Design, deliver, and run enterprise blockchain workloads quickly and easily.
All servers and systems
Should the man complaining about severe back pain be admitted to the hospital or sent to a clinic? Is the patient who was operated on yesterday likely to develop a serious infection? Are there enough vaccines on hand to cope with a projected flu outbreak? How can staff and facility scheduling be handled more cost effectively? These are just a few of the questions asked by healthcare professionals that can be answered more accurately using predictive analytics.
Healthcare needs help, with costs and patient loads soaring. To cope with these profit-sapping trends, a growing number of healthcare organizations are turning to predictive analytics to control expenses in areas ranging from patient care to stock management to staff deployment. A study issued by Orbis Research in January 2017 found that the global predictive analytics market stood at $4.64 billion in 2016 and is projected to reach $19.55 billion by the end of 2022, with a projected CAGR of 22.81 percent between 2017 and 2022.
As the cost and technology barriers blocking predictive analytics adoption fall away, a growing number of healthcare organizations are turning to the technology to improve services and save costs. According to a February 2017 Society of Actuaries report on healthcare industry trends in predictive analytics, more than half of healthcare executives (57 percent) at organizations currently using predictive analytics expect that the technology will allow them to save 15 percent or more of their total budget over the next five years. Another 26 percent forecast savings of 25 percent or more.
The study also found that a majority of healthcare executives (88 percent) belong to organizations that are either now using predictive analytics or planning to do so in the next five years. Additionally, 93 percent of healthcare executives believe that the technology is important to the future of their business.
Most business leaders already have a good handle on analytics and the ways data patterns can be studied to improve business performance. Predictive analytics, on the other hand, remains something of a mystery to many managers and executives. Yet the difference between analytics and predictive analytics is really pretty simple: While traditional analytics is typically used to obtain insights on the here and now, predictive analytics helps users peer into the future to forecast likely trends and behaviors. "On the healthcare provider side, users are interested in obtaining specific predictions," says Brian Murphy, a health strategies and technologies analyst at Chilmark Research, a healthcare IT market research firm.
Predictive analytics utilizes various statistical modeling and other mathematical techniques to scrutinize current and past data and predict what should happen at a specific time based on the supplied parameters. For hospitals and clinics, predictive models can be used to exploit patterns found in historical and transactional data to spot both risks and opportunities in daily operations and patient care.
Models can be configured to capture relationships among multiple behavior factors to enable the assessment of either the potential or danger associated with a particular set of conditions, guiding decision-making for various categories of supply chain and procurement events. "If you’re a healthcare provider you could, for example, look at all your diabetics, and you would know that for your 10,000 diabetics you’ve got X number of dollars to spend, and then figure out the best way to spend that so that you help these 10,000 patients manage their diabetes effectively," Murphy says.
Much has been written about healthcare predictive analytics over the past few years. Unfortunately, many of the technology's most vocal proponents have tended to inflate the potential savings while minimizing the time and effort required to deploy and manage a predictive analytics initiative. Still, there's no denying the fact that predictive analytics is a powerful and highly useful tool that can lead to impressive benefits. Predictive analytics success stories are already beginning to roll in. (It's also worth remembering that healthcare data is regulated. See the U.S.’s Health Insurance Portability and Accountability Act of 1996 and the U.K.’s Data Protection Act for more information.)
Unnecessary readmissions are rampant in a U.S. healthcare system that frequently leaves discharged patients confused about how to care for themselves at home or obtain necessary follow-up care. Readmissions also place an unnecessary cost burden on a system that has few resources to spare. Reducing readmissions promises to significantly lower costs for hospitals. Yet, despite a rapidly growing pool of actionable data, relatively few facilities are using predictive analytics to remediate the problem.
"The idea of predictive analytics comes in looking for relationships that are consistent with readmission that we would not have predicted or we did not understand before," says Mark Wolff, chief health analytics strategist for SAS Institute, an analytics software developer. "Once we identify those relationships, we can set up protocols on how to deal with this type of patient and manage things to prevent readmission."
While there's no single predictive analytics approach to lowering hospital readmissions and its associated costs, several research teams have already created models that address the problem. NYU Langone Medical Center data scientists have developed a predictive analytics algorithm, based on a wide range of clinical factors, that identifies patients who are likely to spend fewer than two nights in hospital. The tool aims to help physicians know when to place a patient on observation and avoid a claim denial from the federal Centers for Medicare & Medicaid Services.
Data scientists at Boston's Brigham and Women's Hospital (BWH), meanwhile, have developed a risk prediction model that's designed to pinpoint hospitalized patients at the greatest risk of being readmitted for preventable reasons. Using data collected between 2009 and 2010 on approximately 9,200 BWH patients, the researchers were able to identify seven factors that best predicted which patients would be readmitted:
Under the model, each predictive factor is given a point value ranging from one, for less important factors, to two or more for critical indicators. For example, a patient who remains in the hospital for five days or more after admission is given two points.
"In the near term, you can actually have a library of these models that look for relationships between the parameters and the patient as they’re going through their care," Wolff says.
Operating rooms are expensive to build, equip, staff, and support, so it's in every hospital's interest to optimize operating room use without compromising patient health. To achieve this goal, a growing number of administrators are recognizing that predictive analytics can help hospitals better understand the relationships between the many operating room variables that tend to ruin effective scheduling.
These variables include surgeon availability and preference, operating hours, and equipment availability and functionality. With so many factors to consider, creating an optimal surgical schedule isn't easy. Fortunately, predictive analytics can help streamline operating room management. Hospitals are now awash in patient, staff, and facilities data. The trick is to mine the right data, study operating room utilization patterns, and use predictive modeling to match the most appropriate human and support resources with the right operating rooms.
"Demand on an operating room is a kind of forecasting exercise," says Wolff. "We look at historical demand, we look at available capacity and various demographic and macro-economic trends, and then we say whether demand for a particular procedure or a particular variable will go up or down."
Murphy adds, "When it comes to operating room scheduling, there's a lot of demand that is completely predictable, and predictive analytics can help. The difficulty is getting surgeons on board, because for the most part, they want to be able to control their schedule rather than let somebody else control it for them."
One of predictive analytics' most powerful attributes is its ability to help adopters peek around corners and uncover potential opportunities as well as lurking challenges. Based on data collected over several years, hospitals and clinics can create predictive models that anticipate and coordinate inventory acquisitions in response to expected price increases caused by shortages and inflation, bargain prices created by market gluts, increased needs driven by seasonal patient intakes, and numerous other factors. Predictive analytics can also be used to reveal internal savings opportunities created by use and waste patterns as well as inventory consolidation and standardization. "One of the main cost factors is drugs management itself, including wrongly managed expiry dates, which is usually a big source of waste. Predictive analytics help you have the right amount of the right drugs at the right time, and that brings strong savings," says Alvaro Gomez-Meana, worldwide chief technologist for health and public service at Hewlett Packard Enterprise.
The supply/demand situation in healthcare is much more complex than it is in many other industries, according to Wolff. "There are many situations in a hospital—for instance, in terms of supply and demand with expensive drugs or drugs that require either refrigeration or some special consideration—where optimization of supply against demand can produce greater confidence that you have exactly what you need."
One of the best ways to trim labor costs is to forecast demand far enough in advance to match staff and resources, thereby reducing the likelihood of incurring last-minute expenses. Although staff scheduling and time/attendance systems have been widely used for many years, they have had only a limited positive effect on overall labor costs and productivity. This is primarily due to a lack of accurate forecasts that hospitals and clinics can use to establish workforce requirements in advance or in real time.
"It's a matter of moving through a calendar that's optimized against an algorithmic schedule versus, essentially, a couple of people sitting down in a room and sort of guessing how many nurses they’ll need in what ward based on what they did last week," Wolff says.
In emergency rooms and intensive care units, predictive analytics is becoming critical for both quality of care and patient safety. ER and ICU patients are highly susceptible to sudden downturns caused by infection, sepsis, and other critical events. Such incidents are challenging for busy staff members to predict and expensive to treat. As a result, Murphy says, "people are coming around to the idea that predictive analytics leads to more accuracy."
Researchers are now developing ways of using predictive analytics, fed by bedside medical device data, to forecast an emergency medical situation minutes or hours before it becomes a potentially life or death event. At the University of California at Davis, for example, researchers have developed a predictive analytics tool that mines electronic health records (EHRs) to give physicians and nurses an advance alert to patients who are at risk of sepsis, which has a 40 percent mortality rate and is often difficult to detect until it’s too late.
A predictive model developed by Atrius Health is designed to enhance care for high-cost, high-need patients. The model is integrated into the organization's EHR system to pinpoint key clinical factors, generating a color-coded banner that flags patients who are at risk for hospitalization within the next six months.
Physicians can also turn to predictive analytics to help make more accurate on-the-spot diagnoses. A patient entering an emergency room complaining of chest pain, for example, can be evaluated rapidly and efficiently by a physician using a diagnostic tool based on predictive analytics to immediately determine whether or not hospital admittance is required.
Healthcare predictive analytics can also reduce or prevent ICU and ER bottlenecks by analyzing patient flow during peak times, giving administrators an advance opportunity to call in extra staff or make other necessary arrangements before service is adversely affected.
"The common theme here is that there’s a tremendous amount of digital data available in hospitals and in the broader healthcare community that has never been available before," Wolff says. "We have algorithms—statistical, mathematical techniques that produce incredible analysis efficiently with a high degree of confidence—and now we’re using that to tackle the problems we’ve all been dealing with for quite some time in a deeper, more robust way."
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.