Predictive analytics involves extracting information from data and using it to forecast the future based on existing patterns and associations. It has been used to optimize existing processes, better understand customer behavior, identify unexpected opportunities, and anticipate problems before they happen. For managed care organizations, the potential is limitless.
“By pooling all data resources, health insurers can look at a population’s risk factors and make all types of predictions, such as what diseases patients are most likely at risk for in the future,” says Kristen Kelley, MPH, CIC, CLC, director, Infection Prevention, Indiana University Health, the largest healthcare system in Indiana.
As the senior vice president of Healthcare Analytics at Indianapolis, Indiana-based Anthem, Inc., Patrick McIntyre sees significant opportunities to better leverage big data to support optimal decision making. “Payers have historically only used claims data for predictive analytic models to help better guide members’ care,” he says. “However, with technological advances, we can better leverage clinical data (i.e., EHRs, laboratory results, etc.) as well as psycho-socioeconomic data to predict members’ healthcare needs.”
To capitalize on these emerging capabilities, Anthem has launched “member clinical risk modeling.” It applies data science techniques and tools, such as machine learning and pattern recognition analytics, to improve clinical risk assessment results through more precise clinical models. “These new models, when implemented in a big data environment, offer the ability to develop and score predictive analytic models in a matter of minutes rather than weeks,” McIntyre says. “Combining new models with big data capabilities has resulted in more timely and precise clinical models that we now use for our plans to achieve a more proactive and targeted outreach to assist members in making healthcare decisions.”
For example, when a member diagnosed as diabetic hasn’t been filling his prescriptions for insulin or other medications Anthem health plans have this information in their claims data. “We can analyze that data and share the information with the member’s primary care physician to alert them of a potential risk,” McIntyre says. “That physician now has the opportunity to directly contact that member, and perhaps prevent an emergency medical event that could lead to a costly hospitalization or emergency room visit.”
In addition, Anthem has been leveraging big data to help law enforcement and federal agencies recoup tens of millions of dollars lost to fraud, waste, and abuse (FWA). These FWA recovery efforts are projected to grow into the hundreds of millions of dollars over the next several years. In 2016, Anthem implemented real-time analytics leveraging big data tools and techniques to detect potential FWA claims before payments were made, which accelerated its investigative activities.
Richard Clarke, PhD, vice president of advanced analytics and reporting, Highmark Health, Pittsburgh, believes that the most valuable analytics lead to differential business decisions that deliver more value to members. “As an industry, healthcare is data rich, but often insight poor,” he says. “Highmark is focused on ensuring that its analytics (both standard reporting and advanced predictive/prescriptive models) are connected to and integrated with our business processes. We are seeing examples in which tangible value is delivered through advanced analytical techniques that leverage new and diverse data sources. This could be incorporating social determinant data to better predict what intervention will deliver the most impact for a member, blending clinical and claims data to prescribe the best transition of care plans, or predicting which member is most likely to have an avoidable emergency room visit in the near future.”
In the past decade, health systems have made a major push to get a handle on their clinical data. “As the Affordable Care Act fostered risk-based arrangements, health systems have increasingly developed the necessary resources to better understand their populations’ needs and issues,” says Adam C. Powell, PhD, president, Payer+Provider Syndicate, an operational and strategic healthcare consulting company.
The Healthcare Information and Management Systems Society’s Electronic Medical Record Adoption Model suggests that many health systems now have relatively mature EHRs. “By having more comprehensive data, they can make more accurate predictions,” Powell says. “We have had recent advances in interinstitutional data sharing, which is increasingly enabling providers to incorporate externally-generated clinical data into their models.”
Despite these advances, Bill Fox, JD, MA, vice president, healthcare and life sciences, MarkLogic, a data integration platform, says the pressure is on health systems to make analytics more actionable. “We are at a stage in which payers, providers, and hospital systems—and even different areas within those systems—might each use different purpose-specific data to meet their unique requirements,” he says.
For instance, one data warehouse may drive cardiac disease management and another warehouse may drive diabetes management, creating two unnecessary silos and a costly and time-consuming project to integrate them, despite the fact that these conditions are often seen together in comorbidity populations. “The end result has been an exponential growth of data silos that create real data governance challenges and undermine the return on investment of predictive analytics solutions,” Fox says. “Health systems can reduce these governance issues and gain more value from their analytics solutions by implementing operational data hubs. Within these hubs, all data reside in one place and can be leveraged for both analytic and operational purposes, thereby improving data integrity, security, and data management challenges—while allowing information to be leveraged more cost effectively across hospital systems and departments.”