With sophisticated real-time and predictive data analytics systems, clinicians are empowered to plan more timely, targeted and effective medical interventions and ultimately improve patient outcomes and reduce avoidable hospital admissions. However, many hospitals and health systems face obstacles to implementing predictive analytics programs.
Programmers have designed advanced algorithms to address a variety of conditions and issues across the care continuum, such as reducing patient readmission rates, identifying patients at-risk of heart disease and preventing hospital-acquired infections. Much of the early results in these areas are promising, but challenges in efficiency, security and interoperability have created barriers to achieving widespread success.
Hospital systems have encountered three main obstacles to implementing predictive analytics solutions.
- Real-time insights. First, delivering information to physicians in real-time is crucial to helping them practice evidence-based medicine. Some algorithms in their current state are challenged to produce real-time or near-time insights when dealing with complicated or multidimensional data.
- Scalability and security. Predictive analytics produce more valuable results if they can be applied across large populations, but the analytics systems must remain in compliance with HIPAA. Keeping vast amounts of data secure from people with malicious intent, however, is still a formidable task.
- Systemwide integration. Integrating health data across multiple databases is challenging because of interoperability issues between IT systems. In healthcare, interoperability describes the extent to which different IT systems and software applications can communicate, exchange data and use that shared information. It is often easier to achieve a high level of interoperability between solutions from the same vendor because they were designed to work together, but getting two different vendor solutions to exchange and use that shared information is much more challenging.
Hospitals and health systems across the U.S. experience these challenges to varying degrees depending on their goals, the data analytics systems they deploy and the scope of their organization, among other factors. Nevertheless, health systems have achieved substantial progress in improving care coordination and outcomes through real-time and predictive analytics.
The following three healthcare organizations have demonstrated some of the ways providers are applying predictive analytics to patient care. Because some of these organizations only recently implemented their data analytics systems, it is too early to tell exactly how much predictive analytics can improve patient outcomes in the long term. However, clinical leaders and IT chiefs are optimistic about the systems' potential.
SCL Health
Broomfield,Colo.-based SCL Health has adopted predictive analytics software to help it combat sepsis, which is of increasing concern to hospitals and health systems throughout theU.S.Sepsis accounts for more readmissions than any of the four conditions CMS tracks for reimbursement purposes, including heart attack, heart failure, chronic obstructive pulmonary disease and pneumonia, according to a research letter published in JAMA. Researchers estimate the average cost per sepsis readmission is approximately $10,070.
Sepsis is a deadly condition, as it often goes undetected until it reaches more severe stages when it is more difficult to treat. Therefore, predictive analytics that alert clinicians to signs of sepsis as early as possible, or ideally even before they occur, could help hospitals save lives and protect the bottom line.
With this in mind, SCL implemented an alert system in its EHR powered by predictive analytics that notifies clinicians when a patient is at risk for developing sepsis. John Middleton, MD, vice president and CMIO of SCL, said the alerts have helped the health system achieve some progress, but notes that obstacles to maximizing the system's potential remain. Particularly, SCL has encountered difficulty using the platform's statistical insights to determine course corrections or best practices to reduce or avoid readmission for a patient.
"We are early in our experience of using vendor-supplied predictive analytics regarding readmission risk," Dr. Middleton says.
Mercy Hospital Fort Smith
Like many hospitals across the country, Mercy Hospital Fort Smith (Kan.) has experienced rising patient demand and crowding in its ED. The hospital's ED serves as the primary access point for patients and community members seeking medical care, generating between 60 percent and 70 percent of inpatient admissions. Swelling patient demand compounded with fundamental inefficiencies in ED operations challenged the hospital's ability to treat and move patients through the system.
The address this challenge, Mercy Fort Smith implemented a data analytics platform to optimize its ED operations, improve patient experience and lower the number of frustrated patients who ditch the ED due to excessive waiting times. The hospital implemented an "air traffic control" platform as a tool to help guide ED staff to best practices through specific algorithms designed to monitor metrics, predict bottlenecks and recommend countermeasures in real time.
For example, algorithms in the software integrated data about patient volumes from the EHR, claims database, weather forecast and other sources to predict ED surges before they happened and send an advance warning to catalyze actionable conversations among key interdisciplinary staff leaders. These conversations take place in a virtual huddle chat room, allowing staff to coordinate operations via text and prevent potential bottlenecks.
Mercy Fort Smith saw creative problem solving, efficient habit formation, and collaboration on the frontline and among interdisciplinary staff in just 5 months. The hospital achieved its goal of improving key performance metrics in the ED. Mercy Fort Smith's left without being seen rate dropped 30 percent. The average length of stay for discharged patients fell to just 24 minutes, a 13 percent reduction, and its door-to-doctor time dropped by 15 minutes, a 20 percent reduction. This enabled the hospital to serve an additional 2,500 patients and generate an additional $1.3 million annually.
Penn Medicine
To help physicians intervene early in patient care, Philadelphia-based Penn Medicine launched an analytics program in February 2017 that forecasts which lung cancer patients are at risk of ending up in the ED.
Penn Medicine's research team is creating a predictive formula that uses data from recent lab tests, radiology visits and patient-reported symptoms to help the health system assess at-risk patients and respond accordingly, such as by introducing home care treatment or recommending more frequent follow-up appointments.
In addition to the predictive analytics, Penn Medicine opened the Oncology Evaluation Center, which creates a specialized treatment plan for each patient based on the information aggregated from the predictive analytics software. Like many of the aforementioned organizations and health systems around the country, Penn Medicine has not yet seen measurable improvement in patient outcomes and spending that could be attributed to predictive analytics. However, its leaders are confident that predictive analytics will be a key component of preventive medicine in the future.
The potential for predictive analytics
Under value-based reimbursement, enhanced focus on population health management practices that emphasize preventive medicine are integral to success. With predictive analytics, providers have access to sophisticated insights that help them identify and target the patient populations at greatest risk for chronic conditions, avoidable hospital admissions and other healthcare challenges.
As health systems increasingly incorporate such systems into their IT infrastructure, data analytics providers are working on ways to improve the usability and efficacy of these systems.
More articles on health IT:
5 must-haves for effective analytics in clinical decision-making
How the right data analytics diminish administrative burden on clinicians
A closer look at 5 health systems' interesting, data-driven population health initiatives