A DETERMINANTS OF HEALTH CONCEPTUAL MODEL FOR FISH AND WILDLIFE HEALTH Open Access

Health is a common but vaguely defined management goal for wildlife populations (Hanisch et al. 2012; Stephen 2013). Health in veterinary medicine, particularly as concerns wildlife, historically focused on adverse pathophysiologic or productivity outcomes or used a disease-centric approach (Gunnarsson 2006; Nordenfelt 2011; Stephen 2013, 2014). Perspectives on health are changing and there is consensus in other fields, such as human public health and herd health, that health is more than the absence of disease (Eriksson and Lindström 2008; Jayasinghe 2011; Nordenfelt 2011). These fields acknowledge that health is not a dichotomous state where an individual or population can be classified as healthy or unhealthy but is rather an aspirational capacity (Arah 2009; Nordenfelt 2011). Being healthy is the ability or capacity to realize full function, satisfy daily needs, and adapt to or cope with changing environments (Frankish et al. 1996; Eriksson and Lindström 2008).

The determinants of health (DOH) approach is the prevailing public health perspective for understanding what makes a population healthy or not (Eyles and Furgal 2002; Cieza et al. 2016). The DOH model includes 12 key determinants of health (Table 1) signifying the interacting and varying contributions that abiotic, biotic, and social elements make to health outcomes (McDowell 2017; PHAC 2018). Advantages of using a DOH approach in public health are: The inclusion of interactions among the many contributing factors that influence resilience, rather than a focus on a physiologic state (Frankish et al. 1996; Nordenfelt 2011); the expansion of the scope of interventions, information, and expertise that can be employed to influence health by considering these multiple factors (Frankish et al. 1996; Stephen 2013); and recognition of factors in addition to adverse outcomes, such as death and disease, which allow for a proactive approach that can be implemented prior to adverse outcomes (Frankish et al. 1996; Stephen 2014). The DOH approach focuses on building and reinforcing the capacity to remain healthy rather than on delaying actions until harms are detected (Stephen 2014).

Some authors advocate for a multifactorial approach that integrates diverse drivers as the next step toward a modern understanding of wildlife health (Deem et al. 2008; Hanisch et al. 2012; Stephen and Duncan 2017). Despite their arguments, those authors did not provide an operational definition of health and therefore a gap exists on how a multifactorial approach to health might be applied in practice. We hypothesized that a DOH model for wildlife populations could be a way to address this gap.

We focused on two well-studied, socially important species for which the management of health and resilience is a priority in Canada: barren ground caribou (Rangifer tarandus groenlandicus) and Pacific salmon (Oncorhynchus spp.). The purpose of our study was to determine if: 1) a DOH model for wildlife health could be derived from an analysis of literature on wildlife health and resilience; 2) the model could be applied to more than one species; and 3) the model reflected how fish and wildlife health managers and investigators conceptualized health.

Thematic analysis can identify patterns and themes within and across qualitative data (Clarke and Braun 2017). Thematic analysis is useful for applied health research, particularly when research questions involve analysis of policy (Braun and Clarke 2014). We followed the six-phase thematic analysis structure described by Braun and Clarke (2006) where themes were based on the functional attributes of each DOH in the public health model (Table 1; PHAC 2013).

We selected two case studies, barren ground caribou and Pacific salmon, to investigate if the DOH model could be adapted to wildlife. A scoping literature review was conducted and summarized according to standard methods (Levac et al. 2010) as outlined in Table 2. We used the definitions and attributes of DOH used in public health to categorize features characterizing health or resilience in the literature that we reviewed. In socioecologic systems, resilience is defined as “…the ability to cope with shocks and keep functioning in much the same kind of way” (Walker and Salt 2012), and therefore overlaps modern perspectives of wildlife health as described in Stephen 2014. If no equivalent to a specific human DOH was found, that DOH was dropped from further consideration whereas if a paper used features to define health differently from the human DOH, a new wildlife DOH theme was created. From this process, candidate DOH themes were created for barren ground caribou and Pacific salmon health. The DOH we identified for the two types of animals were combined to develop our initial wildlife health DOH model.

We conducted two separate expert opinion processes, one focused on caribou and the other on Pacific salmon, specifically sockeye salmon, to test whether the candidate DOH were consistent with how experts perceived wildlife health. We identified experts using peer-referential techniques (Penrod et al. 2003; Christopoulos 2009). The Environment and Natural Resources Department (ENRD; Government of the Northwest Territories) provided an initial list of caribou experts. Each expert was contacted by email with a follow-up phone call and given an opportunity to provide names of additional experts. Because this was a small group of experts, only three levels of referrals were required using snowball sampling (Christopoulos 2009). A total of 34 experts were identified by their peers. Contact information was secured for 31 and 11 participated.

A similar process was used for the sockeye salmon study with an initial list of experts in Pacific salmon biology derived from the staff scientist directory for the Department of Fisheries and Oceans Canada (DFO), Pacific Science Branch, Pacific Biological Station. This list was evaluated by an independent fisheries biologist familiar with the science employees at the Pacific Biological Station to further refine the list to those with background knowledge on salmon health or resilience. Experts were contacted through email to request their participation and to nominate up to three colleagues. A total of 38 DFO staff or contractors in salmon research were contacted to request their participation and 12 completed the exercise.

A diagrammatic approach to network analysis was used to identify the type and interrelationships of determinants of caribou and salmon health deemed important by the experts based on their personal and professional judgment and experience. This approach can provide insight into the opinions of experts on complex issues, specifically as to how different factors interrelate (Campbell and Muncer 2005; White 2008). Caribou experts all participated remotely via email. A face-to-face meeting was offered to the salmon experts, with nine of the 11 completing their assignments at that meeting and two of the 11 communicating only via email.

The participants were each asked to draw a diagram of the direction, interrelationships, and relative impacts of the various factors they believed to determine caribou or salmon health, depending on their expertise. Before beginning the diagram component of the exercise, participants were given a brief description of our perspective of health as the ability, or capacity, to realize full function, satisfy daily needs, and adapt to or cope with changing environments. We also drew their attention to the similarity of this perspective with the concept of resilience. Caribou experts were asked to consider a population of barren ground caribou in the Northwest Territories. Salmon experts were to consider a population of marine adult Fraser River sockeye salmon (Oncorhynchus nerka). For the salmon, a geographic boundary, the Strait of Georgia, was collectively agreed upon by the experts during the in-person session and was shared with the remote participants.

Participants were given a list of factors identified during the thematic analysis for the literature-based model. Experts could include any, all, or none of the provided factors in their diagrams and had the option to add additional factors as they saw fit. With ‘health’ written in the center of the page, the experts wrote down the factors in the surrounding space. The experts were then instructed to draw an arrow between their chosen factors and health in the center to represent the causal relationships they believed to exist. For each relationship, a plus or minus (+, –) was added to denote whether the factor had a positive or negative impact on the outcome. Finally, experts provided an impact score between 1 and 10 for each relationship to denote the size of the impact that the factor had on the outcome (1=small or negligible impact, 10=very large impact). Participants were informed that the impact scores would be used to identify which relationships within the networks were likely to have the largest impact on health. The salmon experts requested the ability to note differences in relationships in the network for juvenile freshwater Fraser River sockeye salmon, if they existed, using an alternate color on the diagram. Finally, the experts were asked on what source of knowledge their answers were based, using the categories of experience in the field, traditional knowledge, scientific literature, intuition, common sense, anecdotal evidence, and not sure. For this question, the experts could select up to three sources.

Data from the two case studies were evaluated independently, using the same technique. If three or more experts mentioned the same relationship, it was included in the network analysis along with the mean impact score. As necessary, some factors were consolidated to account for inconsistencies in terminology. The Fruchterman Reingold algorithm in the open-source network visualization and analysis software Gephi 0.9.2 (Gephi 2018) was used to visualize the relationships between the nodes. The Eigenvector centrality statistic was used to determine which nodes had the most connections. Single-sided t-tests were used to compare the weights of positive vs. negative relationships in the aggregate networks using Stata 15.2 (StatCorp LLC, College Station, Texas, USA). We also used a paired, two-sided t-test to compare the mean impact scores of the salmon adults and juveniles, where experts included juvenile differences.

Six DOH themes were identified in the literature for salmon and caribou health: abiotic environment, needs for daily living, social environment, biologic endowment, direct mortality pressures, and human expectations (Tables 3, 4). Abiotic environment, analogous to physical environment in humans, relates to the health of the natural environment in which the populations exist and includes different factors related to climate and anthropogenic pressures on the environment (Smit and Spaling 1995; Canter and Atkinson 2011; Raby et al. 2015). The needs for daily living, comparable to socioeconomics in humans, included factors related to the equitable distribution of resources which would allow an individual control and discretion over their decisions. Factors connected to habitat, food, and the ability to express natural behaviors in the environment fell into the needs for daily living category (Stephen 2014). Social environment reflects how the community at large can impact an individual's capacity to cope by influencing access to various resources. Population demographics, interspecific competition, and intraspecific competition were aspects pertinent to social environment. Biologic endowment, a fundamental determinant of health that is the inherited or predisposed capacity to cope based on biology made up of factors such as genetics, disease, and stress, is the same in both wildlife and humans. One DOH that impacts wildlife but is not a recognized DOH for human health is direct mortality pressures. Direct mortality pressures are the factors that pose an immediate threat to wildlife survival and include predation and hunting or fishing by people (Munns 2006). The final DOH identified for wildlife was human expectations, which is comparable to health services in humans. Health services are defined as “services… designed to maintain and promote health, to prevent disease, and to restore health and function… to population health” (PHAC 2013). In wildlife, analogous service functions are provided by various policies, management actions, education programs, and their performance levels established by stakeholder expectations. Although the specific components or attributes of each DOH may vary between contexts, such as in different ecosystems, these six categories represented the general themes extracted from the literature regarding caribou and salmon resilience and health. Based on the similarities found in the caribou and salmon literature regarding DOH themes and the types of factors that drive health and resilience, we were able to integrate these two species-specific models into a single model (Fig. 1).

The resulting network model (Fig. 2) of factors influencing caribou health and their interrelationships based on our experts' opinions included all six DOH themes identified in Figure 1. There was no statistical difference in the mean impact scores of the positive and negative relationships (depicted by arrow weight; P=0.663). Recruitment (a social environment factor) and forage quality and quantity (a needs for daily living factor) had the largest perceived positive direct impacts on caribou health. Although disease and stress were included, 87% of the expert-identified drivers of health were represented by other DOH (Table 5). The abiotic DOH were believed to impact caribou health in the most ways, representing 40% of the nodes in the network (Table 5). Experience in the field and common sense were the most-commonly mentioned sources of knowledge specified by the caribou experts (six mentions each). Scientific literature and intuition were mentioned five and four times, respectively. Traditional knowledge and anecdotal evidence were mentioned once each. No caribou expert selected being “not sure” of their knowledge sources while completing the network exercise.

For salmon, five of the six DOH themes were identified collectively in the expert network (Fig. 3). Human expectations did not feature in the network analysis. Pathogens, disease, stress, and genetics were featured in the expert network analysis; however, there was also a large emphasis on habitat quality, including water and food quality. Thirty percent of the nominated DOH themes in the salmon network involved biologic endowment, which included pathogens, disease, stress, and genetics (Table 5). The impact score of the positive relationships was not statistically different from those of the negative ones (P=0.781). The salmon network had fewer positive relationships (six) than negative relationships (nine). The salmon experts had an opportunity to include different relationships or impact scores for juvenile sockeye salmon if they believed them to be different than those for adult salmon. The mean impact scores for the adults and juveniles were not statistically different (P=0.247); therefore, we only modeled the adult group. Experience in the field and scientific literature were the most commonly mentioned sources of knowledge specified by the experts (12 mentions each). Common sense (five) and anecdotal evidence (three) featured less frequently as knowledge sources. No experts selected traditional knowledge, intuition, or not being sure.

In both the salmon and caribou cases, experts reported that they had sufficient information and expertise to complete the assignments and assess a wide suite of DOH. The experts did not identify any additional DOH beyond those found in Figure 1.

A DOH model for wildlife health was adapted from the model commonly described in human population and public health. This model accommodated the suite of drivers described in literature on the health and resilience of barren ground caribou and Pacific salmon and was compatible with how experts perceived the drivers and determinants of health for these species. Expert opinion and the collective health and resilience literature for both barren ground caribou and Pacific salmon supported a cumulative effects health model involving multiple factors. The numerous factors influencing health were seen to extend far beyond the disease or pathogen focus common in wildlife health studies and legislation (Stephen 2013, 2014). The caribou and salmon DOH themes were the same, allowing them to be combined into a single wildlife DOH model. Although human drivers and expectations were included in the final model, both the literature and experts emphasized biotic and abiotic factors.

Resilience in ecology is a complex concept that acknowledges the impacts that diverse factors can have on an ecosystem's capacity to cope with change (Biggs et al. 2012; Walker and Salt 2012). The resilience of an ecosystem is dependent on the functionality and structure of its components (Gunderson 2000; Walker and Salt 2012), the nature, severity, and duration of impacts on the system (Rapport 1998; Biggs et al. 2012), the potential cumulative impact of multiple stressors (Gunderson 2000), and the effectiveness of management measures or interventions (Biggs et al. 2012). These components of ecosystem resilience are congruous with many of the drivers of health identified by the expert participants in our study.

The range of interventions available to wildlife managers, particularly when it comes to health (or more traditionally, to disease), is limited (Stephen et al. 2018). A DOH approach may be a method to identify potential issues that reduce a population's resilience in advance of a harm, or reduces their ability to cope with harms, without needing to rely on the standard disease control approaches used in domestic species. For instance, in public health, factors in the social and physical environments are modified to reduce human exposure to hazards in advance of disease (Cole et al. 1998). Various policy and regulatory measures, for example, aim at exposure as a primary public health target (Cole et al. 1998). The DOH approach may be useful to target not only factors that increase susceptibility to disease (Frankish et al. 1996) but also to direct action on the major drivers of population declines and extinctions such as habitat loss, climate change, unsustainable hunting, poaching, and harvesting, pollution, and invasive species and disease (World Wildlife Fund 2016).

Not only is a multifactorial and proactive approach to health needed for wildlife, but also there is a need for an operational definition of health. Health is a management goal for both caribou (ENRD 2011) and Pacific salmon (DFO 2005), but health is rarely defined in management documents. Without a definition of health for wildlife, it remains an amorphous concept (Nordenfelt 2011; Stephen 2013) making planning, management, and measurement toward health goals very challenging. Based on experience in public health, a DOH approach could help to provide a mode for attaining the goals advocated for by Stephen (2014) and Hanisch et al. (2012). The DOH model helps to identify the external drivers of health, recognizing the complex and interrelated nature of health (Jayasinghe 2011). A multifactorial model of health, like a DOH model, could help make explicit some of the external drivers of health which could in turn help to identify a wider suite of stakeholders, interventions, and policy options to prevent harm and to promote health (Pourbohloul and Kieny 2011; Rapport and Hildén 2013).

Using qualitative methods to adapt the DOH model to two types of wildlife may have affected the structure and content of the model. The literature component of our study, which was conducted following established methods used for thematic analysis (Braun and Clarke 2006; Clarke and Braun 2017), and scoping of literature reviews (Levac et al. 2010), constrained the concepts that could be explored in constructing our model. Selection of additional search terms in relation to our selected species, such as “survival” or “population dynamics,” may have increased the number of returned journal articles and, therefore, possibly more potential factors impacting the species. However, there is information to suggest that this was not a significant bias. The indicators for caribou identified in our scoping review were compared with those found in a comprehensive review of caribou literature (Greig et al. 2013), and their review did not identify any additional factors. Furthermore, the expert opinion networks did not result in added DOH. Our study showed the generic model could be applied to two different species and shared core determinants that were applicable across populations and life stages.

Network analysis is an established method commonly used to investigate opinions on the relationships between various interconnected elements (Campbell and Muncer 2005; Hecker et al. 2013). We elected to use the diagrammatic approach to network analysis (Green and McManus 1995) instead of the more common matrix approach (White 2008), as there are a number of criticisms of the matrix approach including: the production of larger networks with more causal connections than are perceived by any one participant (Muncer and Gillen 1997); an overly complex representation as participants are asked to consider relationships that may not exist (Campbell and Muncer 2005); and there is no opportunity for free choice of factors (Green and McManus 1995). The diagrammatic approach aims to address these issues by providing the opportunity for participants to spontaneously identify the most important relationships and by analyzing the results in a way that represents the average network (Campbell and Muncer 2005).

We concluded that the network exercise was acceptable and understandable because all but one expert who participated in the in-person salmon workshop submitted data. One expert, who contributed data remotely to the salmon network and therefore did not participate in the priming presentation or group discussion, submitted data but not in the network format. The submission by this expert, as well as the low response rate from the caribou experts for the digital survey, may signify that these methods are best implemented in person. There were a few salmon experts who noted that they could not separate in their minds the impacts of factors on adult and juvenile Fraser River sockeye salmon. This perspective may reflect a life course epidemiology perspective, where the accumulation of events over a lifetime impact an individual or population's capacity to be healthy at any given point (Ben-Shlomo and Kuh 2002). The in-person session allowed for the group to clarify the parameters of the network exercise, but our study did not assess the strengths and weaknesses of in-person or email responses. This mixed approach to engaging experts was selected as the most feasible way to facilitate participation of experts living across a large geographic area.

In humans, the DOH model is not typically used as a measurement tool (Diez Roux 2008) but as a framework for planning, policy development, and guiding research (Pourbohloul and Kieny 2011). A hallmark of the DOH model is its adaptability to different contexts, populations, and challenges (Pourbohloul and Kieny 2011; Mayhew and Hanefeld 2014). Health is context specific—for wildlife as well as for people (Arah 2009; Jayasinghe 2011). What may be a critical DOH for one population may be less significant for another one. Figure 1 should be a starting point for those wishing to conduct an analysis such as the one we did with the caribou and salmon. For a general health model to be useful, it is important that it be adaptable to nuances while still operating within the general framework. The network analysis was useful for capturing the opinions of experts and was adaptable to different species in different ecosystems, demonstrating the generality of the DOH model. The DOH model that we produced may be the foundation of a wildlife health planning tool that conceives health as a cumulative effect and helps to strategize and prioritize a suite of actions in a world of interacting determinants of health.

We would like to thank Brett Elkin and Stewart Johnson for their assistance in the expert consultations pertaining to caribou and salmon, respectively. We would also like to thank all the individuals who participated in the population health expert opinion processes.