The role of predictive analytics and clinical decision support in value-based care
The role of predictive analytics and clinical decision support in value-based care13 May 2020 | 8min
Value-based care shifts the approach from sick care to prevention and wellness
The key is to focus healthcare resources on the patients who need them the most, and intervene before the point of no return
Predictive analytics and clinical decision support systems allow clinicians to find actionable opportunities where there’s a chance to make a difference
Faced with the realization that most care models are financially unsustainable, many countries are making a shift to various forms of value-based care. At the heart of these new models is the refocus from sick care to prevention and wellness. The rationale is simple: It is more cost effective to prevent disease than it is to treat it. The challenge for health systems is making the shift in a way that yields results when investable resources are already constrained.
“The healthcare system does not have enough money in population management to take care of every patient, so how can we focus our resources on those patients who need it the most?” asks Dr. Philip Chen, Chief Strategy Officer of Sonic Healthcare USA.
The first step is finding patients before their conditions deteriorate. Predictive analytics are the key.
“Our expertise in lab medicine pathology can look at lab results and help identify these patients who have the most severe problems and can benefit from interventions the most,” Dr. Chen explains.
Targeting chronic disease
When we think of predictive analytics, chronic disease usually comes to mind because it constitutes the biggest opportunity for cost avoidance. The goal is use data to predict which patients are trending in the wrong direction, and trigger an intervention before the point of no return, when costs begin to uncontrollably snowball. This form of strategic prevention, if successful, can yield a sizable and immediate return on investment (ROI).
“Chronic disease comes up because that’s where costs get disproportionately high, when a disease is out of control,” says Dr. Bruce Muma, the Chief Medical Officer, President, and Chief Executive Officer (CEO) of the Henry Ford Physician Network.
Finding the opportunities to make a difference
There are many chronic diseases, but preventative healthcare resources are limited, so it’s necessary to focus on a select few where the greatest impact can be realized.
“We primarily focus on heart failure, chronic obstructive pulmonary disease, and diabetes,” says Dr. Muma. “Those three diseases are the most expensive. The window of opportunity to intervene and keep them out of the hospital is very narrow, but if you can intervene in that narrow window, you can save a lot of money, especially for patients with very advanced disease.”
Predictive analytics must be made actionable
Predictive analytics alone are not enough to influence physician behavior. They must also be made actionable at the point of care by the healthcare provider.
“I’ve used validated and well-respected severity of illness measures,” says Dr. Scott Weingarten, the CEO of Stanson Health. “The measures were very good at predicting a patient’s chance of dying during the hospitalization. The problem was that no one knew what to do with the information. If you don’t do anything differently to help patients, then there’s really no value in the information.”
“When I was doing chart reviews on the EMR in the 90s, we notified physicians of unnecessary lab testing. Despite presenting them data and evidence, it was clear this was not enough to positively influence physician behavior,” Dr. Chen adds.
To make predictive analytics actionable requires access to integrated data in real time. To make it successful requires that providers act on it to drive the intended outcome.
A predictive analytics success story
Despite the initial optimism and excitement around predictive analytics, there are remarkably few success stories. But for those who have ascended the steep learning curve, the results are impressive.
Dr. Chen recalls an analysis he ran on the most expensive patients in his Accountable Care Organization (ACO) population. The year prior, he noticed that those same expensive patients had very low expenditures.
- Problem: “Their costs weren’t low because they were healthy,” explains Dr. Chen. “The large majority had chronic disease. We found they were not following up with their doctor.” And that opened them up to major costs downstream.
- Solution: Dr. Chen and his team came up with a plan to get those types of patients back in to see their doctors before their conditions worsened and their expenses skyrocketed. “We developed an automated technology to use phone calls, email, and texting to reach out to these patients on behalf of their physicians and schedule a follow-up appointment,” he says.
- Results: “Our response rate was 44%. These were patients that would have ended up in the hospital or emergency room without an intervention,” says Dr. Chen.
The role of clinical decision support in real-time decision-making
Clinical decision support (CDS) is the preferred method to analyze vast stores of data, identify insights, and send actionable alerts, all in real time. But despite wide use over the last 15 years, CDS has often failed to live up to its potential. And in some cases, it can even contribute to cognitive overload and clinician burnout ⎯ exactly what it was intended to prevent. Why?
“We’ve created so many alerts and decision support tools that we’ve got too many of them now, and now we’ve overloaded the capability of a doctor and nurse to respond,” says Dr. Muma.
“Alert fatigue contributes to physician burnout,” agrees Dr. Weingarten. “It’s caused by CDS systems that cannot read free text and therefore create alerts that are false positives. This creates more work for doctors, instead of less.”
Clinical decision support system design and implementation best practices
Both Drs. Muma and Weingarten are CDS pioneers. To design and implement your own CDS system, they recommend the following:
- Risk-stratify your population, so you can find those most at risk and act, if necessary
- Make sure those identified as risks are those you can actually do something about
- Integrate all data and information in one place, accessible with a single sign-on
- Create a dashboard for patients that allows them to interact with their healthcare team
- Interpret discrete and free text information to reduce false positives and negatives
- Make sure CDS recommendations are clear and concise
- Provide a link to the evidence behind each alert, so clinicians understand the scientific basis for the recommendation
- Review order sets and preference lists to make sure they are not driving overuse of testing
- Provide feedback to providers on how often they are following the CDS, what the benefits of following it are, and the harms of not following it (mortality, costs, morbidity, quality of life, etc.)
The lab’s role in supporting advanced test selection
There is an explosion in the number of advanced molecular tests, which no physician can reasonably synthesize and adopt on their own. Without CDS, doctors will not use the tests properly. Without help interpreting the results, these tests won’t be incorporated into clinical practice.
“With more advanced molecular testing, a barrier to adoption and potential benefit down the road will be that healthcare providers won’t know what to do with the information,” Dr. Weingarten says. “Being able to provide guidance on this new world of molecular diagnostics is absolutely critical to gaining the necessary patient value.”
“The result cannot just be, ‘This is what we found…this is what we think the patient has,’” adds Dr. Chen. “The result really needs to start guiding what we think the treatment should look like and also what the follow-up care should look like.
The next generation of CDS
The future of CDS is in virtual assistants that will use machine learning to analyze data and make recommendations in real time. Not only will they provide better and more personalized care to patients, they will also dramatically reduce the burden on providers.
The key to their success will be their ability to serve up relevant information. This will require that they utilize not just discrete data, but non-discrete data sources as well, like free text in a physician’s note on the patient’s chart. The ability to read both types of data produces a more complete clinical picture and supports more meaningful and actionable insights.
“Tremendous investments are being made in ambient listening devices like Amazon Alexa or Google Home for use in the exam room to listen to provider/patient conversations, interpret the information, and provide guidance,” Dr. Weingarten says. “Not only to the provider – ‘Mrs. Smith is overdue for a mammogram’ – but also to remind the patient at home to take a long walk and cut down on their sodium intake.”
Physicians now spend as much as 50% of their time documenting care – so virtual assistants could theoretically automate this function and give back valuable time.
“To reduce burnout, we need to reduce the amount of typing on the part of physicians and providers and enable them to simply review transcribed and interpreted information,” says Dr. Muma. “We’re not there yet, but I think it will be here before we know it.”
Value-based care requires team-based care
Chronic disease management requires that all disciplines tightly integrate care, so that patients don’t fall into care gaps and reemerge with costly exacerbations. This presents opportunities for various disciplines, once siloed apart from one another, to work more closely together and support each other.
“I think the culture of medicine is changing from us all practicing as individuals – perhaps not communicating as well as we possibly could – to team-based care,” Dr. Weingarten says. “Once you start going at-risk for financial and clinical outcomes, it really accelerates the interest of implementing team-based care.”
Philip Chen, MD, PhD is a board-certified pathologist and the Chief Strategy Officer at Sonic Healthcare USA. Dr. Chen has been practicing pathology and running laboratories in commercial, non-profit, and academic settings. He founded Cognoscenti Health Institute in 2002, which was acquired by Sonic Healthcare. He also ran the clinical laboratories across two health systems: The University of Miami, and Jackson Health Systems in South Florida. Throughout his career, Dr. Chen has been exploring ways to practice laboratory medicine outside of the laboratories. In recent years, he has been focusing on using clinical and financial informatics to deliver value-based services and constructing new, value- and outcome-based reimbursement arrangements. Dr. Chen obtained his MD and PhD degrees from the University of Alabama at Birmingham and received pathology training at Brigham and Women’s Hospital and Harvard Medical School.
Bruce K. Muma, MD, FACP serves as the Chief Medical Officer of the Henry Ford Physician Network. In May 2019, he assumed the roles of HFPN President and CEO, after serving in an interim capacity in those roles since 2017. Dr. Muma leads the network’s clinical integration efforts, which includes monitoring clinical performance and developing clinical support programs to advance the HFPN’s value-based delivery model of a high-performance physician network. Dr. Muma received his medical education at Wayne State University and completed a combined residency in internal medicine and pediatrics at Henry Ford Hospital. He is a fellow in the American Academy of Pediatrics and the American College of Physicians.
Scott Weingarten, MD, MPH recently joined Premier Inc., retaining his role as CEO of Stanson Health. Prior to this, Dr. Weingarten served as the senior vice president and Chief Clinical Transformation Officer at Cedars-Sinai Hospital. In addition to his long-standing tenure as a practicing physician and executive, Dr. Weingarten is also an entrepreneur and inventor, holding three software patents. Prior to forming Stanson Health, Dr. Weingarten co-founded Zynx Health, a highly successful leader in order sets and care plans for electronic health records. After graduating from UCLA’s medical school, Dr. Weingarten completed his internship, residency and fellowship in internal medicine at Cedars-Sinai. He later participated in a National Center for Health Services Research Fellowship at the RAND/UCLA Center for Health Policy Study. During the fellowship, he also earned a Master of Public Health degree at the UCLA Fielding School of Public Health.