A path to combatting the opioid crisis may lie with millions of members and their medical claims, Health Care Service Corporation’s medical directors suspected.
It was 2018. By then, the U.S. opioid epidemic already had claimed nearly 400,000 lives, with an average of 130 people dying daily, the Centers for Disease Control and Prevention reported. Guided by federal and state regulation and legislation, health care providers began limiting prescription doses and amounts, treating almost every patient as a potential overdose victim.
HCSC leaders wanted to find a way to identify members most likely at risk for an opioid overdose or adverse event. They believed data would lead to the answer.
That belief paid off. Claims data helped them validate a screening tool that could accurately predict members at highest risk for an opioid-related event.
Their findings were published in the online edition of the Pain Medicine Journal. The analysis was a first-time collaboration between HCSC’s data scientists and behavioral health, pharmacy and medical experts using predictive analytics to target a specific medical condition.
“To really impact a member’s quality of life, that’s really something.”
“Having data scientists working with clinicians could lead to much more efficiency,” says Dr. Frank Webster, HCSC senior executive medical director. “To really impact a member’s quality of life, that’s really something.”
Confirming accuracy
The data scientists began their work by reviewing a study published in 2015 that focused on opioid overdoses and events among Veterans Health Administration patients.
The VA study used a risk index created to predict opioid overdose likelihood based on health conditions and opioid medication characteristics.
HCSC’s team applied the methodology to members of the company’s plans in Illinois, Montana, New Mexico, Oklahoma and Texas to determine its accuracy.
“It was 90% accurate” for the VA patients, says Lan Vu, HCSC principal data scientist. “So, we said we can take this, use it for our population and see if it will still work.”
Vu and her team worked with HCSC pharmacy experts to identify nearly 1,100 members who had an opioid overdose. They applied the risk index to these members and a control group and looked at emergency department use, hospitalizations and changes in prescribed opioid use within a six-month period.
They found the VA’s risk index was equally successful at identifying HCSC members at risk of an opioid overdose.
“Everybody was really excited,” says Vu of the results. “We wanted to help our members. It was very tedious. But we wanted to do anything we can do to make member outreach simpler.”
Analyzing data to improve care and efficiency
Nationwide, health care providers and payers are analyzing the millions of pieces of data they collect to become more efficient, as well as improve quality of care and patient outcomes.
In its 2019 survey of health care payers and providers, the Society of Actuaries found more than half of the respondents use predictive analytics, and most respondents said they use or plan to use predictive analytics in the next five years.
According to the survey, almost half the respondents said predictive analytics help their organizations improve patient satisfaction, while nearly 40% reported cost savings.
“There’s always more we can do,” Vu says, adding that she and her team are eager to perform more analytics projects from which members could benefit.
Using data to guide outreach
HCSC’s findings have become part of its Risk Identification and Outreach program (RIO).
RIO, comprised of behavioral health, medical and pharmacy professionals, work with company data scientists to identify risk patterns and offer outreach and interventions to members and providers. The team uses the risk index to identify members in danger of having an opioid-related event. Since RIO started in July 2019, the program has identified more than 5,060 members for assistance and enrolled at least 2,090 of them in its case management program.
Dr. Conway McDanald, HCSC vice president and chief medical officer of behavioral health, is hopeful more data analysis can be performed to help members with other conditions or diseases.
“This was a great opportunity for us to analyze, organize and visualize clinically actionable data for at-risk member populations, and this model can be applied in a variety of clinical settings,” he says.