When Mark Hays joined UHealthSolutions Inc. as chief technology officer (CTO), it quickly became clear that the not-for-profit affiliate of the University of Massachusetts Medical School needed to implement a new data warehousing project and predictive analytics initiative.
The company, which is based in Worcester, Mass., and provides health care software and consulting services to UMass Medical School and other institutions, realized that to improve patient care, many of its clients would need a better system for analyzing health care data and predicting problems before they occur. But it wouldn't be easy.
The complexity of health care data was exploding as physicians and hospitals began switching from paper charts to electronic medical record (EMR) systems. At the same time, that data was arriving faster over the Internet. Quickly gaining valuable insights from the information meant moving away from monthly batch reports and focusing more on real-time data processing.
Mark HaysCTO, UHealthSolutions Inc.
Hays and his team ultimately launched a new "health care delivery intelligence platform" from scratch. The new Healthcare Explorer system, as it became known, boasts ease of use, flexibility and the ability to quickly deliver much-needed reports with "zero training time," the company claims.
In this SearchDataManagement.com email interview, Hays talks about the scope of the Healthcare Explorer project and the key challenges his team faced along the way. He also gives some sage advice for any organization mulling a similar business intelligence, predictive analytics or data warehousing project.
What kind of problems were your company and UMass Medical School experiencing that led to this project?
Mark Hays: Various business units had developed independent databases-- silos that did not share an efficient foundation or common set of tools. This led to increased costs and a lack of flexible reporting for management. Some groups were building batches of reports by hand, every month, with Microsoft Excel. Automation and standardization were needed. This led to one of the initial challenges, [which was] the need to integrate a series of legacy system databases.
Could you give me an idea of the timeline for the project?
Hays: We launched our Healthcare Explorer project in early 2009, with a heavy focus on design. The first version was delivered internally in June of 2009, and the first client implementation went live in February of 2011 for a total design [and] delivery cycle of roughly two years. A new release that integrates digital fax files, scanned documents, workflow automation and call center data will go live in December [of this year].
What would you say was the most interesting or unique aspect of this project from a data management perspective?
Hays: The need to integrate a flood of new data sources was definitely the most interesting challenge. [For example, we needed to create] near-real-time data feeds from EMR systems. This impacts the basic design of the data warehouse itself, and the data loading process. The fundamental design of the database schema needed to support the relationship between a patient and all of the organizations in a coordinated care environment, which also change over time. Secure Internet-based [electronic data interchange] interfaces were needed, in addition to traditional monthly batch files, plus the ability to export data via EDI to new care management systems. In effect, the foundation of our solution had to provide a “health information exchange” (HIE) that can connect to any external system, in any format, at any time.
What software vendors did you use?
Hays: The basic architecture of the system is built around the Microsoft SQL Server platform and compatible tools, for performance and cost. The evaluation process was much more challenging for the [extract, transform and load (ETL) and EDI] system, and the business intelligence (BI) reporting layer. We selected [Pervasive Software Inc.] and their iHub system to automate data intake, QA and data exchange. Pervasive offered an automated QA system to ensure data quality and reduce operational costs, and the EDI interfaces available with iHub are very helpful for new links to EMR and [health care information systems]. On the BI [and] reporting side, [we were] fortunate to find two key vendors [that were] willing to offer special not-for-profit pricing to meet our unusual ‘broad distribution’ needs.
What are some of the key technologies you developed during this project?
Hays: We developed two key innovations in-house: PIDA, a new predictive analysis solution, and our EMPI (enterprise master patient index) system. Both are based on advanced artificial intelligence technology. [The new] predictive solution is 200% to 300% more accurate than standard commercial products. This jump in accuracy allows us to forecast risk for individual patients and to drive more effective care and disease management. This is what hospitals and health plans need to improve care and reduce costs. With patient data flowing in from many different locations and systems, EMPI is a requirement. Patient records from different systems typically do not share a common identifier or ID code. The EMPI system uses artificial intelligence to match records and [enable] a comprehensive view for each patient.
Could you describe the scope of the data warehousing project?
Hays: Our new ‘health care delivery intelligence’ solution went live in February of 2011, managing years of pharmacy claims and operational data for a state pharmacy assistance program in Massachusetts that serves elders and families across the commonwealth. We are currently in the process of deploying this solution with another major program, and working with clients and prospects in states across the nation.
What would you say are some of the key ‘lessons learned’ from the data warehousing portion of the initiative?
Hays: Never underestimate the challenges you will face with an innovative effort. There are many unknowns, both internal and external, that can impact your schedule and budget and threaten the success of the project. This is particularly true if your client or organization is not familiar with advanced data warehouse [and] business intelligence systems. A lack of shared understanding can breed concern and resistance. It is better to over-communicate and schedule repeated demos for stakeholders, even when it seems that the goals and benefits of your project should be obvious. Always explain what you are presenting for those in the audience who may not be familiar with basic technology.
What is one piece of closing advice you would have for a technology professional undertaking a similar data warehousing and predictive analytics initiative?
Hays: Start by looking beyond the proverbial box and standard assumptions. With all of the changes in competition, risk and technology in virtually every industry, what worked five years ago is not sufficient today. Design to provide insight into what is coming, not retrospective analysis of results from last year. Your clients are changing too. Think mobile and iPad, not just PC and PDF. These opportunities create new challenges.
Find out how Disney discovered a whole new world of data warehouse best practices
Learn to create a data warehouse strategy that supports "big data" analysis
Read a Q&A with Forrester analyst James Kobielus on traditional data warehouse concepts vs. Hadoop