Business intelligence (BI) has rapidly become a mission critical necessity for healthcare organizations, and is now foundational to the delivery of value-based care including: reducing costs while increasing quality, managing population health, improving process efficiency and patient outcomes.
It is not uncommon to see healthcare organizations unknowingly restrained by their limited understanding of the potential of mining BI data to support front line clinical staff and finance professionals.
Challenges that existed merely a few years ago - acquiring BI analysts, mining data, crunching and presenting data - no longer exist for organizations at the forefront of BI. Other, broader challenges now exist in leveraging data as an enterprise asset: how to deliver on the vision of producing contextual BI organization-wide, getting the right data to the right person at the right time. To make good on such a vision, an organization must have systems in place to bring forth highly precise, intuitive and actionable insights that teams and individuals need to efficiently improve the healthcare delivery.
Three prevailing barriers exist, and must be overcome, to realize the vision of contextual BI for clinical and financial teams, and employees. It has to be easy, low-cost, and fast. Historically, these adjectives have rarely coexisted in the same sentence with BI.
That is, until recently when I sat down and spoke with Michael Dulin, MD, PhD, Chief Clinical Officer for Analytics and Outcomes Research for Carolinas HealthCare System CHS. CHS' advanced analytics capabilities are leading the way, turning BI into a highly visible discipline within the organization. CHS is formalizing its BI center of excellence under the brand of Dickson Advanced Analytics (DA2). DA2 is disrupting conventional wisdom, and developing best practices by demonstrating just how contextual BI can be easy, low-cost, and fast.
Q: Dr. Dulin, could you tell me the story of how Dickson Advanced Analytics came into existence and arrived at its current charter today?
Dr. Dulin: About three years ago when we saw the concussion waves of disruptive payment models coming our way, we knew our business model required rapid transformation. The challenges ahead were deciding how to improve the delivery of care to individuals, and families, without having helpful operational metrics on their clinical or financial experiences. During this time, we had our glimpse into the future and realized that having near real-time analytics at the point-of-care was an essential requirement needed for our system to advance the triple aim: better care, better health, and better efficiency [lower cost].
We knew the solution would require major investments in both tools and human capital to give us the next-gen analytical capabilities we needed to transform our culture and business model. So, in 2011, Carolinas HealthCare System made an unwavering commitment to address the bigger picture, and created the Dickson Advanced Analytics group with the goal of building the centralized analytical expertise and capabilities that our organization needed. Fortunately, we started with a number of excellent teams members from across the organization that had analytical talents. For example, within the Dickson Institute for Health Studies, which had been focused primarily on improving quality, we had 20 analysts and statisticians. This number quickly expanded to over 100 people who had background skills in a full host of analytical and BI tools.
Our first step was to design and build a new Enterprise Data Warehouse (EDW) that would be created from the ground-up to support our current and future data needs. The Carolinas HealthCare System EDW brings together financial, clinical, operational, and social data to produce the operational metrics and indicators from which we can harvest and deliver actionable insights to every team member.
Q: I understand that Dickson Advanced Analytics Group has several key priorities, with the overarching goal of bringing better and cost-efficacious care to patients and families. Can you tell me how your operational performance metrics are beginning to make an impact on the experience and outcomes of care?
Dr. Dulin: In the past, operational metrics were static reports and did not reflect things like cost and efficiency to the degree needed. To deliver high quality, high value patient care, we would need to understand not just the number of discharges, but the true cost of providing care to our patients; understand the drivers of that cost; even just to understand the number of patients that were actively being cared for within our primary care medical homes. Our focus now is to address the overall cost of care, number of patients we can serve – those are the types of business questions that are driving our data models. We also look at the cost of care so that we can evaluate financial practices as they relate to clinical performance.
Q: What about population health management? Are you able to leverage your data warehouse and analytical capabilities to get your arms around identifying, tracking, and managing health at the population level?
Dr. Dulin: At the start of the Dickson Analytics group, we knew some easy wins would be important. For example, we needed to clearly define the number of unique patients receiving care across the system and who the attributed care provider was for each of those patients. From there, we could analyze how effective each provider and clinic was making sure that evidence-based care was being provided. Since each primary care, specialty clinic, and acute care facility has a shared electronic health records system, the EDW can integrate the clinical data with prevention-specific, claims, and registry data. This allows us to inform providers about how they are doing against other preventive care and screening targets. Along with this score, we can also deliver a list of patients along with their preventive care and screening gaps. This enables providers to achieve their preventive care metrics with the greatest precision and least expenditure of labor.
Q: The Institute for Healthcare Improvement worked intensely with 60 healthcare organizations for over two years to drive out waste, reduce operating expenses, and improve quality. Interestingly, they found that the one success factor that was common across organizations most successful at reducing waste and operating expenses while maintaining or improving quality was collaborative partnerships between clinical and financial teams. Have you found that DA2, in some way, has been a catalyst that has forged new partnerships between financial and clinical teams to drive out waste and reduce operating costs?
Dr. Dulin: Forging clinical and financial partnerships was part of our team’s design from the beginning. We deliberately put the clinical and some components of financial data in one place to allow us to better understand both our opportunities and the impact of investments designed to improve health outcomes. The data model that we used for the EDW will provide information on the care episode costs as well as quality metrics. This capability will be of key importance as we start to take on more risk-based contracts with insurance companies.
Q: How did you become aware of Predixion and what did this tool bring to the table? How did Microsoft fit into the equation?
Dr. Dulin: A recent CMS analysis showed that, in the coming fiscal year, about 2,200 hospitals will see reductions of up to 2% of their Medicare payments because they were not able to reduce their 30-day readmission rates. Identifying the patients at risk for unplanned 30-day readmissions is the easy part. What’s difficult is providing precise and actionable information—at the point–of-care while the patient is still in the hospital—I believe that this step is a key component of reducing readmission risk. Our predictive model is providing the needed actionable information that will sustain and enhance our performance in readmission metrics.
Predixion shared this same vision with us. Predixion focuses on the last mile of analytics –delivering the predictive analytics to the care teams that need to act upon them. The company offered an easy-to-deploy, self-service predictive analytics solution that integrates with a majority of business intelligence platforms, business applications and clinical systems.
We saw that we could benefit from Predixion’s expertise in readmission analytics; a partnership with them made perfect sense for our team. Predixion truly understands readmission risk modeling, and has the capability to pull the right variables out so that care teams can understand what additional support the patient needs while they are still in the hospital.
One of the other value adds that Predixion brought to the table was its expertise in Microsoft analytics and business intelligence technologies. Because our goal was to deliver precise and actionable information to providers at the point-of-care, the learning curve had to be short and flat. We had to use tools that frontline clinicians and managers used every day and were already familiar with – like Outlook, SharePoint, Excel, and Lync. In a lot of ways, the Microsoft’s analytic tools were a lot easier for clinicians to use than a lot of the clinical tools we have in place. And on the back end, Predixion leverages the tools that are familiar to our analysts and database administrators—like SQL Server Analysis Services, and PowerPivot.
We benefited from the familiarity of Microsoft’s tools and the speed at which we were able to roll out our readmission prevention analytics program to the seven main facilities that are on the standard EMR system—the acute care hospitals. It took us less than four months to roll this tool out to all seven hospitals. We were able to roll this out faster than any previous system of this magnitude.
Q: How do you integrate intuitive and actionable into the context clinicians’ workflow?
Dr. Dulin: Right now it is a dual interface; the care manager is able to pull up Predixion interface side by side with the EMR. Using the Predixion interface, we present the different variables in the predictive model to the clinicians along with possible interventions. We rank these variables with intervention so the clinicians can precisely apply those interventions and, later, learn which were most effective.
This side-by-side approach enables us to share actionable information at the point-of-care with the web version until we feel we have the process right. We can later go back and embed it in the EMR directly. This way the EMR remains stable and consistent during the iterative process.
Q: Are you considering other use cases for the Predixion solution such as preventing in-patient admissions and ED admissions?
Dr. Dulin: Yes, we’re currently exploring predictive data models within the ambulatory environment to prevent avoidable admissions. For certain chronic diseases and co-morbidities we can see patterns where early interventions can help. Eventually, we will need to answer how we can keep people healthy and out of the emergency room. The team is currently thinking about these use cases and the resources needed to support them.
We’re also looking at other scenarios like patient safety, medical reconciliation, and infection control–where the human and financial costs are high, yet avoidable. We’re also looking at genomics and antibiotic stewardship to see how we can change processes in the hospital to reduce adverse drug events and antimicrobial resistance.