Machine Learning Times
Machine Learning Times
A University Curriculum Supplement to Teach a Business Framework for ML Deployment
    In 2023, as a visiting analytics professor...
The AI Playbook: Providing Important Reminders to Data Professionals
 Originally published in DATAVERSITY. This article reviews the new...
Decode the Algorithm: Navigate the World of Machine Learning in Business with ‘The AI ​​Playbook’
  This article reviews the new book, The AI Playbook, by...
To Deploy Machine Learning, You Must Manage Operational Change—Here Is How UPS Got It Right
 Originally published in Harvard Data Science Review. For more...

9 years ago
Four Use Cases for Healthcare Predictive Analytics, Big Data


Predictive analytics in healthcare has long been the wave of the future: an ultimate goal to which everyone aspires but few can claim success.  While the landscape is changing for healthcare predictive analytics as more organizations figure out how to harness big data and implement the right infrastructure for generating actionable insights from a slew of new sources, some providers may still be wondering how the pie-in-the-sky world of big data can actually work for them.

Luckily, a number of pioneering organizations have taken it upon themselves to test the waters of healthcare predictive analytics, generating use cases that spur interest and help carve a path through the wilderness.

This article explores some of the ways healthcare organizations have already found success by turning big data into a strategic asset that can help providers react quickly and effectively to the ongoing challenges of quality care delivery.

Hospital quality and patient safety in the ICU

The ICU is another area where predictive analytics is becoming crucial for patient safety and quality care.  The most vulnerable patients are prone to sudden downturns due to infection, sepsis, and other crisis events which are often difficult for busy staff to predict.  However, a number of organizations have been working on integrating bedside medical device data into sensitive algorithms that detect plummeting vital signs hours before humans have a clue.

At the University of California Davis, researchers are using routinely collected EHR data as the fodder for an algorithm that gives clinicians an early warning about sepsis, which has a 40 percent mortality rate and is difficult to detect until it’s too late. “Finding a precise and quick way to determine which patients are at high risk of developing the disease is critically important,” said study co-author Hien Nguyen, Associate Professor of Internal Medicine and Medical Director of EHRs at UC Davis. “We wanted to see if EHRs could provide the foundation for knowing when aggressive diagnosis and treatment are needed and when they can be avoided.”

PAW Healthcare AD

At Massachusetts General Hospital, an analytics system called QPID is helping providers ensure that they don’t miss critical patient data during admission and treatment.  The system is also used to predict surgical risk, helping match patients with the right course of action that will keep them safest during their care. “Surgeons, even the world-renown surgeons, do not want to operate on a patient who’s going to die on the table,” explained Dr. David Ting, Associate Medical Director for Information Systems at the Massachusetts General Physicians Organization.  “The last thing they want to do is do harm to a patient or do something inappropriately.  The system automates searches using national guidelines, and then it essentially shows the results in a dashboard with a red, yellow, or green risk indicator for the surgeon to see.”

Precision medicine, personalized care, and genomics

“Precision medicine” entered the healthcare industry’s lexicon in a big way earlier this year during President Obama’s State of the Union address.  The President’s vision for a nationwide patient databank sparked hopes of a renewed commitment to genomic research and the development of personalized treatments, but the NIH isn’t the only one who has been using big data to predict the course of diseases related to a patient’s genetic makeup.

Healthcare predictive analytics has been particularly instrumental in the fight against cancer, and has also helped to target the development of preventative measures related to heart disease, diabetes, and even food poisoning based on genetic research.

Population health management, risk stratification, and prevention

Population health management is as much about prevention as it is about treatment, and healthcare predictive analytics equip providers with the tools they need to be proactive about their patients’ needs.  Targeting patients based on their past behaviors can help to predict future events, such as a diabetic ending up in the emergency room because he did not refill his medication or a child with asthma requiring a hospital admission due to environmental triggers of her disease.

By harnessing EHR data, providers can even identify links between previously disparate diseases.  A risk score developed by Kaiser Permanente researchers in 2013 allows clinicians to predict diabetic patients who are likely to develop dementia in the future, while the Army is attempting to curb the rampant rate of veteran suicides by leveraging a predictive risk model to identify patients who may be likely to harm themselves.

“We could save four lives for every hundred people we treated” with better data-driven care coordination and follow-up after a hospital stay for a psychiatric episode, said Lt. Gen. Eric B. Schoomaker, a former surgeon general of the Army and a professor of military and emergency medicine at the Uniformed Services University of the Health Sciences. “This would be unparalleled, compared to almost any other intervention we could make in medicine.  This study begins to show the positive effects big data can have, when combined with administrative health records.”

Healthcare predictive analytics can even prevent bottlenecks in the urgent care department or emergency room by analyzing patient flow during peak times to give providers the chance to schedule extra staff or make other arrangements for access to care.

“Emergency department crowding is a complex problem affecting more than 130 million patient visits per year in the US,” writes Joshua E. Hurwitz, lead author of a study detailing the effects of an online patient flow simulator. “In the current world of scarce resources and little margin for error, it is essential to rigorously identify the specific causes of crowding, so that targeted management interventions can have maximal effect.”

Reducing preventable hospital readmissions

As hospitals begin to feel the financial pinch of high 30-day readmission rates, they are turning to predictive analytics to keep patients at home.  At the University of Pennsylvania, informaticists can look at prior hospitalization histories to flag patients who may be returning to the inpatient setting within 30 days.

Real-time EHR data analytics helped a Texas hospital cut readmissions by five percent by drawing on nearly 30 data elements included in the patient’s chart. “This is one of the first prospective studies to demonstrate how detailed data in EMRs can be used in real-time to automatically identify and target patients at the highest risk of readmission early in their initial hospitalization when there is a lot that can be done to improve and coordinate their care, so they will do well when they leave the hospital,” said Ethan Halm, MD, MPH, Professor of Internal Medicine and Clinical Sciences and Chief of the Division of General Internal Medicine at UT Southwestern.

Meanwhile, the Kaiser Permanente system has been working to refine its readmissions algorithms in order to better understand which returns to the hospital are preventable and which are not, a crucial distinction for value-based reimbursements.

“Classifying readmissions as potentially preventable or not preventable can be used to improve hospital performance,” wrote the authors of the study comparing an algorithm to human review of readmissions cases. “Administrators can sort potentially preventable readmissions into categories that are actionable for improvement. They can identify trends over time or across reporting units. Classifying readmissions as potentially preventable or not preventable can also be used to establish accountability across reporting units and reward top performers.”

For more on  healthcare applications of analytics, see  PAW healthcare, September 27-Oct 1, 2015.

By Jennifer Bresnick
Originally published at

Leave a Reply