Intel’s Prescription for Healthcare of Tomorrow: Predictive Analytics

You wouldn’t know by reading the news that has come out of Washington, D.C. on healthcare lately, but innovation and ingenuity to tackle today’s most pressing healthcare issues are still front and center. The digital transformation of healthcare is underway.  Onstage today at the Renaissance Hotel in our nation’s capital, I had the great honor to present alongside some of the most influential healthcare CEOs at US News & World Report’s Healthcare of Tomorrow keynote. I shared how Intel is using its technology to help healthcare delivery organizations use historical data and build models and algorithms to predict future events. This improves healthcare delivery through optimized clinical, operational and financial performance.

One thing is crystal clear: value-based care is here to stay. Being able to drive changes that lead to better care at lower costs is a bi-partisan issue. It improves health. Plain and simple. More than 85 percent of US healthcare organizations have an electronic health record (EHR) system in place as well as other electronic data systems, capturing a variety of clinical, administrative and other data. These systems are making available an unprecedented and ever-growing amount of data. Now it’s just a matter of getting those data to surface new insights, using those insights to engage clinicians and patients in more meaningful ways, improving clinical care and overall health and wellness while at the same time creating a health system that is efficient and cost-effective.

We’re Helping Drive Healthcare Innovation

In my presentation at the U.S. News Healthcare of Tomorrow keynote, I shared various ways Intel is helping healthcare providers unleash the data they have to drive clinical, operational and financial improvements using predictive clinical analytics.

Intel recently funded a study with Forrester that surveyed leaders across 300 healthcare delivery organizations in the US and China. Of those surveyed, 86 percent indicated that they would be adopting predictive analytics over the next 12 months. Many already are taking steps to utilize predictive analytics. For example:

  • Advanced analytics like predictive clinical analytics has helped Penn Medicine improve clinical care. Penn Medicine increased its identification of septic patients from 50 percent to 85 percent up to 30 hours before those patients became septic.
  • Sharp HealthCare addressed operational efficiencies by evaluating how predictive analytics could optimize the response time for its rapid response teams addressing medical emergencies in the hospital. Intel helped Sharp build a predictive model using EHR data that can help identify patients who would need emergency team intervention within the next 60 minutes (with 80 percent accuracy.)

Predictive Analytics Technology

Predictive analytics is part of a healthcare organization’s overall analytics journey – it’s not the end state. In today’s world of big data, next-generation predictive models using artificial intelligence capabilities like machine learning will help providers advance their analytics maturity at scale as they advance towards prescriptive and cognitive analytics.

I also highlighted four layers of technology implementation that are needed to pave the way for predictive clinical analytics:

Core infrastructure - the compute, storage and networking that allows select hospitals to ingest, store and process very large data sets; both structured and unstructured data that may come from outside.

Data Layers - needed to combine these data sources into one environment, such as a data lake or data mart or data warehouse while providing access to the data for those who need it.

Analytics Layers - supporting many different types of frameworks. This flexibility to incorporate new tools and enable collaboration among IT, data scientists and clinicians in the creation of predictive models is critical.

Application Layer – (visualization layer) the user experience surfacing the results of the analytics and providing actionable insights to the end users.

Building the right structure - applications, analytics, infrastructure and data platforms

Enabling Predictive Clinical Analytics

Lastly, I made an announcement about how Intel is open-sourcing tools that will make predictive clinical analytics more accessible for healthcare organizations. We’re calling these tools “Predictive Model Toolkits” and they consist of four ingredients:

  1. Pre-defined Models
  2. Test Data
  3. Implementation Guides
  4. Data Science Notebooks

Out of the gate, we’re offering up the model that was developed at Sharp Healthcare to predict which patients have the highest likelihood of requiring a rapid response team intervention. Think of this as our attempt to try and offer an “easy button,” or at least an “easier” button, to healthcare organizations that are interested in predicting patients at-risk of requiring emergency intervention.

This is a supervised machine learning model. I’m using the term “supervised” because in this model, a specified set of historical data (ex. patients whose vital signs, lab tests, or demographic profile might put them at risk for an intervention) is used to train a model that can be used to generate a “risk score” for new patients. Now, if the prediction that you want to make is different than Sharp’s, or you have a different business case with a business impact that you feel is more important in helping you build momentum and expertise to expand into other predictive areas, you can still benefit from using these tools.

All of the data engineering, subject matter expert interviews, and feature selection that went into creating the model for Sharp HealthCare — including an implementation guide, synthetic data set to test your model implementation, and data science notebooks that provide the scripts and explain how to prepare the data and select the algorithms — are included. While most organizations will want to experiment with additional features based on the unique qualities of their patient population or clinical workflow, having this content as a starting point is a significant accelerator.

We’d love to hear from technical professionals, data scientists and others within healthcare organizations who download this toolkit, and use that feedback to help them get a start on the predictions they’d like to make. This will be the start of what we hope is a wide-ranging library of open-source healthcare algorithms that address some of the most important analytics use cases in healthcare today, and into the future.  Learn more about this and get access to these tools at

Published on Categories ArchiveTags , , , ,
Jennifer Esposito

About Jennifer Esposito

Jennifer believes that technology has the power to accelerate the transformation of healthcare and to improve health, quality of life, safety, and security worldwide. She works with companies, organizations, and governments around the world to help make this digital transformation real, today. With 20 years of experience, Jennifer brings deep healthcare, life sciences, and biotechnology industry expertise along with a foundation in information and communication technology, and a view that spans across multiple other industries. Jennifer has a graduate degree from the Dartmouth Institute for Health Policy and Clinical Practice at Dartmouth College, where she focused on Epidemiology and Biostatistics. During her time at GE Healthcare, she became a certified Six Sigma Black Belt and remains a full member of the American Association of Physicists in Medicine. Jennifer is the co-chair of the Global Health Security Agenda Private Sector Roundtable and sits on the boards of Digital Square and USA Healthcare Alliance. Follow her @Jennifer_Espo and @IntelHealth.