The NEJM Catalyst Insights Council members responding to our recent Insights Report on data in health care say that genomic data will be among the top three most useful sources of health care data in 5 years — a substantial increase from how the survey respondents view genomic data today, as the least useful data source.
But for some Council members who are familiar with genomic data, a 5-year time frame for broad use of genomic data is wishful thinking.
“We use mostly the usual forms of data: claims reimbursement, utilization, and hospital admissions. We are nowhere near using genomic data,” says Mark Xuereb, MD, Senior Medical Director at Anthem BlueCross BlueShield in New York City.
Xuereb says genomic testing is not mature enough to be a trustworthy source of data to inform insurance decisions. “We use data sets to create, update, or change local health care guidance and guideline protocols,” he says. In using genomic data, “that’s where the problem can come in.”
In New York City, with so many different ethnic populations and unique socioeconomic factors, insurance companies must be very careful in studying each neighborhood’s disease states and utilization of health care resources. For instance, guidance and guideline protocols for neighborhoods with ethnic concentrations must account for endemic medical conditions. Xuereb says the cost to treat plan members based on their unique genetic code would be too high considering genomics’ nascent status.
Farther on the horizon — perhaps 10 years — Xuereb believes health care will be driven by external data such as population data coupled with genetic data. “The combination would allow individual providers and care givers to customize treatment according to the individual’s genetic makeup and environmental status,” he says.
Marco Huesch, MBBS, PhD, Vice Chair of Radiology Research and Director of the Center for Optimizing Radiology Value at Penn State Hershey Medical Center in Hershey, Pennsylvania, also is wary of predicting genomic data’s widespread use in 5 years.
Radiology at the medical center uses genomic testing in breast cancer risk stratification, diagnosis, and treatment, and the Penn State Hershey Institute for Personalized Medicine has a robust biorepository infrastructure with intensive genomics projects across clinical areas such as trauma, oncology, diabetes, and surgery. But until “organizations are collecting gene samples from everyone in a consistent manner,” Huesch doesn’t believe the use of genomic data will go mainstream.
Instead, he primarily uses clinical data from the electronic medical record system for predictive analytics — for example, gauging the likelihood of a patient to develop sepsis or the likelihood of a successful outcome for candidates for ECMO (extracorporeal membrane oxygenation) procedures. Because radiology touches 85% of the medical center’s patients, Huesch says his department is in a good position to gather and analyze data, as well as provide actionable insight for other medical teams.
Using 4 years of data, Huesch can model, with an 88% success rate, who is at high risk for sepsis. Sepsis, a dysregulated immune response to infection, occurs in 4% to 5% of patients and, of those, one in 8 or 10 will die. “It’s a high-profile condition and important for our institution — we hope to avoid terrible outcomes,” he says.
While Huesch might sprinkle in some external claims data (to compare outcomes against institutions across the state) and social media data (to understand sentiment about topics such as mammogram screenings), he has found the sweet spot is in using the organization’s own clinical data to improve the care of patients.
Christopher Dale, MD, MPH, Medical Director for Quality and Value at Swedish Medical Group, which is part of Providence St. Joseph Health, is carefully watching advances in genomic data in other parts of the large Seattle-based nonprofit health system. Swedish Medical Group is using genomic data to determine who would respond to certain cancer treatments. “Beyond that, our team has no plans to incorporate genomic data in the next year or two,” he says.
Swedish Medical Group’s current go-to data sources are in line with respondents to the NEJM Catalyst Insights Report, who name clinical data (chosen by 95%), cost data (56%), and claims data (45%) as their top three. Dale specifies, though, that because of the complexity of cost data, his organization uses item counts rather than their actual cost. For clinical data, Swedish Medical Group incorporates data from various registries in areas like obstetrics and surgical quality improvement.
Dale describes the organization’s big data vision as “reliable delivery of person-centered, high-value health at scale” but says it will be a journey to get there. He believes patient-generated data — which in the survey rises from fourth place to third in 5 years as a most useful source of health care data — will come into play much sooner than genomic data. “Wearables and patient-reported outcomes are a lot closer,” he says.
Right now, he’s focused on what most of his Insights Council peers are: making data and technology systems more interoperable. “There is a lot of work that needs to be done to realize the promise of big data,” he says.