Analytics in Action News

10 High-Value Use Cases for Predictive Analytics in Healthcare

Predictive analytics can support population health management, financial success, and better outcomes across the value-based care continuum.

Predictive analytics in healthcare

Source: Thinkstock

By Jennifer Bresnick

- As healthcare organizations develop more sophisticated big data analytics capabilities, they are beginning to move from basic descriptive analytics towards the realm of predictive insights.

Predictive analytics may only be the second of three steps along the journey to analytics maturity, but it actually represents a huge leap forward for many organizations.

Instead of simply presenting information about past events to a user, predictive analytics estimate the likelihood of a future outcome based on patterns in the historical data. 

This allows clinicians, financial experts, and administrative staff to receive alerts about potential events before they happen, and therefore make more informed choices about how to proceed with a decision.

The importance of being one step ahead of events is most clearly seen in the realms of intensive care, surgery, or emergency care, where a patient’s life might depend on a quick reaction time and a finely-tuned sense of when something is going wrong.

READ MORE: How to Choose a Population Health Management Company

But high-value use cases for predictive analytics exist throughout the healthcare ecosystem, and may not always involve real-time alerts that require a team to immediately spring into action.

Provider and payer organizations can apply predictive analytics tools to their financial, administrative, and data security challenges, as well, and see significant gains in efficiency and consumer satisfaction.

How are healthcare organizations deploying predictive capabilities across the enterprise to extract actionable, forward-looking insights from their growing data assets?

Risk scoring for chronic diseases, population health

Prediction and prevention go hand-in-hand, perhaps nowhere more closely than in the world of population health management. 

Organizations that can identify individuals with elevated risks of developing chronic conditions as early in the disease’s progression as possible have the best chance of helping patients avoid long-term health problems that are costly and difficult to treat.

READ MORE: Using Big Data Analytics for Patient Safety, Hospital Acquired Conditions

Creating risk scores based on lab testing, biometric data, claims data, patient-generated health data, and the social determinants of health can give healthcare providers insight into which individuals might benefit from enhanced services or wellness activities.

“Across all [reimbursement] models, the identification, stratification, and management of high-risk patients is central to improving quality and cost outcomes,” says the Association of American Medical Colleges (AAMC)

“The use of predictive modeling to proactively identify patients who are at highest risk of poor health outcomes and will benefit most from intervention is one solution believed to improve risk management for providers transitioning to value-based payment.”

Avoiding 30-day hospital readmissions

Hospitals and health systems are subject to significant penalties under Medicare’s Hospital Readmissions Reduction program (HRRP), adding a financial incentive for preventing unplanned returns to the inpatient setting.

In addition to improving transitions of care and deploying care coordination strategies, predictive analytics can warn providers when a patient’s risk factors indicate a high likelihood for readmission within the 30-day window.

In a 2016 study from the University of Texas Southwestern, researchers found that certain events occurring during a hospital stay, such as a C. difficile infection, vital sign instability upon discharge, and overall longer length of stay, resulted in a significantly elevated chance of a 30-day readmission. 

Analytics tools that can identify patients with traits that produce a high impact on the likelihood of readmission can give providers an extra indication of when to focus resources on follow-up and how to design discharge planning protocols to prevent speedy returns to the hospital.

Getting ahead of patient deterioration

While still in the hospital, patients face a number of potential threats to their wellbeing, including the development of sepsis, the acquisition of a hard-to-treat infection, or a sudden downturn due to their existing clinical conditions.

Data analytics can help providers react as quickly as possible to changes in a patient’s vitals, and may be able to identify an upcoming deterioration before symptoms clearly manifest themselves to the naked eye. 

Machine learning strategies are particularly well suited to predicting clinical events in the hospital, such as the development of an acute kidney injury (AKI) or sepsis.

At the University of Pennsylvania, a predictive analytics tool leveraging machine learning and EHR data helped to identify patients on track for severe sepsis or septic shock 12 hours before the onset of the condition, a 2017 study explained.

A separate initiative at Huntsville Hospital in Alabama found that combining predictive analytics and clinical decision support (CDS) tools could reduce sepsis mortality by more than half.  The analytics-driven strategy exceeded the accuracy of existing gold-standard tools.

Forestalling appointment no-shows

Unexpected gaps in the daily calendar can have financial ramifications for the organization while throwing off a clinician’s entire workflow. 

Using predictive analytics to identify patients likely to skip an appointment without advanced notice can improve provider satisfaction, cut down on revenue losses, and give organizations the opportunity to offer open slots to other patients, thereby increasing speedy access to care.

EHR data can reveal individuals who are most likely to no-show, according to a study from Duke University.  A team found that predictive models using clinic-level data could capture an additional 4800 patient no-shows per year with higher accuracy than previous attempts to forecast patient patterns.

Providers may be able to use this data to send additional reminders to patients at risk of failing to show up, offer transportation or other services to enable individuals to make their appointments, or suggest alternative settings and times that may better suit their needs.

Preventing suicide and patient self-harm

Early identification of individuals likely to cause harm to themselves can ensure that these patients receive the mental healthcare they need to avoid serious events, including suicide.

EHRs once again offer a treasure trove of data to support suicide risk detection, says Kaiser Permanente.  In a 2018 study conducted by KP and the Mental Health Research Network, the combination of EHR data and a standard depression questionnaire accurately identified individuals who had elevated risk of a suicide attempt.

Using a predictive algorithm, the team found that suicide attempts and successes were 200 times more likely among the top 1 percent of patients flagged.

The strongest predictors of a self-harm attempt included mental health or substance abuse diagnoses, previous suicide attempts, the use of psychiatric medications, and high scores on the depression questionnaire.  

"We demonstrated that we can use electronic health record data in combination with other tools to accurately identify people at high risk for suicide attempt or suicide death," said first author Gregory E. Simon, MD, MPH, senior investigator at Kaiser Permanente Washington Health Research Institute.

Predicting patient utilization patterns

In addition to helping organizations get ahead of no-shows, predictive analytics can give providers a heads up when the clinic is about to get busy.

Care sites that operate without fixed schedules, such as emergency departments and urgent care centers, must vary their staffing levels to account for fluctuations in patient flow.  Inpatient wards must have beds available for patients who need to be admitted, while outpatient clinics and physician offices are responsible for keeping wait times low for patients.

Using analytics to predict patterns in utilization can help to ensure optimal staffing levels while reducing wait times and raising patient satisfaction.

Visualization tools and analytics strategies can model patient flow patterns and highlight opportunities to make workflow adjustments or scheduling changes.

At Wake Forest Baptist Health in North Carolina, analytics tools helped the oncology infusion center anticipate peak utilization times and adjust its scheduling practices accordingly, said Karen Craver, Clinical Practice Administrator.

By analyzing typical utilization rates, the infusion center found that popular mid-day appointment times were creating unsustainable spikes in capacity, while early morning and late afternoon spots went unfilled.

Altering certain scheduling procedures helped to “flatten out the bell curve” and create more even distribution that reduced burdens on nurses and improved patient satisfaction, she said.

“In the morning, we have greater utilization than we did before, but then we can maintain our appointment rate throughout the day, which is a much easier way to work than the steep incline and decline we used to deal with,” Craver remarked.

Managing the supply chain

The supply chain is one of a provider’s largest cost centers, and represents one of the most significant opportunities for healthcare organizations to trim unnecessary spending and improve efficiency.

Predictive tools are in high demand among hospital executives looking to reduce variation and gain more actionable insights into ordering patterns and supply utilization.

Only 17 percent of hospitals currently use automated or data-driven solutions to manage their supply chains, said Cardinal Health in 2017. 

In the same year, Global Healthcare Exchange ranked predictive analytics for supply chain management as the number one item on the executive wish list – a follow-up survey in 2018 found that adopting data analytics tools remained a top priority.

Using analytics tools to monitor the supply chain and make proactive, data-driven decisions about spending could save hospitals almost $10 million per year, a separate Navigant survey added.  Both descriptive and predictive analytics can support decisions to negotiate pricing, reduce the variation in supplies, and optimizing the ordering process.

Ensuring strong data security

Predictive analytics and artificial intelligence are also anticipated to play an important role in cybersecurity, especially as the sophistication of attacks continues to increase. 

Using analytics tools to monitor patterns in data access, sharing, and utilization can give organizations an early warning when something changes – especially when those changes indicate an intruder may have penetrated the network.

Predictive tools and machine learning techniques can calculate real-time risk scores for specific transactions or requests and respond differently depending on how the event has been scored, explained David McNeely, a Fellow at the Institute for Critical Infrastructure Technology (ICIT), in an ICIT report.

“Once the risk score has been determined in real-time, the system can use this during a login event to either grant the access for a low-risk event or to challenge for Multi Factor Authentication (MFA) or possibly block the access for high-risk events,” he said.

“In this way, the system enables IT to apply MFA more liberally across infrastructure and applications since the machine learning system will make decisions of risk which determine if MFA will actually be applied or not.”

This strategy could be particularly effective for preventing ransomware from affecting a healthcare organization, added ICIT Senior Fellow James Scott.

“Early adoption of sophisticated algorithmic defenses such as machine learning or artificial intelligence solutions will transform healthcare cyber defenses beyond the capabilities of average attackers.”

Developing precision medicine and new therapies

As precision medicine and genomics gain steam, providers and researchers are turning to analytics to supplement traditional clinical trials and drug discovery techniques. 

“In silico” testing is a promising way to reduce the need to recruit patients for complex and costly clinical trials while speeding up the evaluation of new therapies.

“FDA’s Center for Drug Evaluation and Research (CDER) is currently using modeling and simulation to predict clinical outcomes, inform clinical trial designs, support evidence of effectiveness, optimize dosing, predict product safety, and evaluate potential adverse event mechanisms,” said FDA Commissioner Scott Gottlieb, MD, after the passage of the 21st Century Cures Act.

“To take just one example of the benefits of these approaches, as we enter an era of drug individualization, modeling and simulation that incorporates aspects of individual physiology and genetics in drug metabolizing enzymes is being used to identify patient subgroups that need dose adjustments.”

In silico models are being used to create control groups for trials related to degenerative conditions such as Parkinson’s disease, Huntington’s disease, and Alzheimer’s, the FDA added. 

“We’re at the beginning of a transformative era in science and medical technology,” Gottlieb said.

Predictive analytics and clinical decision support tools are playing key roles in translating new drugs into precision therapies, as well. 

CDS systems are starting to be able to predict a patient’s response to a certain course of treatment by matching genetic information with the results from previous patient cohorts, allowing providers to choose the therapy with the highest likelihood of success.   

Doing so may improve outcomes and allow researchers to better understand the relationships between genetic variants and the effectiveness of particular therapies.

Bolstering patient engagement and satisfaction

In addition to supporting chronic disease management strategies, cutting wait times, and targeting therapies to produce better outcomes, predictive analytics can keep patients engaged in other aspects of their care, as well.

Consumer relationship management has become a vital skill for both providers and insurance companies looking to promote wellness and reduce long-term spending – and predicting patient behaviors is a key component of developing effective communications and adherence techniques.

“We need to know what works and what doesn’t in our engagement programs, and how to anticipate and predict the best outcomes given very complex characteristics of our membership sub-populations, which span every single segment of the US population,” said Patrick McIntyre, Senior Vice President of Health Care Analytics at Anthem.

“Our goal is to grow out our consumer engagement skills, because we are shifting into much more of a service-oriented, consumer-oriented industry.”

Anthem is using its data analytics tools to create consumer profiles that allow the payer to send tailored messaging, improve customer retention, and discover what strategies are most likely to be impactful for each individual.

Providers, too, are using behavioral patterns to create meaningful care plans and keep patients engaged with their financial and clinical responsibilities.

“Both payers and providers have a wealth of information that they can use to build models. Healthcare providers can also acquire some other sources, like the social determinants of health, for example, that will really help the strength and accuracy of their models,” said Lillian Dittrick, Fellow of the Society of Actuaries.

“When we use predictive models to look at all the variables, it helps us prioritize those patients who are really going to be receptive to changing something in their lifestyle, such as nutrition or exercise.”

Using predictive analytics to inform care management decisions and develop stronger, more motivational relationships between patients and providers can improve long-term engagement and reduce the risks associated with chronic diseases.

“We're seeing more and more that automation and machine learning tools really help with sorting through and processing these very large amounts of data,” said Dittrick.  “There is some kind of predictive modeling that could help improve processes in just about any facet of healthcare.”