Using Data to Save Lives
John Showalter, MD, is a student of medical history. And one of the reasons he's passionate about analytics is his sense that a new chapter is being written. "I firmly believe that predictive analytics is going to be as significant as penicillin and anesthesia in the history of medicine," he says. "And the truth is, we're barely scratching the surface of what's possible."
But Showalter is scratching as hard as he can. As the chief health information officer at the University of Mississippi Medical Center, he has led an effort to apply analytics to significantly enhance the health system's ability to predict heart attack risk. For that effort, Showalter has been chosen as the recipient of Health Data Management's Clinical Visionary award for its Analytics All-Stars recognition program. (The full list of recipients can be found below).
The program recognizes organizations and individuals implementing analytics in innovative ways to improve the health of their patients and the financial performance of their organizations. The Analytics All-Stars program showcases those who are providing pioneering leadership, spurring innovation and driving improvement across their organizations. The recognition program reviewed projects and achievements from 2014 and is built upon the themes of HDM's Healthcare Analytics Symposium, being held July 13-15 in Chicago.
Skin in the game
Showalter has a lot of skin in this game. As a practicing hospitalist, he treats many cases of heart disease, which is a plague in Mississippi: It is the leading cause of death in the state, according to Mississippi's Department of Health. In addition, numerous studies have found that the state leads the nation in percentage of cardiovascular deaths among its citizens.
What's especially frustrating, Showalter says, is that he's always treating patients after their disease has put them in dire straits. In addition, even after the health system treats heart attack victims and educates them on better lifestyle choices and available treatments, 1 percent of all patients leaving the University of Mississippi Medical Center will have a heart attack within a year of discharge.
UMMC, like many of its peers, risk-stratifies its patient population using guidelines and processes from the Framingham Heart Study. But that tool is built to assess the 10-year risk of having a heart attack, and Showalter feels more urgency must be applied. "It's hard for the medical profession to keep motivating patients for a 10-year time horizon."
As a clinician, Showalter knew UMMC needed to vastly improve its predictive capabilities for heart attacks. But as an informaticist, he wasn't confident he had the data quality and analytics firepower to do so.
Conversations with Suwanee, Ga.-based Jvion convinced him otherwise. The company's RevEgis clinical predictive tool uses the same type of correlation/matching techniques and algorithms used by Google and Facebook. Using the cloud-based software on retrospective data, Showalter narrowed the high-risk group to 1.5 percent of UMMC's patients, and then burrowed into that group to identify those patients who were likely to have a heart attack within the next year. A retrospective analysis found that of the patients RevEgis had put in the highest-risk group, seven out of 10 had a heart attack within 12 months. In the group the software deemed to be at more moderate risk, more than 10 percent had a heart attack within a year of discharge. Using the predictive insights, UMMC is able to provide interventions in a tactical way by targeting those with the most immediate risks with the most intense treatment.
Those results were shocking, even to Showalter, who after serving as CIO and CMIO at the health system before taking the role of chief health information officer, felt himself pretty well-versed on the current state of information technology. "I assumed when we started this project that we would have to spend an enormous amount of time cleaning up the data, but I was wrong-Jvion took our dirty EHR transactional data and ran it through the software," he says.
"I've always assumed that we as a healthcare organization had the dirtiest data," he adds, "but when you think about it, Facebook and others are taking texts and snippets of conversation and all kinds of other pieces of data to profile their customers, and compared with those data sets, our information is relatively clean. I think there's a misunderstanding in our industry about the capabilities of these new analytics tools to be applied to our data."
Once he overcame his initial skepticism, Showalter began the arduous journey of convincing his clinical colleagues, especially cardiologists, about the efficacy of the predictive analytics system. At this juncture, it still seems like magic to many of them, Showalter says. The initial test of the predictive analytics system was a retrospective look at 10,000 patients; the next stage calls for a study of 25,000 patients before the system moves into the live clinical environment.
"Our clinicians work in an environment that increasingly relies on embedded, evidence-based practices, and all that evidence is based on cause-and-effect analyses that the medical profession is familiar with," Showalter says. "But what we're doing with predictive analytics is not casual, it's correlatory, and it's a big leap for many clinicians because it flies in the face of what we've been doing for the past 20 years.
"Part of the conversation we're having with the clinical staff is making the argument that if we know there are two factors that have a high correlation for heart attack risk, do we really need to completely understand the causation?" he says. "At this juncture, telling a cardiologist that we can run the data and give them a very high degree of predictability of heart attacks within 12 months-as opposed to 10 years-sounds like science fiction. But we have very powerful tools to do that now, and I don't think anyone, including me, knew just how far we could expand our capabilities today. Even with an informatics background, I initially thought capabilities like this were five years away."
There is a sense of urgency on Showalter's part to get the predictive analytics to the front lines. "While we can't definitely say it will work prospectively, we'll be using the same data, so I'm confident we're going to be successful. Even if we don't find the same degree of correlation, it's still going to be exponentially more accurate than the risk models we're using now. And the patients whose data we used to create the risk stratification models are still out there, and they're still at risk."
While Showalter knows his efforts are only scratching the surface of what predictive analytics can do, he's confident he'll have the resources to keep at it. UMMC recently established a center for analytics and informatics, which Showalter will lead, and designated $2.5 million annually in operational funds for the effort. The center will have a team of 15 analysts who initially will focus on a few key clinical and financial opportunities the health system has identified.
"Even with the limited resources we previously had for analytics, we found $25 million in annual costs savings that we felt could be tackled quickly if we invested in understanding them better and with a clear strategy on what needs to be done. Even if we don't hit that savings goal of $25 million annually, even half of that makes the investment a slam dunk."
The initial targets include antibiotics administration, self-pay readmissions and healthcare-acquired conditions, which account for more than half of the potential $25 million in savings. For antibiotics administration, Showalter's team found that the health system's protocols were keeping patients on broad-spectrum antibiotics longer than necessary, increasing costs while also reducing clinical quality, because patients would respond better to more targeted drug therapies after diagnosis. "Unlike many analytics efforts, which rely on grants and third-party funding, we're funding our effort through operations, and to justify that investment we wanted to initially focus on practical things," Showalter says.
While the predictive analytics effort might still feel a bit like science fiction, Showalter learned early in his career how the practical application of analytics can be a game changer. During his residency at Penn State Milton S. Hershey Medical Center, Showalter also completed an informatics and internal medicine fellowship. One of the projects he worked on during that fellowship was a deep analysis of sepsis patients in the medical center's intensive care unit. A key finding was that the protocol for treating sepsis used an antibiotic that took an hour to get from the pharmacy, a dangerous delay. By changing antibiotics and processes, mortality was quickly reduced by 35 percent.
"I realized right then that using analytics to find the answers to clinical challenges-just one effort-could end up having an impact on potentially hundreds of thousands of lives," Showalter says. "What we're doing today is going to be as fundamental to medicine in the future as a lab test, and at some point we're going to be using genomics to determine an individual's risk for an entire spectrum of disease states.
"But what's exciting to me is how much is possible even today with the data in the state it's in and analytics tools on the market," he adds. "It's hard not to be excited about the direction the industry is heading."
HDM Announces 2015 Analytics All-Stars
Health Data Management announces the recipients of its second annual Analytics All-Stars program, which recognizes organizations and individuals implementing analytics in innovative ways to improve the health of their patients and the financial performance of their organizations.
Nominations were accepted in the spring for the recognition program, for projects and achievements from 2014. A panel of HDM editors reviewed the nominations and selected winners for the program, which is built on the themes of HDM's Healthcare Analytics Symposium, being held July 13-15 in Chicago.
The individuals and organizations honored this year include:
Chief Information Officer (CIO) Visionary: A. Thomas McGill, MD, CIO and vice president of quality and safety, Butler Health System, Butler, Pa.
Under McGill's guidance, Butler Health revamped its business intelligence efforts with Information Builders' WebFocus, to enable Butler's providers to see where they stand one week versus another on various metrics, such as the number of electronic prescriptions they filled or the number of times they ordered appropriate care. This automated measurement and feedback and information access was instrumental in improving efficiency and ultimately patient care. Butler recently participated in a program to improve patient throughput in the emergency department. To measure the progress, McGill's team developed a reporting portal to connect information about patients at an aggregate level, featuring milestones and times by month. The system could even drill down into the details of week, shift and hospital unit. After one year of using the new analytics environment in the emergency department, Butler was able to reduce its arrival to admission times by 15 percent and arrival to discharge times by 10 percent, while increasing in-patient throughput volumes by 7 percent-all without increasing staff or rooms.
Project of the Year, Accountable Care: Integrated ACO, Austin, Texas
Integrated ACO, an accountable care organization serving medically underserved populations in west Texas, developed psychographic analytical tools to customize care approaches to the personality profiles of patients. The technology innovation includes not only the key elements essential for population management and generally available technology, but also disruptive technology and components that drive quality improvement and cost reduction.
Through clinical analytics, the ACO has been able to determine whether clinical practice is in compliance with nationally accepted clinical guidelines or deviates significantly from it. Performance report cards based on these analyses are effective tools to drive behavioral changes in clinical practice, to drive cost reduction and quality improvement. The ACO also developed predictive algorithms to identify patients at risk of becoming hospitalized in the next six months. The predictive models target conditions such as congestive heart failure, pneumonia, and acute and chronic diseases for which good ambulatory care can prevent expensive hospitalizations. The prediction model for congestive heart failure admissions was evaluated using the area under the curve of the model's receiving operating characteristics and the lift for the model at 1 percent, 5 percent and 10 percent of the cases labeled. Nearly 90 percent of true positive predictions were observed while accepting only 10 percent of false positive predictions.
Project of the Year, Population Health: Blue Cross Blue Shield of Tennessee, Chattanooga
The insurer developed an analytics platform (Custom 360) that provides end users with data to analyze financial trends, review healthcare utilization statistics, attend to population health risks and monitor healthcare outcomes throughout the member population. It develops insights for health intervention and for marketing campaigns that are effective and measurable and deliver results. This strategic platform delivers key capabilities including quality analytics, consumer analytics, provider engagement analytics, provider reimbursement analytics, internal analytics and employer group reporting.
A clinician/provider dashboard displays quality ratings, Medicare dollars distributed, spending on specific diseases, and all accompanying tests, procedures and preventive efforts associated with those diseases. This dashboard supports health quality initiatives at BCBST, such as identifying and reducing gaps in care, and aligns them with financial incentives to the participating clinicians/providers.
The platform also utilizes an enterprise data warehouse and integrated marketing software for creating marketing, health and wellness campaigns for members, and applications for stakeholders including provider groups, employers and internal constituents. The insurer has closed more than 422,000 gaps in care and generated approximately $3 million in savings.
Project of the Year, Patient Safety: UnityPoint Health, Waterloo, Iowa
Recent medical research suggests traditional blood transfusion practices among physicians are excessive, resulting in not only financial waste but, more critical, harm to patients. Historically, the most common patient risk associated with blood transfusion was infection, but new studies indicate that eliminating unnecessary transfusions is associated with lower patient mortality, decreased length of stay and a reduction in 30-day readmissions.
UnityPoint Health developed an analytics solution that allows for the observation of system-wide trends, as well as the risk adjustment of every transfusing physician's patient case mix. The solution is "trained" using random forecasts on all inpatients based on their demographics, diagnosis information, and hemoglobin levels retrieved from their Epic electronic health record. Models are applied to each patient receiving a transfusion, and both the conditional expected number of units given transfusion and expected transfusion probability are computed using R statistical programming. Using software from Tableau, UnityPoint is able to aggregate the expected transfusion rates and units conditioned on patient attributes for all of a physician's patients, providing the total number of expected units based on enterprise behavior and a risk adjustment for every physician's case mix.
Excessive transfusion practices can then be ascertained by comparing expected transfusion behavior with actual behavior. By maintaining patient-level detail in the underlying data, UnityPoint can instantly risk-adjust the case mix of any group, from specific physicians to service lines to hospitals to geographic regions, and identify excessive transfusion behavior based on the practices of the entire organization.
Project of the Year, Revenue Cycle Management: Mercy Health, St. Louis
Mercy is using analytics to transform medical documentation workflow and improve the accuracy of provider documentation through the automated selection of patient charts for secondary review. The Mercy analytics team partnered with medical documentation specialists to produce a secondary diagnosis report to integrate business intelligence and clinical workflow processes to simplify the accurate coding of specific inpatient health conditions by allowing information in the report to be embedded in the workflow, enabling front-line caregivers to turn insight from the report into immediate action.
Mercy identified more than 18 secondary diagnoses conditions, including sepsis, anemia, acute kidney injury and hypertensive emergency. Targeted conditions are frequently missed opportunities that affect the overall clinical picture of severity of illness and risk of mortality in addition to providing high value from a revenue perspective.
This article courtesy of Information Management's sister brand, HealthData Management.