5 steps toward a learning health system
Without good data, patient-centeredness is just a buzzword. And without a patient-centric focus and proper organization, data can be rather useless.
[See also: AMIA looks to workforce of the future]
That was the message Sunday from Amy Abernethy, MD, a hematologist/oncologist and palliative care physician, who delivered the opening keynote address at the American Medical Informatics Association's annual symposium.
"Patients are the anchor for our work," Abernethy, the CMO of Flatiron Health, a New York City-based oncology analytics company, told this gathering of mostly academic medical informatics specialists. Patients' stories help clinicians work better with data, she said.
[See also: The cancer doctor]
Comparing the flood of data now available in healthcare and medicine to the mighty, untamed Amazon River, Abernethy said: "We need a vision, a vision for a way forward, and we call that a learning health system. It needs to be powered by a river of data, our river. And we need a North Star." Her North Star is "Janet," a 37-year-old patient with melanoma.
In informing Janet of the risks of interferon treatment, Abernethy pulled in records from numerous data streams that formed rivulets, which dumped into a large river of data, "our Amazon."
Informaticians – the preferred appellation of the AMIA crowd – face two challenges: exploring this river of data and explaining it to practitioners, administrators, payers and, most of all, patients, according to Abernethy, who is on leave from jobs as director of the Duke University Center for Learning Healthcare and the Duke Cancer Care Research Program.
The vernacular in informatics doesn't align with that in other parts of healthcare, Abernethy said. There needs to be a lingua franca. As healthcare professionals think about trying to bridge the communication gap, people find it safer to keep on their sides and not talk to each other.
By mining data, Abernethy was able to find that sexual distress was common in patients with breast, gastrointestinal and lung cancers. Yet, clinicians tend to sidestep the issue of sexual dysfunction, which could make them miss a more serious condition such as cancer.
"We bridged across that river of data to partner with the psychologists and the clinical trialists," Abernethy recalls from her time at Duke. The university was able to get clinicians to inquire about sexual distress as a standard of care within 18 months, not the 14 years or more it typically takes new evidence to make it from the academic literature to the exam room.
"Its roadmap is applicable across many different treatments," Abernethy said. However, even 18 months is not fast enough.
"We've got to get better" at bringing data to standard clinical practice, Abernethy said. "Together, we call that a learning health system."
To get there, Abernethy made five recommendations:
"First, fundamentally, it's about putting the patient at the center," she said.
Second, she said, have meaningful data by putting the information into the right context. It should be taken as more than just a snapshot in time. Duke had a concept called "research-quality clinical data" so data could be repurposed for research purposes, Abernethy explained. "Over time, data completeness goes up."
Third, improve data quality by putting that data to use. "Let's make sure the data are accurate by using them," Abernethy said. "We need to put data to work for all of us. … We care about its quality, its sanctity its security and that it does the right thing in taking care of patients." After all, patient data often changes over time.
Data also must be trustworthy so users can make good decisions about security, privacy and collaboration. "To get information from patients over time, we need systems of trust," Abernethy said. "Trust is the glue that holds it all together." Abernethy noted that Janet wanted clinicians to use her data. "Far from being the Big Muddy, the more that we use data, the clearer the river of data gets," the Duke oncologist said.
And finally, "Data within the context of this story must be interoperable and liquid," Abernethy said. "Even if Janet's disease comes back, through her data, she lives on." Abernethy noted that the Cancer Biomedical Informatics Grid (CaBIG) failed in part because communication broke down. "It was really difficult to communicate the language and culture of informatics to clinical care at the time."