Aligning Dementia Care Science With the Urgent Need for Dissemination and Implementation

With more than 6 million individuals living with dementia in the United States, 11 million family caregivers, and a majority of persons with dementia ultimately residing in an assisted living community or nursing home,¹ providing quality dementia care across disease stages and settings is a public health imperative. Family caregivers are called on to provide extensive assistance to relatives with dementia, long-term care staff typically lack sufficient knowledge to most effectively attend to residents’ cognitive and psychosocial challenges, and optimal models of dementia care often involve providers and services that are not effectively incorporated into the health care delivery system.² Thus, there is an immediate need to dramatically change how we care for all people living with dementia and support their caregivers. But how?

Since the passage of the 2011 National Alzheimer’s Project Act (NAPA), the National Institute on Aging’s Alzheimer’s disease and Alzheimer’s disease related dementia (AD/ADRD) research portfolio has grown 8-fold.³ This welcome investment has led to a significant expansion in the national scientific infrastructure to support descriptive and intervention studies to advance the science of dementia care. At the same time, new conceptual models have emerged to guide research and intervention development, most notably the National Institutes of Health (NIH) Stage Model. This 5-stage model aims to promote the availability and widespread uptake of effective interventions through 5 activities: basic/descriptive science (stage 0); feasibility and safety testing and refinement (stage I); efficacy testing (stages II and III); effectiveness testing (stage IV); and dissemination/implementation (stage V).⁴ Within this model, a primary focus of government funding has been on stages 1 to III. A critical gap identified in national research summits and other systematic reviews of the state of the science, however, is the lack of scientific emphasis on and funding support for dissemination and implementation of dementia care interventions.^5–7

Among the more vexing issues facing dementia care scientists and funders is the critical question of whether an intervention is “ready” to proceed to dissemination and implementation. The answers to such a question are potentially provided by systematic reviews or meta-analyses. Multiple recent and comprehensive reviews commissioned by federal agencies (which use criteria applied to pharmacologic trials) suggest that owing to a number of methodological issues there are few interventions with proven efficacy for people living with AD/ADRD or their caregivers.⁸ For those who elevate such findings as definitive, the conclusion that there are few if no interventions ready for dissemination and implementation is understandable.

But are such conclusions warranted? We argue they are not. We make this argument based on the idea that meta-analytic and systematic review techniques typically do not answer the question of whether there exist any individual interventions ready for wider-scale dissemination and implementation. Meta-analyses often rely on the assumption that interventions are able to be collapsed into distinct types within which empirical findings are pooled and compared. Such an approach is reasonable when the interventions under study are easily defined and classified. However, because of intervention complexity, the nature of person-centered care, and various other issues, collapsing dementia care interventions into distinct “types” is difficult. This challenge has led to considerable inconsistency across meta-analyses and systematic reviews as to how individual dementia care interventions are classified (eg, intervention A is labeled as psychoeducational intervention in one meta-analysis and as multicomponent intervention in another), which makes any conclusions as to whether certain types of dementia care interventions are more effective than others opaque, at best.⁹ We see no indication that even more recent, comprehensive reviews have fully avoided this problem, and indeed it is perhaps unavoidable in dementia care.

A somewhat related issue is that meta-analytic and systematic review techniques essentially use a “between subjects” method to answer a question that essentially requires a within-subjects strategy. Traditional meta-analytic methodology fails when a series of studies consistent with the NIH Stage Model establishes the evidence base and dissemination potential of an individual intervention (eg, a pilot study followed by 1 or more randomized controlled trials, followed by translational studies) only to become obscured in systematic reviews’ inclusion or exclusion criteria and “risk of bias” ratings. The practice of systematic reviews does not capture the within-body evidence of a given intervention program that can better inform whether individual interventions are ready for implementation. Indeed, we have seen firsthand this tension when nationally commissioned evidence reviews are presented to the scientific community and individuals voice concern that the review did not appropriately characterize the evidence of an individual intervention program that was established over years of testing and refinement that then led to successful dissemination and implementation.¹⁰

There are numerous key concerns when relying too heavily on systematic reviews to guide dissemination and implementation potential.¹¹ Most notably, because randomized controlled trials are typically conducted in highly controlled circumstances with unrepresentative samples or settings (often to maximize internal validity), recommendations generated by high-quality systematic reviews may lack relevance to public health agencies and other providers because they are not aligned with day-to-day needs and practice considerations (ie, external validity).^12–14 In addition, dementia care interventions, which often incorporate multiple components and may face barriers ranging from human subjects regulations to logistical challenges, may not survive risk of bias assessments and so become excluded from reviews and meta-analyses. Recent barriers include those evidenced in promising randomized controlled trials of an online education and cognitive behavior therapy that had a 67% dropout rate,¹⁵ of the effect of music on agitation that lacked blinding,¹⁶ and of function- and behavior-focused care that indicated recruitment bias.¹⁷ In addition, systematic reviews themselves may introduce error, which may or may not have implications for conclusions of efficacy.^18–20 Owing to all of the issues raised above, relying solely on systematic reviews as a definitive characterization of whether certain dementia care interventions actually “work” or not and are primed for dissemination and implementation is inappropriate.

At the other end of the spectrum, there is emphasis on advancing evidence that is “informed” by intervention studies while allowing the intervention to be adapted so as to be suitable for a given setting. From this vantage point, progression through the NIH Stage Model is less linear and more dynamic, with adaptation and modification viewed as an expected part of the scientific process to facilitate adoption by tailoring the intervention for their particular population and environment. Ultimately, pragmatic trials are designed to evaluate interventions in real-life situations,²¹ but even in these trials the extent of generalizability will always be limited because no one population or environment is identical to another.

How then should the dissemination and implementation of interventions be guided? Here we argue for an adaptive approach and provide an argument for its support. Implementation science expert Enola Proctor addresses the question of implementation readiness in a compelling and direct way. Simply stated, she suggests, “We implement the best evidence we have.”²² Building on that point, we acknowledge the many merits of systematic reviews and pragmatic trials, and the necessity of conducting robust research at early stages of the NIH Stage Model to bolster the evidence supporting dementia care theories, conceptual models, and intervention mechanisms that result in more consistent efficacy and effectiveness. However, we also believe in a more dynamic, circular model of intervention development that does not necessarily progress in linear fashion through the NIH Stage Model in order to proceed to dissemination and implementation; indeed, even the graphic of the NIH Stage Model allows for a circuitous path. Moreover, we approach adaptation as a natural part of intervention development, albeit a nuanced one, and essential to the full development of an intervention. Implementation science offers comprehensive frameworks to guide various levels of adaptation in an intervention,²³ and perhaps most critically offers insights as to those elements of an intervention that are adaptable (eg, “surface” elements that do not alter the core elements of an intervention), and those aspects of an intervention that, if adapted, fundamentally change its nature.²⁴

In successful implementation, efficacy or effectiveness is only 1 component of a more complex formulation. Considered more broadly, an efficacious intervention may be too expensive to implement in the real world, require too much training, depend on a workforce that is scarce or unprepared, involve technologies that are costly or outdated, have overly complex procedures, or not matter to various stakeholders. Alternatively, an intervention with low or moderate efficacy or effectiveness may be valued by caregivers, people with dementia, health care systems, or community-based care organizations for numerous reasons: their current dementia care is suboptimal or harmful; the intervention is easy to use and deemed acceptable, appropriate, and feasible; and training requirements to implement the intervention seamlessly fit everyday operations and workflows. We emphasize that adopting a more contextual-based approach—one that considers factors other than P values in classic efficacy trials, or meeting the risk of bias bar in systematic reviews—to determine which interventions are “ready” for dissemination and implementation can truncate the long and leaky intervention pipeline and ultimately result in true changes in dementia care.¹¹

Based on this point of view, we put forward 3 guiding principles when considering dissemination and implementation readiness of dementia care interventions:

• The evidence base of dementia care interventions requires ongoing development. Until NAPA, support for the design and evaluation of dementia care interventions was limited. An ongoing focus on stage 0–stage III research, as is now readily available through an expanded federal research infrastructure, is welcome and we hope will continue to build the evidence base. National research recommendations to enhance the rigor of dementia care intervention science are available^5,6,25 and are guiding this critical, ongoing work.

• We must incorporate implementation science principles throughout the life cycle of intervention development. Considering intervention development in strictly linear fashion is not conducive to creating dementia care innovations that are scalable, sustainable, and adaptable. There exist various methodological innovations that can facilitate and expedite the dissemination and implementation process and should be integrated throughout the life cycle of intervention development (ie, stage 0 to stage V).¹² These foci include robust stakeholder engagement,²⁶ health equity frameworks, implementation science models including those that can guide adaptation,^23,24 and advanced intervention designs that optimize an intervention for subsequent implementation in different health care settings and racially and ethnically diverse populations.^27,28

• The heterogeneous nature of intervention evidence is to be embraced, not shunned. We argue that a fuller accounting of available research, including translational and observational studies, is needed when taking stock of the evidence supporting dementia care interventions. Limiting scientific and lay discourse to results of systematic reviews and meta-analyses does a disservice to the needs of providers, families, and persons living with dementia and omits important contextual information.^10,14 It calls to mind the proverbial “throwing out the baby with the bathwater.”

Adopting a broader contextual perspective when designing, evaluating, and disseminating or implementing dementia care interventions is necessary to ensure that people living with AD/ADRD, their caregivers, and professional care providers have access to the most promising programs available.²⁹ Indeed, given the considerable investment of time, resources, and people that has occurred to advance dementia care intervention science, creating more effective pathways to dissemination and implementation is not simply a scholarly endeavor but an ethical one.