A resilience view on health system resilience: a scoping review of empirical studies and reviews

The past 10 years have challenged health systems globally with several public health emergencies (COVID-19, SARS, Ebola Virus Disease), environmental disasters (flooding, hurricanes and earthquakes), as well as war and conflict shocks. Although health providers often deal with crises, these large-scale emergencies have revealed a need to deal with unprecedented care demand and shortages of staff, supplies and infrastructure. Beyond the direct effects and disruptions of each crisis, there are cascading effects such as delays in routine and non-emergency care. When these crises start to spread across political borders, there are governance, financial and cultural barriers that affect the ability to provide care equitably to all those in need.

These sudden shocks and disruptions have prompted policymakers, practitioners, and scholars to learn how to increase the resilience of health systems. In recent research, perhaps prompted by COVID-19, there is has been a surge of interest in the concept of resilience, and many reviews have already been conducted on this topic, exploring the presence of resilience in empirical studies, or whether we can learn from the concepts of resilience generally to improve health system performance. For example, Biddle et al. 2020 provide a narrative review of resilience concepts in empirical studies, but provide little in-depth characterisation of resilience characteristics and their relevant capacities, or their applications across different health systems, which could provide useful learning on how to operationalise resilience in different health system contexts [1]. As Forsgren et al. (2022, this issue) suggest, this focus on the theoretical has led to a lack of knowledge about which strategies for building resilience have been successful, a gap they seek to close [2]. Similarly, Khalil et al. (2022) point out the need to operationalise resilience concepts into healthcare practice [3]. However, what is missing is a thorough analysis of what resilience paradigms and concepts outside of the health domain can offer health systems research and practice. Wiig and O’Hara offer a particularly useful starting point with their analysis of the impact that resilience engineering concepts are having on health systems research [4]. Our aim here is to augment existing reviews on resilience in health systems, and to specifically compare the way resilience is operationalised in empirical studies to the core paradigms and conceptualisations of resilience from a broader set of domains (specifically, engineering sciences, socio-ecological sciences, organisational resilience, and community resilience). We thus take up the unique task, to review the paradigms and approaches to resilience, and map them on the different subsystems of the health system to which they are applied. Rather than bypass the possible variations in conceptual frameworks, we sought to track them; and thereby identify the relationship between conceptual origins showed in the literature, which parts of the health system they focus on, and the aspects of resilience used.

To achieve this objective, we had two specific sub-objectives which guided our search and review method:

The word resilience stems from the Latin resilio, or ‘to jump back’ [5]. Resilience has double roots in socio-ecological systems and in psychology. In psychology, research has focused on the ability of individuals to deal with trauma and extremely adverse events (Comes et al., 2019 [6]). Studies focus on personality traits that help defend against exposure to extreme stress; this includes aspects such as meaningful purpose, agency and a growth mindset [7]. Taking a systems- rather than an individual perspective, in this paper we do not focus on psychological resilience.

In socio-ecological systems, the concept of resilience was introduced in Holling’s seminal paper to characterise the ability of a system to evolve and adapt under shocks and stresses [8]. Holling’s work inspired a rich body of work in fields ranging from resilience of ecosystems to climate adaptation [9, 10]. Resilience in this realm is often defined as ““the capacity of a system to absorb disturbance and reorganize while undergoing change so as to still retain essentially the same function, structure, identity, and feedbacks” [11], or the ability to persist by changing. This evolutionary perspective stresses the emergence of systems properties, and the rise of rare events, which “can unpredictably shape structure at critical times or at locations of increased vulnerability” [12]. This notion of resilience, however, is in contrast with the need for planning and decision-making, and does not fit the long lead-times and life cycles that are typical for decisions in health care planning and management.

In engineering, resilience is largely used to refer to the ability of some system to rapidly return to a single state of equilibrium or stability after a disruption, thereby ensuring continued functioning and rapid recovery from stress and disturbances [13, 14]. Often, engineering resilience is associated with robustness, broadly understood as the ability to withstand shocks and ‘bounce back’ to the same performance level. These principles have been instrumental in identifying critical points in complex, interdependent infrastructure networks [15], analysing disruptions [16] or for designing new infrastructures and tools [17]. However, these approaches do not explicitly incorporate aspects of adaptation and transformation that socio-ecological systems or community resilience encompasses.

From the perspective of social-ecological resilience, which puts forward the concept of resilience as a non-equilibrium notion, there are important links to participation and self-organisation that are common to community resilience and the social capital theories of resilience [18, 19]. These concepts focus on the networked sets of capacities that a community or organisation can generate, in order to be(come) resilient [20]. While community resilience focuses on the adaptive capacity of communities and their abilities to respond [20], organisational resilience seeks to explain the features of highly resilient organisations [21, 22]. These concepts are also prominent in the disaster resilience literature, where resilience has been promoted as a way to analyse and understand the reaction of a system to a hazardous event, promoting activities such as improving coping capacities and livelihoods [23].

Overall, we observe three major principles that are meant to improve resilience: reducing impacts or consequences (robustness), reducing recovery time (absorption), and reducing future vulnerabilities (adaptation and transformation). These principles are represented to a different degree in the four paradigms that we are investigating here: socio-ecological systems resilience [8], engineering resilience [14, 24], community resilience [20], and organisational resilience [21, 22].

For our study, we sought to look for instances in which these concepts were used, addressed, or referred to, either in the empirical studies (aim 1 above) or in the reviews (aim 2 above).

We describe our work as a scoping review of empirical studies and reviews, which we explain as follows: We first searched for any studies relating to resilience and its accompanying definitions in the health literature, and applied inclusion/exclusion criteria (see full details below). We then divided our included studies into two categories to match our sub-objectives above: empirical or non-review studies were analysed according to the characteristics described in sub-objective (1) above, while all reviews were reviewed separately and key resilience themes and concepts were extracted for sub-objective (2). The results from both components were analysed together to satisfy our overarching objective. These components are conceptualised in Fig. 1.

We expanded the search string and method used by Turenne et al. 2019 [25], who provided a conceptual analysis of health systems resilience, by adding broader search terms that enabled us to include regional, local or care provider-based empirical studies. We also limited our search to PubMed, as our focus was to identify specifically how the health literature conceptualises and uses ‘resilience’ and compare this analysis to other disciplinary domains in our analysis and discussion. We conducted our search on 16 June 2021 which yielded 2773 results (see Supplementary file A for PRISMA flowchart); and the full search string is provided below:

To remain up to date with the latest literature, and in response to helpful feedback from one of our reviewers, we updated our search in June 2023 which yielded a further 796 results. These are not included in the empirical study review (sub-objective (1) in Fig. 1), but we did include the extra n = 23 reviews we found in the full-text review analysis (sub-objective (2)).

For the analysis of empirical studies (sub-objective (1) in Fig. 1), we included studies if they met the following criteria:

For eligibility in the full-text review of reviews (sub-objective 2 above), we applied the same criteria, except for study type, since this had to be a review with an included search and analysis method (e.g., scoping review, systematic review, narrative review).

We used the platform ‘Rayyan’ for screening the titles and abstracts. Three reviewers piloted the screening based on the inclusion/exclusion eligibility criteria. Two reviewers rescreened these relevant/possibly relevant records and we resolved the disagreements in group meetings. We followed arbitration by a third reviewer.

Following study selection, the studies were divided into two categories (Fig. 1): (a) empirical studies that concerned resilience of the health system, and (b) reviews (any method included). This resulted in n = 87 articles included in part (a) and n = 30 articles included in part (b) the review of reviews (see Supplementary file A for our adapted PRISMA-Scoping reviews flowchart). The following steps (data extraction and analysis) were conducted separately and their associated methods are reported in two parts below.

While the studies included in these statistical analyses were not reviewed in full, the following data from these articles were extracted:

Further, we followed Meerow et al.’s [27] approach in analysing resilience of what and to what. In the ‘of what’ category, we analysed which parts of the health system were studied, ranging from specific departments within a hospital, to the public health system. Under ‘to what’ we analysed the shocks or stresses that the system under consideration was supposed to, and the drivers for resilience. To understand how the different concepts are applied across paradigms and domains, we developed heat maps that show the frequency of co-occurrence of different terms.

We designed and tested the data extraction form in a spreadsheet shared via Google Sheets to enter: author-title of the review, year and location(s), country in which the empirical study was conducted, threat (to what?), the part of the health system that was studied (of what?), the underlying theoretical paradigm used, the resilience phase, and the concepts that were referred to as means to improve resilience. We also indicated if the related choices were not made specific or could not be inferred from the manuscript. To capture intersections between the concepts, and understand how different capacities are co-evolving or co-used, we analysed and counted all concepts that a paper touched upon, i.e., if a paper referred to e.g., absorption and adaptation, we counted it under both categories.

Where possible, we inferred paradigms and concepts as mentioned from the abstract. If that was not possible, the full papers were scrutinised for additional information. If the category could not be detected, we labelled the paper as ‘not specified’. In addition, because of the breadth of the field, we grouped papers where possible. For instance, papers that referred to ‘rapidly bouncing back’ were categorised under the resilience engineering paradigm.

We included a total of n = 87 articles in the descriptive analysis, available in full in a table in Supplementary File C.

We included a total of n = 30 articles in the review of reviews, included in Table 1 (n = 17 up to June 2021, matching the date of the empirical studies analysed in (a), plus a further n = 23 reviews published after June 2021 up to June 2023). These included systematic reviews (n = 11), scoping reviews and other or non-systematic. Not surprisingly, most of the reviews were published more recently: 2022 (n = 10), 2021 (n = 5), 2020 (n = 4). Most of them identify what constitutes ‘resilience’ in health systems, but only few refer to conceptualisations in non-health domains such as engineering and other sciences for example, Biddle et al. 2020 [1] and Hess et al. (2012) [86]. Many of the earlier reviews refer to the definition of resilience by Kruk et al. (2017), which includes being adaptive, self-regulating, diverse, aware, and integrated [87]. These characteristics draw on socio-ecological systems paradigms but this allusion is not explicitly mentioned in the reviews using them as a conceptual framework. In later reviews, authors mostly appealed to the domains of resilience identified by Blanchet et al. (2017) [88], which includes the concept of transformation, but that aspect is not in fact widely used (e.g. [1]).

Most of the reviews refer to multiple levels or locations at which resilience can be addressed and assessed, and the importance of including stakeholders from service providers to governments to patients themselves in design and planning processes in order to build resilience. The Resilience in Healthcare group [113] in particular notes that patients are part of resilient responses, and draw from their analysis of multiple levels where ‘resilience characteristics’ can be found– ‘individual, team, management and organizational’. Other approaches take up the healthcare system components via the WHO, thus using a modular approach to the system as a whole (eg. Bayntun et al. 2012 [90]).

In terms of resilience aspects or phases, about half the reviews prior to 2021 draw out ‘preparedness’ as a notion covered by their included studies. Hess et al. 2012 [86] take up ‘adaptive capacity’, specifically in relation to climate change; and ‘adaptation’ is further covered by Zhong [91], Chamberland-Rowe et al. 2019 [95], Haldane et al. 2021 [79], while Turenne 2019 [25], Foroughi et al., [105, 107], Thu [109], Fridel et al. 2020 [98], Fleming [110] and Ismail [112], also make explicit mention of ‘transformation’ in addition to adaptation. Transformation and preparedness is also covered in the scoping review by Nuzzo et al. 2019 [94]. However, despite more recent reviews mentioning and identifying the concept of transformation in the studies they reviewed, they note that its application and use is less present in empirical work.

The key messages that stand out from these reviews is the lack of a common definition of resilience [25, 94, 98], and the scarce or underdeveloped use of learning and transformation as concepts operationalised in the empirical literature, which is also in line with our findings from the empirical literature. Furthermore, Nuzzo et al. 2019 point out the lack of implementation frameworks to translate resilience capacities into something that health system actors can employ in response to crises [94]. However, the more recent reviews have started to focus on the importance of operationalising resilience in practice (e.g. [2, 3]). Despite the turn toward community resilience (learning, empowerment) and socio-ecological resilience (self-organization, transformation) concepts being recognized as important—in contrast to the more linear, engineering resilience approaches—this theoretical turn has not (yet) resulted in the adoption of those concepts in practice.