Gillings researchers use big data, such as insurance claims and medical records, to generate real-world evidence on the safety and effectiveness of prescription drugs.
At most medical appointments, doctors and nurses ask their patients a number of questions and enter the answers into a computer or tablet. What happens to all that information?
Translating Science into Practice: Doing great public health research means putting study results to work. Whether examining molecular-level data or facilitating large-scale organizational changes, we’re Gillings. We’re on it! Gillings faculty go beyond the literature and the lab to make real changes in health-care practice and policies — and real differences for patients.
UNC Gillings School of Global Public Health researchers are using those data to generate real-world evidence to improve knowledge about how prescription drugs affect patients.
Before new drugs reach the market, they must endure years of testing. But knowledge is limited about the benefits and harms of these treatments because of the small number of patients tested and how they are selected for trials. There also is a lack of information on long-term effects and potential interactions with other treatments.
That means drugs are being approved, prescribed, and used without a complete picture of their safety and effectiveness in the real world.
Til Stürmer, PhD, MD, MPH, Nancy A. Dreyer Distinguished Professor and chair of epidemiology; Jennifer Lund, PhD, associate professor of epidemiology; and Michele Jonsson Funk, PhD, associate professor of epidemiology and director of the Center for Pharmacoepidemiology (the branch of science that studies the effects of drugs in large numbers of people), are working to change that through rigorous research using big data such as insurance claims and electronic medical records, which can provide information on millions of patients compared to the smaller clinical samples in drug trials.
“These big, or real-world, data allow us to get timely answers to important clinical questions that cannot be answered quickly enough through any other means,” Stürmer says. “Being able to use these data is a prerequisite for assessing the real-world benefit and harm of drugs.”
For example, about 80 percent of pregnant women report nausea. When the generic version of Zofran, a drug developed to combat nausea in cancer patients, hit the market a few years ago, it was being prescribed to about 20 percent of pregnant women — despite inadequate data on its safety.
“Using big data to conduct robust research … is truly a team science effort.”
Michele Jonsson Funk, PhD
Associate Professor of Epidemiology and Director of the Center for Pharmacoepidemiology
Jonsson Funk and one of her graduate students, Elizabeth Suarez (who graduated in Fall 2019), analyzed de-identified health-care data and found that women using Zofran (or its generic form) during pregnancy did not have a higher risk of miscarriage, preterm birth, or gestational hypertensive disorders compared to women who used other drugs to treat their nausea.
“By using these big data, we can see the effects these treatments have on a diverse population of patients in the real world, not just the highly selected ones in the clinical trials,” says Jonsson Funk, who has studied several women’s health issues including medication use during pregnancy, pelvic floor disorder treatment, sex differences in statin benefits, and effects of diabetes medications on breast and endometrial cancer risk.
UNC is home to a uniquely rich resource: de-identified data from a nationwide 20 percent random sample — about 4.5 million individuals — of Medicare beneficiaries ages 65 years and older, including all medical encounters, procedures, and pharmacy-dispensed prescription drugs over time until the death of the beneficiary. Originally funded by a Gillings Innovation Laboratory, this dataset is now available to all UNC researchers.
The interdisciplinary diabetes working group led by Stürmer and Dr. John Buse, MD, PhD, chief of endocrinology at the School of Medicine, uses these big data to better understand the effects of various diabetes treatments. Stürmer also has used Medicare data for research on treatments after heart attacks, and on potentially inappropriate prescriptions that could put older patients at risk of adverse health outcomes.
“The introduction of Medicare drug plans (Part D) in 2006 was a game changer. UNC’s epidemiology department was at the forefront of using these data,” Stürmer says. “More recently, our ability to link multiple data sources, including insurance claims with clinical data from electronic health records, allows us to combine rich longitudinal records with clinical detail that was previously unavailable.”
Addressing questions of drug safety and effectiveness is not just about putting the right data together — it’s also about putting the right people together. Fortunately, the interdisciplinary environment at UNC is one that nurtures collaborations.
“We bring together pharmacoepidemiologists who understand the particular study designs that can increase the reliability of findings, clinicians who understand the condition so that the questions we ask are relevant, biostatisticians who ensure that our analytic tools are appropriate to the task, programmers who can efficiently manipulate terabytes of data, and patients who share which health outcomes are most important to them,” Jonsson Funk says. “Using big data to conduct robust research and produce meaningful results is truly a team science effort.”
Opioid Use
Opioid abuse in North Carolina has reached epidemic levels: In 2017, more than five North Carolinians died each day from an unintentional opioid overdose. To reduce the supply of unused, misused and diverted opioid pills, in 2017 the state legislature enacted the Strengthen Opioid Misuse Prevention Act, or STOP Act. Among other provisions, the law limits the legal prescription of opioid pain medications to a five-day supply for acute injuries and a seven-day supply post-surgery. It does not restrict opioid prescriptions for chronic pain.
While many states have implemented prescribing limits based on the number of days supplied, Gillings graduate student Jessica Young used national insurance claims data to examine other dimensions of opioid prescribing, such as the number of pills dispensed, and overall dosage (morphine milligram equivalents [MME]) dispensed. Her results show that these dimensions can yield a different perspective on patient use. Notably, one out of every 10 patients receiving opioids for postoperative pain received over 500 MMEs, putting them at 21 percent higher risk of having long-term opioid use following surgery compared to those receiving a dosage under 500 MMEs.
“The law uses ‘day supply,’ but there are other ways to characterize people who might be getting more opioids than needed,” said Young’s adviser, Michele Jonsson Funk, PhD, associate professor of epidemiology and director of the Center for Pharmacoepidemiology.
Young has received a $142,967 dissertation award from the National Institute on Drug Abuse to support completion of her doctoral dissertation in the field of drug use research.
