Today’s healthcare organizations face increasing pressure to achieve better care coordination and improve patient care outcomes. To accomplish these results, organizations are turning to predictive analytics.
This area of statistics deals with the use of data and machine learning algorithms, predicting the likelihood of future outcomes based on past data. Predictive analytics can be used in healthcare to “identify pain points throughout the stages of intake and care to improve both healthcare delivery and patient experience,” says Lauren Neal, a principal at Booz Allen Hamilton.
“The combination of analytics and human-centered design can ensure that healthcare providers address inefficiencies along the patient journey and tailor services to meet the unique needs of the patient population,” says Neal.
According to a 2017 study by the Society of Actuaries, 93 percent of health organizations say predictive analytics is important to the future of their business, with 89 percent of providers currently using predictive analytics or planning to do so in the next five years.
MORE FROM HEALTHTECH: Find out how predictive analytics applications are changing oncology.
Neal says that the Office of the National Coordinator for Health Information Technology, in partnership with the University of California, San Francisco, has already begun applying machine learning algorithms to predict outcomes for patients with kidney disease, helping to keep people healthy and cut costs.
“With 20 percent of Medicare’s budget going to the treatment of kidney disease, predictive modeling can provide clinicians with additional insights into the risks and benefits of treating patients earlier, with the goal of reducing the number of Americans developing end-stage renal disease,” Neal says.
Though expectations around future capabilities remain varied, plenty of healthcare organizations are already seeing benefits from predictive analytics in the way of patient care. Here’s a look at a few of those instances.
Penn Medicine Looks to Predictive Analytics for Palliative Care
Philadelphia-based healthcare system Penn Medicine began harnessing predictive analytics in 2017 to power a trigger system called Palliative Connect.
The program gleans data from a patient’s electronic health record and uses a machine learning algorithm to develop a prognosis score. The generated score, which is based on 30 different factors, helps clinicians determine a patient’s likely prognosis over the next six months.
The program ultimately works by “identifying patients who are at the highest risk of a bad outcome when they come into the hospital,” Dr. Katherine Courtright, assistant professor of medicine at the Perelman School of Medicine at the University of Pennsylvania, explains. “It helps our palliative care team recognize those patterns and proactively reach out. It’s a proactive approach.”
Predictive analytics works particularly well for this type of patient identification, Courtright says, because it isn’t reliant on a clinician’s witness of warning signs.
“We know one of the barriers to getting these services to seriously ill patients, particularly in a hospital setting, is the focus in hospitals on the acute problem,” says Courtright. “When clinicians are so busy, they’re focused on what the patient came in from. It’s hard to step back and see the whole person as a trajectory.”
Palliative Connect initially ran as a pilot program at one of Penn Medicine’s hospitals from December 2017 to February 2018. The program assisted in identifying 85 patients for consultation, compared to 22 that would have been identified in a similar patient population without predictive analytics — a 74 percent increase. In July, Courtright published the full results of the pilot program in the Journal of General Medicine.
Since the inception of Palliative Connect, Penn Medicine has expanded its use to increase the reach of expert palliative care for the seriously ill. Researchers have also begun conducting a second pilot program at another one of the system’s hospitals, this time with an increased number of patient participants.
READ MORE: Discover how predictive analytics can play an influential role in operating rooms.
Predictive Analytics Makes Headway in Imaging
When it comes to medical imaging, predictive analytics is already making waves in speed and accuracy.
CheXNeXt, an artificial intelligence algorithm being trained and studied by researchers at Stanford University, is able to screen chest X-rays in a matter of seconds to detect 14 different pathologies with an accuracy rivaling that of radiologists. CheXNeXt researchers hope to be able to use the algorithm to help with the diagnosis of urgent care or emergency patients who come in with a cough.
Although the algorithm has yet to be introduced in a clinical setting, Dr. Matthew Lungren sees this technology changing the way care is offered by prioritizing patients based on predicted outcomes.
“The algorithm could triage the X-rays, sorting them into prioritized categories for doctors to review, like normal, abnormal or emergent,” says Lungren, assistant professor of radiology at the Stanford University Medical Center, in a Stanford Medicine article. “We need to be thinking about how far we can push these AI models to improve the lives of patients anywhere in the world.”
Predictive analytics in medical imaging is set to have a big impact on cancer care too, says Anant Madabhushi. For pathologists, it will mean using predictive analytics to improve identification of specific things on images.
“They are using AI to find things like lymph node metastases,” Madabhushi, bioengineering researcher and director of the Center for Computational Imaging and Personalized Diagnostics at the Case School of Engineering at Case Western Reserve University, says. “These things are challenging, and AI can play a role in alleviating that challenge.”
Predictive modeling will inherently help oncologists make better-informed decisions regarding patient care. Instead of conducting tissue-destructive tests or relying on genomics, AI algorithms can harness information from images to identify patients with a more aggressive disease who are therefore in need of more aggressive treatment. It could also let physicians know which patients have less aggressive cancer and might be able to avoid the side effects of chemotherapy.
And though research in predictive analytics for patient care is still developing, Madabhushi says that it will ultimately become a significant tool for radiologists and oncologists in their roles treating cancer.
