At AcademyHealth’s Concordium 2015 meeting Sept 22 in Washington, D.C., three physicians leading accountable care organization (ACO) efforts at academic medical centers described the challenges they face, ranging from build-vs.-buy decisions for technology platforms to gaps in their data to a shortage of data scientists.
William Borden, M.D., is associate professor of medicine and the director of healthcare delivery transformation at the George Washington University. In addition to practicing cardiology, Dr. Borden leads GWU’s ACO and other care improvement efforts.
Now in year two of its ACO efforts, GWU’s data challenges are perhaps elevated because it is in the process of switching EHR vendors from Allscripts to Epic, with a go-live of February 2016. But it already has a wealth of claims data in an Epic data warehouse. But speaking to the “build vs. buy” question, he said so far GWU has chosen to build its analytics tools. “The build process has been beneficial because we can customize to our needs the claims analytics and data warehouse,” Borden said. “We have learned form researcher colleagues who have deep experience working with Medicare claims data.”
He added that for organizations new to the ACO realm, it helps to get your feet wet by figuring things out on your own at first. “We are much better consumers than we were a year ago. We are talking to vendors and know what questions to ask because we better understand our operational needs,” Borden said.
Borden identified two data gaps GWU would like to close. One is around patient-reported information. “We have CAHPS [Consumer Assessment of Healthcare Providers and Systems] data, but it is not actionable. I would like to see us get more patient-reported information. The second gap is around, discrete socioeconomic data. For instance, as GWU does transitions of care and discharge for the most vulnerable populations, there is no easy way to identify patients who struggle with homelessness, he said.
Sreekanth Chaguturu, M.D., is vice president for population health management at Partners HealthCare in Boston, and is responsible for Partners’ ACO efforts. He said that although Partners was a pioneer in developing population health analytics tools, it has chosen to sunset those applications and work with vendor Health Catalyst because it wants to focus its efforts on providing care, not being an IT company. Nevertheless, he added, there are lots of gaps in current offerings and you still need a strong IT shop to customize the systems and make them work.
Chaguturu said a key to success has been to create a governance structure. “One of the questions you have to ask is how to organize yourself to make the most of the assets you have,” he said. Partners created a governance structure that reviews, IT, business and clinical priorities. But once Partners analysts began reporting out data, they had to make adjustments because providers and administrators said they were getting overloaded with data, he said. “For instance, we backed off from weekly reports to middle management to monthly ones and then finally quarterly ones with a 12-page narrative explaining the data,” he said. “We were proud of that, but they said that 12 pages was too long. We have to constantly tweak how we present it to make the data actionable by different users.”
One current challenges he is working through is trying to figure out what data and analytics capabilities to centralize vs. having available locally. “What do you keep at the physician level or hospital level, and where do you keep the predictive modeling capabilities?” he asked. There are limits to data self-service, he added. “We have found that clinicians and middle management have challenges looking at it on their own. We now keep regular office hours so people can come in and ask questions.”
Lindsay Jubelt, M.D., is medical director of population health at Mount Sinai Health Partners in New York and an assistant professor of medicine at the Icahn School of Medicine at Mount Sinai. She started out her presentation by describing saying Mount Sinai used to be a “feed-the-beast” hospital focused on doing lots of procedures in a fee-for-service model. But since 2013 it has been undergoing a complete transformation to becoming a population health center. “It is an exciting time, but we have never done this before. Population health is new, and there is lots of data we haven’t worked with before,” Jubelt said. "Our first approach was to buy IT platforms because we didn’t have the bandwidth to create things. But we have found no external solution can understand us, so we build on top of those tools.” For instance, they built a registry of patients with chronic kidney disease within Epic that can survey lab data and notes for patients at different stages of kidney disease. “But Epic’s capability ended there,” she said. Mount Sinai fed that registry information into their data warehouse to flag patients that need to see a nephrologist or be contacted by a care coordinator..
Jubelt said Mount Sinai learned it had to better integrate IT and clinical teams. Previously IT and data reporting was a completely separate entity. If you wanted a report, you put in a ticket, and got in a queue. “The problem with that is that it is incredibly inefficient. Often the report would not be what the user wanted. We decided to break down those silos and integrate IT folks. The two groups have become one and they can push data out to front line users, which has led to more innovation at the front line.”
Jubelt said the timeliness of data is crucial. Mount Sinai started out following the traditional approach of focusing on the small percentage of patients who account for approximately 20 percent of all costs. The problem with that is that patients don’t stay at one level of risk, she said. “It became a waste of time and energy. We were chasing our own tails. We needed to have more timely data.” Now they have shifted focus a bit. If a patient has had three hospitalizations in the last year and is in the hospital now, that is when that person is more receptive to services, and meeting them in person makes a difference, she said. “We have to figure out what is the real-time data we need to interface with patients.”
When asked about other data challenges, Jubelt said that every payer has different claims data formats. “Understanding how that data is structured and putting it in a data warehouse in a way to make it equal across the platform is frustrating.”
GWU’s Borden said claims data is powerful but often delayed. He described emergency department clinicians and cardiologists looking up data about patients using the Maryland-based CRISP health information exchange as the best type of actionable, real-time data. But both Jubelt and Partners’ Chaguturu said that although they are excited about the potential of using HIE data, so far the HIE data in their areas is limited. “We have not seen real benefits yet,” Jubelt said. “It is not as comprehensive as we would want it to be.”
The providers also mentioned the challenge of presenting data to physicians to get them to change their habits or focus. “When we first started, we had an army of people go to physician practices, explaining the ACO and the data, and the reception was mixed,” Jubelt said. “I would say the impact of that data was not very much. For physicians it was a small percentage of their panel. Second, we were following metrics they didn’t believe in or care about.” She said they continue to talk to practices but seek to shift data presented to what matters most to front-line clinicians.
“We have a long way to go in getting data to front-line clinicians,” Partners’ Chaguturu said. Instead of creating report cards to show people their avoidable readmissions, Partners has sought to create clinical programs to proactively address issues, so they might see system-level improvement as opposed to using metrics at the physician level to drive change.