The foundational science of learning health systems

Abstract

The National Academy of Medicine's Future of Health Services Research report¹ called for researchers to better address the system-level priorities facing health-care organizations, patients, and communities to achieve meaningful impacts on population health outcomes. The report strongly recommended that future research initiatives support close partnerships with health-care organization leaders and the end-users they serve (e.g., patients, providers, and communities) to ensure their priorities are addressed. When aligned with health-care organization and end-user priorities, health services research has great potential to facilitate the translation of research into practice that supports population health goals (e.g., access, outcomes, equity, experience, and value) in health-care organizations and communities.

Moreover, U.S. health-care organizations are increasingly embracing learning health systems^2-4 and the Quintuple Aim framework (improve access, outcomes, equity, experience, and value of care) to better align research with health-care organization goals.^{5, 6} The National Academy of Medicine defined learning health systems (LHSs) as when “science, informatics, incentives, and culture are aligned to support continuous improvement, innovation, and seamlessly embed knowledge and best practices into research and care delivery.²” An LHS is led by a learning community comprised of health-care organization leaders, investigators, end-users, and other affected groups who identify shared health priority goals and support continuous, standardized processes that curate multi-level data for ongoing research and quality improvement initiatives that support meeting these goals.

Funding priorities therefore need to be aligned with these NAM recommendations to incentivize investigators to conduct partnered LHS research focused on Quintuple Aim goals. Nonetheless, LHSs are impeded by a lack of funding sources to enable their development and investigator involvement.⁷ Differences in research and quality improvement timelines and regulations also impede their ability to sustain research and quality improvement initiatives over time.⁸ LHSs require upfront administrative and data analytics infrastructure to enable involved parties to set priorities, plan and conduct the research and quality improvement initiatives, curate and maintain primary and secondary data over time, and ensure compliance with research and quality improvement regulations.⁹ While investigators can garner external research funding, this typically supports hypothesis-driven scientific inquiry but not the required infrastructure to enable ongoing administration and data curation to optimize LHS functions. Moreover, research and quality improvement efforts are siloed due to different regulations regarding data curation and intervention monitoring (e.g., research clinical trials require ongoing monitoring and pre-specification of interventions and analyses). Research can take years to achieve results whereas health-care organization leaders need answers more quickly, impeding opportunities to directly inform improvements in health system goals.⁸

To better align national research funding with population health goals, an updated scientific agenda is needed to support innovations in the development and sustainment of LHSs. Foundational research that optimizes the LHS infrastructure and capacity enables health-care organizations to better align research with quality improvement goals to improve Quintuple Aim goals. Current LHS-focused funding mechanisms focus on training opportunities in LHS core competencies for early-career investigators,^10-12 such as in implementation science, informatics, systems science, and end-user engagement. Yet for health-care organizations to fully take advantage of trainees' new knowledge, new funding mechanisms are needed that are grounded in scientific inquiry but also can be directly applied to improve LHS infrastructure, ultimately sustaining a repeatable process of generating new research and quality improvement initiatives to improve population health.

In light of the NAM Future of Health Services Research and Learning Health System reports, we propose an updated scientific agenda for funding agencies to support innovation in the development and sustainment of learning health systems, to align research with health-care organizations and Quintuple Aim goals. This scientific agenda encompasses LHS foundational methods derived from the Agency for Healthcare Research and Quality LHS core competencies^{10, 13} and funding mechanisms and priorities based on the U.S. Department of Veterans Affairs (VA) Office of Research and Development (ORD) Health Systems Research program.^14-17 LHS scientific agenda foundational methods include implementation science, data science, engagement science, systems science, and policy analysis. Each method has a research and quality improvement goal, for example, research validating implementation strategies informs how providers directly implement treatments in routine practice, thus enabling LHS research innovations to be applied directly to quality improvement in health-care organizations.

ORD's Health Systems Research program is applying the LHS scientific agenda¹⁷ to better align the research priorities of investigators with the needs of VA clinical operations partners to ensure substantial real-world impacts on Quintuple Aim outcomes for Veterans. As an embedded research program within the VA's national health-care system, ORD funds VA-employed investigators to conduct research across the translational spectrum on Veteran-focused health priorities. Specifically, ORD's Health Systems Research program funds investigators to conduct innovative research on the overall organization, financing, and delivery of care in the VA health-care system to improve Veteran Quintuple Aim goals (access, outcomes-quality/safety, equity, experience, and value) using LHS foundational methods (implementation science, data science, engagement science, systems science, and policy science).^{16, 17} Health Systems Research also supports the Quality Enhancement Research Initiative (QUERI), which is funded separately through VA clinical services, enabling VA investigators to conduct quality improvement projects and curate data more quickly in the VA health-care system.^{14, 15} Hence, ORD provides a unique opportunity to align separate funding sources to support research and quality improvement efforts to achieve common Quintuple Aim goals using LHS methods.

Table 1 describes the foundational methods that encapsulate the LHS scientific agenda, with examples of research and quality improvement initiatives for implementation science,¹⁸ data science,¹⁹ engagement science,²⁰ systems science,^21-23 and policy science.^{24, 25} The examples provided are cross-cutting rather than disease- or condition-specific per recommendations from the NAM Future of Health Services Research report to promote scientific advancements that impact the health system as a whole.

Figure 1 describes how these foundational methods support LHS infrastructure and processes needed to achieve the health system and Quintuple Aim goals. These methods can inform innovations in infrastructure, and, through repeatable cycles of research and quality improvement initiatives, achieve health impacts for a given priority. The figure is based on ORD's application of the Learning Health System Learning Cycle^{4, 9, 15} in research and quality improvement funding mechanisms.¹⁴ LHS foundational methods inform the infrastructure through the engagement of the learning community to set priorities, guiding the LHS data curation process using data science/informatics innovations, and by studying and applying systems, implementation, and policy strategies that promote the discovery and uptake of evidence-based interventions into practice.

The LHS cycle begins with the formation of a learning community, comprised of health-care organization leaders, investigators, and end-user representatives (e.g., patients, providers, and community members), who agree on a shared priority goal. Examples of priority goals that have been studied in ORD using LHS methods include suicide prevention, opioid use/pain treatment, access to virtual care, and precision oncology.^{16, 26, 27} Initial work starts with the performance-to-data (P2D) phase where the learning community commissions investigators to work with health-care organization leaders and end-users to develop a population-based cohort using existing data to ascertain gaps in Quintuple Aim measures. In the data-to-knowledge (D2K) phase, investigators identify determinants impacting gaps in Quintuple Aim measures, incorporating external evidence (e.g., systematic reviews) and generating new evidence (e.g., new clinical trials, longitudinal studies, or informatics interventions). This phase involves the collection of new data that are added to the population-based cohort for future research and quality improvement initiatives. In the third phase, knowledge-to-performance (K2P) involves the application of foundational LHS methods such as implementation strategies to scale up and evaluate evidence-based interventions and determine their overall population health impact, often yielding new information on Quintuple Aim gaps for the cycle to be repeated.

Learning cycles can be short-term to achieve a health-care organizational goal, or longer-term when there is a need for a more definitive clinical trial. An example of a short-term (6 months) VA-based Learning Cycle²⁸ included assessment of gaps in vaccination rates, mixed-methods data collection to identify individual perspectives regarding vaccine preferences, and implementation and evaluation of communication strategies to encourage vaccination among VA patients. Another example involved ascertainment of gaps in pediatric surgical outcomes using electronic health record data (P2D), assessment of effective strategies that reduce complications (e.g., with wound closures), and applying those strategies in routine practice (K2P).²⁹ Longer-term cycles typically involve research protocols and include, for example, the detection of a health-care disparity, understanding the origins of the disparity, and testing an intervention mitigating the disparities in a given practice setting.³⁰ Assessment of outcomes from the intervention study may also discover new observed gaps in quality and equity among subpopulations that in turn inform a repeated cycle of investigation and improvement.

Ultimately, the Learning Cycle enables health-care organizations to identify, rigorously study, and apply effective tools or strategies that are derived from LHS foundational methods to improve Quintuple Aim goals. Repeatable cycles of research and quality improvement, in turn, can also inform innovations in the LHS foundational methods. For example, foundational research in implementation and engagement science informs how health-care leaders and investigators can better cultivate and incorporate end-user perspectives into the co-design of interventions and implementation strategies. Investigators who design a new patient engagement process for eliciting input on clinical interventions can also apply a similar process to identify new priorities for the next iteration of the Learning Cycle. Data and systems science methods help optimize clinical workflows to enable more efficient study of new interventions or quality improvement practices (e.g., patient-level precision health data to inform clinical trials eligibility and ambient dictation into the electronic health records to mitigate provider workload). Innovations in the data, systems, and implementation sciences can improve the automation of quality improvement processes such as audit and feedback or process mapping. Policy science can inform how new health-care payment models or regulatory relief can improve the quality of care, strengthening health-care organization leadership support for ongoing research intervention, or quality improvement initiatives.

The following are two examples of U.S. health-care organizations that have implemented LHSs. Both have lessons for building and sustaining LHSs using different funding sources and enterprise-wide infrastructure using LHS foundational methods.

Vanderbilt University Medical Center leveraged funding from its health-care organization and the NIH's National Center for Accelerating Translational Science Clinical and Translational Science Award Program to fund the LHS Platform.³¹ Using LHS data science methods, the Platform uses electronic health record data to deploy pragmatic trials of new medications and other interventions such as community-based care coordination for opioid use disorder,³² or implementation strategies to improve clinician prescribing.³³ The LHS platform also uses engagement and systems science, where end-users select priority topic areas (P2D) and help design the research and ensure outcomes and methods fit with clinical workflows (D2K). Investigators and clinical partners also apply point of care randomization in the overall study execution to maximize rigor (K2P).

In VA, the Women's Health Research Network (WHRN) involves over 75 clinical practice sites that support research and quality improvement to enhance the delivery of evidence-based care tailored to women Veterans.³⁴ Using engagement science, WHRN gathers rapid Veteran feedback on gaps in access and quality of care to inform shared research and quality improvement goals among VA leaders and investigators (P2D). WHRN also supports investigators in developing new research interventions (D2K) and implementation strategies to disseminate research findings into practice (K2P).³⁵ While the WHRN supports the infrastructure needed for multi-site research initiatives, investigators also are involved in the QUERI-funded Enhancing Mental and Physical Health of Women through Engagement and Retention (EMPOWER) program, which supports ongoing data curation and quality improvement initiatives addressing WHRN priorities.³⁶

The LHS scientific agenda articulates the core foundational methods that enable investigators, health system leaders, and end-users to work together to study and apply tools that improve population health for a given health-care organization priority. These foundational methods along with the Learning Cycle also provide funding agencies with an LHS research roadmap for how discovery simultaneously supports innovation, health system improvements, and Quintuple Aim goals. By identifying performance gaps in Quintuple Aim goals using existing data, LHSs enable health-care organizations and investigators to curate more comprehensive data and knowledge (evidence) supporting interventions that address these gaps, and then scale-up to improve overall performance and population health goals.

Ideally, well-funded LHSs support investigators and clinical operations teams to conduct research and quality improvement projects focused on a shared health priority goal, facilitating extraction and analyses of data from electronic health records and other sources (e.g., health information exchanges and mobile technologies). To this end, LHSs leverage both internal health-care organizations and external (e.g., federal and private foundation) sources to support the learning community to gather input and priorities from affected users as well as the personnel to support the infrastructure to manage and sustain the LHS over time.⁷ Internal funding sources are essential to ensure leadership buy-in, notably through administrative support of learning communities and for seamless data curation for the benefit of health systems operations and research. Joint funding enables adequate support for a stable group of LHS personnel including statisticians, informaticists, clinical trialists, and mixed-methods experts, who generate the data, work with investigators to help design research and quality improvement study protocols, conduct analyses, and work with the learning community to disseminate/publish results. Well-funded LHSs also support system-level studies including pragmatic trials that use outcomes from electronic health record data,^31-33 as well as develop and evaluate novel clinical decision support or artificial intelligence tools. With population-based data, LHSs can also measure Quintuple Aim goals efficiently and with sufficient numbers across subpopulations.

Overall, there is an urgent need for funding agencies to invest in research that enhances discovery and innovation in these foundational methods so that LHSs can be more effective and efficient in improving Quintuple Aim goals and overall population health. Both research funders (e.g., VA, AHRQ, and NIH) and health-care organizations need to co-invest in the infrastructure and processes that enable LHS functioning to ensure innovation and discovery of new tools, methods, and processes so that patients can benefit from the most up-to-date treatments, technologies, and information in a continuous, equitable fashion.³⁷ More stable sources of funding for LHS foundational methods also can incentivize investigators to partner with health-care organization leaders to achieve population health impacts on shared priority goals, thereby addressing the persistent barriers to translation to practice. Health-care organizations in turn benefit from the application of well-established LHS methods to ultimately improve Quintuple Aim goals for those they serve.

AMK drafted the manuscript and provided content on key aspects of the framework; AEB, RWO, MZB, and MMG edited the manuscript and provided content on relevant research and operations programs; and AEB and MZB made edits and wrote core components of the sections. All authors reviewed and approved the manuscript.

No conflicts of interest.