Cigna’s proprietary algorithms are aided by the use of machine learning, a subfield of artificial intelligence that refers to the science of getting computers to act intelligently without being explicitly programmed.
“If we’re doing work in the opioid space that’s unparalleled, we’ll be the preferred partner for the customer, client and the physician,” Dr. Boxer said. In addition to adding competitive advantage, the algorithms are potentially saving lives and contributing to a decrease in health care costs for the patients that might have otherwise become addicted to opioid painkillers, he said.
Dr. Boxer spoke about Cigna’s artificial intelligence initiatives ahead of the health-insurance company’s impending $54 billion merger with Express Scripts. He declined to comment on how the deal would impact his role.
In 2016, more than 40% of 42,250 opioid overdose deaths in the U.S. involved a prescription opioid, with the most common drugs being methadone, oxycodone and hydrocodone, according to the Centers for Disease Control and Prevention.
In 2017, the Department of Health and Human Services declared the opioid crisis a national public health emergency. Insurers over the past two years have announced policies that aim to help curb opioid-related overdoses.
Deaths from heroin and synthetic opioids have surpassed prescription opioids as a cause of overdose death, WSJ has previously reported.
Cigna in 2016 committed to reducing opioid use among its customers by 25% within three years, Dr. Boxer said. “We have taken a leadership role in addressing this crisis,” Dr. Boxer said.
As part of that goal, the company that year began building a predictive model, using machine learning and predictive analytics, that identifies customers likely to overdose within the next month.
A combination of 16 datasets are used to inform the algorithms, including data about patients’ behavioral health claims, chronic disease history and interactions with pharmacies. The algorithms were built with the help of in-house staff. Over the past few years, Cigna has hired more than 1,000 data scientists, software engineers and analytics experts, Dr. Boxer said.
When the patients are identified, a behavioral case manager reaches out to them. The ultimate goal is to proactively change behaviors, Dr. Boxer said.
During a yearlong pilot program that ended in June 2018, Cigna identified about 1,130 customers for outreach.
The algorithms are still in use today.
In addition to helping prevent opioid abuse through the use of algorithms, Cigna also requires clinicians to analyze claims data to find opioid use patterns that suggest possible misuse, and to notify the health care providers of the potential risk, according to a Cigna spokesperson. It has also identified more than 2,600 prescribers of high-dosage opioid medications to ensure the dosage is appropriate, medically necessary and safe for the patient, the spokesperson said.
The high-risk opioid customer outreach initiative is one example of how the company is using artificial intelligence to help distill insights from massive datasets. “It’s getting increasingly hard to detect the signals and patterns from all that data. Separating the signal from the noise in health care can create tremendous value,” Dr. Boxer said.
The company has also developed a machine learning-enabled tool called One Guide that analyzes data from medical claims, medical procedures, biometric data, benefits and pharmacy claims to anticipate customer needs. For example, the tool can help identify customers that have not used financial incentives for annual checkups or free health coaching. The tool can help create personalized messages for customers, who are then notified through the myCigna mobile app.
Dr. Boxer said artificial intelligence initiatives such as these have helped improve the customer experience. “We want to be relevant to the consumer in a way that differentiates us,” he said.