Improving Outcomes, Value, and Data Collection in Oncology

Peter E. Gabriel, MD, MSE

Informatics, the science behind compiling and making use of high-quality data, could revolutionize how clinicians understand cancer and treat patients over the next decade, according to one expert.

Peter E. Gabriel, MD, MSE, the chief oncology informatics officer for the Penn Medicine Abramson Cancer Center and director of informatics for the Department of Radiation Oncology at the University of Pennsylvania, discussed the need to collect better, actionable data at the 10th Annual NANETS Symposium in Philadelphia. Making strides in that realm could do more for the field than drug development and even immunotherapies, he said. But so far, cancer centers have done a poor job at taking advantage of data infrastructure.

“I think that this whole area—being better at managing data and information—is really the key to what I would call the next era of advancement in oncology,” Gabriel said.

He zeroed in on 3 areas where informatics could drive change in oncology: precision medicine, value-based care, and research.

As the body of evidence grows, clinicians struggle to keep up with new breakthroughs. Instead of relying only on their minds, they must begin to leverage information contained in electronic health records (EHR), Gabriel noted. EHRs can help providers better understand the individual and the specifics of the disease, opening the door to personalized care, he said.

Further, computer-generated predictive analytics enable providers to understand how a path of care will likely affect a patient. “Clinicians tend to think that we know that stuff by intuition, experience, and judgment,” Gabriel said, “but we don’t. Computers can do a much better job if they have the data.” That well-grounded insight can arm physicians with the knowledge to make decisions that result in better outcomes.

Meanwhile, the prices for cancer treatments are skyrocketing. By 2020, for example, the total cost of cancer care is projected to hit $175 billion, a 40% increase from 2010. That rise has yet to improve outcomes, Gabriel said. He pointed to the body of research, which shows that oncologists “fail to deliver the right care to the right patient at the right time way too often, and we have too many mistakes that cause harm.” The industry must strive for value, he added; otherwise, health insurance companies will impose solutions, which could prove prohibitive for clinicians.

Informatics allow a hospital to measure the quality of its care, Gabriel said. In turn, they may target areas in need of improvement, develop action plans, and identify outlier providers who aren’t using the right therapies, he said.

Finally, large-scale data collection and sharing stands to yield more accurate findings than clinical trials. Gabriel said those trials enroll roughly 3% of the overall population, a proportion that leaves plenty of data on the table. “That’s not a good thing. It means that the research we do is difficult to generalize,” he added. “We need to find a way to learn from what we’re doing every day.”

If a number of medical institutions were to collaborate to develop a flow of normalized, combinable data, they could lay the foundation for sweeping analyses. Data registries, he said, can also help meld EHR and other data to build a base from which hypotheses can be tested. “It has the potential for much more rapid accrual and much lower costs than in clinical trials,” Gabriel said.

But that task would require oncology to overcome several challenges. Although nearly 100% of hospitals have some sort of EHR system, the data contained there are often “messy,” Gabriel said. The components are often unstructured, meaning that they might be written text that is not easily boxed and sorted. And structured data is sometimes incomplete, he added.

Thus, institutions must steer clinicians away from free-text notes. Innovators, on the other hand, must look to enhance natural language processing so that it can sift important nuggets from unstructured data. Informatics professionals can also help process and interpret data, Gabriel said.