Psychological impairment and extreme weather event (EWE) exposure, 1980-2020: A global pooled analysis integrating mental health and well-being metrics

Extreme Weather Events (EWEs) impose a substantial health and socio-economic burden on exposed populations. Projected impacts on public health, based on increasing EWE frequencies since the 1950s, alongside evidence of human-mediated climatic change represents a growing concern. To date, the impacts of EWEs on mental health remain ambiguous, largely due to the inherent complexities in linking extreme weather phenomena with psychological status. This exploratory investigation provides a new empirical and global perspective on the psychological toll of EWEs by exclusively focusing on psychological morbidity among individuals exposed to such events. Morbidity data collated from a range of existing psychological and well-being measures have been integrated to develop a single ( “ holistic ” ) metric, namely, psychological impairment. Morbidity, and impairment, were subsequently pooled for key disorders, specifically PTSD, anxiety and depression. A “ composite ” (any impairment) post-exposure pooled-prevalence rate of 23% was estimated, with values of 24% calculated for depression and ⁓17% for both PTSD and anxiety. Notably, calculated pooled odds ratios (pOR = 1.9) indicate a high likelihood of any negative psychological outcome ( + 90%) following EWE exposure. Pooled analyses of reported risk factors ( p < 0.05) highlight the pronounced impacts of EWEs among individuals with higher levels of event exposure or experienced stressors (14.5%) and socio-demographic traits traditionally linked to vulnerable sub-populations, including female gender (10%), previous history (i.e., pre-event) of psychological impairment (5.5%), lower socio-economic status (5.5%), and a lower education level (5.2%). Inherent limitations associated with collating mental health data from populations exposed to EWEs, and key knowledge gaps in the field are highlighted. Study findings provide a robust evidence base for developing and implementing public health intervention strategies aimed at ameliorating the psychological impacts of extreme weather among exposed populations.


Introduction
Extreme Weather Events (EWEs) have a substantial global public health and socio-economic burden, equating to an estimated 6 million deaths and 50 million injured since 1950, with economic losses of US $640 billion between 1970 and 2019 also projected (CRED, 2020;IFRC, 2020;WMO et al., 2020).The myriad interactions between EWEs and human health are complex, with impacts generally classified into two main categories: (i) direct, due to finite physical climatic manifestations (e.g., storms, floods), and (ii) indirect, often caused by changes in biogeophysical processes influenced by climatic phenomena (e.g., water quality, land-use change) (Watts et al., 2015;Forzieri et al., 2017).These two impact categories actively interact with social factors (e.g., demographic profile), and potentially modify (e.g., amplify, mitigate) the intensity of subsequent impacts, ultimately influencing mental health and well-being (Watts et al., 2015).
Current evidence suggests that the frequency and intensity of EWEs has increased significantly since the 1950s in concurrence with a ⁓0.8 • C global temperature rise (IPCC et al, 2021).While evidence explicitly linking anthropogenic climate change with EWE frequency and/or intensity varies with respect to phenomena type, indicators suggest human-mediated global warming has likely resulted in an increase in compound EWEs since the 1950s (IPCC et al, 2012;2018;2019;2021).A recent report from the Centre of Epidemiology of Disasters (CRED) indicates a ten-fold increase in the number of climate-mediated disaster events recorded in the last ⁓70 years (CRED, 2020), with events associated with a climatic origin accounting for ⁓79% of all disaster typologies (e.g., technological, conflict-related) over the last 50 years (IFRC, 2020).
Since 2016, an annual, global, multi-disciplinary effort to track the links between climate change and public health, including monitoring of key "progress" indicators, has been led by The Lancet Countdown (htt ps://www.lancetcountdown.org/), a taskforce seeking to provide policy-makers with evidence-based feedback relating the impacts of climate change on public health.While initially identified as a key emerging public health concern, a series of limitations have precluded successful integration of climate-associated mental health indicators into progress metrics (Watts et al., 2015(Watts et al., , 2018(Watts et al., , 2019(Watts et al., , 2020)).Similar constraints apply to the integration of EWEs and mental health fields, with main obstacles ranging from ambiguity regarding attribution (i.e., cause-effect, psychological disorders potentially exhibiting compounded and distal origins), to the inherent complexity of psychological disorders, including co-morbidity, and symptom variability as a product of "resilience" and life-course epidemiology (Kuh et al., 2003;Goldman and Galea, 2014;Watts et al., 2018).Additional constraints include under-reporting and differing diagnostic standards depending on geographical location and socio-demographic background (Watts et al., 2015(Watts et al., , 2021;;Berry et al., 2018;Hayes et al., 2018;Habrok et al., 2020).Accordingly, to date, much of the research emphasis in the context of EWEs and mental health has been placed in exploring associations between temperature extremes and metrics associated with mental health at a (macro-) population level (Watts et al., 2019).While informative, this approach precludes identification of direct links between individual exposure and mental health outcomes.This lack of research focus contrasts with mounting evidence for increased risk of psychological impairment in response to extreme weather, potentially amplified by increasing event frequency, duration and intensity, and concurrently, projected estimates of a substantial, widespread, and cumulative psychological burden (Trombley et al., 2017;Obradovich et al., 2018;Clayton 2020Clayton , 2021;;Liu et al., 2021).
Notwithstanding the aforementioned difficulties, key quantitative insights into the psychological burden associated with EWEs may be obtained from the disaster-psychopathology literature.Hydrometeorological events, which can be tentatively associated with climatic change (e.g., floods, storms, droughts), are frequently linked with an increased risk of developing psychopathological disorders (Bourque and Willox, 2014;Trombley et al., 2017;Habrok et al., 2020;Palinkas and Wong, 2020).However, to date, much of the disaster literature has focused in evaluating PTSD, with a need for a more "holistic" approach aiming to incorporate different mental health metrics (Goldman and Galea, 2014).As such, mental health morbidity data derived from an iterative, pooled or meta-analytical approach, which aims to condense psychological/well-being data in response to extreme weather events, may prove particularly insightful.As of yet, no empirical literature review has focused exclusively on EWEs at a global scale, with available studies often regionally-specific, adopting an overarching approach towards "public health impacts" (both physical and mental) or the concept of disaster and thus including non-climatic natural events (Rubonis and Bickman, 1991;Norris et al., 2002;Galea et al., 2005;Neria et al., 2008;Rataj et al., 2016;Lowe et al., 2019;Cruz et al., 2020;Weilnhammer et al., 2021).
Within this context, the current study represents the first empirical attempt to collate, integrate, and analyse psychological morbidity data from populations exposed to EWEs.Key geographical and socioeconomic factors interacting with (modifying) the cause (EWE) and effect (psychological disorder) relationship are extracted and analysed, thus enabling a greater understanding of the associations between extreme weather exposure and mental health disparities and inequity.Ultimately, study findings seek to provide an evidence base for policymakers and other stakeholders for designing and improving intervention and/or mitigation strategies, aimed to guide resource allocation efforts before and following extreme weather events.

Review Scope and bibliographic databases
Given the exploratory and multi-disciplinary scope of target literature, the literature identification protocol employed was adapted from the Population/Concept/Context (PCC) Framework for "scoping" literature reviews (Peters et al., 2020).Additionally, a range of published scoping reviews and pooled analyses, focusing on topics related to public health, (social-)epidemiology, and disaster-psychopathology (e.g., Rubonis and Bickman, 1991;Norris et al., 2002;Galea et al., 2005;Sargeant et al., 2006), were consulted to inform the review process.The following research question was developed to direct the literature identification process: "What was the global prevalence of psychological impairment among populations exposed to Extreme Weather Events during the period 1980-2020 and what risk factors were associated with impairment?" Literature searches (conducted July 1st, 2020) were confined to Scopus, Web of Science and PubMed bibliographic databases.Notably, the review was potentially restricted in scope in terms of database searches, with an exclusive focus on the three databases deemed to be most pertinent in the context of EWEs.Additional and potentially relevant databases (e.g., PsycInfo) were omitted from the review protocol.Employed literature search terms (Table A1; See Appendix) followed pre-established classifications derived from the Population-Agent-Outcome Model (PAO) model used for previous scoping reviews (e.g., Hynds et al., 2014).Search terms relate to the two primary (in-)direct impact classifications of EWEs (e.g., direct = "hurricane"; indirect = "water quality"), and the primary psychological outcomes associated with population exposure to EWEs (e.g., PTSD, anxiety, depression).All database searches used Boolean positional operators (e.g., "AND", "OR").

Eligibility criteria, article Screening and data extraction
A total of 1218 records were identified through bibliographic database searches with de-duplication reducing this to 923 (Phase 1-2; Fig. 1).All records were initially screened for suitability based on article title and abstract content, with forward-selected articles subject to fulltext screening (n = 321) (Phase 2-3).At each stage, article screening and inclusion followed a set of pre-established suitability criteria (Table A2).A senior postdoctoral fellow led the review (CC).PH and JOD independently assessed all abstracts for suitability/relevance, where a disagreement arose, the authors (CC, JOD, PH and SL) conferred to reach agreement (i.e., majority vote).A central objective of the review was to quantify the psychological and well-being impacts of EWEs on exposed populations.The concept of EWE is inclusive of extreme weather and climatic events, which can be grouped under the term "climate extremes", and follows IPCC terminology, i.e., an abnormal, above-below threshold, and (temporally) irregular weather/climatic phenomena (IPCC et al, 2012).Exposure was predicated on respondent direct (or "lived-in") EWE(s) experience.Importantly, exposure was dependent on individual study design, with a degree of spatio-temporal ambiguity in relation to individual EWE exposure, particularly in the context of large-scale and distance-based studies (e.g., online, mail surveys).Event types were grouped into four main categories to facilitate analysis with further details provided in Appendices B-C.Specifically, "quantification" refers to a psychological diagnosis (individual case/outcome), and/or evidence for well-being impairment (or lack thereof) among evaluated population samples.Both the term psychological impairment (cf.Rubonis and Bickman, 1991), and the data presented within, incorporate a range of pre-established (non-)clinical and study-specific (or "generic") psychopathological, mental health and well-being measures.Thus, literature inclusion was not constrained to investigations employing standardised psychological disorders (e.g., DSM, ICD-10).The underlining criteria and rationale employed for article inclusion and established data extraction field (sub-)categories are outlined in Appendices B-D.

Quantifying psychological impairment
All included investigations provided a number or percentage of the population meeting predetermined criteria deemed sufficient to attribute psychological impairment.Thresholds were determined and followed the discretion of individual authors.In an effort to integrate all data into a single "composite" metric representative of (overall) psychological impairment, reported psychological disorders were condensed into four nosological domains outlined by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV, 4th Edition) (viz.Rubonis and Bickman, 1991), as follows: (i) PTSD (including other "stress" measures), (ii) Depression, (iii) Anxiety and (iv) Substance Use Disorder.Additionally, a fifth domain classification, denoted as "Additional Distress", was used to amalgamate disorders (and domains) not frequently encountered (n < 3) among identified studies (e.g., Somatization, Schizophrenia).The outlined approach enabled calculation of one composite and five domain-specific estimates of psychological impairment with emphasis given to data estimated at a composite level and for PTSD, anxiety and depression; the latter representative of the three main psychopathological domains in the disaster-psychopathology literature (cf.Norris et al., 2002;Neria et al., 2009;Goldman and Galea, 2014).In the case of longitudinal studies, i.e., investigations based on psychological evaluation at multiple points (post-event) in time (or waves), all impairment data extracted and pooled were restricted to the first evaluation.Additional details on the extraction of impairment data and pooling categories are provided in Appendices B-E.

Reported risk factors
Where reported, all variables (i.e., potential confounders and/or modifiers) statistically associated (p < 0.05, irrespective of employed statistical test/method) with psychological impairment within C. Chique et al. individual studies were extracted and collated.Due to the large (cumulative) number of risk factors investigated across studies, where possible, risk factors were nested into one of three main risk categories commonly employed in the disaster-psychopathology literature (Maguen et al., 2009;Goldman and Galea, 2014), namely "pre-event", or intrinsic risk factors (primarily socio-demographics); "peri-event", or variables pertaining (and confined to) a specific event and its timeframe (e.g., perceived intensity, stressors); and "post-event", i.e., variables associated with conditions and settings following exposure to extreme weather (e.g., access to aid/relief).

Study-Specific and pooled odds ratio calculations
Odds ratios (ORs) were calculated as a relative measure to evaluate the strength of association between EWE exposure and psychological impairment.Study-specific ORs were estimated for each domain including composite data, with a "pooled" (or adjusted) OR (pOR) calculated using the Mantel-Haenszel method (Mantel and Haenszel, 1959).pOR, which represents a weighted metric for aggregating study-specific (non-)exposed populations and standard deviations, provides an overall measure of association between EWE exposure and specific domains.OR calculation relied on availability of case-control data and was therefore restricted to those investigations employing independent control groups (n = 12).Only one investigation reporting "actual pre-post" (i.e., pre-and post-event) control data met the criteria allowing OR calculation.This investigation was treated separately from those employing independent control groups in terms of pOR calculation.Investigations reporting retrospective pre-post data were also excluded from calculations.Subject to adequate data reporting, ORs were calculated for studies reporting composite (n = 7) and domain-specific impairment (n = 8).The latter was only calculated for the three main disorder domains, i.e., PTSD (n = 7), anxiety (n = 7) and depression (n = 8).
Cross-sectional study types, i.e., investigations based on evaluation at a single point in time, dominated the dataset (51/59; 86.4%).Among the few longitudinal studies identified (8/59; 13.6%), a majority were confined to only two waves of psychological evaluation (7/8; 87.5%).Importantly, the reported time lag between EWE exposure and (postevent) psychological evaluation was highly comparable between investigations employing cross-sectional and longitudinal study designs (Appendix F).While most investigations were based on nonrepresentative sample selection criteria (34/59; 57.6%), i.e., a study sample not reflective of the general population, several incorporated representative (or random) sampling designs (26/59; 44.1%) (Table A3).Diagnostic methods generally consisted of self-reported measures, accounting for 71.8% of the total examined population, with clinically established cases less common (23.5%).A substantial number of investigations lacked a control population (41/59; 69.5%) with just 12/59 (20.3%) studies employing independent control criteria, while 7/59 (11.9%) reported pre-and post-event impairment data.The latter included four studies reporting retrospectively acquired pre-event data.Overall, 75.2% (n = 45,576) of respondent impairment data lacked any control parallel.

Estimates of psychological impairment
Comprehensive summaries of pooled psychological impairment data, i.e., morbidity data extracted and collated (specifically) at a subcategory level, for both composite (any impairment) and domainspecific categories are provided in Table 1-3Summary statistics for key (sub-)categories are presented in Figs.3-4, with pooled data at a national level presented in Table A4.Additional impairment data are presented separately in Table A5.Overall, a composite (post-event) psychological impairment rate of 23.2% (10,052/43,385) was estimated (Table 1).However, a considerable proportion of investigations failed to provide sufficient data to derive composite impairment (20/59; 33.9%).Lack of reporting of co-morbidity data among studies employing multiple measures represented a recurrent limitation in the calculation of composite values.Highest rates of impairment were estimated for depression (24%), with lower values calculated for PTSD and anxiety (⁓ 17%) (Table 1).Compiled data also enabled the determination of composite values solely based on standardised psychopathology (e.g., DSM), with a composite (i.e., PTSD, anxiety, depression) prevalence rate of 18.6% (5889/31,613) calculated.Estimated psychological impairment exhibited marked variability among identified studies, with the largest range reported for PTSD (0.7%-100%) (Table 1).
Populations exposed to events classified as "wildfires" (35.1%) and "drought/heat" (31.3%) exhibited the highest composite prevalence rates (Fig. 3).Respondents experiencing events within the "storm" and "flood" categories exhibited comparable prevalence rates (20.4%-23.8%).Populations exposed to "multiple" (more than a single event) weather events also reported higher composite prevalence (30.9%) in comparison to those experiencing single ("one-off") events (18.6%) (Fig. 3).Temporally, composite impairment rates exhibited a decadal monotonic increase (11.1%-39.5%;cf.event year in Table 2).Higher values were also observed for individuals evaluated within one-month post-event, particularly at a composite level (49.6%) (Tables 2-3).The estimated recency and relevance of shorter-term sequelae may be supported by a higher prevalence observed among individuals recruited from relief centres (81.9%), with diagnoses in these settings generally generated within a two-month window (Tables 2-3).

Reported risk factors
As shown (Table 4), over half of reported risk variables occurred within the pre-event risk category (156/290; 53.9%).Here, female gender was most frequently associated with risk of psychological impairment (29/156; 18.9%).Specifically, female gender was often linked with increased occurrence of PTSD (15/73; 20.5%) and anxiety (6/23; 26.1%).Additional (frequent) risk factors included lower socioeconomic status and reports of previous mental health symptoms/disorders; each accounting for 10.3% (16/155) among pre-event risk factors.Within the peri-event category, event exposure was the variable most commonly associated with any type of psychopathology (42/69; 60.9%) with levels of fear/perceived threat and (in-)direct experience of physical injuries or somatic conditions (i.e., personally or through family members) identified as common risk factors for PTSD (Table 4).Event exposure was also the most important risk factor across the three adopted risk typologies (42/257; 14.5%) (Table 4).In relation to postevent factors, respondents experiencing property damage/loss and financial stress were most at risk of developing psychological impairment (⁓ 20%).

Pooled odds ratios
Odd ratios calculated from reported composite data are provided in Fig. 5, with domain-specific estimates presented in Figs.6-8.Key

Table 1
Estimated psychological impairment values at composite and domain-specific levels.The total population subject to psychological diagnosis (N) and corresponding number of positive cases (n) per category are also provided. 1 Values excluding data derived from non-standardised mental health measures contained within the "Additional Distress" domain.*Estimates presented are subject to accurate (full) study co-morbidity reporting and thus differ from domainspecific values.characteristics for investigations incorporating control groups are also provided in Table A6.Inconsistent reporting of co-morbidity data precluded (composite) OR calculation in 5/12 (41.7%) studies (Table A6).
Similarly, at a domain-specific level, two investigations failed to provide morbidity data necessary for pooled OR calculation.
Overall, the odds of developing any psychological impairment were approximately 90% higher among individuals exposed to extreme weather events in comparison to control populations (pOR = 1.9;CI = 1.7-2;Fig. 5).At a domain level, odds were highest for PTSD development (pOR = 4.8), with a population sample exposed to hurricanes (cf.Kar et al., 2007;Tucker et al., 2017) reporting the highest probability of PTSD development (OR = 7.4-7.8)(Fig. 6; Table A6).Similarly, pORs indicate exposed populations were approximately twice as likely to exhibit depressive symptomology (pOR = 2.04) (Fig. 7).Anxiety was the only domain exhibiting a negative calculated association with EWE exposure (pOR = 0.6) (Fig. 8).Here, it is important to note that estimates are particularly influenced by control group data provided from Brown et al. (2019) with higher odds of anxiety diagnosis in contrast to those exposed to extreme weather (Fig. 8).Brown et al. (2019) reports data from school children exposed to wildfires, with an explicit study caveat represented by introduction of a post-event (and pre-measure) school-wide student mental health support programme.As such, reported anxiety data were possibly influenced by a successful intervention.

Composite estimates of psychological impairment and odds ratios
To the authors' knowledge, this represents the first global collation and empiric integration of mental health and well-being data exclusively from populations exposed to extreme weather events.Calculated ORs indicate a high probability (+90%) for development of any form of psychological impairment following exposure to weather/climatic phenomena (pOR = 1.9;Fig. 5), providing strong evidence for the detrimental effects of EWEs on human health.Notwithstanding, estimates presented, particularly pertaining pooled data, need to be interpreted cautiously due to a number of inherent methodological limitations which are discussed in detail in the limitation and recommendation section below.
Accounting for potential limitations, comparisons of prevalence data with available mental health "baselines" may prove particularly insightful in ascertaining the relevance of pooled estimates presented.For example, data recently curated by the Institute for Health Metrics and Evaluation (IHME), indicate a global prevalence rate for mental disorders of ⁓ 13% (GDB, 2019).This is considerably lower than estimated composite impairment rates within the current study (18.6%-23.2%)(Table 1), representing the potential severity of EWEs on mental health.Likewise, global (12-month) prevalence rates for mental disorders provided by Steel et al. (2014) (⁓ 17%), may be indicative of a  Total n (%) 137 ( 100) 42 ( 100) 64 ( 100 analogous with reported "general" population estimates of psychological impairment following EWE exposure (⁓ 20%) (Clayton 2021), supporting the presented prevalence rates.Following WHO baselines for disorder prevalence following natural (i.e., (non-)climatic) disasters in general (6%-11%) (Berry et al., 2010), the estimated composite rates are indicative of (i) a significant impact which can be specifically attributed to extreme weather, or (ii) an underestimate of previous baselines.In either case, these highlight the relevance of the presented prevalence data, in improving our understanding of the psychological burden of EWEs.

Psychological impairment and domains
PTSD is the only assessed psychological disorder requiring exposure to a trauma, and thus is less likely to be methodologically constrained in terms of population, exposure level, and background (pre-event) impairment (Lowe et al., 2019).This conveys a degree of support for pooled PTSD estimates (Tables 1-3); comparisons with reported  (background) epidemiological figures may prove more particularly pertinent in the context of PTSD.For example, comparison of PTSD figures (17.7%) with general population estimates (2%-10%) (Kessler et al., 1995(Kessler et al., , 2005;;Galea et al., 2005;Atwoli et al., 2015) indicate significantly increased rates of PTSD following EWE exposure, with this finding further emphasized by high estimated ORs (pOR = 4.8; Fig. 6).
While PTSD featured as the most frequently encountered psychopathological measure, depression exhibited a higher prevalence (24%) among pooled cohorts (Table 1).Observed trends concur with recent findings reported by Lowe et al. (2019), with depression cases generally higher than PTSD among 20/26 (76.9%) peer-reviewed investigations focusing on populations exposed to "environmental" disasters.Once again, comparison with epidemiological data, which have placed the global prevalence of depression disorders between 3% and 12% (Kessler et al., 2009;Steel et al., 2014;GBD, 2019), are indicative of an elevated (domain-specific) burden linked with EWE exposure.Nonetheless, depression is one of the most prevalent mental health disorders among the general public, and consequently, the possibility of empiric inflation by pre-event morbidity should be highlighted (Kessler et al., 2012;Goldman and Galea, 2014).For anxiety disorders, potential baselines of 4%-11% have been reported among the general public (Kessler et al., 2009;Steel et al., 2014;GBD, 2019), thus indicating a potentially significant impact within the context of EWEs (17.3%;Table 1).

Socio-economic, geographical and temporal trends
Analyses indicate two main regional trends, namely, higher impairment among residents of low-income regions at a composite level (47.8%;Fig. 3), and conversely, a higher proportion of EWE-related anxiety (19.4%) and depression (21.1%) in high-income settings (Table 3; Fig. 4).In low-income regions, a combination of key drivers including high poverty levels, increased exposure to extreme weather, and restricted access to "recovery" resources are likely associated with higher risks of psychological impairment following EWEs (McFarlane and Williams, 2012;Rataj et al., 2016;Berry et al., 2018;Morganstein and Ursano, 2020).The relevance of lower socio-economic status, as inferred from prevalence data obtained from investigations based on low-income countries, is also reinforced by its recurrence among (pre-event) risk factors identified in the literature (Table 4).Importantly, associations between low socio-economic background, extreme weather, and a higher risk of psychological impairment, were reported in both high-and low-income settings (e.g., Galea et al., 2007;Bandla et al., 2019) highlighting its importance at (inter-)national scales irrespective of regional income.
The higher prevalence of anxiety and depression disorders in highincome settings mirrors recurrent trends (Kessler et al., 2009;Witcchen et al., 2011;Kessler and Bromet, 2013), linked to typically "western" cultural traits (Koplewicz et al., 2009).For example, Kessler et al. (2005Kessler et al. ( , 2012) ) estimate epidemiological anxiety baselines between 18% and 25% in the USA.These values are considerably higher than calculated estimates (Table 1), emphasizing the potential bias introduced by regional trends.However, calculated ORs did not demonstrate any particular geographical trends (Figs.5-8; Table A6), with the caveat that only one investigation reported control data from a typically non-Western setting (Felix et al., 2011).Anxiety was the only domain exhibiting a negative (OR) association with extreme weather exposure (Fig. 8), this highlighting the complexity of interpreting mental health data.
Globally, Asia consistently experiences the highest proportion of major hydro-meteorological hazards, including (event) incidence, associated mortality and economic loses (CRED, 2020;IFRC, 2020;WMO et al., 2020).This was reflected in the current study, with Asian populations exhibiting the highest pooled psychological impairment rates (Table 1 and A5; Fig. 3).Similarly, Asian studies were characterised by the highest composite ORs (Fig. 5; Table A6).This has been attributed to elevated regional population density, with the potential to exacerbate the effects of EWE exposure (Watts et al., 2015;Stephenson, 2008;Forzieri et al., 2017;IFRC, 2020).In this context, presented data may demonstrate the reported (cumulative) spatio-temporal impacts of EWEs at a global scale (IPCC et al, 2012;Goldman and Galea, 2014;WMO et al., 2020), with a higher event frequency and gradual (global) population expansion reflected in the monotonic temporal increase of prevalence rates (cf.event year intervals; Tables 2-3).However, the latter may also be an artefact of improving psychopathological diagnoses resulting in increasingly reliable case detection (Aboraya et al., 2006).Identified trends also provide key insights into (post-event) psychological impairment and the effects of recency, with marked impacts among cohorts with immediate (<1 month) EWE exposure (49.6%;Table 2).The importance of short-term impacts, which are also reflected via higher impairment among respondents from relief centres (81.9%;Table 2), highlight the need for swift and efficient mental health intervention following EWEs.Collated data tend to concur with the literature suggesting psychological disorder symptomology typically peaks within 12 months following EWE exposure (Table 2) (Goldman and Galea, 2014).Conversely, some investigations did identify long-term sequelae (>16 years) (Thordardottir et al., 2016;Dai et al., 2017), indicating efforts towards providing long-term recovery resources also merit consideration.However, investigations evaluating cohorts over extended timeframes were relatively rare (4/59; 6.9%), representing a critical focus area for future research.
A higher prevalence rate among respondents from rural settings (38.8%;Table 2) points to the influence of "location" on mental health outcomes.Several investigations have focused on the impacts of EWEs in rural settings (e.g., Morrisey and Reser, 2007;Berry et al., 2008;Vins et al., 2015;Ellis and Albrecht, 2017;Hayes et al., 2018;Hrabok et al., 2020).Here, a number of archetypal factors of rural populations, linked to community culture and health perception, and which are often compounded by limited access to aid, may lead to impaired health-seeking behaviour and (post-event) adaptation capacity.These include rural stoicism, self-reliance, and prevailing stigmas associated with psychological disorders (Morrisey and Reser, 2007;Allen et al., 2012;Vins et al., 2015;Hrabok et al., 2020).Furthermore, isolation or lack of awareness around (or the benefits of) recovery resources is prevalent in rural and remote locations (Berry et al., 2008;Vins et al., 2015).Pooled data suggest rural populations are particularly prone to PTSD (Table 3), potentially underscoring issues with access to immediate emergency relief and/or mental intervention following EWEs, and/or impaired social support (due to isolation) in rural or remote locations (Goldman and Galea, 2014;Vins et al., 2015).Reported results contrast with baseline epidemiological evidence indicating a higher psychological burden in urban areas (Peen et al., 2010;Allen et al., 2012), suggesting rural inhabitants are particularly vulnerable to the impacts of extreme weather.

Socio-demographicsrole of gender, ethnicity and age
The predominance of pre-event risk factors (156/290; 53.9%) (Table 4), emphasizes the crucial link between intrinsic sociodemographic factors and a predisposition to psychological impairment following EWEs.Generally, a higher mental health burden is associated with "marginalized" population groups which often lack the resources to adequately cope with traumatic events (Galea et al., 2005;Cianconi et al., 2020;Hrabok et al., 2020;Clayton, 2021).The concept of "allostatic" (over-)load, i.e., accumulation of chronic stress and life events (Guidi et al., 2021), can help elucidate the role of lower social standing, predisposition towards psychological vulnerability, and likelihood of impairment following environmental challenges.The latter is clearly reflected among risk factors identified in reviewed investigations (Table 4), as lower educational attainment, previous history of mental health impairment and/or trauma, and a (overall) lower socio-economic status are frequently reported.A marked gender imbalance was also found.Female gender has been consistently associated with a higher risk of psychological impairment following EWEs (WHO, 2014;Berry et al., 2018;Obradovich et al., 2018;Lowe et al., 2019).This has been attributed to a combination of cultural, socio-economic, and physiological factors, which can result in a gendered "disadvantaged" status, often irrespective of income (Norris et al., 2002;Kessler et al., 2005;Lopez et al., 2006;Tang et al., 2014;Watts et al., 2015).Specific factors linked to higher impairment rates among females vary substantially and may include reduced access to education and aid resources (e.g., preand post-EWE), poor nutrition in contrast to male counterparts (both generally and during periods of food scarcity), stress/burden associated with a traditional caregiver social role (e.g., wife, mother), and likelihood of experiencing post-disaster violence (Norris et al., 2002;Kimerling et al., 2009;WHO, 2014;Watts et al., 2015).
Evidence also suggests females perceive and react more negatively to natural disasters and are considered more susceptible to the effects of these events (Norris et al., 2002;Bonano and Gupta, 2009;Tang et al., 2014).While interpretations remain tentative, the composite prevalence rates presented, which monotonically increase with sample female representation , indicate a higher burden among female gender with potential domain-specific implications (cf.depression, PTSD; Fig. 4).As expected, the role of gender is also predicated by its recurrence among risk factors (29/290; 10%) (Table 4), second only to levels of event exposure.Conversely, just two investigations reported male gender as a risk factor (2/290; 0.7%) representing a high level of gender disparity.Overall, females represent a critical sub-population in the implementation of pre-and post-EWE mental health intervention strategies (e.g., counselling, outreach) and recovery resource allocation, all of which can play a key role in ameliorating the psychological burden of disaster events (Cohen 2002;Morganstein and Ursano, 2020).
Identifying as a racial/ethnic minority was also a recurrent risk factor linking EWE exposure and psychological impairment (Norris et al., 2002;Galea et al., 2005;Goldman and Galea, 2014;Lowe et al., 2019).The importance of racial background is reflected in higher estimated composite and PTSD prevalence rates .Similarly, belonging to a minority was a frequently identified risk factor (4.1%; Table 4).A higher likelihood of impairment among minorities has been attributed to disadvantaged social standing and associated vulnerability to trauma (Norris et al., 2002;Galea et al., 2005;Berry et al., 2010;Hayes et al., 2018;Lowe et al., 2019).Assessment of racial/ethnic background as a potential risk factor was exclusive to US-based investigations (n = 30), and thus, not representative of wider global risks associated with individuals from minority backgrounds.
A degree of subjectivity in relation to age and impairment is mirrored in the comparable recurrence of young and old age among risk factor typologies (Table 4).The results highlight a degree of ambiguity in the nexus of disaster events, age, and mental health sequelae, which is likely driven by locally-specific cultural and socio-economic attributes (Kohn et al., 2005;Cook and Elmore, 2009;Parker et al., 2016;Lowe et al., 2019).For example, several previous studies report that older (particularly elderly) populations exhibit high resilience in the face of natural disasters (Norris et al., 2002;Kohn et al., 2005;Cook and Elmore, 2009); a trend tentatively supported by prevalence rates among older adults (>51) found in the current study .Here, cumulative (past or pre-event) life experience may confer a degree of protection from stresses imparted by EWEs (Knight et al., 2000), with children conversely yet to develop coping mechanisms to counteract stressful situations (Goldman and Galea, 2014).Notably, the data reported by Kar et al. (2007), resulting in the highest PTSD OR (Fig. 6), derives exclusively from a sample of school children exposed to the cyclone Orissa.However, calculated prevalence rates were highest for the adult (31-59) sub-category (38.6%;Fig. 3), representing a key age sub-group for consideration in the context of mental health intervention.Overall, the results support evidence of middle-aged individuals being particularly vulnerable in post-disaster settings, attributed to high levels of life stress and responsibility burden (e.g., dependents) (Norris et al., 2002;Goldman and Galea, 2014).

Limitations and recommendations
The authors consider that a number of inherent impediments exist which require careful consideration in the interpretation of presented pooled estimates.Importantly, a number of methodological limitations stem from the comprehensive approach towards EWEs and individual exposures necessary to produce a comprehensive "scoping" review.Data akin to a meta-analytical or systematic review were not attainable given the wide range of EWE types, levels of exposure, variations in sample size, and range of mental health outcomes analysed.Additionally, due to the study design associated with some investigations, and particularly the lack of participant controls, it was not possible to calculate studyspecific effect size(s) for several studies.Crucially, extracted psychological impairment data (Tables 1-3) more closely represent "prevalence" rates, i.e., positive cases at a single (post-event) point in time, than longitudinal "incidence", or a finite metric of psychological impairment linked to EWEs (pre-and post-event).As such, few studies have established a "direct" temporal link between EWE exposure and adverse mental health outcomes, representing a key knowledge gap with respect to our current understanding of the intersect between socioepidemiological mechanisms in the aftermath of climatic extremes.Most disaster-focused investigations lack pre-event ("background") mental status data, or refrain from incorporating independent control groups to measure or predict the trajectory of outcomes (Norris 2006;Yzermans et al., 2009).These factors, and concomitant limitations, are clearly reflected in the compiled dataset with a majority of investigations lacking any control criterion (69.5%;Table A3).The uniform distribution of event-evaluation time lag characterizing the pooled dataset (Table A3; Appendix F) also highlights potential issues associated with "resilience", temporal variability of symptoms, and the nature of diagnosis (Norris et al., 2002;Norris 2006;McFarlane et al., 2009).In conjunction with a paucity of longitudinal studies, the temporal trajectory of mental health sequelae in the aftermath of EWEs represents a critical knowledge gap and focus area for further research.
The accuracy of presented estimates is also potentially influenced by the spatial adjacency of studied EWEs relative to affected (or targeted) study populations, with impacts and outcomes ultimately proportional to levels of event exposure (i.e., dose-response) (Norris et al., 2002;Goldman and Galea, 2014;Harville et al., 2015).The importance of event exposure and experienced stressors is clearly reflected in its predominance (14.5%) among reported risk factor typologies (Table 4).Additional peri-and post-event risk factors linked to levels of event "exposure", including perceived threat/fear, injury/somatic conditions, and property damage, all commonly associated with PTSD (Norris et al., 2002;Neria et al., 2008;Goldman and Galea, 2014), were also frequently reported in the reviewed literature (Table 4).Notably, disaster victims tend to report higher somatic conditions and concerns in comparison to non-exposed counterparts, potentially highlighting important links between psychological and physiological stress in the aftermath of EWEs (Norris et al., 2002;Ursano et al., 2009;Yzermans et al., 2009).
Further, psychological measures are also invariably constrained by the range of diagnostic criteria employed, and the potential incorporation of cultural bias subject to their regional origin, which may limit cross-cultural application (Galea et al., 2005;McFarlane et al., 2009;Lowe et al., 2019;WMO, 2020;Moore et al., 2020).Inevitably, compiled data will also be subject to variability according to study-specific geographical, cultural and socio-economic variables.In this context, efforts should be directed towards (i) implementation of culturally-relevant psychological and well-being measures and (ii) generation of data at a regional level facilitating comparisons among samples with similar socio-demographic backgrounds.
The need for clear, detailed, and homogeneous data reporting of pertinent study aspects (e.g., methods, outcomes) is critical, and highlighted by the proportion of investigations failing to fully report comorbidity data (20/59; 33.9%), and those providing insufficient information for OR calculation.Further, pooled analysis inevitably relied on integration of morbidity data from disparate events in terms of intensity, forecasting and (pre-and post-event) risk communication levels.Both forecasting and risk communication strategy can be determinant for event preparedness and influence psychological outcomes (Ellis and Albrecht, 2017;Morganstein and Ursano, 2020).For example, a higher composite impairment in "wildfire" (35%) and (possibly) "drought/heat" (31%) event categories (Table 2), could be related to their relative unpredictability and temporal ambiguity with lower associated levels of event anticipation and preparation.However, given the range of spatio-temporal impediments potentially influencing EWE exposure levels and resulting "impairment", comparisons of calculated estimates for individual EWEs categories are challenging.As previously outlined, this represents a key methodological limitation of the comprehensive "scoping" approach necessary to produce an all-encompassing metric of psychological impairment.In this context, forthcoming investigations can provide additional insights through the implementation of a "focused" set of eligibility criteria in terms of (EWE) individual exposure, diagnosis, space and time.
A key limitation of the review is also represented by the low number of investigations focusing on "climatic" (gradual or sub-acute) event types identified (Table A3).Characterized by higher prevalence rates (Table 2), data extracted from these studies may support indicators of a higher (and more complex) public health and socio-economic burden associated with slow-onset in contrast to acute (or fast-onset) climatic phenomena (Vins et al., 2015;Watts et al., 2015;Obradovich et al., 2018;Morganstein and Ursano, 2020).However, the limited number of reviewed studies focusing on sub-acute events prevents further insights.Similarly, few identified investigations focused on "wildfires" events (Fig. 3), a factor potentially affecting applicability of the results for EWEs with increased frequency and intensity in recent years.Given the paucity of research in concurrence with an increasing global frequency, this is a key area for further research and which would benefit from a similar empirical approach employed within.

Conclusion
The pooled analyses presented in this study provide a concise empirical link between individual extreme weather exposures and adverse psychological impacts.This comprehensive "scoping" review serves as a framework for future systematic reviews focusing on specific EWE types and mental health outcomes while elucidating knowledge gaps and limitations.Despite inevitable methodological limitations, the prevalence rates presented in this study provide (novel) global, regional, domain-and category-specific "standards", representing highly relevant reference points for forthcoming investigations.Crucially, psychological impairment estimates presented tend to be above available epidemiological baselines, demonstrating the psychological toll associated with extreme weather exposure.Overall, a prevalence of pre-event risk factors, identified from both individual studies and through collective pooled data, and which often relate to socio-demographic variables linked to marginalized groups, highlights the need for stakeholders to adopt and/or improve bespoke anticipation and preparation interventions.Further, sub-populations which should be prioritized in the context of pre-and post-disaster intervention and recovery resource allocation are outlined.

Fig. 1 .
Fig. 1.Schematic of the systematic literature review protocol employed.

Fig. 2 .
Fig. 2. Global distribution of reviewed investigations.The number of articles per country of origin and the composite metric for psychological impairment are also provided.N/A = Not Available.

Fig. 3 .
Fig. 3. Plot of composite psychological impairment values stratified by key data sub-categories.The number of reporting studies (n) in each sub-category is also provided.Black squares indicate the mean value estimated per sub-category with horizontal black lines providing value range.The vertical dashed grey line indicates the "global" mean value.N/R = Not Reported.

Fig. 4 .
Fig. 4. Plot of psychological impairment values by key data sub-categories per domain (PTSD, anxiety, depression).The number of reporting studies (n) in each subcategory is also provided.Black squares indicate the mean value estimated per sub-category with horizontal black lines providing value range.The vertical dashed grey line indicates the "global" mean value.N/R = Not Reported.

Fig. 5 .
Fig. 5. Forest plot showing calculated ORs for studies reporting composite psychological impairment.The population column provides the total number of positive cases (n) among exposed and control groups (N).Black squares indicate ORs which are also provided at the right along with 95% confidence intervals.The pOR is provided at the bottom of the plot.The vertical black lines indicates the point (or threshold) of null effect.

Fig. 6 .
Fig. 6.Forest plot showing calculated ORs based on PTSD studies.The population column provides the total number of positive cases (n) among exposed and control groups (N).Black squares indicate ORs which are also provided at the right along with 95% confidence intervals.The pOR is provided at the bottom of the plot.The vertical black lines indicates the point (or threshold) of null effect.

Fig. 7 .
Fig. 7. Forest plot showing calculated ORs based on depression studies.*Study based on pre-post control data and not included into pOR calculation.The population column provides the total number of positive cases (n) among exposed and control groups (N).Black squares indicate ORs which are also provided at the right along with 95% confidence intervals.The pOR is provided at the bottom of the plot.The vertical black lines indicates the point (or threshold) of null effect.

Fig. 8 .
Fig. 8. Forest plot showing calculated ORs based on anxiety studies.The population column provides the total number of positive cases (n) among exposed and control groups (N).Black squares indicate ORs which are also provided at the right along with 95% confidence intervals.The pOR is provided at the bottom of the plot.The vertical black lines indicates the point (or threshold) of null effect.

Table 2
Estimated values for composite psychological impairment stratified by selected data sub-categories.The study n column provides the number of investigations (among sub-category total) with adequate (full) data reporting to derive pooled values.

Table 3
Estimated psychological impairment values per domain (PTSD, anxiety, depression) stratified by data sub-categories.The study n column provides the number of investigations (among sub-category total) providing adequate (full) data reporting to derive pooled values.