Ten Reasons Why Learning Health Systems Will Have a Transformational Effect on Health and Health Care

Abstract

The description of each feature, to follow, is accompanied by at least one citation, out of multitudes that are eligible to be referenced. A fully referenced paper goes beyond the scope of this commentary; but in almost every case, the works referenced below point to a more complete set of citations supporting the assertions offered here.

There are many reasons why innovations are adopted, but the process typically begins when the idea behind the innovation attracts the attention of potential adopters [2]. Beginning with the 2007 seminal workshop report from the (then) Institute of Medicine [3] with 574 citations and later with the Friedman et al. 2010 article cited 622 times [4] and Greene et al. cited 430 times [5], the concepts and methods associated with Learning Health Systems have garnered significant attention. Gro wth of the LHS, both as a concept and set of methods, is difficult to quantify, but the two indices presented in Figure 1 offer evidence of steadily increasing attention. The first is the number of retrievals through a PubMed search for articles with the title or abstract containing “Learning Health Systems” or “Learning Healthcare Systems.” (The actual number of relevant publications is likely much higher since the search as conducted retrieved only 52% of the articles published in this journal.) The second is the number of full-text downloads of articles published in Learning Health Systems, which is fully online.

In the authors' own subjective experience with LHS, dating back to 2009, no one has said that LHS is a bad idea. One conjecture about the appeal of the idea goes to its name: very few people oppose “learning” and almost no one opposes “health.” While “system” is a somewhat more controversial concept, when combined with the other two terms, “system” takes on a more positive connotation. Indeed, the challenge and the corresponding opportunity associated with realizing learning and transformation at the system level can attract diverse thinkers who recognize that system problems demand system solutions.

There are many descriptions of “improvement cycles” in LHS; but, to a reasonable approximation, all such models include transitional elements of practice to data, data to knowledge (evidence), and knowledge (evidence) back to practice. While cyclical activity to drive improvement is hardly a new concept, the LHS brings inclusion and co-creation—similarly referred to as co-production—to the level of imperatives. An extensive literature describes the hyper-collaborative nature of Learning Health Systems [11], and a scoping review [12] offers examples of how LHS models incorporate the related concept of co-production. The imperative for inclusion, however, derives primarily from common sense. It is intuitively obvious that including everyone with a stake in a health problem in learning communities that are dedicated to solving that problem will lead to a solution that is more likely to be implementable and effective. This fundamental idea is reflected in several of the LHS core values and commitments as cited above.

An emphasis on infrastructure as a key component of Learning Health Systems appeared early in the life of the concept [13]. A strong emphasis on infrastructure, in the form of a comprehensive set of shared socio-technical services, is a strong differentiator of the LHS from other approaches to health and care improvement [13, 14]. It can be said that infrastructure is what makes the LHS a system as opposed to a disconnected or only loosely connected set of health improvement projects. Consistent with the concept of LHS as a system, LHS infrastructure can be described as a set of 10 socio-technical services as shown in Table 2, each primarily supporting a particular stage in the improvement cycle [15]. Infrastructure lends resilience to the LHS through the provision of services by specialist staff with purposefully redundant skill sets; it lends adaptability through the potential for enhancement of services to benefit all improvement activities; and it lends scalability through the low marginal cost of connecting a new improvement effort to these services. Coordination of the comprehensive services comprising a complete infrastructure also ensures that the work of each service will remain mission-focused and be minimally redundant with other services.

Improvement cycles directed by inclusive communities will collect and analyze data internal to the boundaries of an organization, a network of organizations, or a jurisdictional region. These data and the results of analyzing them will generate evidence that is specific to local circumstances. This internal evidence is then complemented by review and integration of evidence generated by studies conducted elsewhere [16]. It is often the case that local studies are conducted with institutional resources and employ smaller samples, enabling them to be performed more rapidly. External studies are often grant-supported, using larger samples. While external studies often have taken longer to complete, their results, if published, are readily accessible in a world that increasingly offers open access to published knowledge. The powerful combination of rapidly conducted studies reflecting local circumstances and published studies reflecting circumstances existing elsewhere enables learning communities to rapidly generate interventions with confidence that they are basing these interventions on the best evidence available.

The potential for AI methods, if thoughtfully and ethically implemented, to profoundly improve health and care is widely appreciated [17, 18]. At present, however, the rate of creation of AI models, embodying this potential, is far greater than their rate of implementation. Metaphorically, just as it is much easier for an airplane to leave the ground than it is to land on it safely, it is much easier to develop an AI model than to deploy it in a way that truly improves health. Since the LHS improvement cycle embraces discovery as well as implementation, the services provided by a mature LHS infrastructure, as seen in Table 2, provide what is needed to get an AI model “safely on the ground”—that is to say, effectively deployed into health promoting and care delivering environments. While some AI proponents may suggest that advanced AI will provide an alternative to Learning Health Systems, a more realistic scenario is that AI and LHS can mature together and reinforce each other.

The literature on innovation adoption reminds us that innovations that are “trialable” on a small scale are more likely to be successfully adopted [2]. Learning Health Systems can expand through a process best described as “organic infiltration” as opposed to “inoculation,” which requires a substantial up-front commitment of resources. An LHS can take root with a small number of improvement cycles, each initiated by diverse individuals sharing a passion to solve a health problem. At the same time, a primitive infrastructure can be cobbled together with services already available. Successful experience will begin to change the culture. This process will engender bidirectional positive feedback, as shown in Figure 2, whereby a positive change in any one of the three elements will engender positive changes in the other two.

Agreement about what constitutes a Learning Health System remains at a high level of abstraction. Currently, there are at least four distinctive approaches to LHS [19] that would meet, at least as a goal, the original IOM definition. As complex adaptive systems, these approaches will inevitably evolve and possibly branch [20]. While some may clamor for specificity around an LHS, this may prove to be a fruitless pursuit. Current efforts to build capability/maturity models [21, 22] may generate a satisfactory level of specificity about one type or approach to LHS, but there will always be others. Going forward, LHS will become better at learning and applying knowledge; that is, “meta-learning.” By “learning how to learn better” and figuring out how to un-learn and de-implement what isn't working, they will enhance their ability to improve health [23]. This will occur introspectively through self-assessment, through discussions at meetings as professional associations continue to develop LHS affinity groups, and through informal discussions among members of the same organization.

As an LHS matures, as the mechanism illustrated in Figure 2 operates over time, improvement cycles will become more numerous. Focused improvement will change from exceptional to routine activities. At this point, the institution, network, or region will be learning from the experience of every person [24]. The socio-technical infrastructure supporting the improvement cycles will have grown to support routine learning, and the culture of the institution will have achieved consistency with the core values and commitments. The ideas that first captured the imagination of the pioneers who began the journey will continue to capture the imaginations of those who followed.

Taken together, the 10 reasons enumerated above make a compelling “why” for the LHS. Those seeking to apply this model in their own settings can use the reasons as a high-level roadmap to bring organizational leaders to the table and help distinguish the LHS from other approaches for translating knowledge into action. Through LHS approaches, the ephemeral successes and incremental improvements cited by Mate and colleagues can become lasting and transformational.

The authors declare no conflicts of interest.