chris tignanelli

Improving COVID-19 Care in Real Time

How a UMN Learning Health System project developed evolving, evidence-based guidelines to improve hospitalized COVID-19 patient outcomes — and what it means for the future of care.

By Mona Rath | November 22, 2021

A man is admitted to the hospital with symptoms of COVID-19. His attending physician interprets his lab work and decides on a treatment plan. The doctor has been busy caring for patients for months, and isn’t aware of new research suggesting that this particular patient (based on his body mass index and blood levels of a protein formed by dissolving blood clots) will have better odds of staying out of the ICU with a stronger dose of blood thinning medication.

But at just the right moment, an alert appears in the patient’s electronic health record, notifying the doctor of the revised recommendation: “Reminder: Patient meets criteria to receive a stronger dose of anticoagulation.” The doctor makes the adjustment, the patient receives the recommended medication at the right dose, and he avoids escalating to intensive care.

Versions of this scene played out at hospitals across the M Health Fairview health system in Minnesota during the ongoing COVID-19 pandemic, thanks to a University of Minnesota (UMN) Learning Health System project. A team from the University worked with the Agency for Healthcare Research and Quality (AHRQ) Evidence-based Care Transformation Support (ACTS) initiative — a national, collaborative group of implementation scientists, informaticists, clinicians, and healthcare delivery systems — to develop a clinical consensus anticoagulation guideline for COVID-19 patients.

The Institute of Medicine defines clinical practice guidelines (CPGs) as “systematically developed statements to assist practitioner and patient decisions about appropriate healthcare for specific clinical circumstances.”[1] CPGs offer instructions on which tests to order, which services or medications to provide, duration of hospitalization, or other details of clinical practice.

Researchers at UMN and the University of Michigan had previously published work showing that critically ill patients with H1N1 influenza often develop blood clots and have improved outcomes if they’re treated with anticoagulation medication. When early reports from China in the spring of 2020 indicated that COVID-19 patients were developing blood clots, the UMN team approached ACTS with a proposal to develop an anticoagulation CPG to improve care for COVID-19 patients.

The proposed guideline would be developed in the context of a “learning health system.” Dr. Christopher Tignanelli is a Minnesota Learning Health System Mentored Career Program (MN-LHS) scholar and an acute care surgeon who worked on the ACTS LHS project. He explains that learning health systems embed knowledge generation processes and researchers in daily practice to produce continual improvement in care. They also increase the rapid discovery and translation of research into practice.

After a period of collaboration among academic U.S. hematologists facilitated by ACTS and UMN, the anticoagulation CPG was implemented on May 14, 2020 at Bethesda Hospital in St.

Paul, Minn. Ten days later, it was scaled across M Health Fairview’s 12 Minnesota hospitals. The clinical decision support system includes passive cognitive support for physicians, as well as interruptive decision support (pop-up alerts and notifications) to guide anticoagulation care.

A study of the CPG implementation was published in September 2021 in JMIR Medical Informatics. It looked at 2,503 patients and found that those who received care aligned with the new guideline had a 61 percent lower risk of progressing to severe disease (defined as transitioning to the ICU) within 48 hours, and a 47 percent lower risk of developing severe COVID-19 disease at all. Their risk of death was reduced by 33 percent, and they observed a 61 percent reduction in bleeding complications.

Learning health systems are a “living process”

Clinical practice guidelines exist for many conditions, but COVID-19 presented unique challenges in both its novelty and the rate at which new evidence on effective treatment was emerging. “Papers were coming out at a very rapid rate for COVID-19 — every week or so a new paper would come out,” says Tignanelli. “So you had to have some process to keep up with the anticoagulation literature and make sure the guideline was staying updated to the latest evidence.”

The project team partnered with the UMN School of Public Health AHRQ-funded Evidence-based Practice Center (EPC) to develop a review process. (UMN has one of nine national federally funded EPCs, dedicated to reviewing all relevant scientific literature on a wide spectrum of clinical and health services topics to produce evidence reports.) To develop and maintain the anticoagulation CPG, a team of UMN EPC medical librarians, physicians, biostatisticians, informaticists, and others regularly review the literature to see if the evidence on anticoagulation practices has changed. They create evidence summary tables, hosted on a webpage developed in collaboration with the UMN School of Design and shared with the guideline committee at M Health Fairview. The committee then decides whether to change the guideline based on the new evidence.

“So far, we’ve made four modifications to the guideline over the past year and a half,” says Tignanelli. “It’s being reviewed again as we speak, to determine if it should undergo a fifth modification.”

It wasn’t just external research that informed the evolving guideline. The LHS project team also tracked how it affected outcomes in the hospitals where it was implemented, and modified recommendations based on ongoing analysis. “Taking data and converting it into practice is the first step in the learning health system,” Tignanelli says. “The next piece is taking that practice and converting it back into data. Is the guideline working? Are there subgroups of patients it doesn’t work for who are still getting blood clots?”

Indeed, the LHS team was able to identify early on that the guideline didn’t seem to be performing well for ICU patients. One of the first modifications they and the M Health Fairview committee made to the guideline was to adjust the dosage for patients in the ICU.

In a typical research process, one might ask a question, answer it, and implement a guideline into practice. And that’s where it ends. “A lot of health systems don’t have the resources to continuously evaluate implemented guidelines, or they might evaluate a single metric, such as whether it’s reducing mortality or whether people are following it,” Tignanelli says.

Learning health systems offer a different model. The three LHS steps — evaluating current data to convert it into knowledge (known as data to knowledge: D2K); converting the knowledge into practice (knowledge to practice: K2P); and finally evaluating the practice to identify room for improvement (practice to data: P2D) — represent a continuous iterative framework for maintaining best practice, especially in the face of rapidly changing medical evidence. “It’s a cycle that goes around this circle, and keeps updating itself,” Tignanelli says. The learnings are also shared with the broader scientific community through publication, adding to the shared knowledge base for all.

Trial design within the LHS offers a more pragmatic approach than typical clinical trials, which tend to be tightly controlled, with rigorous enrollment criteria and results that don’t always translate to real clinical practice. “In pragmatic trials, we enroll people in real, actual practice and see if a certain intervention improves care or not,” Tignanelli says. “It’s much more agile.” The LHS model also enables an ongoing synthesis of evidence, yielding an evolving knowledge base that can improve real-world practice and outcomes in an iterative way.

The leading edge of possibility

Ultimately, the vision is to enable the creation of CPGs that can be downloaded — and automatically updated — like apps on your smartphone. An evidence-based medicine unit, medical society, or institution could develop, analyze, and maintain a guideline for, say, c. difficile infection. And hospitals across the country could simply download it, enable automatic updates, and be assured they’re acting on the most up-to-date, evidence-based best practices for that condition. Given the limited resources of many hospitals, many of which don’t have the informatics, IT builders, researchers, and subject matter experts needed to do this work internally, this stands to be a game-changer in improving care nationally.

Much work is needed to make that vision a reality. An important next step is ensuring that CPGs can be accessed and implemented across all electronic health record (EHR) platforms. For instance, the LHS ACTS anticoagulation guideline was developed to run on the EHR platform used by the M Health Fairview system. But other health systems use different platforms. The team is now collaborating with the COVID-19 Digital Guideline Working Group (C19 DGWG) and a health IT vendor to make the guideline interoperable, and therefore accessible to more providers. “It’s as if we did everything on Android, for example, and now we need to make it work across Apple and Microsoft as well,” Tignanelli says.

Already, the LHS ACTS anticoagulation guideline project, which was recently presented by Dr. Tignanelli at the American Medical Informatics Association 2021 Annual Symposium, is being viewed nationally as a leading edge of what’s possible. “We did the whole process of evaluating evidence for a guideline, implementing it, tracking it, updating it as the evidence changed, and maintaining it for years,” Tignanelli says. “Now the next step is to expand what Minnesota did, but with four or five health systems and in an interoperable fashion. And hopefully scaling the project and being a model nationally for others.”

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1114973/

© 2015 Regents of the University of Minnesota. All rights reserved. The University of Minnesota is an equal opportunity educator and employer. Privacy Statement