aamc.org does not support this web browser.

    My First-hand Experience with an Evolving Learning Healthcare SystemMelanie Spencer, PhD, MBA


    My ringing phone startled me out of a pre-coffee, early morning stupor, and I raced across the room to answer it.  Anyone responsible for elderly parents knows that receiving a landline call before 7 a.m. often means there is a crisis.  This call produced particular fear.  The last time my phone rang this early was two weeks before, when my father was frantically and unsuccessfully trying to revive my mother, who had suddenly collapsed.  When I picked up the phone, I could hear the anxiety in Dad’s voice as he described a gastrointestinal bleed that I knew would need emergency treatment and lead to hospitalization.  I did not foresee that this was just the beginning of a long summer of seemingly unrelated hospital stays that, in retrospect, might have been preventable.

    In 2012, The Institute of Medicine published Best Care at Lower Cost:  The Path to Continuously Learning Health Care in America, with a lofty vision to apply the best evidence and practice “to produce high-quality healthcare that continuously learns to be better.”(1) The book’s authors describe a long list of problems in U.S. healthcare, their potential sources and solutions.  Their recommendations are comprehensive and concrete, covering infrastructure needs, goals for improving care and policy changes that would support adoption of the healthcare learning environment and improve care for patients. Unfortunately, progress toward widespread adoption of a Learning Healthcare culture has been limited, not by lack of vision or motivation, but by the enormity of creating consistent change in a fragmented healthcare system. 

    At Carolinas HealthCare System (CHS), we recognized that our goal of delivering consistently exceptional care was challenged by complexity and fragmentation in our own system.  To address these issues, CHS purposefully aligned incentives to focus on care and quality performance metrics, while investing in cross-continuum care coordination.  Yet, to achieve Learning Healthcare that continuously and systematically improves, a healthcare system must also leverage data collected in clinical care to create practical knowledge that can be readily applied. (1)  Fortunately, CHS has consistently made substantial investments in data, technology and analytics that support Learning Healthcare.  Our integrated electronic medical records (EMR) system has been in place since 2008 and includes consolidated patient records across the care continuum, from our network of primary care practices to acute care hospitals to traditional and free-standing emergency facilities to post-acute care facilities and beyond.  These data are stored in an Enterprise Data Warehouse that is updated daily.  It has become a rich resource for research and analysis that supports CHS as a laboratory for Learning Healthcare on a manageable scale.  However, realizing the full value of these capabilities requires a systematic approach to harvest embedded knowledge, identify the best opportunities to improve patient care, rigorously test solutions in a real-world setting and deploy what works at each and every patient encounter.  This type of applied research is explicitly structured to improve care locally and contribute to scientific knowledge broadly by generating valid and generalizable information, while also determining the effects of an intervention in an actual care setting.  Objective data, routinely collected in care, are what is needed to identify study populations and outcomes, allowing studies to be seamlessly woven into day-to-day clinical operations. 

    My father’s care experiences illustrate how difficult this work can be.  Three days after his first hospitalization, Dad was readmitted.  One day into his stay, an unknown hospital employee entered his room and invited him to enroll in an electronic home health monitoring program designed to identify problems early and prevent readmissions.  He was alone and very ill with pneumonia.  Predictably, he refused to participate.  His physician and nurses were not aware of the program and could not answer his questions. In recounting his experience, Dad was proud that he had “put a stop to it” before “they” replaced his much-loved home caregiver with a machine.  His limited understanding and the lack of program integration with his principal care team prevented his informed decision.  Unfortunately, he was readmitted to the hospital three weeks later.

    This example highlights multiple opportunities that were missed because a structured approach was not in place to evaluate and learn.  Could the system have identified through analytics that my father was high-risk after his initial hospitalization due to previous health problems and the recent death of a spouse? Might his second hospitalization have been averted if he received wrap-around services at discharge?  Did someone examine which parts of the system broke down?  Asking and answering all of these questions are fundamental to a Learning Healthcare culture. 

    Well-intentioned quality improvement programs, like this example, are ubiquitous at hospitals across the country.  Unfortunately, they often lack the rigor necessary to support broad, evidence-based change even though the need to understand and measure the effectiveness of healthcare practices has never been more urgent. (2)  Financial pressures to reduce costs and government incentives to improve quality are imposed into a healthcare environment without clear evidence to guide which changes are effective and how they should be implemented to achieve both goals simultaneously.  While standard quality improvement programs can address limited scope problems, this approach is not sufficient to achieve Learning Healthcare.  Because such programs are typically a response to an urgent problem, they are often implemented quickly, with limited planning, narrow focus, haphazard evaluation and scarce internal funds.  They potentially add expense to care delivery and yield no proven linkage to improved outcomes.  These types of projects sit squarely at the intersection of research and quality improvement, which is fertile ground for continuously Learning Healthcare.  As the Institute of Medicine concluded, we can do this differently and better. (1)

    The home health monitoring program that attempted to enroll my father is a good example of quality improvement that had great potential to improve care and lower costs.  The good news is that my father did receive home monitoring after his third hospital visit and has not been readmitted since.  The bad news is that there was no rigorous evaluation. We will never know if his success was because of this program, nor will we understand which patients might benefit, how best to implement informed enrollment, or which program components are most effective.  As a result, the evidence that might have supported how to use home monitoring for other patients in hospitals across the country does not exist.

    It is clear that Learning Healthcare Systems can benefit both patients and institutions.  This fundamental principle has become a driving force for change in CHS research infrastructure development over the last few years.  Transformation has been the result of deliberate efforts to influence a culture that previously considered research slow, theoretical and non-essential to operations.  As a first step, CHS aligned research priorities with institutional and patient priorities to ensure focus on things that mattered.  A key goal was to demonstrate that work supporting large systems initiatives could be better and faster when it used a structured approach that also created evidence to guide implementation decisions.  A second goal was to move from using data for reactive analysis to leveraging data to inform strategic planning and future institutional priorities, such as identifying health disparities in the community. These early efforts resulted in successful projects and external funding that provided traction for further growth.  In 2016, the CHS Center for Outcomes Research and Evaluation (CORE) was officially launched to catalyze and support continued growth and dissemination of applied research across CHS.  CORE is a multidisciplinary team with system knowledge and the diverse expertise required to plan, operationalize and evaluate the applied research that is integral to a Learning Healthcare System.  CORE directly supports the academic mission of CHS and links the system’s quality improvement and strategy work. CORE’s most successful studies are defined by collaborations between the CORE team, physician faculty, quality improvement and system leadership. The studies described below illustrate these points.

    AIRTIGHT (Aiming to Improve Readmissions Through InteGrated Hospital Transitions) (3) is a non-blinded, pragmatic, controlled trial with two parallel groups that aims to evaluate if implementation of recent recommendations for hospital transition programs is effective at reducing 30 day readmissions. An automated risk scoring system randomly allocated referral of 1876 adult patients at high risk for readmission to Usual Care or a new program called Transition Services. The Transition Services program bridges inpatient, outpatient, and home settings, providing patients virtual and in person access to a dedicated multidisciplinary team for 30 days. For this pragmatic study, all outcomes data are retrieved electronically from administrative medical records. After reapplication of inclusion and exclusion criteria at hospital discharge, analyses will follow intention to treat, such that patients will be analyzed based on initial randomized referral group.  Specifically this study design and analysis were crafted to evaluate the real-world effectiveness of implementing a new care delivery paradigm on a population’s outcomes.  Additionally, because the study also aims to inform healthcare system improvement, interim implementation metrics are reported to guide quality improvement. Metrics include patient capture rates and observed versus expected readmission rates for patients participating in the intervention. AIRTIGHT provides an example of a situation where CHS recognized the need to improve 30 day readmissions and invested in a novel intervention, as well as in the time and resources to structure a rigorous evaluation of the effect on outcomes. The study is complete and the publication is under review.

    CHOSEN (Carolinas HealthCare Outpatient Antimicrobial Stewardship Empowerment Network) CHOSEN seeks to address the increasing problem of antimicrobial resistance by establishing evidence-based provider and patient centered standards for appropriate outpatient prescribing, an area with limited information on what is effective for reducing inappropriate prescribing.  This work extends the scope of an existing acute care Antibiotic Stewardship Network to include CHS ambulatory facilities across metropolitan, suburban and rural communities.  The CHOSEN program enhances data access for providers, offers provider- and patient-designed tools to support appropriate prescribing and identifies key drivers to improve outpatient antimicrobial stewardship.  The project has co-Principal Investigators – a physician and a researcher – to ensure integration of research into practice based care.  This project was initiated as a partnership between providers, quality improvement specialists, pharmacists, practice directors and the applied research team.  CHOSEN uses mixed methods research approaches to understand what is important to patients and providers, to assess the design and effectiveness of tools and to evaluate appropriate prescribing outcomes.  This project is underway and has resulted in two oral presentations (4-5) and two publication currently under review.(6-7)  CHOSEN was funded by The Duke Endowment.

    1. Hagland M. 2012. IOM report: 'the path to continuously learning healthcare in America'. Healthc Inform 29:30-33.

    2. Lieu TA, Platt R. 2017. Applied Research and Development in Health Care - Time for a Frameshift. N Engl J Med 376:710-713.

    3. McWilliams A, Roberge J, Moore CG, Ashby A, Rossman W, Murphy S, McCall S, Brown R, Carpenter S, Rissmiller S, Furney S. 2016. Aiming to Improve Readmissions Through InteGrated Hospital Transitions (AIRTIGHT): study protocol for a randomized controlled trial. Trials 17:603.

    4. Davis ME, Davidson L, Spencer M, Schmidt M, Liu T, Taylor YJ, Connor C, Buehler K, Yates TD, Burns R, Rossman W, Handy E, Macon M, Scotton J. Exploring patient awareness and perceptions of the appropriate use of antibiotics: a mixed-methods study. Accepted for Oral Presentation at the International Conference on Communication in Healthcare & the Health Literacy Annual Research Conference, Baltimore, MD, October 8 - 11, 2017.

    5. Yates T, Davis M, Davidson L, Spencer M, Taylor Y, Buehler K, Connor C, Rossman W, Burns R, Handy E, Macon, M. Not a magic pill: Provider perspectives on barriers to antibiotic prescribing in the outpatient setting. Accepted for Oral Presentation at the International Conference on Communication in Healthcare and Health Literacy Annual Research Conference, Baltimore, MD, October 8-11, 2017

    6. Schmidt ML, Spencer MD, Davidson, L. Patient, Provider and Practice Characteristics Associated with Antimicrobial Prescribing Rates in the Ambulatory Care Setting Across a Large Integrated Healthcare System. Infect Control Hosp Epidemiol. (In Revision)

    7. Marion Davis, Tsai-Ling Liu, Yhenneko Taylor, Lisa Davidson, Monica Schmidt, Traci Yates, Janice Scotton, Melanie Spencer.Exploring Patient Awareness and Perceptions of the Appropriate Use of Antibiotics: a Mixed Methods Study. Antibiotics. (In Revision)