hospital works with over 350,000 patients annually, with thousands of data points on average per patient.
Although our data is large, we've been able to compress it greatly. Since deploying the new big data analytics solution, we have reduced our SQL Server footprint for the data warehouse and business intelligence applications from 31 to 15 servers, repurposing hardware where appropriate and retiring aging servers. Our data analytics team has also turned off 30 data loads that were used in the old solution.
What are some challenges in managing data that specific to healthcare?
One main challenge in managing data specific to healthcare is ensuring its quality and accuracy. That is essential in understanding our patients, their treatment options, and our care pathway protocols. The data is useless if it can't provide us with accurate, relevant insight.
Complying with HIPAA regulations regarding data access is another challenge. It has also been difficult for us to integrate unstructured data from doctor's notes into our EDW. We have volumes of text notes to parse through but haven't yet gone too far in that direction.
What kind of system were you using before this, and what were some of the issues or bottlenecks?
Prior to implementing the IBM solution we are using now, we relied on a Microsoft SQL Server data warehouse that used complex SQL Server Integration Services (SSIS) to support the various BI projects, including CSW, created across the hospital. These SSIS packages often prolonged development cycle times, restricting the number of new BI projects the hospital could implement to approximately 12 in five years.
Eventually, as our data continued to grow, the system expanded to over 30 individual servers that became increasingly difficult and expensive to maintain. In fact, we were consistently outpacing our storage capabilities.
As part of our long-term goal for an easily managed solution that was scalable and efficient, we looked to upgrade to a more sophisticated system that met the hospital's self-service demands. That's where Brightlight and IBM came in.
What kinds of backend data systems are you pulling data from? What kinds of challenges does that present?
The platform pulls data from 10 different sources, including the electronic medical records (EMR) system, billing, and general ledger. With our previous system, it was time-consuming to develop and maintain all the different SSIS packages. This, in turn, slowed down our internal processes and stifled the hospital's ability to develop new data projects. As our data grew, we also experienced great difficulty requesting and acquiring additional storage. By the time we reached 31 servers, the system had become rather challenging to manage.
In contrast, our new IBM system is scalable, meaning that we can grow our data stores quickly and efficiently without having to request and purchase additional server space. The solution also eliminates the need for us to manage it as closely as the previous system; there's no need for our IT staff to create complicated indexes. The new system was up and running with EMR, billing, and supply chain data within five months.
As users access data in the new system, how important is speed?
Speed is extremely important, not only in accessing the data but in the ability to analyze it. On- demand access to information has a direct impact on patients and business processes. By being able to quickly and accurately gain insight into patient data points, doctors and hospital administrators can better determine courses of care, including treatment prescription and care protocol.
Who are users of the data? How much training did they need?
Because of the IBM solution's self-service capabilities, there are more than 90 users interacting with the data warehouse solution directly. Requests for additional access come in nearly every other day. This includes the hospital research organization, doctors in surgery, and the strategic planning group. Although extensive training isn't needed to access the data, users must have some basic knowledge of our business intelligence tool, which connects to the enterprise data warehouse. To comply with HIPAA guidelines, users will also need to obtain approval to access the data based on specific use cases.
What else would you like to do with the new system?
Now that we have the capability to complete 70 projects in five years, as opposed to just 12, we are adding new data on a project-by-project basis. So far, we have completed a dashboard project for the emergency department, created physician profiles to meet regulatory requirements, worked closely with the surgical department to support power users and their dashboards, and supplied new data to the finance and strategic planning departments, as well as to clinicians and quality assurance staffs.
Our next moves will be to begin integrating barcode data for medication administration and to support our clinic access goals. We are also working to replace our existing ETL tool, as well as set up for disaster recovery.
What made the IBM system you chose a good solution?
The IBM PureData System for Analytics provides for fast, efficient analytics, and has also enabled us to consolidate our system, bringing simplicity and extensibility to our infrastructure.
In addition, although there is a learning curve during the implementation process, from an IT perspective, PureData for Analytics is known for being one of the easier data warehouse appliances to learn. It fits with our already established BI solutions and overall IS infrastructure. It also supports connections with our other data analysis tools, making it both a sustainable and maintainable solution, as well as one that can easily be accessed by our staff.
With the help of professional consulting services from Brightlight, the firm we've worked with on the implementation, we've also been able to take advantage of Brightlight's Netezza Data Integration Framework (nzDIF), a pre-packaged solution that keeps all of the data processing in the data warehouse environment. nzDIF has helped us expedite implementation.
PureData for Analytics has also provided us with the ability to create "sandbox" environments within the platform. The sandboxes allow business analysts and power users to write SQL code, perform data profiling and modeling, and hand their work to a developer to integrate into production data. This partnership will enable us to dramatically increase the number of projects we'll be able to implement over the next several years. In fact, one of our CSW analysts was able to add around 50 inpatient and 50 emergency populations to our portfolio through development in the sandbox. Previously, complex integration scripts delayed deployment and stifled project growth.
In addition, prior to PureData for Analytics, complex queries would take minutes or hours to answer, causing delays and limiting the ability to use data to improve operations and care. Now, some queries that were running in five minutes are taking just four seconds. Other queries that used to take several weeks for an analyst to compile manually into spreadsheets are now accomplished on demand.
Our renewed ability to access such vast amounts of integrated data has not only sped up query response time but provided us with insight into patterns and indicators not previously available. This has helped us to more accurately define patient populations and better tailor treatment patient protocols.
With IBM PureData for Analytics, we will continue to bring in more big data and support hospital and research efforts with a focus on providing the best care possible to our patients for the long term.