Unlocking the promise of learning from everyone with cancer

Locked behind the firewalls of proprietary systems sits a treasure trove of data that could help diagnose heart disease, diabetes, cancer, and other conditions faster and more accurately and better treat people with them. But there it sits, largely untapped, because the electronic health record infrastructure was never designed to let organizations easily share data.

Electronic health records were first developed in the 1960s but didn’t become mainstream until about 12 years ago when the federal government provided incentives for their use. At the time, expectations were high that they would be the solution for seamlessly and securely collecting and sharing valuable patient data. EHRs would reduce the need for faxing records from one doctor’s office to another and end the practice of manually inputting the same information into multiple databases.

The country isn’t there yet. The rush to develop electronic health records produced proprietary and competing data systems that are customized for each health provider organization, like hospitals and medical offices. In addition, it’s taken years to develop software standards to enable the sharing of data across systems. Thanks to organizations such as Health Level Seven International and the Office of the National Coordinator for Health Information Technology, some progress has been made.

The chaos of Covid emphasized the lack of standards and what electronic health records can’t easily do, like quickly make shareable patient data available so doctors can evaluate which treatments worked for which patients.

At the height of the pandemic, as physicians at the Mayo Clinic were battling to keep patients alive, the medical staff frequently had to hit pause to fill out lengthy REDCap surveys to inform Minnesota state health officials about the number of patients they were seeing with Covid-19. The state-mandated surveys lived outside Mayo’s regular EHR system, and so required painstaking manual work to record Covid case information on giant Excel spreadsheets. “When we were going through our surges, we were drowning and burdened” by all the paperwork, Priya Sampathkumar, an infectious disease and critical care specialist at Mayo, told my team at MITRE, the nonprofit research and development organization I work for. System A couldn’t talk to System B. “Filling out these pieces of paper just added insult to injury.”

Given the ways electronic health records are currently structured, it’s difficult, if not impossible, to share and analyze high-quality data from millions of patients in disparate health systems to drive research, improve current treatments, and inform meaningful discussions between patients and their providers.

For example, most of the data that lead to novel cancer treatments today come from clinical trials. That’s a problem, because these trials involve less than 6% of adult Americans living with cancer. The percentage for children is even smaller. That means there is limited information about what treatments work for which patients. Clinicians and researchers don’t have ready access to data from the vast majority of cancer patients, data that could potentially identify what treatments have worked well for, say, a 52‑year‑old Hispanic woman who has diabetes as well as breast cancer or a 70-year-old man with Stage 3 lung cancer.

The quest to expand standards and interoperability across electronic health record systems to improve the quality, safety, and effectiveness of patient care is, fortunately, not starting from scratch. The Fast Healthcare Interoperability Resources (FHIR) standard makes it easier to share data by defining how health information can be exchanged among different computer networks, regardless of how it is stored in those systems.

In 2019, MITRE and several other nonprofits launched an effort to develop a common standard and language for cancer care that could be incorporated into EHRs and used to capture the characteristics, treatments, and outcomes of every person with cancer. We built our standard on existing best practices, such as HL7’s experience developing the Fast Healthcare Interoperability Resources standard.

The result, mCODE (short for minimal Common Oncology Data Elements), is now being tested by more than 60 health organizations and other stakeholders — including EHR vendors — who see the potential of learning from millions of patients’ experiences. We chose cancer to test the hypothesis that only a minimal amount of critical information is necessary to produce valuable results comparable to those found in clinical trial reports.

We also learned the necessity of involving the community to build consensus around new standards and drive them forward. mCODE was developed by a multidisciplinary group of subject matter experts, including cancer clinicians, informaticists, health services researchers, experts in data standards and interoperability, people living with cancer, and others under the auspices of MITRE and the American Society of Clinical Oncology.

Rather than focusing on standards for exchanging data, mCODE aims to standardize health records so diverse stakeholders can share information in a meaningful way to achieve large-scale outcomes, such as more efficient research, faster trial matching, and more personalized medicine. Using mCODE’s common data language and open-source, nonproprietary model, organizations can access and analyze data from various EHR systems, including essential data that can be hard to find today, such as patients’ cancer stages or the outcome of specific treatments.

In its first pilot project, the mCODE team collaborated with a clinical trial group that is testing a new use of an existing drug to treat breast cancer. Initial results on preliminary data, not yet published, indicate that the accuracy of mCODE’s results match those of the clinical trial team’s 95% of the time. Since then, mCODE has been incorporated into several other clinical trials, and is being tested for other uses, such as cancer registries and prior authorization of treatments. The team is also freely sharing its expertise and open-source technology with organizations involved in cardiac diseases, genomics, and dementia who want to use the mCODE approach to develop and test standards for their specialties.

With a standards-based approach like mCODE, every doctor would have valuable information about a patient’s disease and possible treatments at their fingertips at the point of care. The insights have the potential to improve patient care and shared decision-making, drive innovation, and set the foundation for a national cancer health learning system.

The possibilities to improve patient care and research through sharable data are endless. But it will take the whole community to make it happen: electronic health record vendors, health systems, payers, researchers, and patients. It also will take a change of incentives. Many of today’s players, such as EHR vendors and health systems, have made their patient data proprietary, believing it gives them a competitive advantage. That’s not the case anymore, since no one organization can ever have enough data on its own to solve big problems.

As doctors, scientists, and patients desperate for effective treatments see the obvious benefits of gaining access to data at scale, they will drive this approach forward. Unlocking the potential of these proprietary systems could not be more important.

Jay J. Schnitzer is a pediatric surgeon and the senior vice president, chief medical officer, and chief technology officer at MITRE.

First Opinion newsletter: If you enjoy reading opinion and perspective essays, get a roundup of each week’s First Opinions delivered to your inbox every Sunday. Sign up here.