In an increasingly digitized healthcare world, it is widely understood that data and analytics are the backbone to many clinical and operational improvement efforts. At Mission Health, a seven-hospital health system based in Asheville, North Carolina, senior executive leaders have created a culture of continuous improvement grounded in analytics.
“To drive continuous improvement using analytics, the equation involves the people, the process in which those people are working and the data and the technology that support the people and the process. The analytics are really only just one part of the story, but at Mission Health, I think we’ve shown that they can be a critical component, a catalyst if you will, to drive the reaction of improved outcomes, clinically and operationally,” Chris DeRienzo, M.D., chief quality officer at Mission Health, says.
Mission Health serves communities across 18 counties in western North Carolina with 800 employed providers across 140 practices as well as an accountable care organization that includes hundreds of physicians and more than 90,000 patients. By leveraging analytics, Mission Health has made improvements in areas such as readmissions reduction efforts, improving sepsis, stroke, and heart failure outcomes, and scaling the preventative care needed to succeed in ambulatory population health.
Foundational to this work, DeRienzo says, is having a reliable enterprise data warehouse and analytics environment, as well as clinical program leadership and a team of Lean engineers. The first step, he says, is to align analytics efforts with the organization’s core purpose, what Mission Health executives refer to as the “Big(ger) Aim”— “getting every person to their desired outcome, first without harm, also without waste and always with an exceptional experience for each person, family and team member.”
During an interview with Healthcare Informatics, DeRienzo outlined several analytics projects that demonstrate the organization’s progress, to date, in creating a culture of data-driven continuous improvement and harnessing analytics to drive clinical and operational performance.
And, as a result of this work, Mission Health has seen, as of February 2018, an across-the-board drop in readmissions, a 58 percent increase in sepsis detection; a 32 percent reduction in severe sepsis mortality rates; a 20 percent increase in on-time surgery starts; 12 lung cancer deaths avoided and a 37 percent in lung cancer screening; and other improvements across population health outcomes—all while realizing a 64 percent reduction in staff hours to collect data and prepare reports.
Data-Driven Care Process Models Drive Results
In an effort to address clinical variation, Mission Health clinical and executive leaders set out several years ago to develop care process models (CPMs). “We started by bringing together the clinical teams to help identify the best practice, the informaticists to help bake those best practice workflows into the electronic health record (EHR) in as frictionless a way to use as possible, and then an analytics team to measure, not only utilization of the pathway but also the outcomes that we’re trying to drive. That’s the core recipe of a CPM,” DeRienzo says.
In the first nine months of 2017, 24 CPM teams were created that involved clinical leaders, both physicians and nurses, as well as members of the Mission Analytics team, performance improvement experts and committees to review care plans. Those 24 CPMs focused on clinical issues such as COPD (chronic obstructive pulmonary disease) exacerbation, breast cancer screening, depression, hypertension and chest pain, just to name a few. As part of this work, the analytics team created dashboards that provide real-time data and analytics on a provider and patient level and to track performance.
“As we’ve begun to scale that work, the analytics have really helped us to drive the kind of conversations that we need to have, both to improve our care process models themselves as well as to drive adherence and then show how our CPMS are helping to change outcomes for patients,” DeRienzo says.
Speaking to just the COPD exacerbation CPM, DeRienzo says, “Once folks use the COPD exacerbation CPM, they have 100 adherence to the goal guidelines. We’re now up to over 80 percent utilization, and we have modeled that if we got to 100 percent utilization, we would reduce a number of ED visits and a number of inpatient hospitalizations, and yield something on the order of just under $200,000 in reductions in direct costs.”
Mission Health now has 60 CPMs live across inpatient and ambulatory operations and is on track to have 80 models live by the end of the year. “We’re bringing one new CPM live about every two weeks,” he says. “The result has been thousands more patients screened for cancers, reductions in mortality and reductions in direct costs. It’s across the board,” DeRienzo says.
DeRienzo credits Mission Health’s board and senior leadership for its ongoing investment in analytics to enable the organization to drive forward with these efforts. “They are believers in how technology can be leveraged to improve outcomes, and it’s required millions of dollars of investment in infrastructure, in an enterprise data warehouse, and in people time to build the clinical and administrative team needed to do something with the data,” he says.
He also credits Mission Health’s success to “an alchemy of people, processes and technology,” adding, “Our clinicians, the PI team, the informaticists, the analytics people, everybody here is here for the right persons, they really care. My role is to help equip them with the teams and the technologies needed to get to where they want to go and that creates an enormous capacity for change.”
Readmissions Prediction Tool Leveraging Machine Learning
With an eye toward keeping patients healthier after hospital discharge and to reduce readmissions, Mission Health leaders initiated a project to use data and analytics to automate the calculation of a risk model for 30-day inpatient readmission.
“This was our first foray into the machine learning space; trying to leverage the power of data to better target patients who had been discharged from the hospital for our care managers to focus on and to try to keep them safe at home,” DeRienzo says. “This gets to where I think the power of technology is going, which is leveraging data and analytics to decrease the amount of time that humans have to spend doing things that humans don’t absolutely have to do in healthcare.”
Mission Health’s data science team was tasked with creating a risk model that “beats” the LACE index (LACE is based on four factors, length of stay, acuity of admission, co-morbidities and emergency room visits). The project team also wanted a model that would provide clinicians with predictions by 8 am for every patient discharged in the past 24 hours indicating patient’s likelihood to be readmitted compared to a baseline model. Once the model was created, project leaders worked with the care management team to implement it back in February, DeRienzo says. Throughout the pilot phase, the risk model and the user interface in the dashboard have been continuously refined based on user feedback.
“Six months from now, we will have worked through the implementation science around using it as optimally as we use it in practice, and my strong suspicion is that six months from now, we’ll see a meaningful impact on our outcomes,” he says, adding, “Care managers spend a good portion of their time right now figuring out which patients to focus on, and this model helps them spend less time focusing on which patients to help and actually spend more time helping patients. And in the short term, as we piloted that in the first 90 days, they are now seeing a whole new universe of people who were opaque to them before.”
DeRienzo says this project exemplifies the potential of machine learning technology in healthcare. “To me, technology can help return some humanity to the way we practice medicine. And, when it does that best is by reducing the amount of time that humans spend doing things that we don’t necessarily have to do,” he says. “When you strip away all the things that humans don’t have to do in healthcare, really what you’re left with is sort of the raw things that make us human, such as my interactions with you as a patient.”
As health systems and provider organizations look to move forward with machine learning and data analytics work, DeRienzo recommends that healthcare leaders should recognize that analytics is only one part of the equation. “The power of your analytics is directly related to the people who are going to use them, the process in which you are working and you have to begin with that end in mind. In our experience, starting our analytics journey with work that was core to our mission, core to our purpose, really gave us enough runway to then try and fail, try and fail and then try and succeed and success breeds more success.”
And, he says, “Technology and process exist in healthcare to help humans do more of what only humans can do. That analytics and the tools that we can use, the potential for leveraging AI, all is incredible, but fundamentally, at its core, healthcare is about people caring for other people. And I think if folks remember that and if that’s at the center of how they approach their technology decisions, there will be a lower risk of them losing their North Star.”