Background
Public health interventions are designed to address a range of modifiable risk factors of non-communicable disease; however, they often yield modest improvements in population health [
1‐
4]. Furthermore, the effectiveness of interventions is often reduced as interventions are evaluated in more naturalistic contexts. For example, a systematic review of obesity prevention programmes found that those interventions tested in more real world (‘pragmatic’ trials) contexts did not significantly reduce child body mass index (− 0.09 kg/m
2; 95% CI, − 0.19 to 0.01) while those under taken under more controlled research environments (explanatory trials) did (− 0.21 kg/m
2; 95% CI, − 0.35 to − 0.08) [
5]. Similarly, a meta-analysis of childcare-based physical activity intervention reported significant effects for trials evaluated under research conditions (SMD 0.80; 95% CI, 0.12 to 1.48) but not more real world environments (SMD 0.10; 95% CI, − 0.13 to 0.33) [
4].
A number of factors have been suggested to contribute to the disappointing impact of many non-communicable disease interventions, particularly those evaluated in more naturalistic environments, including differences in the characteristics of participants and the availability of expertise and resources between efficacy research and evaluations undertaken in community contexts [
4,
6]. Suboptimal implementation of interventions, however, has been frequently identified as a fundamental contributor to their variable and sometimes limited effect [
7‐
9]. Implementation strategies are methods or techniques used to enhance the adoption, implementation and sustainability of an intervention [
10]. They may include strategies such as educational meetings, audit and feedback, local technical assistance, or building coalitions [
11]. However, reviews of the effects of such strategies indicate that, to date, they typically result in only small improvements in the fidelity of intervention implementation [
10,
12‐
15]. Such findings have been consistent across clinical and community settings for a variety of public health and clinical conditions [
12‐
15]. It is perhaps unsurprising then, that interventions of modest effectiveness, delivered in real-world contexts using strategies with modest impact on implementation may fail to achieve intended improvements in public health.
A further complicating factor to the translation of public health research evidence into community health improvement is that many tested public health interventions, and strategies to implement them, may not be suitable for widespread application in usual service delivery contexts [
6]. As a result, adaptations are often made to ensure interventions and implementation strategies are suitable for the characteristics of the local population and can be delivered within the existing skills, resources and infrastructure of provider organisations [
16]. While the process of ‘adaptation’ has been variously defined in the literature, broadly, it is understood to involve modifications to the intervention or to the approaches for their implementation to improve ‘fit’ with local contexts and capacity [
17]. Reviews of the impact of adaptations, however, suggest that they can have a beneficial or detrimental effect on the impact of health initiatives [
18].
While the purpose of adaptation is to improve ‘fit’, processes of repeated, purposeful modification (or adaptation), routinely occur in other fields, such as engineering and information technology, for the purpose of ‘optimising’ the performance of products through the accumulation of incremental improvements. Similar concepts are implicit in continuous quality improvement approaches in medical care [
19,
20]. Ongoing, purposeful adaptations to interventions or implementation strategies may similarly represent a promising approach to ‘optimise’ the potential impact of public health interventions in achieving public health objectives. Such an approach may be particularly beneficial when undertaken in the context where the intervention is to be implemented and by, or in partnership with, the agency responsible for its delivery (and other end-users). It is also consistent with recommendations that health services generate and use data for service improvement [
21].
There are a number of recent examples of systematic and iterative approaches to optimising the effectiveness of public health interventions and their implementation. The multi-phase optimisation strategy is a process recommended for developing and evaluating e-Health interventions through identifying and refining active intervention components and their dose prior to undertaking a confirmatory randomised trial [
22]; it has been applied to a variety of public health issues, including obesity, smoking cessation and HIV to maximise the effects of these interventions [
22‐
24]. Similar to the focus of quality improvement and continuous quality improvement methods in medicine [
19,
20], other processes of iterative, data-informed modifications in public health have targeted the enhancement of the impact of implementation strategies. For example, sequential randomised evaluations of three strategies to implement school nutrition policies improved the incremental cost effectiveness ratios for implementation of the policy in schools (versus usual care) from $4730 to $2627 and facilitated its subsequent implementation ‘at scale’ [
25,
26]. While such examples exist, optimisation processes applied across phases of intervention development to large-scale delivery appear uncommon. In Canada, for example, government and private foundations have funded thousands of public health pilot projects that are rarely further developed, improved and integrated into public health services – an outcome described by a former health minister as a tragic ‘waste of time, talent and energy’ [
27]. Further, while evidence may be used to inform selection of public health interventions, the effectiveness of approaches to their implementation or their effects on community health outcomes once adopted as a health service are rarely evaluated, precluding the opportunity for ongoing, evidence-based evolution of the programme [
28].
While examples of approaches to iteratively enhance the impact of a public health initiative exist, and employ similar methods, they do not appear to be bound by a unifying, clearly defined concept. Work in the area has also tended to focus either on improving the impact of an intervention or its implementation strategy. Both, however, represent important determinants of public health impact. To progress optimisation in public health, a standardised terminology is needed to provide clarity of concepts and facilitate communication and shared understanding among those working in interdisciplinary fields. It can also help avoid definitional issues that often plague emerging disciplines [
29]. Defining the concept and key parameters of optimisation as it applies to public health, therefore, will provide a basis for subsequent work to develop conceptual understanding, methodologies, techniques, measures and practical guidance to advance the science of practice of optimisation for public health improvement. In in the context of public health we aimed to (1) generate a consensus-based definition of optimisation and (2) describe key considerations for optimisation.
Discussion
To our knowledge, for the first time, this study provides a consensus-based definition of optimisation in the context of public health. It did so by employing the expertise of a group of international researchers, public health policy-makers and practitioners representing leading organisations across a range of disciplines. The key elements of the final consensus-based definition of optimisation were it being a process that was data driven, iterative, targeting an impact that is stakeholder defined and conducted in the context of finite health resources. Such elements align well with the evidence-based medicine paradigm [
37], suggesting that the process is consistent with the underlying values of the field and may represent a promising approach of improving the health and welling of the community. Importantly, the study also explored seminal issues related to the application of optimisation in public health, including whether, when and how such processes should be undertaken. In doing so, the study provides greater conceptual clarity and a broad base for further work in the field.
A number of aspects of the definition are similar to optimisation processes in other fields [
38‐
41], in particular its iterative and data-driven nature. There are also parallels to related concepts such as quality improvement cycles and other improvement frameworks in healthcare [
19,
20]. The practice of optimisation is also not new in public health. There are a number of examples in public health of processes that would be consistent with the definition of optimisation proposed in this study [
22‐
26]; however, these have typically focussed on approaches to improve the effectiveness of interventions during the intervention development stage or approaches to improve the effectiveness of strategies to improve programme implementation. A definition encompassing a range of stages of the translation continuum from intervention development to large-scale implementation appears unique in the context of public health and may provide a unifying concept for current work in the area. The explicit role of stakeholders in defining optimisation impacts may also be a distinguishing feature of optimisation relative to other related concepts in the field [
42].
Participants identified a number of challenges to optimisation in public health that need to be considered prior to embarking on optimising, including the availability of good quality data to optimise implementation strategies for existing evidence-based interventions and the stability of funding to enable optimisation to occur over long periods. As such, there appears most opportunity for optimisation when the outcomes assessed can make use of routinely collected data sets such as administrative records, clinical records, public health surveillance systems or information technology. For optimisation processes to flourish in public health, novel methods of data capture or identifying sources of routinely collected robust outcome data will likely be required. Optimisation is also difficult if the underlying core components and mechanisms of a programme are not known or made explicit. In a field such as implementation science, there remains very little empirical evidence to support an understanding of implementation processes and impacts [
43,
44]. Advances in mechanistic evaluation of implementation strategies will improve the viability of optimisation processes applied to implementation strategies in public health.
Nonetheless, the findings of this study suggest that the broad application of optimisation processes in public health is likely to represent a considerable challenge. As well as the practical considerations identified by participants, including access to routinely collected data, the public health workforce may require significant capacity-building or processes to engage those with expertise in health economics, research trial methods, mechanistic programme evaluation, adaptive interventions and research designs. Examples of where optimisation has been applied to improve the impacts of public health initiatives have typically been in the context where such expertise is available and has been applied [
26,
45]. As such, strategies to support partnerships between researchers and public health policy-makers and practitioners, including embedding of researchers in public health service agencies, may represent one means of enhancing expertise, capacity and infrastructure to facilitate optimisation. Furthermore, public health decision-making is influenced by a range of social, political and organisational factors, of which research evidence is one [
46]. Optimisation, particularly of public health policy, may be difficult to achieve in the context of these other considerations, which may favour policy stability (rather than change), the introduction of ‘new’ programmes (rather than optimisation of existing programmes) or investments in public health programmes that are short term. While the challenges are considerable, optimisation processes offer enormous potential to efficiently and expediently improve the impact of public health initiatives.
There are also some methodological aspects of the study that warrant consideration. The modified three-round Delphi approach with a highly interactive face-to-face component [
47] was found appropriate to address the study aims. The day-long workshop was found particularly useful in engaging multiple stakeholders. The workshop also enabled multiple qualitative techniques to be applied, which was instrumental in eliciting participants’ opinions and gathering rich qualitative data that reflected both the individual contributions and the opinions that were formulated via group processes. We suggest that a traditional Delphi survey method would not have produced the highly nuanced data we were able to collect, or the type of evidence to question and expand on existing definitions of optimisation.
Participants were purposefully sampled to provide diverse expertise and broad representation of relevant public health professional associations, using existing networks of the research team to identify individuals that were well positioned to provide input into the research. It is possible that some participants may not have felt able to express their views freely if they had an existing professional association with a member of the research team. However, the extent to which this may have occurred and any bias it may have introduced are unclear. Nonetheless, participants arrived at a consensus definition of optimisation. A further limitation of the study was that it explored a number of key issues and concepts but, in many instances, this occurred at a high level. Furthermore, several discussions, such as methodological considerations in defining the outcome of optimisation and the levels at which optimisation in public health may occur (micro, meso and macro), were initiated but not well developed. These emerging topics were relevant to study participants and may warrant further investigation.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.