Since I just wrote a Tech Trend article about the foundations of a learning health system, I thought it was a good time to check in with John Cuddeback, M.D., Ph.D., chief medical informatics officer for the Anceta Collaborative, the clinical analytics subsidiary of the American Medical Group Association (AMGA).
Anceta has grown to almost 40 large medical groups contributing data from the electronic health record systems. Physicians receive performance reports with comparative data. Working along seven disease lines, medical groups share data to identify opportunities for improvement and to recognize best practices. For instance, medical groups can do a query to see which medications their physicians are prescribing for glycemic control in any subgroup of patients with diabetes, and see comparative data for similar patients of other medical groups participating in the Anceta Collaborative Data Warehouse.
Cuddeback described progress on several fronts. First, he said, Optum’s acquisition of Humedica has strengthened the platform Anceta uses, he said. It brought Optum’s capability to handle adjudicated claims. The Optum One model offers a set of apps you can choose on top of the same platform. “Fortunately, Optum encouraged Humedica to build everything on the same platform rather than trying to integrate everything behind the scenes. It actually is very well integrated.” As provider organizations are taking risks on populations, they are getting adjudicated claims for those populations either from CMS or from commercial payers. “There is a new registry module that updates many of the data elements daily, because obviously if you are trying to use this for care coordination, that timeliness is crucial,” he said. “When Humedica started out, having data that was six weeks old was a triumph. But if you are trying to use it for anything involving active outreach to the patients, it has to be daily.”
Anceta has boosted its research focus in collaboration with Optum Labs, Cuddeback said. Optum Labs, based in Cambridge, Mass., now has 22 partners. “We are able to bring together clinical data from Humedica, claims data from United Healthcare and the pharmacy benefit data from OptumRx,” he explained. “About 14 percent of patients who are in the clinical data also have significant periods of coverage by a United plan or by OptumRx, so we have not just what was happening in the EHR, but also prescriptions people were filling.” Anceta has added a PhD and a pharmaceutical epidemiologist. It is building a research team to apply rigorous methods to draw causal inferences from observational data. It is linking people from its large clinical database with Optum Labs’ claims data without revealing the protected health information on either side, he said.
The focus of that research? “We see some variation from group to group on outcomes among medical groups in the collaborative, but what we see much greater variation in are processes of care, like prescribing patterns and the way specialist referrals are used,” Cuddeback explained. “It is kind of good they are getting similar outcomes for their patients, but the fact that it is costing a lot more, for instance, because of significant differences in the use of expensive medications is a problem.” For instance, there is a huge disparity in how many patients with Type 2 diabetes are on newer, expensive medications such as DPP-4 inhibitors and GLP-1 agonists. “Basically everyone is getting about the same improvements in glycemic control, but some people are giving those drugs to fewer than 10 percent of those patients, and some are giving them to about half of their patients,” Cuddeback said, “and you can tell that the people who have worked on protocols for this are the ones using those drugs most selectively. So I think there is a real opportunity where we see variations in process of care but are essentially getting the same outcomes.”
Anceta members also are working on the use of predictive analytics, he said. When you bring all the data together and normalize it to compare apples to apples, you also can create predictive models, he said. “With clinical data you can not just predict a readmission, but predict an initial admission,” he noted. Humedica’s models take advantage of clinical data and are much stronger than claims-based models. The model gives members a list of all active patients with a certain condition such as COPD or heart failure in descending order of likelihood of hospital admission over the next six months.
Cuddeback said 15-hospital Aurora Health Care in Wisconsin has presented its results at Anceta and AMGA meetings. Aurora has focused on the top 20 percent of heart failure patients with this predictive model, and has managed to reduce rate of admissions in that group by 60 percent.
For high-risk patients, Aurora shifts from a fundamentally reactive model of care to a proactive one, using a new role, a health coach RN, assigned to help that patient manage their care. “The proactive approach has to work in the context of a typical clinic,” he said. “It is a substantial challenge to implement and a challenge for others to replicate.”
“The other groups have heard Aurora’s story. They get the point, but there is a lot that has to be done,” Cuddeback said. It involves defining new roles and how they work with everybody else in the system. “Three dozen groups have access to these same predictive models, yet very few of them are doing anything as impressive.” To help other practices replicate Aurora’s efforts, AMGA has hired a director of clinical translation, Jill Powelson, RN, MBA, previously with Baptist Medical Group in Memphis.