Predictive analytics in the limelight

Five years ago, virtual care was the bright-and-shiny new concept in healthcare delivery. In virtual care, clinician-patient interactions are conducted remotely, leveraging digital technology such as video and IoT devices to achieve the same outcome as a personal visit. Using virtual care technology, providers can monitor patients no matter where they are.

The healthcare industry was enthusiastic about the promise of virtual care in maximizing healthcare resources and extending much-needed care to remote locales, but providers soon got a reality check as they realized the significant technological challenges implementing such a system would pose. Today, virtual care continues to grow and mature as healthcare providers adopt elements of this delivery model.

In the meantime, healthcare industry attention has refocused on another promising technology advancement: applying artificial intelligence and machine learning to enable predictive analytics. The potential of predictive analytics has captured the industry’s imagination in the same way virtual care did five years ago: a thrilling new application for AI that can literally save lives.

Predictive analytics at work in the real world

Drawing on great volumes of patient data, predictive analytics can augment traditional clinical decision-making with data-enabled determinations at both the individual and population level, resulting in more positive patient outcomes. The benefits of analytics can also extend to other parts of the healthcare organization: for example, operational analytics can be applied to real-time dynamic staffing to ensure that the appropriate qualified expertise is in place at the right location at any given time.

Analytics was a hot topic at the recent HIMSS 2018 healthcare IT conference; of the many use cases that were shared, here are three of special interest to WWT.

Use Case #1 – Predicting patient decline at Sharp HealthCare

At HIMSS, Intel shared results from an eight-week proof of concept using predictive analytics at Sharp HealthCare in San Diego. Sharp wanted to improve its rapid-response effectiveness for medical emergencies by augmenting human diagnostic capabilities with predictive analytics to spot potential crises. Using machine learning to analyze electronic medical records (EMR) data, Sharp created a predictive model that would detect patient decline and identify cases at risk of requiring medical intervention within the next hour. The model drew on patient stats such as blood pressure, temperature and pulse rate.

When tested against historical data, the model was 80 percent accurate in predicting patients at risk of needing rapid-response team support. With distributed analytics and scale-out Intel® architecture, Sharp found an actionable way to derive enhanced value from its EMR investment, helping the hospital deliver more proactive, efficient interventions.

Use Case #2 – Technology backbone for data strategy

Wanting to improve patient care while lowering costs, a Missouri-based healthcare provider has been partnering with WWT on several initiatives.

In an ongoing engagement, we have been helping design a technology backbone and data ecosystem that will allow them to achieve their strategic data goals. The long-term goal includes leveraging disparate data sources to provide a full 360-degree view of each patient in order to identify optimal treatment paths.

The healthcare provider’s comprehensive care approach has been able to reduce costs and improve patient outcomes for high-risk patients, including:

  • 50 percent reduction in preventable readmissions.
  • 35 percent fewer days spent in the hospital.
  • 60 percent fewer septic shock deaths among remote patients.

In addition, our software development team created a custom application with an easy interface for care teams and patients which serves as a unified interface between the various provider clinical systems and the patient. The integrated data sources, clinical systems and unified interface enable a predictive model and data ecosystem to support health system’s exploratory, operational and financial needs.

Use Case #3 – Predictive models to prevent Opioid abuse

A state agency needed a clear understanding of the challenges it was facing from opioid addiction. The agency, which was tasked with reducing death rates due to abuse, engaged with WWT to develop predictive models that would help combat the issue.

Using big data analytics methodologies, we began by merging disparate datasets from previously disconnected sources, aligned and integrated the data, then de-identified the datasets using custom tools built by our scientists. Predictive models were then created to examine demographic profiles, identify predictive factors of opioid abuse and build a classification model to score patient risk.

As a result, never-before-linked datasets now offer the agency actionable insights that were previously unattainable. And, a data ecosystem has been created that supports public policy decisions and intervention.

Reliance on EMR data drives a pressing need for interoperability

For predictive analytics to yield meaningful results, it must have access to great volumes of raw data to draw on. To collect and store this data, different devices must also be able to talk to each other – but lack of interoperability turns out to be a significant barrier to progress. Surprisingly, even some EMR systems are not fully operable within themselves.

In response, WWT and other ecosystem leaders are working on health IT tools and solutions to promote seamless interoperability to move real-time data. Partnering with customers, health systems and application developers, we’re helping them create the right tools and infrastructure to achieve and scale out interoperability.

Just as virtual care has earned gradual but growing acceptance as a practical means of maximizing healthcare resources, predictive analytics is proving itself in real-world use cases, leveraging data to improve patient care.

Leave a Reply

Your email address will not be published. Required fields are marked *