Technology and care coordination helped a cerebral palsy program reduce healthcare costs per child by 210 percent.
An interdisciplinary team at Nationwide Children’s Hospital, Columbus, Ohio, has developed a “local learning health system,” leveraging data from electronic health records (EHRs) to drive improvements in care while reducing unnecessary care utilization in cerebral palsy patients.
In one year, their work reduced healthcare costs per child from 176 to 210 percent, according to the study, published in the journal Developmental Medicine and Child Neurology.
Leaders tested the idea on a cohort of 131 children with cerebral palsy, who are often high-volume healthcare utilizers. The hospital invested approximately $225,000 in technology and care coordination FTEs during the program’s first year and achieved savings of $1.3 million in reduced healthcare charges—a return of $6 of savings for every $1 invested.
During the first year, the hospital also achieved a 43 percent reduction in total inpatient days, a 27 percent reduction in inpatient admissions, a 30 percent reduction in emergency department visits, and a 29 percent reduction in urgent care visits. In addition, the hospital cut healthcare costs by $6,500 to $7,000 (176 to 210 percent), compared with two different control groups.
“These immediate benefits came from standardizing care based on available evidence and coordinating care,” says William E. Smoyer, MD, vice president and director, Center for Clinical and Translational Research, The Research Institute at Nationwide Children’s Hospital. “The enduring benefits will come from the accumulation of combined clinical and research data that results directly from the routine provision of care for every patient.”
A New Model for Care
Leaders at Nationwide Children’s developed what they call a “Learn from Every Patient” program model to deliver systematic improvements in care. In this model, clinicians use each patient encounter to collect data that is transferred to a data warehouse, where it is integrated with other patient-related data sets, such as proteomics, genomics, pharmacy and radiology data, or patient-reported outcomes. Researchers can then query this large, integrated data set to answer important clinical and research questions to improve patient care, reduce costs, or both. This research generates new knowledge that can be systematically applied to standardize evidence-based care improvements and reduce costs for all subsequent patients, Smoyer says.
“By building a system like this and harvesting data that has been collected as part of the routine documentation of care, you can systematically improve care for all patients,” Smoyer says. “The idea is that if a patient comes back two years from now, you could assure the patient of something that is only rarely possible now—that the patient’s care would be better than it is today because it will be based on peer-reviewed evidence.”
A learning health system requires data collection to be integrated with the standard provision of care. Specifically, informatics teams build highly specific discrete clinical and research data fields directly into the EHR to capture patient data, rather than allowing physicians to make notes using free text. As a result, leaders can capture significantly more data. “That is what allows the systematic learning to occur,” Smoyer says. “Every time clinicians provide care, they are actually generating research-caliber data from which they can learn and then use to improve care in the future.” Smoyer calls this “evidence-generating medicine.”
Here’s how it works: Smoyer challenges clinicians to develop three to five research questions that they believe would clearly improve the care of their patients. For example, one project aimed to determine whether routine hip X-rays were useful in annual screenings of all children with cerebral palsy. The clinical and research informatics teams collaborated to embed the data required to answer those questions in the EHR. Over time, clinicians were able to analyze the data and determine that such X-rays were not necessary for patients with mild cerebral palsy. This allowed them to reduce healthcare costs as well as the children’s radiation exposure, which systematically improved their care.
Similar to this example, during the first 12 months of the pilot, the team collected data for five physician-driven research projects designed to improve routine clinical care. Some clinicians already have had their research published in peer-reviewed literature, offering an additional incentive for clinicians to participate.
Lessons Learned
Smoyer offers advice for piloting a learning health system.
Assemble a coalition of the willing. “Lack of support from a few key stakeholders could keep the entire model from working,” Smoyer says. He recommends recruiting enthusiastic physicians, nurses, and program directors. Other important stakeholders include research informatics and clinical informatics teams, the data warehouse team, and finance leaders, who can direct funds to support needed FTEs.
Host meetings of key stakeholders every other week. Regular meetings can help team members acclimate to change. For example, one of the biggest challenges at Nationwide Children’s was gaining physician buy in to change the way they work with the EHR. Most physicians prefer to write “stories” about patients’ illness histories, as they were taught in medical schools. But instead of using free text to note observations about medical histories, the EHR at Nationwide Children’s cerebral palsy clinic now includes a series of data fields with distinct data choices within the fields. For example, clinicians can select specific values using drop-down boxes. This makes the data far more searchable for research purposes.
Eventually, these meetings helped get physicians on board with the pilot because they realized this approach offered them an opportunity to improve outcomes for their patients. “It’s motivating to almost all physicians if you offer them a realistic chance to improve care, not just provide it,” he says. “In exchange for giving up some autonomy in documenting the medical record, physicians have an opportunity to answer specific clinical and research questions they believe are important to improve the lives of their own patients.”
Invest in an expert to ensure data integrity. In a learning health system, “It is critical to collect a medical record that is both complete and accurate,” Smoyer says. That means no blank data fields and no incorrect data. To achieve these goals, program leaders hired a partial FTE—specifically, a physical therapist who was a content expert in cerebral palsy—to do quality control on the data entry after it was entered.
Focus on care coordination. Nationwide Children’s assigned a care coordinator to each child in the program to help families navigate the system and reduce gaps in care. Care coordinators used EHR-structured notes to create care plans, which were shared with other care team members.
Testing the Model More Broadly
Back in 2007, the Institute of Medicine—now the Health and Medicine Division of the National Academies of Sciences, Engineering, and Medicine—first challenged organizations to create learning healthcare systems that apply evidence for continuous improvement. But progress toward this goal has been slow, in part because behavior change can be difficult for providers, Smoyer says. However, it is not impossible. In fact, such fundamental changes are required to truly reform health care and maximize value, according to a viewpoint article Smoyer co-authored in the Journal of the American Medical Association. Smoyer describes how learning health systems can be developed at the local level to improve quality and reduce costs from the bottom up. The rise of value-based payment models will likely drive payer and provider interest.
Smoyer recently presented the results of the pilot program with cerebral palsy patients as part of a continuing education session sponsored by the Centers for Medicare & Medicaid Services. Based on the early results of his pilot program, he is hopeful that more research will determine if the Learn From Every Patient model can be replicated on a broader scale.
For organizations pursuing local learning health systems in the future, the adoption of IT can help, but even more important is a commitment to change. “Changing the culture of how stakeholders provide care is the only reliable way to systematically improve the value of care,” Smoyer says.
Interviewed for this article:
William E. Smoyer, MD, is vice president and director, Center for Clinical and Translational Research, The Research Institute at Nationwide Children’s Hospital, Columbus, Ohio.
