Dr. Marsolo is an Associate Professor in Population Health Sciences. His research focuses on infrastructure to support the use of EHRs and other real-world data sources in observational and comparative effectiveness research, along with standards and architectures for multi-center learning health systems
Electronic Health Records (EHR) were hailed as ground-breaking, would you say they’ve lived up to the hype, and what do you think will be the next frontier in healthcare data?
If you look at the grand promise and vision, we aren’t where we should be. With a fee-for-service healthcare system, we aren’t paying anyone to collect high-quality patient data, and there is no financial incentive, on anyone’s part, to share it. To illustrate, if a patient has a lab test at Duke, then sees a doctor at UNC, it’s financially beneficial for UNC, or any health system for that matter, to redo the same test. The 21st Century Cures Act, which mandates standards for data sharing, should improve this somewhat, but it also creates the challenge of ensuring the EHR data are compliant with the new standards. This will require that we redesign workflows and record information in structured ways so that it’s tagged appropriately for both patient care and research purposes.
To the second part of your question, right now I think the next frontier is fighting COVID-19, but previously I would have indicated the shift towards value-based care and population health activities, which are being driven by many health plans in both the private and public sector. Healthcare systems have to share a lot of data to keep a patient well, so with value-based care, information is collected once, then follows the patient. Population health activities, including data from Fitbits, Patient-Reported Outcomes, and the explosion of telehealth visits will also feed into the broader picture of patient health. Duke is heavily invested in the value-based care space, and I believe they will continue to lead in this field.
Since I’ve been at Duke, I haven’t done specific research on value-based care, but while at Cincinnati Children’s Hospital, I did something analogous, where we worked with improvement networks to define model care guidelines for children with inflammatory bowel disease. Without any funding from health plans, we determined the data to collect so we could provide the best care, then configured templates so we could capture the data in the EHR for care management and improvement purposes.
Using EHRs for research was supposed to lower healthcare costs, have you seen this in any of your studies?
To answer this question, you would have to design a study to specifically look at costs, and I haven’t done that, mainly because until recently, funding agencies like the Patient-Centered Outcomes Research Institute were prohibited from paying for studies that investigate cost. In some of my previous work, we’ve seen that using EHRs for research and quality improvement activities (e.g., calculating patient outcomes, generating reports for population management, and pre-visit planning) can lower the cost of doing those activities—but that’s not necessarily lowering the cost of care. It can get a little tricky because in theory, if you’re improving outcomes and keeping people out of the hospital, you’re lowering costs, but it depends on whose costs you’re lowering. For example, if you give someone a drug that they weren’t on previously, you’ve just shifted costs.
Concerning patient care, a classic example of cost savings would be in reduced testing. But once again, you come up against incentives—as long as we have fee-for-service healthcare, there is a financial incentive to repeat tests rather than figuring out how to pull values from another system. We’re better in areas where repeated testing harms the patient, for example with x-rays and imaging, but the only way to ever fully change the process is to align financial incentives with desired behaviors, which means paying for value, not services.
Have EHRs helped advance healthcare research?
EHRs were not designed to support research, but I’d say they have evolved to help meet the needs of researchers. Capabilities have been added, such as creating flags for potential trial participants who come into the ED or inpatient ward, and processes have been developed for obtaining consent and delivering questionnaires. With patient information all in one place, it’s easier to design workflows and set study criteria. At the Duke Clinical Research Institute, we’ve worked to create guidance and tools on how to leverage the EHR as part of pragmatic trials, such as ADAPTABLE and PREVENTABLE (e.g., defining alerts, creating screening reports). But there are still challenges to getting it right—EHRs are legal records, and we have to build modules with administrative permissions so only certain staff are given access.
I believe we’ve somewhat turned the corner with recent federal legislation around data sharing and application programming interfaces. FHIR (Fast Healthcare Interoperability Resources) is now codified as the standard, which gives us another opportunity to improve what I see as the “state of the state.” So, with the legislative push on the federal level and the value-based care push on the insurance side, I’m cautiously optimistic things will improve.