With the increase in available health data, scientists can make predictions about who will get sick, when they will fall ill, and what diseases they will have, experts said at the Health 2.0 Annual Fall Conference 2017 in Santa Clara, California.
“Five years ago, people were talking about big data and healthcare, but it was difficult to get ahold of datasets,” said Iya Khalil, PhD, cofounder and chief commercial officer of GNS Healthcare. “We were looking for data, but couldn’t find them.”
But the tide has turned, she reported. Data are now readily available from a variety of sources, including electronic health records and organizations that collect biospecimens, such as the Michael J. Fox Foundation.
These data can be used to develop tools to make predictions about population and individual health, she explained.
GNS is using data analytics to predict the likelihood of disease progression and to model therapeutic targets, molecular characterization of patient subpopulations, and disease biology, Dr Khalil reported during her presentation.
With data from melanoma patients, for example, “we are able to predict who is likely to respond,” which is valuable information because researchers have predicted that 10% to 15% of melanoma patients will not respond to any treatment, she pointed out.
And it is now possible to determine the cause of a health problem in a statistically meaningful way.
In fact, machine-learning technology was used to identify the low-density-lipoprotein disease pathway associated with coronary artery disease in preliminary studies. This proof-of-concept study — conducted by Dr Khalil, her GNS colleagues, and researchers from the Global Genomics Group — demonstrated how discoveries like this could be accelerated in the future.
Predicting Who Will Get Sick
Invisible patients are the proverbial “needles in a haystack” in a population, said Jason Pyle, MD, PhD, who is chief executive officer of BaseHealth, a company that is using analytics to pinpoint where patients are on their health trajectory.
The prevention of adverse health events is of increasing interest to provider networks. The idea is to find people who look reasonably healthy on paper but who are headed for a tipping point, where they will “face one adverse health event after another,” Dr Pyle explained. “With the right medical intervention at the right time, we can change their healthcare trajectory.”
“You can’t rely on claims to pinpoint at-risk patients; you have to look at everything else,” he said. Researchers use a database of information from 55,000 peer-reviewed scientific papers on the 41 disease categories that account for about 90% of all healthcare expenses, and look at individual records for clues that a patient might be headed for problems.
Hypertension, advanced coronary artery disease, borderline blood-sugar problems, inactivity, prediabetes indications, and risky behavior, such as drinking and smoking, are all indicators.
“The core of our technology is an expert system that takes into account the entire peer-reviewed literature and about 70 million patient lives,” said Dr Pyle.
Identifying “Cost Bloomers”
This will allow healthcare providers to offer preventive services before the onset of costly health problems, which can save the system “big dollars,” said Nigam Shah, MBBS, PhD, from the Center for Biomedical Informatics Research at Stanford University in California, who is cofounder of Cardinal Analytx.
Cost is driving most of the market for the business of health predictions, he pointed out.
Dr Shah and his colleagues have been working to identify “cost bloomers” — people who will become sick in the next year — and then suggest precisely targeted interventions. They have been providing individual healthcare expertise to physicians so that they can make better personalized decisions for their patients.
Physicians can launch a consultation with the Stanford team by providing a blood sample and a patient’s chart.
“We then produce a report showing what has happened to patients with similar problems,” Dr Shah told Medscape Medical News. So far, the team has done 32 consults.
The team also looks at questions not asked by the referring physician. For example, “if the patient is taking six drugs and you add a seventh drug, what happens to their GI bleed? What is their chance of ending up in emergency?” Dr Shah explained.
Health 2.0 Annual Fall Conference 2017. Presented October 2, 2017.
Follow Medscape on Twitter @Medscape and Ingrid Hein @ingridhein