According to the National Academy of Medicine (formerly the Institute of Medicine), the U.S. health care system spends almost a third of its resources — $750 billion annually — on unnecessary services and inefficient care. New predictive analytics tools promise to reduce waste and improve care by forecasting the likelihood of an event — for example, a patient being readmitted to a hospital or developing a life-threatening infection — and allowing providers to tailor treatments and services accordingly. These tools are now being used across the continuum of care, from disease surveillance to chronic disease prevention to identifying patients who are at risk of deterioration.
But despite the tools’ power to improve care, most health care institutions are not yet using them. Among the impediments to adoption are the bewildering array of options providers face, from mobile applications to web-based tools to programs that integrate with electronic health records. To better understand what stands in the way of adoption, and what facilitates successful implementation, we interviewed 34 key figures from leading U.S. health systems, policy makers, and predictive analytics vendors. Among our most important findings: Success depends less on the tool itself than on getting buy-in at all levels from the start.
Here are three lessons:
Regardless of whether a provider is developing predictive analytics in-house, as many large academic medical centers have done, or purchasing tools off the shelf, managers should make sure they are involving the right people throughout the entire process. Homegrown tools require special development expertise, and both these and commercial tools require validation, implementation, evaluation, and ongoing improvement. It’s necessary to have a multidisciplinary team, with clinical, analytics, data science, information technology, and behavior change skill sets available from start to finish.
A common reasons these tools are underutilized is that frontline employees don’t fully understand their value. Thus, successful programs start with a problem where predictive analytics can make a clear difference. For example, 50% of newborns with untreated sepsis (blood infection) will die. Therefore, healthy babies are often given antibiotics presumptively — “just in case” — which can lead to complications and increased antibiotic resistance. Clearly, it would be desirable to identify newborns at low risk for infection and spare them the presumptive antibiotics. Kaiser Permanente in Northern California has done just this, using a predictive tool to reduce the use of antibiotics by half without an increase in sepsis-related complications.
Demonstrating the clinical impact of a predictive tool can go a long way toward engaging those who will use them. This is particularly important for clinical staff who may otherwise be skeptical of “black box algorithms,” whose inner workings remain hidden from them. Bringing clinical staff on board early allows team members to influence which predictive tools are implemented and how, and to see early results. While this can be time-consuming, the benefits cannot be overstated. This applies to both commercial tools and those developed in-house. Commercial vendors, in fact, may have to work even harder with staff to develop trust in their products.
Without a clear implementation plan and staff skilled in supporting behavior change, implementation of a predictive tool can stall. We’ve found that health care organizations that regularly used implementation experts to support change and improve quality across a range of IT and other types of projects had a head start when implementing predictive analytics. These individuals work alongside clinicians to map workflows and identify what might need to change when introducing a new process or tool. They may have a clinical background or one in service redesign or quality improvement.
Clinical champions have often proved to be essential in successful predictive analytics implementation — and health IT implementation generally. Any group of change agents should include a subset of well-respected clinicians or other thought leaders in the organization. These individuals should actively reach out to promote the tool, demonstrating its use and educating people about its expected benefits. At one leading public hospital in the Southern U.S., a small number of physicians helped promote the use of predictive models throughout the hospital. Their work gave rise to a center for predictive analytics, and today the institution uses these tools in numerous ways, including to reduce readmissions and to identify patients at risk of sepsis or returning to the intensive care unit.
Just as important as frontline buy-in is engagement from the top, especially from the CEO. Organizational leaders are often unfamiliar with advanced analytics technology and applications. Educating leadership about a tool’s expected benefits is critical in generating support. One large U.S. academic medical center did this by including tool performance measures in the executive dashboard, making its benefits clear to top management. A tool’s value may be quantified in terms of quality improvement, improved patient or clinician satisfaction, or efficiency gains.
One measure that is likely to resonate for management is reduced readmissions among Medicare patients, as hospitals may be financially penalized for readmissions. Models aimed at reducing readmissions among high-risk patients are understandably popular; one model, for example, was shown to reduce the likelihood of readmission for heart failure patients by 26%.
Ongoing attention from senior management is vital for the long-term sustainability of predictive tools; the models decalibrate over time and require regular maintenance. Successful organizations take a lifecycle approach to managing and maintaining these tools, which requires budgeting for long-term resource requirements, including investments in improving data quality and infrastructure, recalibration, and in-house data science and technology capability. Where commercial tools are purchased, costs such as software licenses, consulting, or other vendor-related fees also need to be factored into long-term budgets.
Implementing predictive analytic tools in health care is a means to an end — where the end should represent an improvement in health or health care outcomes, including lower costs. Fully realizing the benefits from a specific tool requires a structured and thoughtful approach, involving the right people, with the right skills sets, at the right time. As we’ve shown, the key to successful implementation has little to do with the model itself. Success depends on the time, effort, and resources set aside for communication, change management, and making the tool a seamless part of user workflow. Clear, committed leadership and a culture strongly supportive of change and learning are also critical factors. Done well, the result can be an increase in high-value care — that is, targeting appropriate health care to those who need it.
Meetali Kakad, MD
Meetali Kakad, MD, is a public health physician and until recently Head of eHealth (CMIO) for a large Norwegian regional health authority. Dr. Kakad was the 2015-2016 Norwegian Harkness Fellow in healthcare policy and practice, based at the Brigham and Women’s Hospital & Harvard Medical School. She is currently a doctoral researcher at the Centre for Health Services Research at Akershus University Hospital Trust in Norway.
Ronen Rozenblum, PhD, MPH, is a healthcare executive, researcher, lecturer and entrepreneur. He is an Assistant Professor at Harvard Medical School and Co-Founding Director of the Unit for Innovative Healthcare Practice & Technology and Director of Business Development of the Center for Patient Safety Research and Practice at Brigham and Women’s Hospital in Boston. He is an expert in patient-centered care, patient experience and engagement, and health information technology.
David Westfall Bates, MD
David Westfall Bates, MD, is the Medical Director of Clinical and Quality Analysis, Information Systems at Partners HealthCare System, Inc. Dr. Bates is Chief of the Division of General Internal Medicine and Primary Care at Brigham and Women’s Hospital. He is a Professor of Medicine at Harvard Medical School and Professor of Health Policy and Management at the Harvard School of Public Health, where he is the Co-Director of the Program in Clinical Effectiveness. He served as External Program Lead for WHO’s Global Alliance on Patient Safety and is former board chair of the American Medical Informatics Association.
This article originally appeared on HBR.org and is being brought to you by Medtronic.