December 17, 2019
Artificial intelligence and big data offer some of the most exciting prospects for medicine. But making good use of the data is another story.
Data available for research is growing at an unprecedented pace, in part driven by a decade-long move toward electronic records. Big data in health care holds promise in reducing care costs and hospital re-admissions, improving care and preventing adverse drug effects, according to a report published in 2018 and presented at the American Medical Informatics Association’s annual meeting.
Even Alphabet’s Google is getting in on the action, testing out machine learning algorithms that will read chest X-rays with the accuracy of a radiologist, the website Venture Beat reports. The radiologists aren’t waiting for Google to come in and do all the work, though. The American College of Radiology earlier this year announced a collaboration with computing company NVIDIA that will allow thousands of radiologists to use artificial intelligence for diagnostic purposes, using their own data in their own facilities to protect patient privacy.
Researchers are using huge data sets to learn more about the functions of genes to improve precision medicine, which creates therapies based on a patient’s unique genetic code. Dignity Health is leveraging big data and analytics to spot sepsis in its earliest stage, an inflammatory response to an infection that kills about a quarter of a million people in the U.S. each year. Express Scripts is using big data to spot chronic illness or a patient with a high risk of getting addicted to painkillers.
But when it comes to data, quality and quantity are paramount. Ownership confusion, privacy protections and technological issues have all been obstacles in recent years to using the mountains of data that’s available.
Health care providers have poor understanding of the “black box” aspect of many machine learning programs, where it may be difficult to understand why the algorithm generated the results that it did, even if they end up being fairly accurate. That makes many health care providers nervous about applying machine learning’s suggestions to an actual patient.
It’s also difficult to use data in clinical research. Most data is collected for a specific reason: a pharmacy wants to record inventory or a doctor’s office wants to bill insurance companies. But clinical research needs more specific information, such as exactly the time a particular drug was administered and its outcome. Hospitals and other health care providers also must invest in technology that will de-identify patients, so the data can be shared with researchers without compromising privacy. The industry needs financial incentives and clear guidelines on the use of data analysis in patient care, according to the authors of the 2018 report.
Even though health care’s use of data has lagged other industries, more investment in the space should help. Retailers, for example, already know consumers pretty well. Why shouldn’t health care providers understand their patients too?