This healthcare big data project’s tools and predictive models involve real-time monitoring of patient data and are expected to be available soon to other to providers
One healthcare big data project has begun to report progress on using predictive analytics to improve patient care in the diagnosis and management of such health conditions as sepsis and heart failure. This pioneering effort is being done at the University of Pennsylvania Health System’s (Penn Medicine’s), Institute for Biomedical Informatics (IBI).
What will be of high interest for pathologists and clinical laboratory executives is how this big data project incorporates lab test results into the effort.
Recently, Penn Medicine announced Penn Signals, a big-data project that, in part, relies on the lab data housed in the academic medical center’s laboratory information system (LIS) as well as its outpatient and inpatient data house in its electronic health record (EHR) system.
Early Penn Signals Accomplishments
Penn Medicine includes the Perelman School of Medicine, the Hospitals of the University of Pennsylvania-Penn Presbyterian, and other hospitals, hospice and outpatient services, and physicians. According to a Pennsylvania Medical Society (PAMED) article, the Penn Signals project has, thus far, accomplished:
1. A sepsis early warning system; and
2. Detecting patients trending toward heart failure.
Goal to Leverage Available Data
The Institute for Biomedical Informatics (IBI) was established by Penn Medicine in 2013. It oversees research and use of big data techniques. Penn Medicine’s latest goal, according to an article in Health Data Management, is to harness the power of big data to develop predictive analytics models and methods for spotting illnesses in their earliest stages, or even preventing them from occurring in the first place.
The IBI describes its Penn Signals big-data initiative as a system for real-time monitoring of patient data for clinical decision support.
“Our goal is to build an infrastructure that can scale up to handle a huge variety of data sources within our system that contain information about the health of our patient population. We started with the obvious candidates—our electronic health records and labs—to try to develop predictive models for severe sepsis and heart failure,” stated Corey Chivers, PhD, Senior Data Scientist, Penn Medicine, University of Pennsylvania Health System, in the Health Data Management article.
Penn Medicine plans to release Penn Signals, an open source real-time application platform, to other organizations in June 2016, Information Week reported.
Build of the Data Warehouse
More than three million patient records over a 10-year time span comprise Penn Signals’ data warehouse, noted FierceHealthIT.
Data for the warehouse originates from these clinical systems, according to Health Data Management:
• an outpatient Epic-brand EHR used by 1,800 affiliated doctors;
• the inpatient Allscripts EHR, used by Penn Medicine’s five hospitals;
• and the Cerner enterprise LIS.
Penn Medicine data scientists use these data to build care-pathway prototypes, test them with patients, and reflect those results in algorithms, explained an article in CIO.
(From L to R): Penn Medicine’s Brian Wells, Associate Vice President of Health Technology and Academic Computing; Michael D. Feldman, MD, PhD, Associate Professor of Pathology and Laboratory Medicine; and, Mike Restuccia, Vice President and CIO, following the Healthcare Informatics (HCI) 2016 Innovator Awards Program. HCI’s editors selected Penn Medicine as “Co-third-place Winning Team” for the healthcare system’s contribution to the field of Translational Medicine. “[Pathologists] can dissect health system processes and refine those because [they] now have the ability to look at that business data in combination with genetics,” Feldman told Healthcare Informatics. “It will allow us to refine how we deliver care based on genetic variation within disease groups. I think that is a pretty novel way of thinking about precision medicine,” he concluded. (Photo copyright: Healthcare Informatics.)
For example, the algorithm to detect indicators of sepsis risk recognized the blood infection’s thresholds of temperature, heart rate, respiratory rate, and white blood count, CIO noted.
The Penn Signals predictive model also acknowledged 200 clinical variables and enabled Penn Medicine to detect 80% of severe sepsis cases within 30 hours of the onset of symptoms, Health Data Management reported.
“We’re creating machine-learning predictive models based on thousands of variables. We look at them in real-time, but we train them up over millions of patient records,” explained Mike Draugelis, Chief Data Scientist, Penn Medicine, in the CIO article.
As to heart failure, an algorithm made it possible for Penn Medicine to find 20% more patients likely to trend toward the disease, and people who are five times more likely to be readmitted to one of the system hospitals after heart failure, reported the PAMED article.
Penn Medicine’s Michael Draugelis and Intel’s Ron Kasabian delivering the keynote address, “Improving Medical Decision Making with Predictive Analytics on Big Data,” at the 2015 Strata + Hadoop World data conference in New York City. (Video copyright: O’Reilly Media.)
Reducing Readmissions in Carolinas
Penn Medicine is not the only leading healthcare system eyeing big data’s importance to preventing acute and critical episodes and inappropriate readmissions. Carolinas HealthCare System (CHS) in North and South Carolina is also mining large quantities of clinical data to identify useful patterns for clinical intervention.
In fact, Dark Daily explored the Carolinas’ efforts to leverage technology and mine data about two years ago. (See Dark Daily, “Hospitals Mine Clinical Data to Help Reduce Costs and Avoid Readmissions, Creating Opportunities for Clinical Laboratory Pathologists to Contribute to Improved Patient Outcomes.” April 4, 2014.)
The Dark Daily e-briefing reported on work by the Dickson Advanced Analytics Department, CHS’ in-house data arm, which launched in 2012.
CHS has reportedly used big data to: 1) rate hospitalized patients based on their risk and to change care delivery as data suggest; and 2) intervene with more than 150,000 patients based on risk score.
Labs Data Invaluable to Analytics
As analytics projects at Penn Medicine and the CHS suggest, lab data are key to important decisions that ultimately affect clinical care and patients’ lives. Pathologists and laboratory administrators must ensure they have the staff and IT resources necessary for capturing and coordinating patient diagnostics data throughout the care continuum.
While some healthcare systems are beginning to build electronic data warehouses and hire data scientists, others are busy tapping information and saving lives.
It’s important pathologists and medical laboratory administrators acknowledge the value and progress in application of lab data.
As coordinator of diagnostic data, a clinical laboratory is a powerful place. The big data projects by these leading healthcare systems suggest opportunities indeed exist for labs to leverage data and contribute to improved patient care as well as reduced costs.
—Donna Marie Pocius
Related Information:
Penn Health Sees Big Data as Life Saver
Using Big Data to Save Lives in Pennsylvania
Penn Medicine Taps Big Data to Save Lives
Analytics a Catalyst for Clinical Evolution at Penn Medicine
Big Data Makes a Difference at Penn Medicine
Carolinas Healthcare Using Big Data to Better Care for Patient Populations
The 2016 Healthcare Informatics Innovator Awards: Co-Third-Place Winner—Penn Medicine
ACOs Are Learning to Use Big Data
Pathologists May Be Healthcare’s Rock Stars of Big Data in Genomic Medicine’s ‘Third Wave’
New York’s Mount Sinai Medical Center Using Big Data to Improve Clinical Care
Labs have been doing this on a smaller scale for years. We use lab results from the local population to make our test ranges and and yearly statistics on antibiotic susceptibilty testing to provide physicians with antibiograms as a decision-making tool. I think it’s great that others involved in the healthcare system are starting to think like us.