Population health can feel like a nebulous buzzword, but health systems around the country — armed with analytics — have started to make good on the promise of big data, using it to drive tangible improvements to care.
These five health systems have successfully used data analytics in specific, practical ways — to drive operational efficiencies in cancer infusion treatment, for example, or to map and tailor pediatric burn treatment to the communities that need it most.
Here are five interesting ways hospitals and health systems are using data analytics to drive population health and the patient experience.
1. Geisinger Health System (Danville, Pa.). As an early EHR adopter, Geisinger Health System has collected patient data for more than two decades, according to a Harvard Business Review article penned by four Geisinger physicians and IT leaders. Two years ago, the health system began combining EHR data with other sources of data, like patient satisfaction surveys and health and wellness apps.
Geisinger cleans the data, normalizes it and breaks it down to the most granular level possible. The system then puts it into what it calls the Unified Data Architecture, which is an enterprise-wide analytic structure that enables the system to break down big data and apply it to practical, point-of-care issues. For example, if a patient is injured in a car accident and taken to a Geisinger emergency room, he or she will be treated for immediate injuries, while underlying issues may go undetected. However, the UDA is able to take in information from clinical notes and diagnostic imaging reports and detect things like abdominal aortic aneurysms.
This advanced analytics ability — scanning information and identifying anomalies — saves patient lives.
2. Children's National Health System (Washington, D.C.). Children's National was one of the first health systems to combine EHR data with geographic information system technology and apply it to population health. One of the system's first applications of this was to identify pediatric burn trends in Washington, D.C., and target interventions by neighborhood.
"Using GIS, we're able to provide clinicians with data-driven answers rather than anecdotal intuition that can positively impact patient care," said Jefferson McMillan, former manager of business intelligence and clinical analytics at Children's National. "We're able to identify health trends or areas of need within a particular region or neighborhood and provide clinicians with data to design interventions to reach those targeted populations."
Children's National's IT department teamed up with its trauma and burn surgery division to map children's burn injuries by location, demographic and characteristic of burn. The team pinpointed an issue in a handful of communities that was leading to an unusually high number of pediatric burn injuries — high water heater settings. They were then able to partner with a community organization to promote water heater safety tips within those communities, and the ED recorded a marked drop in pediatric burn cases.
3. UCHealth (Aurora, Colo.). The University of Colorado Cancer Center in Denver, part of UCHealth, operates 10 infusion center locations in the state and was experiencing double-digit growth in patient volume year-over-year. However, patient volume was not consistent — the centers experienced an influx of patients in the middle of the day and slower traffic in mornings and evenings — meaning patients would sometimes have to wait hours for a chair to open up, according to a case study.
To help address these issues, UCHealth decided to deploy machine learning technology at its infusion center in October 2015. The software combines data science and advanced mathematics to analyze operations and find efficiencies. UCHealth started small — at one center with 28 chairs, according to the case study. "The platform immediately began producing remarkable results," Ashley Walsh, UCHealth's perioperative business manager wrote for Becker's Hospital Review. "Patient wait times dropped by 60 percent during 'rush hour' and, throughout the day, were an average of 33 percent lower. This patient wait time reduction occurred while patient peak time visits increased by 16 percent and overall patient volumes increased by 7 percent. As an added benefit, staff overtime hours in the infusion center dropped by 28 percent."
Once it realized the potential of the program, the health system spread the technology to six other infusion sites that included 104 more chairs, according to the case study. Beginning in May 2016, UCHealth applied the data analytics solution to find other operational efficiencies in its operation room scheduling. It hopes to expand the solution into other areas such as radiology, pharmacy and clinical labs.
4. Banner Health (Phoenix). Banner Health has successfully combined data analytics and telehealth to improve the health of its most at-risk patients — those with five or more chronic conditions. The health system rolled out a pilot Intensive Ambulatory Care program in 2013 to 132 patients with five or more chronic diseases. Each patient received a HIPAA-compliant tablet that enabled them to remotely communicate — via audio and visual messaging — with a care team led by an intensivist primary care physician.
The care team tracked patients' physiologic data remotely with biometric sensors for blood pressure, oxygen saturation, weight and heart rate, as well as fall detection. This enabled care teams to provide daily remote support and intervene at critical moments. The pilot program reduced hospitalizations by 45 percent and drove down the overall cost of care by 27 percent. Since then, the health system has continued to grow the Intensive Ambulatory Care program. The latest in-house study found a 49.5 percent reduction in hospitalizations and 34.5 percent reduction in the overall cost of care.
"We have always taken a long-term view of our business and realized very early on that the current healthcare system was not sustainable," Peter Fine, president and CEO of Banner Health, said in a statement. "With legislation driving reform, we knew that we needed to manage population health and essentially keep people healthy and out of the system to reduce costs, while ensuring better patient outcomes.
5. UNC Health Care (Chapel Hill, N.C.). University of North Carolina Health Care has started to put an emphasis on developing more sophisticated forms of data analytics to inform its move toward value-based care, improved quality and population health. One of its data-driven projects began about two years ago, when it completed the first phase of a project aimed at reducing readmissions for high-risk patients. The health system developed a predictive model based on historic and current clinical data and utilization data. UNC Health used this model to pinpoint what factors would most likely increase a patient's health risk and outcomes.
"[W]e were able to assemble a very talented group of industry experts who worked through over 40 different advanced analytical models and literally hundreds of potential factors to develop a new technique for avoiding hospital readmissions," Jason Burke, system vice president and chief analytics officer of the system's enterprise analytics and data sciences division, said in 2015. "The new technique outperforms every other model we've tested." The technique combines EHR information, patient zip codes and socioeconomic data to assign patients a risk score. In a 2016 interview with Information Management, Mr. Burke said analytics helped improve their ability to estimate a patient's risk of readmission by about 30 percent. "The answer was a resounding yes. We found that those analytic methods provided a dramatic improvement in our ability to predict risk," he said.