Skip to Main Content

INFORMATION FOR

Big data for big questions

Yale Medicine Magazine, 2016 - Spring

Contents

More than a million Medicare beneficiaries receive hospice care every year, at a cost of more than $13 billion. The benefit is meant to provide comfort and palliative care in the last six months of life. But does it work? That’s what second-year medical student Shayan Cheraghlou wanted to find out in his research with Thomas Gill, M.D., professor of geriatric medicine and director of Yale Program on Aging.

As a participant in START@Yale, Cheraghlou spent the summer before his first year of medical school conducting research with Gill. The effort earned him a first author credit on the paper “Restricting symptoms before and after hospice” published online last March in the American Journal of Medicine. Cheraghlou also presented a poster at this year’s annual meeting of the American Geriatrics Society in May.

But it wasn’t an interest in hospice that drew Cheraghlou to spend his last summer before medical school conducting research. It was big data. “I had read a lot about how data are being used in medicine to guide practice,” Cheraghlou said. “I wanted to get involved in a project where I could learn how to code and use big datasets to ask and answer relevant clinical questions.”

Gill had just the right kind of dataset. He has followed a cohort of 754 adults over age 70 for more than 18 years, with monthly telephone interviews and comprehensive home-based assessments every 18 months. Linked to health care utilization data from the Centers for Medicare and Medicaid Services, the combined resource comprises the Yale PEP Study, a project sponsored by the National Institute on Aging and conducted at the Yale Program on Aging/Pepper Center. “It’s a unique resource and has provided an increasingly large number of students and other junior investigators extraordinary opportunities to evaluate clinically meaningful questions in a high-priority population,” Gill said.

With these data, Cheraghlou learned how to program and code using statistical analysis software, completing a set of complex analyses in the 10 weeks before the academic year began.

From the cohort of 754, Cheraghlou analyzed data on 241 participants who entered hospice between 1998 and 2013 and who have since passed away. He had access to a wealth of information on the seniors. In the year before and three months after each of the 241 entered hospice, Gill and his research team interviewed and assessed them for restrictions in their daily activities, including 15 symptoms that could be causing these restrictions.

“What we found was that these symptoms increased pretty drastically in the few months prior to their admission in hospice,” Cheraghlou said. The study revealed an increase in both the prevalence of restricting symptoms, defined as spending more than half of the day in bed or cutting back on their daily activities, and the number of these symptoms. Hospice is meant to help alleviate this burden, but Cheraghlou wanted to know whether it actually does.

“After admission to hospice, some symptoms came down close to, equal to, or even below where they were in the year prior to admission, so it’s like they got that functionality at the end of those three months,” Cheraghlou said.

Hospice was indeed working, but it could do more. Medicare offers the benefit for six months, but Cheraghlou found that patients only take advantage of it for about two weeks. On average, they died 15 days after admission. “The fact that doctors are waiting until the very end to send patients to hospice suggests that we might not be using it as effectively as we could,” he said. “If the trends that we’re showing are true and it’s really helping individuals deal with these symptoms, it would be helpful to do it earlier.”

Such big datasets make it possible to answer big questions like whether a nationwide government-funded health benefit actually works. Access to this type of information is only increasing, Cheraghlou says.

“Data collected in health care are now accessible in large sets. Then, outside of health care, you have Internet search histories, online shopping patterns, etc. There’s a wide variety of information that’s relevant to medicine and to the way people lead their lives in general that wasn’t available before.”

His work with Gill complete, Cheraghlou is now querying big data on big questions about head and neck malignancies with Benjamin Judson, M.D., associate professor of surgery. “I don’t know exactly the area of medicine I’ll go into, but I’d definitely like to continue data-driven research in the future.”