Monday, January 21, 2013

Healthcare, big data and analytics

Before I began my MBA program here at the University of Utah, I worked for three and a half years at Epic, one of the leading vendors of electronic medical records supporting their relational database/reporting platform.  There, I worked with customers to set up and maintain extracts from the live system into RDBMS platforms for reporting.  For the last six months, I have been working as a consultant doing reporting and database administration for Singing River Hospital in Mississippi.  I have supported customers of all different sizes: from the relatively small ones like St. Joseph's Hospital West Bend Clinics with databases of only a few gigabytes, to huge ones like Cleveland Clinic, Kaiser Permanente, and Sutter with databases of several terabytes.  Since I work in the healthcare industry, I wanted to discuss how healthcare organizations are currently using big data and analytics and where they are headed in the future.

Healthcare reporting up to nowd

Analytics and big data in healthcare are in their infancy.   When I was at Epic, most customers were doing one form of reporting or another, varying in their levels of sophistication.  Some new customers were just starting to use the data for basic statistical reports while more experienced customers were deeper into the data to identify, monitor, and improve care for different patient populations.  Studies that previously would have taken months and cost hundreds of thousands of dollars could now be done in a few minutes by one doctor who was familiar with SQL.  However, no customer that I knew of was using the data for any kind of predictive analytics. 

Epic's biggest selling point has been its integrated system [1], which combines clinical and financial data into one medical record, which physicians and nurses across the entire organization could access.  Instead of the patient having a chart in each office or hospital with different, each patient has one chart containing his or her entire patient history.  This in and of itself is great, and maybe I'm biased because I work on the reporting end of things, but I think going forward the greatest benefit EMR's will give to healthcare organizations is the trove of data they generate and the insights that the organizations will be able to glean from it. 


One of the reasons that healthcare organizations haven't leverage big data are the challenges that healthcare data presents.  For those who are not familiar with big data, it is generally described using the “3 v's” model [2]: volume, variety, and velocity.  In other words, data sets that are so large, varied, and change so quickly that they are impossible to work with using traditional data marts. While it doesn't have to volume or velocity of web data, healthcare data is extremely varied [3].  Last time I checked, Epic's relational database contained over 15,000 tables and about 130,000 columns.  Many healthcare organizations have data coming from multiple systems as well, further complicating things.

Recent innovations

Organizations are beginning to find solutions to these problems and make use of big data.  Explorys, a spinoff of Cleveland Clinic, combines clinical, operational, and financial data in a cloud based architecture with massively-parallel data processing [4] to provide its customers high-speed data access [5].  They currently serve 13 major healthcare systems [6].

Recently, UPMC [7] selected Oracle [8] as its software vendor in a $100 million analytics project [9].  Dr. Steven D. Shapiro, UPMC's chief medical and scientific officer explained “Today’s healthcare institutions have access to unprecedented types and volumes of data that have the potential to unlock the secrets of human health and disease leading to new and highly personalized care pathways.”

Some organizations have already used predictive analytics to reduce hospital re-admissions [10].  Parkland Health and Hospital system in Dallas developed a predictive algorithm to identify in real-time heart failure patients at high risk for readmission or death.  With this, they have been able to reduce heart failure readmissions by 20%.

Future direction

This is just the beginning.  As time goes on, more and more healthcare organizations will adopt similar techniques and build upon the work that has previously been done, increasing the overall knowledge base of diseases and treatment.  As computing power and storage capacity increase, genomic data will be able to be combined with medical data for analysis [11], allowing doctors to personalize treatment to each patient based on his or her genes.  I'm excited to see what the future holds.