Health and migration in the context of a changing climate: a systematic literature assessment

For the purposes of this synthesis research, 'climate change' was defined as a change in the state of the climate that can be identified by changes in the mean and/or the variability of its properties, and that persists for an extended period, typically decades or longer (IPCC 2018). This assessment included studies exploring climate variability and any climate hazard or extreme weather event that could be plausibly linked to anthropogenic climate change and where authors referred to climatic change. 'Health' was defined broadly to include physical, mental, and social components and due to the population health nature of this assessment, was extended to include food security⁹ and water security,¹⁰ and water, sanitation, and hygiene (WaSH). The definition was not extended to the social determinants of health,¹¹ which were examined separately.

Studies were included whether health was considered as a driver or an outcome of climate change-related migration. 'Migration' was defined as the movement of persons away from their place of habitual place of residence, either across an international border or within a state regardless of legal status, degree of choice, causes of movement, or length of stay (IOM 2019a). This included all possible mobility responses such as forced displacement, migration, and planned relocation (The Nansen Initiative 2015). The population studied was required to engage in or be affected by a mobility response. This included populations at the origin, mobile, and host communities. Immobility was also considered a mobility response to ensure trapped populations were not excluded. In summary, the assessment did not narrowly focus on the health of 'climate migrants/climate refugees' but rather included research that studied health and migration in the context of a changing climate.

This nexus crosses several research fields and intersects with both natural and human sciences, and therefore the evidence originated from diverse disciplines including demography, climatology, public health, policy, international relations, and geography. With this in mind and consultation with a subject librarian the following four academic databases were selected; CABI Direct—Global Health (1973 to present); Ovid Medline (1946 to present); PubMed (1966 to present) and Scopus (1970 to present). Also, Google Scholar was searched with the same search terms. Grey literature was not included in this assessment due to the focus on empirical evidence and a recognition of the multitude of conceptual papers and editorials on this topic. Nevertheless, the search strategy was designed to capture primary studies included in important grey literature reports.

The RepOrting Standards for Systematic Evidence Syntheses (RoSES) Pro-forma and flow diagram was used as the quality criteria framework for reporting because it was specifically created for environmental management and conservation research (Haddaway et al 2017) which integrates diverse methodologies.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) (Moher et al 2009) was not appropriate due to its focus on clinical trials and emphasis on meta-analysis.

Types of studies: All research designs that included empirical data whether primary or secondary, were considered for inclusion. Due to methodological ambiguity in the search results, the 'empirical' criterion was further refined to require a methodological explanation and a description of the data used. Modeling and qualitative case studies were included but commentary, editorials, and literature reviews were not.

Search Terms: Table 1 outlines the search terms which were categorized using the PEO approach (modified PICO) under three concepts and the search was conducted across (AND) and down (OR) with synonyms for the key search terms: Climate Change, Health, and Migration. The search strategy consisted of free-text and MeSH (i.e. Medical Subject Headings) terms and was tailored to the search requirements of each database (figure 1: Example search string Scopus). Several preliminary pilot searches were conducted from September—November 2018 to test the search strategy. Test searches revealed very few publications on the topic before 1990. The chosen timeframe was thought to capture the most current and relevant evidence about the research aims and specific objectives. Test searches also revealed a multitude of studies examining plant and animal health and migration in the context of climate change and led to a reconsideration of the selected databases. The actual search was conducted in December 2018 and was limited to studies published in English or German.

Zoom In Zoom Out Reset image size

Table 2 outlines the inclusion and exclusion criteria, which were strictly applied to each study. The assessment included research investigating populations worldwide engaging in or affected by mobility responses. These responses were on a spectrum from displacement, encompassing forced movements, to migration, which is considered predominantly voluntary, and also immobility. Populations from places of origin as well as mobile and host populations were included. The assessment required a climate hazard as at least one driver of the mobility response. If a migration driver was economic, political, social, and/or demographic and there was no explicit or implied link to climate change or climate variability the article was excluded. If the hazard was not related to climate change (e.g. volcano, earthquake) the article was excluded. The assessment required studies to focus on health including measurable health outcomes (e.g. prevalence of disease) or immediate determinants of health considered by this assessment to be food and water security. Studies that referred to social determinants of health as outcomes (e.g. income, employment) were excluded.

Authors were not able to make unilateral decisions to include their own work because each study was reviewed by at least two independent reviewers. This process was also blinded using the Covidence platform so that 'votes' to include or exclude could not be seen by the partner reviewer.

Reference lists of included and excluded studies were scanned for relevant studies that were manually added to the Covidence platform for full-text review to reduce the likelihood that eligible studies were missed. Figure 2 presents the screening process and the key reasons for exclusion.

Zoom In Zoom Out Reset image size

Data extraction codes were predetermined and agreed upon by reviewers who have expertise in climate change and migration and climate change and health. Codes were aligned to the aim and specific objectives.

The data were extracted in an excel sheet (supplementary material (available online at stacks.iop.org/ERL/15/103006/mmedia)). Two authors (PNS, IM) reviewed all 50 selected studies in full-text and conducted a quality appraisal and data extraction simultaneously and collaboratively. When consensus could not be reached regarding quality appraisal or extraction coding, a third independent author (CM, KB) was consulted. When data were incomplete, unclear, or missing, an attempt was made to contact the authors of eligible studies.

The quality appraisal tool MMAT version 2018 was chosen a priori and used to identify threats to internal and external validity for all eligible studies. MMAT provides five custom questions for each study design (qualitative, quantitative randomized controlled trials, quantitative non-randomized, quantitative descriptive, and mixed methods), pertaining to both study and reporting quality (Hong et al 2018). Quality appraisal results were recorded in the results spreadsheet along with all other coded findings. Three modeling studies could not be appraised by the MMAT and this remains a limitation of this systematic assessment.

All eligible studies were included in the analysis. A meta-synthesis was undertaken which is considered advantageous when dealing with broader research questions with disparate outcomes (CEE 2018). Extracted data were synthesized using Collaboration for Environmental Evidence (CEE) methodology (CEE 2018). More specifically, once data was extracted from the selected papers according to predefined codes, an appropriate analysis method was selected. This consisted of simple descriptive statistics (example for the codes: 'research setting', 'study design') and thematic analysis in the case of more detailed narrative data (example for the codes: 'research question', 'key findings'). In the latter case, we used the six steps outlined by Braun and Clarke (2006) namely, 1. Familiarizing yourself with your data 2. Generating initial codes 3. Searching for themes 4. Reviewing themes 5. Defining and naming themes 6. Producing the report which included meaningful data presentation and visualization. The quality of the included studies was considered a result and therefore was a component of the meta-synthesis.

The search strategy revealed 1904 studies with 64 duplicates that were removed. We screened 1840 studies by title and abstract and excluded a further 1660 because they did not meet the inclusion criteria or met an exclusion criterion. The main reasons for exclusion at 'first pass' (title and abstract) were: Not empirical evidence (e.g. editorial); Focus on a dyad (either climate change and health or climate change and migration) not the nexus (climate change-health-migration); Animal or plant health/migration not human health/migration. The resultant 180 studies were screened in full-text and 130 were excluded for the reasons outlined in the screening process flowchart (figure 2). The assessment yielded a total of 50 eligible studies. The main reasons for exclusion at 'second pass' (full-text) screening stage were: Not empirical evidence (e.g. editorial), focus on a dyad (two concepts)not the nexus (all three concepts); no health outcome. Less common reasons included; Mobility driver that was unrelated to climate change, or a population that was not engaged in or affected by mobility.

The 50 eligible studies were conducted in 27 countries. If one study included more than one country, it was duplicated in the analysis so that there were 59 study sites altogether. Figure 3 presents the geographic location of the identified studies by WHO region. Most studies were conducted in the South East Asian region (n = 21) followed by the African region (n = 18), the Pan American region (n = 13) the Western Pacific region (n = 4), the Eastern Mediterranean region (n = 2) and Europe (n = 1). Notably, just under one-quarter of the eligible studies took place in Bangladesh (n = 12). 85% of the study settings were in the Global South. Studies were conducted between 1992 and 2018 with the majority covering the period between 2013 and 2016 (n = 29). Details about the eligible studies can be found in table 4.

Zoom In Zoom Out Reset image size

The assessment revealed a diverse range of research questions, aims, objectives, and hypotheses. The assessment authors characterized these aims using a common research framework in climate change and health that is also used by working groups 2 and 3 of the IPCC; (1) Health impacts of climate change; (2) Adaptation to climate change for health including the health (and other) co-benefits of mitigation (IPCC 2014). Studies looking at the health impacts of climate-related migration asked the broad question: What are the relationships between migration and health in the context of climate change? (See table 4 studies: 3, 5, 6, 7, 10, 16, 20, 21, 27, 37, 41, 47). Others measured health outcomes in climate-displaced communities (table 4 studies: 9, 11, 12, 14, 25, 28, 29, 33, 34, 40, 43). Of course, not all research aims fit neatly into these categories and some looked at both 'impacts' and 'adaptation' (table 4 studies: 35, 38). Others were quite different and explored the association between climate vulnerability scores and net migration (table 4: 19) or how climate change influences the so-called 'healthy migrant effect' (table 4: 22).

Studies fitting under the 'adaptation' theme aimed to assess perceptions of climate change and implications for mobility with health being a finding (table 4 studies: 2, 8, 17, 24, 30, 32). Others considered the extent to which climate-related mobility was adaptive or maladaptive using health outcomes as an indicator (table 4 studies: 36, 42, 44, 48, 49). A few focused on relocation and explored the barriers and facilitators to success some of which related to health (table 4 studies: 4, 15). No eligible studies looked at the health co-benefits of mitigation in reference to climate-related migration and health.

The eligible studies (n = 50) were categorized, with reference to MMAT, into five types of study design: quantitative descriptive studies (15; 30%), quantitative non-randomized studies (7; 14%), qualitative case studies (13; 26%), mixed-methods studies (12; 24%); and modeling studies (3; 6%).

Each study was individually appraised for quality using the MMAT tool and we condensed a finding of the overall quality of evidence by methodological group (see table 4 and supplementary material) (Hong et al 2018). For individual studies, the quality score was attained by using the five custom questions for each study. We used a ranking system according to the articles' adherence to each of the criteria i.e.—% time we could answer 'yes' to the quality appraisal question (*****100%, ****80%, *** 60%, ** 40% * = < 20%). In mixed methods studies there were 15 quality appraisal questions (quantitative component, qualitative component, mixed methods general) so 15 rather than 5 was used as the denominator. For the overall rating we also used the % time we could answer 'yes' to the quality appraisal questions overall for that particular study design (see explanation below table 4 for more details). Below we present a critical appraisal in narrative form that aims to characterize the nature of the included studies, explore quality issues, and provide an explanation of the key strengths and weaknesses identified.

The eligible quantitative descriptive studies (n: 15) mainly consisted of surveys and case reports. These studies aimed to describe the existing distribution of variables without regard to causal relationships and are well suited to generate hypotheses. Of the quantitative descriptive studies two received a rating of *, three received a rating of **, four received a rating of ***, three received a rating of ****, and three received a rating of *****. Overall, the quality was high (63% QA questions answered 'yes'¹² ). We observed a lack of methodological detail, including sampling strategy, resulting in difficulty determining representativeness and limiting reproducibility. Yet measurements were largely appropriate and the risk of non-response bias was generally low. Analysis ranged from simple descriptive statistics to regression analyses.

The eligible quantitative non-randomized studies (n: 7) mainly consisted of cross-sectional analytic studies that were used to study climate exposures without using randomization to allocate units to comparison groups. These types of studies were considered a 'step up' methodologically from quantitative descriptive studies as they usually compared migrating and non- migrating households and resultant health outcomes, or climate-vulnerable and less climate vulnerable households. Of the quantitative non-randomized studies three received a rating of **, three received a rating of *** and one received a rating of *****. Overall, the quality was high (63%). Quality issues included temporal-spatial scales of the hazard, mobility response, and health outcomes rarely matching, representing an important methodological limitation. Some studies inadequately matched comparison groups demographically or paid limited attention to confounding.

Eligible qualitative studies (n: 13) mainly consisted of case studies and narrative research. These comprised a significant portion of the selected studies even without considering the qualitative component of eligible mixed-method studies. Of the selected studies, qualitative methods were used to explore and explain in-depth the issues intrinsic to a particular case of climate-related migration and health. Of the qualitative studies two received a rating of *, two received a rating of **, two received a rating of ***, two received a rating of **** and five received a rating of *****. Overall, the quality was high (68%). Qualitative studies with higher MMAT scores described data collection in detail and derived findings clearly from the data (e.g. using quotes to substantiate themes). Those with lower MMAT scores did not demonstrate coherence between data sources, collection, analysis, and interpretation.

The included mixed methods studies (n: 12) combined quantitative descriptive studies (mostly using surveys) with qualitative methods (mostly using focus groups or interviews). Mixed methods seemed to be well suited to the topic due to its methodological complexity and the fact that one method can complement the other and provide further details and explanations. Stronger versions balanced the emphasis on each method per the research question and adhered to the quality standards for both qualitative and quantitative components. Of the mixed methods studies, four received a rating of *, two received a rating of **, three received a rating of ***, three received a rating of **** Overall the quality was high (67%). Mixed methods studies with higher MMAT scores rationalized the approach and the choice of convergent or sequential design. They also meaningfully integrated the data types and identified and addressed divergences and inconsistencies between qualitative and quantitative results. Those with lower MMAT scores superficially applied one method and lacked integration i.e.—quantitative and qualitative results were reported and interpreted separately.

The modeling studies (n: 3) were included as there was no reason to exclude them according to the protocol however, MMAT could not be used to appraise their quality and this remains a limitation of this systematic assessment.

The key concepts of this assessment as defined by the PEO approach (Populations participating in, or affected by mobility responses; the Exposure of climate hazards; and health Outcomes) were coded, extracted, and synthesized into a narrative, tables, and appropriate data visualizations. A conceptual map is shown in figure 4 integrated the results of this assessment into the four conceptual frameworks outlined in the introduction.

Zoom In Zoom Out Reset image size

Figure 4 shows that climate change impacts are mediated through a range of socio-economic and other factors with different consequences for migration and health and that food and water security play an important role in these complex, multi-faceted and interacting pathways. Some of these pathways depicted in figure 4 may be empirically difficult to estimate due to under-representation in the literature.

A Sankey diagram (figure 5) further characterizes the intersections between mobility responses and health in the context of climate change. The graph depicts the evidence captured by this assessment. The mobility responses (left side) are linked with health issues and outcomes investigated in the eligible studies (right side). The width of the flows is proportionate to the number of relevant studies. Multiple connections within a study are depicted with 101 individual connections from 50 included studies. The nature of these relationships is not depicted in the Sankey diagram but was both positive and negative. In other words, certain mobility responses in the context of climate change both benefited and compromised health, yet the predominant relationship was negative in the eligible studies. Predominant patterns of mobility include forced displacement, rural-urban migration, relocation, and seasonal migration. Predominant health outcomes include food insecurity, access to healthcare, infectious disease, and mental health issues. Forced displacement and relocation was associated with food insecurity, access to healthcare and mental health issues whilst rural-urban migration and seasonal migration was associated with food insecurity, access to health care, and infectious disease. The terminology in the diagram reflects the language used in the research papers and demonstrates numerous connections (see section 4.2).

Zoom In Zoom Out Reset image size

First, within the extensive literature on climate change and health, proportionately few studies explore it concerning migration. Similarly, within the climate change and migration literature, proportionately few studies explore health (n: 50) (figure 2: Screening process flowchart). Within this evidence base there seems to be a transition from attempting to isolate sole drivers of migration and to quantify the number of people likely to be on the move under different climate scenarios, to a more nuanced discussion about how multiple factors interact to shape migration choices. Such research looks at the circumstances under which people move in a changing climate and the consequences for health. It is now important to extend this research further to reveal insights on how human health can be protected and promoted in the context of new patterns of human mobility as the climate crisis accelerates. This is crucial as people who are displaced due to climate change do not have the international protection afforded to refugees, and people internally displaced from climate impacts may fall through the gaps of national protection frameworks.

Considering that migration frameworks encourage 'whole of government approaches¹⁴ ' (Abubakar et al 2018, IOM 2019b) and climate change and health frameworks rely on inter-sectoral collaboration (IPCC 2014, WHO 2016), the research community will need to keep pace by adopting systems thinking to explore this type of nexus research and provide policy-relevant insights. As a minimum, climate change and health can be incorporated into migration research, and migration considered a determinant of health, as well as an important consideration within climate change adaptation frameworks.

Second, relationships between climate change, migration, and health are strongly heterogeneous. These relationships appear to depend on the differential exposure, sensitivity, vulnerability, and adaptive capacity of populations and are mediated by several social, economic, political, and demographic factors. Consequently, as the conceptual map demonstrates (figure 4), it is difficult to form general conclusions about how these relationships work, especially under variable climate scenarios, potential mitigation, and adaptation activities, and climate and social tipping points.

Since there is no definitive relationship between climate change, migration, and health, each case must be carefully assessed, taking health and the determinants of health into account and recognizing that migration is not uniformly an adaptive solution to climate change risk (Warner and Afifi 2014, Adams 2016, IOM 2020).

In recognition that mobility responses to climate change can (to different degrees) be adaptive or maladaptive in terms of health and other dimensions, approaches are required to address the vulnerabilities of communities at risk of (and engaged in) forced migration and to support people engaged in and affected by mobility as a form of adaptation to climate impacts. This heterogeneity may require policy and practice to be broad and responsive enough for the context of rapid change and uncertainty so that narrow views about what the future may bring do not cause further harm or reduce the range of adaptive possibilities (Mertz et al 2009, Philibert et al 2013).

The heterogeneity in the literature essentially means that there are strict limits on what can be synthesized. There is a need to further develop evidence-synthesis methodologies that can better deal with diverse bodies of evidence, especially in light of growing transdisciplinary research. Accordingly, the policy implications of this assessment need to be cautiously interpreted with these limitations in mind. In sum, we interpret the evidence to indicate that policy and action should be responsive enough to protect health—and health determinants, especially food and water security—regardless of the climate scenario. Such policies may include those which protect the human rights of migrants, ensure food and livelihood security, support sustainable agriculture, and mobilize remittances to benefit health (table 4: 3, 4, 15, 17, 24, 28, 34, 44; IOM 2020).

The heterogeneity inherent in the current evidence base means that climate hazards can both increase and restrict mobility (Behr and Diaz 2013, Etzold et al 2014, Adams 2016), and that mobility responses can both benefit and compromise health. The included studies indicate that climate hazards can impact areas within a country differently, as well as communities within a region, households within a community, and even individuals within a household. Yet this complexity must not lead to inaction. This finding should rather prompt researchers, practitioners, and policymakers to acknowledge how multiple drivers of migration interact with climate hazards to contribute to causal chains leading to different mobility patterns (Future Earth 2019), and to seek to protect and promote the health of affected populations'.

The third finding of this assessment reveals that health can be a precursor, driver, or outcome of migration. For example, climate change-related health threats can act as mobility drivers as people seek to move to sites of reduced risk and/or in hope of better health and livelihood (Islam et al 2014, van der Geest et al 2014, Mcelfish et al 2016, Grecequet et al 2017).

Alternatively, health risks and diminished mental well- being can emerge among migrant populations en- route and in destination sites (Tschakert et al 2013, Molla et al 2014b, Loevinsohn 2015, Perez-Saez et al 2017).

This finding reinforces Schutte et al's (2018) conceptual framework that connects climate change and migration, depicting a bi-directional arrow between health and mobility because migration is a determinant of health and, conversely, health can shape migration decisions (Abubakar et al 2018, Vearey et al 2019). In other words, migration can be both an adaptive response (e.g. a response in part to health risks) and a 'crisis' (e.g. where migration itself contributes to health risks and other vulnerabilities) (Tacoli 2009). Accordingly, health concerns can arise both from preventing or deterring migration as well as via poorly managed migration (Etzold et al 2014, Dinkelman 2017, Vearey et al 2019). As reflected in this assessment where the health of both host and origin communities was examined in the context of climate-related migration.

The findings highlight the following health concerns for people involved in or affected by migration in a changing climate

• (a)  
  Altered distribution of infectious disease (see table 4 study: study 1, 7, 27, 28, 33, 34, 38, 40, 43).

A few studies examined connections between climatic changes, migration, and HIV/AIDS. In India, drought-affected farmers migrating to urban areas were found to have increased risk of contracting HIV and other sexually transmitted infectious diseases (Anupama et al 2016). In Malawi, a positive and non-linear relationship was demonstrated between HIV prevalence and the extent of rural hunger, in the context of a surge of rural-urban migration during drought, particularly by young women (Loevinsohn 2015). This is consistent with contemporary accounts of increased transactional sex in times of famine, with hunger compromising immune function.

Other infectious diseases were also examined. In Bangladesh, a cross-sectional study found a high incidence of malaria, dengue, childhood diarrhea, and pneumonia among members of climate-vulnerable communities; almost half of the respondents had been homeless for more than a month due to floods and cyclones in the preceding ten years (Kabir et al 2016). Another study in Bangladesh comparing people displaced primarily due to environmental change with people displaced mainly due to political reasons, found poor living and sanitation conditions contributed to significant morbidity from infectious disease, particularly diarrhea and pneumonia. The study concluded that people displaced by climate threats in Bangladesh were more vulnerable to disease than people displaced for other reasons. The difference is explained as unequal exposure to poor household environmental conditions (water and air) (Molla et al 2014a). In Nigeria, increased prevalence of malaria, typhoid, cholera, diarrhea, dysentery, influenza, and tuberculosis amongst fishing communities was observed in the context of migration due to sea-level rise and coastal flooding (Oyekale et al 2013). In Bangladesh, hazard related forced migration led to migrant ill-health, as a result of both intense physical labor and poor living conditions. Permanent migrants faced similar problems due to the longer-term hazards of living in urban slums (Penning-Rowsell et al 2013). Another study in Bangladesh found that slum migration in the context of climatic disasters led to an increased risk of health issues from unhygienic and overcrowded living conditions and a lack of access to WaSH. Climate migrants suffered from undernutrition, micro-nutrient deficiencies (increasing their risk of contracting infectious diseases), and notably, different waterborne diseases when compared to the host community (Rahaman et al 2018).

• (b)  
  Non-communicable diseases (NCDs)

In the context of an increasing burden of NCDs globally, it is perhaps unsurprising that they are also a concern for people on the move in the context of a changing climate. Some studies indicate that dietary transitions that migrants experience in destination sites may increase the risk of diet-related NCDs (Ahlgren et al 2014) supporting similar research (Galbete et al 2018, Danquah et al 2018, Boateng et al 2019) (see table 4: study 4, 12, 29, 33, 34). The authors of one study argue that in the Marshall Islands—as climate change undermines the capacity for traditional food cropping, imported foods contribute to health risks, and people migrate to urban centers—the prevalence of chronic diseases will increase (Ahlgren et al 2014). Other studies reveal difficulties migrants face in managing NCDs in the context of climate-related migration with links to barriers in accessing healthcare. This finding builds on recent research on the syndemic of climate change, obesity, and malnutrition (Swinburn et al 2019) and pathways between climate change and NCDs (Savage et al 2019) (See table 4: study 10, 12, 13, 24, 29).

• (c)  
  Psychosocial health issues emerge where displaced people are separated from social networks (Loevinsohn 2015, Dinkelman 2017), face socio-economic deprivations (Hutton and Haque 2003) and experience the degradation of ecosystem services (Tschakert et al 2013) and other fundamental resources that are key determinants of identity such as land ownership and connection to country (See table 4: study 9, 11, 23, 30, 47, 50).

Following floods in Pakistan, a cross-sectional study detected high levels of depression, aggression, and Post Traumatic Stress Disorder (PTSD) among Internally Displaced Persons (IDPs). Risk factors for mental disorders related to conditions in shelters (Atta et al 2013). In Taiwan, elderly people relocated by the government post-Typhoon Morakot, experienced high levels of PTSD symptomatology (Chen et al 2011). Concerning river-bank erosion in Bangladesh, migrants were found to have significantly higher levels of distress than non-migrants, but rather owing to socio-economic deprivation than to displacement (Hutton and Haque 2003). In a Senegalese farming community where migration is used as an adaptive strategy, poor health was raised as a problem during periods with dust storms and prolonged rain. Further 'reduced solidarity' was identified as an indirect impact of adverse climatic changes (Mertz et al 2009). In Ghana, a study that explored migration in the context of climate change found that environmental change can trigger acute feelings of sadness, fear, anger, disappointment, and helplessness (Tschakert et al 2013). In Shishmaref, Alaska, residents experience ice melt, erosion, and habitual flooding without political commitment for permanent relocation. This uncertainty, in addition to the psychological impacts of floods causes significant stress (Wolsko and Marino 2016).

Mental health issues were identified as prominent in situations of forced displacement, particularly in contexts of increased violence. This is of concern given that some research suggests that deviations from normal precipitation and mean temperatures increase the risk of intergroup conflict (Hsiang et al 2013). Whilst the climate change—conflict relation is not observed in every context, there is substantial agreement that some conditions make this relationship more likely, such as agricultural dependence (Koubi 2019), political exclusion, a history of conflict and high levels of poverty (Uexkull et al 2016)

• (d)  
  Access to health care; Mobile populations face barriers to accessing health care. For example, in Northern Russia due to early ice melt and more open water, herders face extra safety risks and delays in accessing healthcare (Amstislavski et al 2013). Access to health care was considered a key relocation outcome for three small climate-vulnerable coastal communities in the Solomon Islands. Residents of one relocated community reported limited access to healthcare services post-relocation (Albert et al 2018). Access to healthcare was also disrupted by the extreme weather events themselves (Coker et al 2006, Behr and Diaz 2013). Studies recommend reducing financial, geographical, and cultural barriers to accessing healthcare (See table 4: study 4, 5, 10, 20, 21, 25, 27, 21). Such a reorientation of health systems has been referred to as 'migrant- sensitive health systems' that systematically incorporate the needs of migrants into health financing, policy, planning, implementation, and evaluation (Villa and Raviglione 2019). Migrant inclusive health systems might consider epidemiological profiles of the migrant population as well as cultural, language, and socioeconomic factors (IOM 2019b). Recommendations to make health systems migrant inclusive and climate-resilient (WHO 2016, Schwerdtle et al 2018) have synergies with general health systems strengthening and goals towards universal health coverage that aims to ensure all individuals and communities receive essential, quality health services from health promotion to prevention, treatment and rehabilitation and palliative care. Apart from being sensitive to migrant experience and healthcare needs, health systems also need to be climate resilient such that they can better anticipate prevent, prepare for and manage climate-related health risks (WHO 2016).

Health extends beyond the remit of the health sector and many sectors, especially those related to food systems and water management, have health determining functions. This is the basis of the 'health in all policies approach'. The fourth key finding is that food and water security (including access to WaSH) represents an important mediator of both mobility responses and health outcomes in the context of a changing climate. The negative health impacts of climate-related migration for mobile populations and sending and host communities can be mitigated by ensuring access to climate-resilient WaSH infrastructure and strengthening food security (See figure 4).

Extensive and growing literature documents the links between climate change and agricultural yield with newer evidence looking at the impact of climate on the nutritional quality of staple crops (Myers et al 2017, FAO 2019). Agricultural yield is a direct driver of food security and is directly tied to both climate change and migration (Afolayan and Adelekan 1999, Feng et al 2010, Massey et al 2010, Mueller et al 2014, Cai et al 2016). Whilst some of this research was captured in this assessment if there was not a sufficient health focus it was not included. That is to say if the focus was on agriculture, economy, or livelihoods as opposed to food and nutrition security or malnutrition in any form.

To illustrate how food and water security are intertwined in the matrix of migrant health in the context of climate change (See figure 4), Adams (2016) describes immobile populations in Peru that, due to temperature extremes, excessive precipitation, abrupt seasonal weather changes, and drought, are exposed to reduced crop productivity and water scarcity leading to human and animal disease. In this case, health impacts exist because populations do not move. Food insecurity is not acting as a migration driver but rather negatively affecting people in place. Another study in Tanzania (Afifi et al 2014), reported a positive relationship between rainfall shortage and out-migration after taking other important demographic and socio-economic factors into account. In this case, food security is the mechanism through which rainfall variability affects human mobility. Whilst people ranked migration as a key coping strategy and said they 'follow the water', migration also negatively affected families where dependents were neglected for long periods (Afifi et al 2014).

Similarly, Heaney and Winter (2016) found that increased drought frequency and duration in Kenya caused food and water shortages that render the Massai incapable of sustaining their livestock, leading to rural-urban migration. Whilst remittances, invested in food and healthcare, benefitted sending communities, the migrants themselves struggled with mental health issues (stress, unhappiness, loneliness) related to social isolation and experienced barriers to accessing healthcare which they used as a last resort and only for physical complaints (Heaney and Winter 2016).

Drought impacts negatively affecting food and water security in Afghanistan have led to social impacts including water-related conflicts, migration, malnutrition, poor health, and a sense of hopelessness and loss (Iqbal et al 2018). In this context, farmers said their religious beliefs helped them withstand long periods of drought and to refrain from committing suicide. Disproportionately affecting low-income farming households, drought also affected food choices with families reporting an inability to meet the goal of consuming nutritious foods (Iqbal et al 2018).

In Vanuatu, migration increased vulnerability in already vulnerable communities where seasonal migration significantly compromised local agricultural production and food security in sending communities (Craven 2015). Other research shows increased reliance on imported foods may increase the risk of diet-related NCDs (Ahlgren et al 2014, Savage et al 2019). While remittances served to finance health, education, and maintain food security in times of crisis, there is a risk that remittances will be insufficient to cover the loss of sustenance food production. Edwards (2013) found food shortages drove relocation and Etzold et al (2014) discovered the most food-secure people did not need to migrate to adapt to the negative effects of rainfall vulnerability.

These studies highlight the differential health impacts on different populations engaged in or affected by climate-related migration. They illustrate how food and water insecurity can contribute to migration decisions and affect health, and also how migration in the context of climate change can reduce food and water security in sites of relocation with flow-on consequences for health. This supports other research that people may move as an adaptive response to climate vulnerabilities including food and water insecurity or conversely, that they may move into sites of higher health risk (Vinke 2019).

Policy responses differentiated by mobility characteristics with consideration to populations at origin as well as mobile and host communities may mitigate the negative health impacts or at least not contribute to them further.

The fifth finding of this systematic assessment relates to the variable degree to which climate, meteorological and environmental data were integrated into the individual studies to link the migration and health phenomenon to climate change (see table 3). Based on the included studies, the degree to which this data was integrated is categorized into three approaches;

• (a)  
  The study referred to weather or climate change in the narrative, and/or the environmental event can be linked to climate impacts (e.g. hurricane, rainfall variability), but did not display any meteorological data (See table 4: studies 2, 4, 6, 9–13, 15–21, 23, 25, 28, 29, 31, 33, 34, 36, 38–41, 43, 45, 46, 48–50).
• (b)  
  The study extracted meteorological data and conducted their analyses describing weather (short period, or point in time) or climate information independent of migration or health data (data over several years and aggregated) (See table 4: studies 1, 3, 5, 7, 8, 26, 30, 32, 35, 42, 44, 47).
• (c)  
  The study extracted own meteorological data and overlaid and analyzed climate data (for example, as a comparison between drought vs. non-drought years) with migration and/or health data. (See table 4: studies 14, 22 24, 27, 37)
• (d)  
  Whilst the degree of climate and environmental data integration required depends on the specific research question, the variability raises several questions, including the amount of climate data that are required to produce robust evidence linking climate impacts to human migration (and health), how these data are accounted for in evidence synthesis and the basis on which to link data demonstrating environmental changes to anthropogenic climate change (Rosenzweig and Neofotis 2013, Ebi et al 2017).

Whilst there are good examples of robust integration of climate data into mobility and health data (Hsiang et al 2013, Burke et al 2015) to the authors' knowledge, there is no recognized matrix or accepted framework to guide researchers and maintain quality in climate, migration and health research. There is a need to further refine and build upon the three approaches outlined above and to develop a typology to guide future climate change, health, and migration research. As recognition of the impacts of climate change on health expands, there is a risk that quality will be compromised if public health researchers superficially link their findings to climate change without carefully delineating weather from climate and natural variability from climate change and examining the causal evidence.

We find there are diverse methodological approaches to studies of the climate-migration-health nexus. In-depth, qualitative case studies featured prominently in this assessment and align well to this research field considering mobility decisions are influenced, among a range of complex and interacting factors, by experiences and perceptions. Yet, due to the methodological complexity of examining human and natural systems over different temporospatial scales, the question arises as to what design would yield the 'best available evidence' for this emerging field of nexus research? Choice of method will be determined by the research questions and context yet there will be value in; quantitative survey design, for example, comparing health outcomes of mobile with non-mobile households in sites of climate risk (See table 4: study 7, 13, 17, 18, 20, 23, 26, 47); analysis and synthesis of large-scale migration, health and environmental datasets (See table 4: study 14, 15, 21, 37, 49); and qualitative methods that seek to understand site-specific migration drivers and health experiences, including in smaller community settings (See table 4: study 13, 18, 20, 29, 30, 31, 32, 43, 47, 50).

Innovative, transdisciplinary approaches are also required such as incorporating climate data into routine epidemiological surveillance systems, early warning and risk assessment for climate-sensitive diseases, and other types of disasters, epidemics, and nutritional crises. Innovative tools and methods would seek to strengthen data management, observations, develop climate-relevant services (Thomson and Mason 2019), and fill existing migration and health data gaps that constrain evidence-based policy and practice. Apart from innovative approaches, there is also a need to enable this type of research in the funding environment for example, rapid financing of research to track and understand mobility before, during, and after climate hazards and stressors. A large proportion of this research is retrospective which has clear limitations particularly in terms of recall bias.

Some obstacles preventing research engagement in this area may be that the standard epidemiological methods are not fit for purpose for the complexities of health and migration in the context of a changing climate. This leads to difficulties in identifying causal links between climate change, mobility responses, and human health. And yet, as identified by the Global Compact for Migration (UNHCR 2018a), there is a need to collect and use accurate data to inform evidence-based policy. Longitudinal migration and health data such as that from the Health and Demographic Surveillance System (HDSS) could be further employed by transdisciplinary teams to answer important questions about climate-migration-health connections and their policy significance. Linking back to the motivation for this synthesis research, international agreements and frameworks such as the Global Compact for Migration and the SDGs would be strengthened by strategic research to address climate-migration-health knowledge gaps and build on the findings of this systematic literature assessment.

Health outcomes in the context of human mobility and climate change are context-dependent and are shaped by diverse economic, environmental, political, social, and demographic factors. Broad systems thinking is required that moves beyond push and pull factors and the linear origin and destination understanding, and rather interrogate complex causal chains, triggers, thresholds, health consequences, and policy and practice significance (Future Earth 2019). Research that addresses internal and rural-urban migration, as well as migration into sites of climate risk (D'Amato et al 2011, Vinke 2019), will be useful as these are dominant patterns of migration at the current 1 degree of warming. It will be of value to go beyond considering vulnerable populations and health risks, but also collect data on resilience and adaptation that support health and wellbeing for those on the move in a warming world.

The strengths of this research include blinded study selection and quality appraisal as well as the extensive search in four literature databases. In terms of limitations, the inclusion of English and German studies published in academic journals may have led to selection bias.

Concerning study design, conventional systematic literature reviews (SLRs) are predominantly used to analyze quantitative empirical data in the health sciences with a focus on randomized controlled trials (Berrang-Ford et al 2015). There are very different perceptions of what constitutes an SLR in social, health, and environmental sciences, and definitions of minimal standards are contested. However, there is consensus that one size does not fit all (Berrang-Ford et al 2015).

We undertook a focused review of the literature that sought to answer a specific research question using pre-defined eligibility criteria for documents (Berrang-Ford et al 2015). We followed standard methodological steps: (1) define the research question and scope of the study, (2) document selection, including development of inclusion and exclusion criteria, (3) critical appraisal of study quality, (4) analyze and synthesize evidence, quantitative and/or qualitative, and (5) present results (Petticrew and Roberts 2006, Higgins and Green 2008, Barth and Thomas 2012). We called this research a systematic literature assessment due to the need to adapt conventional methods to answer the research question and to account for the broad types of studies captured. Some choices were made that arguably make this secondary research less like a standard SLR such as the exclusion of grey literature, combining heterogeneous studies in a meta-synthesis and the fact that we cannot report on publication bias, due to the inclusion of heterogeneous studies that used different outcomes and different effect measures.

The decision to exclude grey literature may have led to the omission of some important findings yet peer-reviewed studies aligned better with the research aim and specific objectives to synthesize empirical research. Primary studies within grey literature reports should have been captured by the search strategy.

Ambiguity in reporting within eligible studies may have led to over or underestimation of methodological quality. Caution is needed in interpreting these studies because secondary data were often used with other objectives and because some studies used small sample groups (especially qualitative case studies) with findings that could not be generalized to larger groups. The three modeling studies (See table 4: studies 1, 19, 40) were not assessed for quality because it was not possible to use MMAT for modeling studies.

Although the criteria for inclusion in terms of climate change are clearly presented, many eligible studies suggest potential links to climate change rather than demonstrate the link with climate data. Apart from a few particular examples (e.g. Perma-frost melt, Alaska) other environmental risks are assumed to be linked to climate change (e.g. river-bank erosion, cyclone, flood). Whilst the likelihood of their occurrence and/or severity may increase with climate change, other factors may be responsible for or contribute to the climate hazard such as natural climate variability or development activities.

Although the inclusion of all three elements of the climate- migration-health nexus was required for inclusion in this assessment, some studies still focused on a 'dyad' within the nexus, whereby climate change was considered, yet either migration or health was underexplored. Subsequently, extractable data for these underexplored elements was at times inadequate and required interpretation. This limitation was unavoidable, in an assessment that aimed to use systems thinking in a research environment that is essentially still in operating in silos.

This assessment identified some well-represented knowledge clusters that are amenable to synthesis via systematic review including the health impacts of climate-related forced displacement in Bangladesh and the characteristics of effective planned relocation governance. Underrepresented areas of research that would benefit from further primary research include the health considerations for trapped populations (Hirvonen 2016), the mental health impacts of climate-related migration, and the gendered health impacts of climate-related migration. Further research could analyze the question of quality in terms of climate data integration in climate change-related public health research. Finally, research could examine how the syndemic of climate change, obesity, and undernutrition (Swinburn et al 2019) interacts with non-communicable disease management and migration.