The Association between Physical Environment and Externalising Problems in Typically Developing and Neurodiverse Children and Young People: A Narrative Review

The physical environment is of critical importance to child development. Understanding how exposure to physical environmental domains such as greenspace, urbanicity, air pollution or noise affects aggressive behaviours in typical and neurodiverse children is of particular importance given the significant long-term impact of those problems. In this narrative review, we investigated the evidence for domains of the physical environment that may ameliorate or contribute to the display of aggressive behaviours. We have considered a broad range of study designs that include typically developing and neurodiverse children and young people aged 0–18 years. We used the GRADE system to appraise the evidence. Searches were performed in eight databases in July 2020 and updated in June 2022. Additional articles were further identified by hand-searching reference lists of included papers. The protocol for the review was preregistered with PROSPERO. Results: We retrieved 7174 studies of which 67 are included in this review. The studies reported on green space, environmental noise and music, air pollution, meteorological effects, spatial density, urban or rural setting, and interior home elements (e.g., damp/sensory aspects/colour). They all used well validated parent and child reported measures of aggressive behaviour. Most of the studies were rated as having low or unclear risk of bias. As expected, noise, air pollution, urbanicity, spatial density, colour and humidity appeared to increase the display of aggressive behaviours. There was a dearth of studies on the role of the physical environment in neurodiverse children. The studies were heterogeneous and measured a range of aggressive behaviours from symptoms to full syndromes. Greenspace exposure was the most common domain studied but certainty of evidence for the association between environmental exposures and aggression problems in the child or young person was low across all domains. We found a large knowledge gap in the literature concerning neurodiverse children, which suggests that future studies should focus on these children, who are also more likely to experience adverse early life experiences including living in more deprived environments as well as being highly vulnerable to the onset of mental ill health. Such research should also aim to dis-aggregate the underlying aetiological mechanisms for environmental influences on aggression, the results of which may point to pathways for public health interventions and policy development to address inequities that can be relevant to ill health in neurodiverse young people.


Introduction
The physical environment encompasses all aspects of a child's physical world and may be defined as objective characteristics of the physical context in which children spend their time (e.g., home, neighbourhood, school). The influence of children's physical exposures has been summarised differentially by various models, theorems, and theorists over the previous century. Notably, these include the physical environmental elements of children's exposome (a term introduced by Wild [1,2] regarding the non-genetic influences on outcomes across the lifespan) and Bronfenbrenner's bioecological model [3][4][5], proposing that children develop within an environmental milieu of five interconnected systems, spanning aspects from urban design (e.g., presence and structure of sidewalks), traffic density, and design of venues for physical activity (e.g., playgrounds, parks, and school yards), to biologically active chemicals, radiation, the internal chemical environment, and psychosocial aspects [6]. The difficulty with these conceptualisations of child development is that they include both physical environmental and (psycho)social influences. As exemplified in a review of the influence of interior hospital environmental interior conditions, Harris [7] segmented the environment into distinct physical exposure categories: ambient, architectural, and interior design.
In this work, an operationalised definition of "physical environment" was incorporated to identify eligible environmental exposures. This classification was derived from a coalescence of Harris's [7], Bronfenbrenner's [3][4][5] and Wild's [1,2] theorems. This resulted in the inclusion of a diverse array of domains, from ambient exposures (sunlight, sound, meteorology), interior design elements (colour, lighting), architectural features (space/spatial crowding), and biological active agents (i.e., air particulate pollutants), to physical aspects of children's microsystem (i.e., home, school, and neighbourhood characteristics).
A variety of theories have attempted to explain the mechanisms via which environmental domains influence physical and mental health. Although none of these mechanistic models have been fully proven, there are suggestions that positive effects may be the end product of pathways that link several elements, such as mitigation (reduction in air pollution or traffic noise), restoration (stress reduction and attention restoration in alignment with what the Attention Restoration Theory posits) and instoration, whereby attributes of the physical environment, such as greenness in particular, may promote physical activity and social capital and cohesion [8][9][10].
Previous theories have primarily focused on the stress-reducing effects of greenspace, either via a protective influence from harmful environmental stimuli (noise and air pollution) [11][12][13], or via the restoration of attentional resources [14,15]. Recently, it has been posited that greenspaces may provide more direct physiological benefits via increased exposure to phytoncides (plant-derived antimicrobial volatile organic compounds) [16]. Whilst preliminary research into the effects of phytoncide exposure is positive, it is currently inconclusive and additional studies are required [17]. Neuroimaging studies are also shedding insight into potential mechanisms for greenspace exposures potential mechanisms, with one study [18] showing that it can beneficially deactivate the prefrontal cortex in regions linked to depression and rumination.
The literature also indicates that aspects such as ambient air particulate matter exposure may negatively impact development via neuroinflammatory pathways [19][20][21][22][23][24]. Noise pollution may also have detrimental effects via contributions to subjective annoyance and irritation; whilst not necessarily directly causing aggression, noise exposure in those with low threshold for expressing anger may increase its severity [25,26] via draining of attentional and cognitive resources and subsequently leading to increased self-regulatory difficulties [27]. Social-behavioural mechanisms may explain the relationship between behaviour and fluctuations in meteorological effects (such as temperature), e.g., the routine activity theory that proposes that warmer temperatures facilitate more frequent social interaction, increasing opportunity for aggression [28] or that heat increases hostility and physiological arousal and consequently to aggressive behaviour [29]. Theories have posited that high spatial density triggers perceptions of crowding and a subsequent physiological stress arousal response [30][31][32]. Why proximity elicits these responses is still unclear and has been linked to competition for resources and invasion of personal space [33]. Baird et al. [34] reported a beneficial association between household crowding and reduced conduct problems in children with intellectual disabilities. The authors propose several theories about these potentially counterintuitive findings, suggesting that increased avail-ability of and proximity to family members, in intergenerational households, and parental habituation to problematic conduct behaviours are all potential mechanisms underpinning this finding. Using a sensory room unaccompanied may be associated with a sense of autonomy in children and young people which in turn reduces distress [35]. Other pathways may contribute to the impact that music listening has on a broad range of psychological and physiological benefits [36][37][38][39][40][41].
As discussed, social-behavioural mechanisms may explain the relationship between aggression and climate effects, for example the routine activity theory proposes that warmer temperatures facilitate more frequent social interaction, increasing opportunity for aggression [28]. Alternatively, the general aggression model (GAM) is more grounded in a physiological aetiology of aggression, suggesting that heat increases hostility and physiological arousal and consequently aggressive behaviours [29].
From the evidence presented so far, it appears that both physical and social environments, in addition to genetic and epigenetic influences, shape the developmental trajectories of children [42][43][44][45][46]. However, in the main, published research is focused on typically developing rather than neurodiverse children [47]. Previous work has evidenced disproportionate influence of children's early environmental milieu in shaping a range of socio-emotional and cognitive developmental outcomes. Specifically, learning disabled children are more likely to be affected by social adversity, poor housing, and poverty [48]. These children are also exponentially more likely to be exposed to negative environmental exposures such as air pollution [49]. To address failings in supporting these children and their families, an important element is to reduce socioeconomic inequality and improve residential conditions [50]. Furthermore, children with complex neurodisabilities have increased barriers to accessing potentially therapeutic aspects of both the physical and social early environments [51]. Disabled child access to urban greenspaces, for example, is not only infrequent in comparison to their typically developing peers [52], but when significant resources are employed to facilitate access for neurodisabled children, the high-risk nature of visiting these spaces requires rigid structure, impacting on the quality of nature experiences when they do occur [53]. This is one example of the health inequities and disparities experienced by neurodiverse children in comparison to their peers, exemplifying the need for additional research in these domains.
Externalising disorders are characterised by display of a range of behaviours which are associated with poor impulse-control, and include rule breaking, impulsivity, and inattention; in addition, a core component of these conditions is the presence of heightened aggression.
Specific child and adolescent externalising disorders include conduct disorder (CD), oppositional defiant disorder (ODD), and attention-deficit-hyperactivity disorder (ADHD). Of particular concern is this repeated presence of aggressive behaviour in these disorders as it is often associated with referral to services and application of a range of restrictive practices, most commonly antipsychotic medications but also inpatient admissions.
Aggressive behaviours and general behavioural problems such as destructive behaviours have an overall negative influence on carers due to stress and negative interactions between carers and the person they care for, likely resulting in a deterioration of the quality of care [54]. Moreover, behavioural problems are associated with increased service costs because of the impact of behaviours on staff and need for high support levels [55]. Aggressive episodes also provoke concerns about threat to personal safety as well as cause panic and upset [56,57].
Neurodevelopmental disorders (NDDs) are a category of "etiologically diverse conditions" with onset during the developmental period and are characterised by below average intellectual functioning and adaptive behaviour [73]. This classification includes disorders such as intellectual disability (also called learning disability in the UK), autism spectrum disorders (ASD), and other developmental delays (DD). Whilst we appreciate the nuances of the definitions for brevity and clarity, we will refer to those children with NDDs as neurodiversity in this context. About one in one hundred individuals has a neurodevelopmental disorder and there are about 351,000 children with intellectual disability in the UK, often coexisting with other neurodevelopmental disorders [74]. Prevalence of aggressive behaviours in NDDs appears to fluctuate depending on sampling methods and assessment strategies, ranging from 8.3% in community samples [75] to 64% in inpatient care [76,77]. Children with intellectual disability were six times more likely to have conduct disorder measured by the Strengths and Difficulties Questionnaire compared to their typically developing peers [78]. Aggressive behaviours are persistent over time [79], with displays of aggression being consistently linked with neurodiversity [57,[80][81][82][83][84][85] though prevalence rates reported can be inconsistent.
Whilst previous research has examined predictors of broadly defined challenging behaviour in children with intellectual disability [86][87][88], none of the studies has included examination of the influence of the physical environment specifically on such behaviours to date. Here, we build on previous work examining the influence of single domains of the physical environment on aggressive behaviour of typically developing and neurodiverse children by including (1) children across the spectrum of ability and (2) all available objective domains of the physical environment.
Therefore, in this narrative review, we examine the certainty of evidence of the impact of the physical environment on typically developing and neurodiverse children's aggressive behaviours. The outcome of interest was either psychological or biological proxies of aggressive behaviour, annoyance and irritability measured by validated psychometric questionnaires (measures or outcomes which have been empirically evaluated for reliability) or biological markers such as blood pressure, heart rate and skin conductance. The findings are presented by environmental domain (Greenspace, noise pollution, air pollution, meteorology, spatial density, rurality of residence, interior design, and music) and separately for typically developing and neurodiverse children.

Search Strategy
We adhered to the Preferred Reporting Items of Systematic reviews and Meta-analyses (PRISMA) statement checklist [89] in conducting the review, as well as guidance from the Synthesis Without Meta-analysis [90], and the Meta-analysis of Observational Studies in Epidemiology [91] to improve the precision of our reporting. The study protocol was preregistered on PROSPERO (CRD42020160251). Because of the heterogeneity in outcome used and the variation in exposure measures, we were unable to perform a meta-analysis. Instead, we reported the degree of certainty of the evidence available in terms of protective/detrimental, inconclusivity, or no association for each outcome and exposure metric across each domain of the physical environment.
The electronic search strategy comprised 8 bibliographic databases (MEDLINE, Psych-INFO, Web of Science, CHINAHLplus, Embase, Cochrane library, EThOS and ProQeust dissertations and theses) and two grey literature sources (NICE evidence search and Google scholar). The inclusion of the latter sources facilitated the retrieval of additional studies from a more diverse range of sources (including policy and public health), whilst mitigating publication bias and increasing the comprehensiveness of the review [92,93]. The search was carried out in July 2020, and replicated in the update to June 2022 with no year of publication limit. Bibliographies of retrieved articles were searched to maximise retrieval of relevant articles. The search strategy was overseen by a specialist librarian (see Supplementary Materials).

Selection Criteria
Studies were included if they (a) reported primary research, (b) were written in English, French, German, Mandarin Chinese and Spanish which were languages spoken by fellow researchers and therefore could be translated, (c) included human participants aged between 0-18 years, (d) contained a psychometrically valid parent or child reported outcome measure of aggressive behaviours or physiological measures of arousal (identified as a proxy measure of aggressive behaviour) and (e) examined exposure to domains of the physical environment.

Screening and Appraisal Process
All retrieved articles were screened by the first author (A.B.). A sub-sample of titles and abstracts (10%) were co-screened by a senior researcher (A.H.) and a post-doctoral researcher (R.R.). Inter-rater reliability for this initial screening was 87% (0.868). Full text data extraction was conducted by the main author (A.B.) using a modified flexible data extraction template used for non-Cochrane reviews [94] with co-screening conducted for a proportion of studies (59%) by independent researchers (see acknowledgments). Substantial agreement between the primary author and co-screeners was reported (83%, κ = 0.6126) with disagreements resolved by the senior researcher (A.H.) who also crosschecked the extraction table for any inconsistencies.
Risk of bias assessment (RoB) and GRADE protocol were adapted from a systematic review by Clark, Crumpler, and Notley [95] on the evidence relating to effects of environmental noise pollution on mental and physical health outcomes. Four items from this review were used to assess the bias for each paper:

•
Evaluation of the quality and validity of the exposure: whether the paper used established or validated environment metrics. • Bias due to confounding: whether studies included adjustment for potential confounding variables. • Bias due to sampling methodology and reporting of attrition rate.

•
Outcome assessment leading to information bias: whether studies were using validated aggressive behavioural outcome measure(s).
One measure of RoB that was not included in this review was "due to blinding to exposure outcome" as it was not considered appropriate for the methodology of the majority of the retrieved studies which infrequently blind outcome assessors. Overall RoB ratings for each study were aggregates of high, low or unclear across the four domains. We adopted a conservative rating strategy where studies that had equal reports of low and high risk of bias were classified as high.
The GRADE system [96] is a widely used tool recommended by The Cochrane collaboration [97] which provides a ranking of quality for evidence on interventions and relevant outcomes. The modified GRADE approach assigns a priori the highest quality of evidence to longitudinal or intervention studies, and the lowest to cross-sectional studies, subsequently up-or down-grading evidence dependent upon various methodological factors such as RoB, studies not comparing the same variables, inconsistency of findings between studies, imprecision (effect estimate confidence interval containing 25% harm or benefit), publication bias of funnel plot reported, and other considerations (large effect RR > 2, adjustment for all plausible confounding, dose response gradient). As we did not carry out a meta-analysis assessment of GRADE criteria such as precision or publication bias was not possible.

•
Greenspace was measured by land use data percentage of natural space in the neighbourhood (e.g., for the UK, a census output area such as LSOA) or measured within a set distance of the child's residence. Other indices included satellite derived neighbourhood greenspace (e.g., normalised difference vegetation index (NDVI)) and percentage of neighbourhood greenspace. • Blue space was measured by parents reporting on number of days taking their children to a beach. • Environmental noise pollution included road traffic, construction noise, and aircraft noise. • Air pollution was measured by particulate matter, tobacco smoke (nitrogen dioxide: NO2), and elemental carbon attributed to traffic (ECAT).

•
Meteorological variables included seasons, hot or cold weather, humidity and sunlight. • Spatial density and interior home/facility design included space per child in square metres (high/low density), wall paint, sensory room, presence of damp. • Urbanicity and rurality were described by the location of the child's residence or school.

Measures of Aggressive Behaviours
The studies utilised a number of psychometrically valid parent and child reported measures of aggressive behaviour, as well as observer ratings. These comprised the full instrument or conduct, aggression, and externalising behaviour domains as follows: Other outcomes used were observer rated frequency of aggressive behaviour

Results
The two searches retrieved 7434 records. After deduplication, 7174 were screened of which 257 underwent full-text assessment, resulting in the inclusion of 67 papers (details are shown in the PRISMA flow diagram, Figure 1). Six of which reported on the physical environment and aggressive behaviours in neurodiverse participants.
We report RoB separately for studies carried out with typically developing (Table 1) and neurodiverse populations ( Table 2). We follow the same format for the GRADE evidence summaries for the environmental exposures on outcomes of aggressive behaviours for typically developing and neurodiverse children (Tables 3 and 4). We report RoB separately for studies carried out with typically developing (Table 1) and neurodiverse populations ( Table 2). We follow the same format for the GRADE evidence summaries for the environmental exposures on outcomes of aggressive behaviours for typically developing and neurodiverse children (Tables 3 and 4).   Residential proximity to major green spaces: if home address was within 300 m of available greenspace.
School greenness and combined home-school greenness: Average weighted NDVI in a 100 m buffer around school and home locations.
Blue space: Parental report of how many days they accompanied their children to the beach annually.

Low
Beneficial association: residential greenspace increased greenspace (NDVI) in 100 m and 250 m buffer radii around child's residence was significantly associated with decreased conduct problem scores.
No associations: between greenspace playing time, residential proximity to greenspaces, blue space attendance, and combined school and home greenspace (NDVI) and conduct problems. Ward level residential greenspace: Deciles of the percentage of greenspace within the family's UK ward. Access to private garden space: Parent reported child access to private garden space.

Low
No association: between ward-level greenspace and conduct problem trajectories over time.
no association between child access to private garden space and conduct problem trajectories was reported.
Proximity to city parks: Proximity of residence to nearest park.
Residential greenness: Greenspace (NDVI) in a buffer of 300 m around participants home residence.
Parent-completed Conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) 3 .

Low
Beneficial association: children whose mothers reported low educational attainment reported significantly more conduct problems as the distance of home residence to closest greenspace (parks) increased.
Harmful association: increased residential greenspace was associated with increased conduct problems in the high maternal education group and approaching significance in the low education group.  [100]. Low Beneficial association: for children living in an urban environment, a 1 inter-quartile range increase in greenspace was significantly associated with lower externalising behavioural scores.
No association: for children residing in rural or suburban areas, no association was reported between greenspace and externalising behaviours.
Feng and Astell-Burt, 2017 [110] Australia n.a. Longitudinal 4968 4 9 Greenspace measured as the percentage of land-use classified as parkland (domestic gardens not included).
Parent-completed externalising behavioural sub-scale (conduct and hyperactivity scale combined) of the Strengths and Difficulties Questionnaire (SDQ) [98] Low Beneficial association: a non-linear association between increased local greenspace and reductions in children's SDQ scores was reported, proportional to local land use classified as greenspace. Neighbourhood greenspace was defined as the percentage of natural space within groups of census output areas (LSOAs). Private garden access.
Parent-completed conduct problems subscale scores of the Strengths and Difficulties Questionnaire (SDQ) [98] Low No association: between neighbourhood greenspace and conduct problems was reported.
Beneficial association: children's access to a private garden was associated with significantly decreased parent reported SDQ conduct scores. Varying spatial buffer zones of greenspace surrounding children's residence (NDVI) at ages 7 and 12.
Parent-completed externalising subscale scores of the Behavioural Assessment System for Children, Parent Rating Scale, Second Edition (BASC−2) [99]. Scores ≥ 60 classify children as "at risk" for conduct disorder in clinical settings.

Low
Beneficial association: greenspace at age 7 was significantly associated with decreased conduct scores at the 200 m buffer radius only.
No association: no associations were reported between greenspace and conduct scores or aggression scores at age 12 at any buffer radius.
Beneficial association: increased NDVI at 200 m and 800 m buffers at age 7 was significantly associated with lower probability of being "at risk" of conduct problems. Varying buffer zones of green space around participants' home addresses and distance to major green spaces was computed with the normalised difference vegetation index (NDVI).
Parent-completed conduct and hyperactivity subscales (combined to derive externalising behavioural scores) of the Strengths and Difficulties Questionnaire (SDQ) [98].

No association: between NDVI and parent-reported externalising behaviours in White
British or South Asian participants. Greenspace was measured using land use data, which reports percentage of greenspace in the family's ward (excluding gardens).
Self-completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].

Low
No association: between greenspace and conduct problem scores. USA Schools Cross-sectional n.a. 5-18 n.a. Greenspace (NDVI 1 ) in a 100 m buffer zone surrounding schools Incidence rates of attacks or threats with and without weapons in schools. Low Harmful association: increased school greenness was associated with increased incidence of threats and attacks (with or without weapons). Greenspace defined as the % area of total natural space 5 and parks within 500 m of the child's residence.
Private garden access.
Primary caregiver-completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].

Low
No association: between total green space and children's conduct problem scores.
Beneficial association: between children not having access to private garden and increased SDQ conduct scores was reported. Semi-natural, forested, and agricultural areas (greenness) and residential and industrial areas in a 5000, 4000, 3000, 2000, 1000, 500, 300 and 100 m buffer from the residential address Primary caregiver-completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].

Low
No association: between landscape surrounding child's residence and conduct problems were reported. Residential road traffic noise was assessed using modelling different periods of the day, including daytime (Lday), nighttime (Lnight), and weighted 24 h (Ldn).
Parental completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].

Low
Harmful association: weighted 24 h (Ldn) noise exposure was associated with increased conduct problems.
Crombie et al., 2011 [124] UK/Spain/ Netherlands Schools Cross-sectional 1900 9-10 n.a. A continuous noise from aircraft and road traffic measure calculated in dB for each school.
Parental completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].

Low
No association: no association between air traffic noise and conduct problems.
Beneficial association: between increasing road traffic noise and decreasing SDQ problem scores was reported.
Essers et al., 2022 [125] Spain/ Netherlands Residence Longitudinal 7958 18 m-9 years 7.5 Average 24 h noise exposure at the participants' home address during childhood was estimated using EU maps from road traffic noise and total noise (road, aircraft, railway, and industry).

Low
No association: between noise exposure and conduct problems or aggressive behaviours was reported.
Grelat et al., 2016 [126] France Not reported Cross-sectional 517 7-11 n.a. Noise indices were calculated from the front and most exposed façade of the child bedrooms using a noise map.
Child self-report questionnaire on annoyance from various traffic and ambient noise sources. Unclear Harmful association: increased road and general transport noise exposure was significantly associated with increased child annoyance.

Haines et al., 2001a [127] UK Schools
Cross-sectional 340 10 n.a. Exposure of schools to high and low aircraft noise.
Parent-completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].
Child self-report questionnaire on annoyance due to; aircraft, train, road and neighbour noise.

Low
No association: between aircraft noise exposure at school and SDQ conduct problems was reported.
Harmful association: increased aircraft noise exposure was significantly associated with increased annoyance. Parent-completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].
Child self-report questionnaire on noise annoyance.

Low
Harmful association: increased aircraft noise exposure at school was significantly associated with increased annoyance.
No association: between aircraft noise and SDQ conduct scores were reported. Parent-completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) [98].

Low
No association: between aircraft noise and conduct problems was reported.
Harmful association: between increased road traffic noise and higher conduct problem scores.
Night (L night ) and day (L den ) indicators of road traffic noise at child's residence were created using weighted long-term annual average sound levels.
Parental completed conduct problems subscale of the Strengths and Difficulties Questionnaire (SDQ) 4 .

Low
No association: between day or night noise exposure and conduct problems was reported. Cross-sectional 2441 0-6 n.a. Self-reported cigarette smoke exposure (screening questionnaire asking families if anyone in the household smoked).

Air pollution
Diagnosis of Disruptive Behaviour Disorder (DBD) was gained from child's electronic health record. Diagnoses were identified using International Classification of Diseases-ninth revision (ICD-9) [137].

Unclear
Harmful association: childhood smoke exposure increased risk of disruptive behaviour disorder.
Gatzke-Kopp et al., 2020 [138] USA Residence Longitudinal 1096 0.5 6.5 Child salivary cotinine (a metabolic of nicotine) was measured as a proxy for exposure to cigarette smoke.
Primary caregiver-completed conduct problems subscale (SDQ) [98] and the Disruptive Behaviours Rating Scale (DBDRS) [139]. These scores were combined to create a composite conduct problems score.

Low
Harmful association: increased cotinine levels associated with increases in a multi-informant latent factor of conduct problems.

Low
Harmful association: increased PM 2.5 and NO 2 at age 12 was significantly associated with increased odds for conduct disorders at age 18.
Teacher-completed Daily Behavioural and Emotional Questionnaire (DBEQ) [151]. Low Harmful association: between increased humidity during winter and increased aggression.
No association: between other meteorological variables and aggression.  [157].

Low
Harmful association: increased humidity was significantly correlated with increased externalising behaviours.
Beneficial association: increased sunshine was significantly correlated with decreased externalising behaviours.
No association: temperature was not correlated with externalising behaviours.
Observational recordings of school yard aggressive behaviours over an academic year used to construct an aggression intensity index for each participant.

High
Harmful association: frequency of aggression was significantly increased during the warm season.
Beneficial association: increased temperature and humidity were associated with significantly decreased frequency of aggressive events. A monthly time-series of average ambient temperature was constructed, and temperature was further aggregated for the periods 1, 2, and 3 years preceding each CBCL assessment.
Parental completed Child Behaviour Checklist (CBCL) [100]. Low Harmful association: between ambient residential temperature 2 and 3 years prior to assessment and externalising behaviours was reported (this effect remained when controlling for urbanicity, humidity, traffic density and proximity to roads or freeways).

Low
Harmful association: household crowding was significantly positively associated with conduct problems across development.  [164] between ages 5.5-6.

Unclear
No association: between overcrowding in the home (at age 3) and age 4 maternal reported externalising behaviours.

Harmful association:
overcrowding in the home (at age 3) was significantly associated with increased teacher reported externalising behaviours at age 5. Teacher completed Child Behaviour Checklist-Teacher Rating Form (CBCL-TRF) [171].

Low
No association: between parent reported externalising behaviours and children residing in urban and rural environments.
Beneficial association: urban schoolteachers reported significantly more externalising behaviours than rural teachers.

Low
No association: in self-reported violent behaviour between rural and urban participants was reported.
Beneficial association: two violent behaviours; "chased with weapons" and "injured someone with weapons" was significantly more frequently reported by urban adolescents. Anger out and in was significantly elevated in the rural condition. Observer rated aggression using a modified version of the Overt aggression scale [180] in modified and non-modified sensory rooms.

Unclear
Beneficial association: aggression after modified sensory room use was significantly decreased in comparison to non-modified room use.  Sensory room modification.

History of aggression/The Stepping Stones Sensory
Room Questionnaire (SSSRQ) [193] measured distress levels pre-and post-sensory room use.

Unclear
Beneficial association: between reductions in client-reported stress following modified sensory room use and history of aggression was reported.  1 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.
2 Downgraded due to inclusion of cross-sectional study design and inconsistency.
3 Downgraded due to inclusion of cross-sectional study design and inconsistency.
4 Downgraded due to inconsistency.
5 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.

Noise pollution
Residential aircraft noise exposure Residential noise exposure Predicted aircraft and road traffic noise exposure High and low aircraft noise exposed schools Predicted road traffic noise exposure 1 Downgraded due to unclear risk of bias and inability to assess consistency.
2 Downgraded due inclusion of cross-sectional study design and inability to measure consistency-upgraded 1 level due to large effect size.
3 Downgraded due inclusion of cross-sectional study design and inability to measure consistency.
4 Downgraded due inclusion of cross-sectional study design and inconsistency.
5 Downgraded due inclusion of cross-sectional study design and inability to measure consistency.
6 Downgraded due inclusion of cross-sectional study design and inconsistency.
7 Downgraded due inclusion of cross-sectional study design and inconsistency. 1 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.
2 Downgraded due to inclusion of cross-sectional study design and inconsistency.
3 Downgraded due to high risk of bias, indirect physical environmental exposure metric, and inability to assess consistency. 4 Downgraded due to inclusion of cross-sectional study design.
5 Downgraded due to inability to measure consistency.
6 Downgraded due to inconsistency.
7 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.
8 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.
9 Downgraded due to inclusion of cross-sectional study design and inconsistency.
10 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.
11 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency. 1 Downgraded due to high risk of bias, indirect measure of physical environment, and inability to assess consistency.

Summer seasonality
2 Downgraded due inclusion of cross-sectional study design.
3 Downgraded due inclusion of cross-sectional study design and inconsistency.
4 Downgraded due inclusion of cross-sectional study design.
5 Downgraded due to indirectness and inability to assess consistency.
6 Downgraded due to inclusion of cross-sectional studies and inability to assess consistency.
7 Downgraded due to inability to assess consistency. 1 Downgraded due to high risk of bias and inability to assess consistency.
2 Downgraded due to high risk of bias and inability to assess consistency.
3 Downgraded due to high risk of bias and inconsistency.
4 Downgraded due to high risk of bias and inconsistency of results.
5 Downgraded due to inconsistency of results.
6 Downgraded due to inability measure consistency.

Urbanicity (Reference category urban)
Urban vs. rural residence Schools recruited from urban and rural areas Rural and urban children recruited from Head start centres. Urbanicity (density of residences surrounding address)  1 Downgraded due to inclusion of cross-sectional study design and inconsistency.
2 Downgraded due to inability to assess consistency.
3 Downgraded due inclusion of cross-sectional study design and inability to assess consistency.
4 Downgraded due to inability to assess consistency.
5 Downgraded due inclusion of cross-sectional study design and inability to assess consistency.
6 Downgraded due to High risk of bias and inability to assess consistency.
7 Downgraded due to High risk of bias and inability to assess consistency. 1 Downgraded due to High risk of bias and inability to assess consistency.
2 Downgraded due inability to assess consistency.
3 Downgraded due inability to assess consistency.

Music
Alternating 15 m periods of silence and Instrumental music. Aggressive content of child's favourite music artists Observer rated child aggression Low quality 1 -No effect a (1) n.a.
b Coyne and Padilla-Walker, 2015 [183]: Self-completed 5-item questionnaire on physical aggressive behaviour. 1 Downgraded due to high risk of bias and inability to assess consistency. 2 Downgraded due to high risk of bias, indirect measure of physical environment, and inability to assess consistency. b Barger et al., 2020 [188]: Parental reported severity of child's conduct problems retrieved from The National Survey of Children's Health (NSCH) [189]. c Durand and Mapstone, 1997 [190]: Observer rated frequency of challenging behaviours.
1 Downgraded due to inclusion of cross-sectional study design and inability to assess consistency.
2 Downgraded due to high risk of bias and inability to assess consistency.
3 Downgraded due to High risk of bias, Indirectness, and inability to assess consistency.
4 Downgraded due to inability to assess consistency.
5 Downgraded due to inability to assess consistency.
6 Downgraded due to inability to assess consistency.
7 Downgraded due to inability to assess consistency.
Inconsistent evidence for harms or benefits was reported across eight studies [106][107][108][109]112,113,117,121] that examined associations between satellite derived neighbourhood greenspace (NDVI) and parental-reported child aggression related outcomes. Two studies [101,109] examining the association between parental-reported child aggression and conduct problems and percentage of land designated as natural land, reported high-quality evidence. Proximity of the child's residence to greenspace was inconsistently associated with parent reported conduct problems across three studies [106,108,114]. Very low-quality evidence [111,116,120] reported no relationship between percentage of neighbourhood greenspace and both child and parent-reported conduct problems. Moderatequality evidence from three studies [34,111,120] reported inconsistent beneficial effects of access to private garden space on parent-reported conduct problems.

Environmental Sound and Noise
Three longitudinal and eight cross-sectional studies (n = 23,665) assessed the association between environmental noise pollution including road traffic, construction noise, aircraft noise and aggression outcomes [123][124][125][126][127][128][129][130][131][132][133]. These studies were primarily conducted in the UK, Spain, Germany, and the Netherlands, and one study in China. Three of these studies [124,131,132] used data from the multi-national RANCH study examining the influence of high and low road and aircraft noise on the behaviour of pupils who attended schools that were close to main roads or under flypaths. Two studies were judged to be of unclear RoB [126,130], with the majority being rated as low RoB. A very low-quality evidence for harmful association [130] between residential aircraft noise exposure and increased child annoyance was reported. Similarly, low-and very low-quality evidence was found for associations between increased residential noise [126], predicted air and road traffic noise [131], and heightened self-reported child annoyance. Schools located in areas of high aircraft noise were associated with increased child-reported annoyance [127][128][129], but inconsistently correlated with parent-reported child conduct problems ( [127,129], both very low quality). Two studies [124,132] examining the role of residential aircraft noise on the parent-administered conduct problems subscale of the SDQ reported no association (low quality). Five studies [123][124][125]133,185] reported very low quality inconsistent evidence for estimated noise exposure effects on parent-reported child aggression.
Two studies (longitudinal and within-group repeated measures) from the USA (n = 658) assessed the association of music on childhood aggression [183,184]. Both studies were rated as high RoB. Aggressive or sexual music content was associated with increased self-reported aggressive behaviour in adolescents ( [183], very low quality). Low-quality evidence reported no association between alternating periods of instrumental music and observer rated aggressive behaviours [184]).

Air Pollution
Eight longitudinal and six cross-sectional studies from Lithuania, China, Korea, Iran, Canada, USE, and the UK (n = 45,607) explored the influence of air particulate matter on aggressive behavioural outcomes in typically developing children and young people [34,107,116,123,135,136,138,141,142,[144][145][146][147][148][149]. One study was rated as high RoB [146], one as unclear RoB [136], and the remaining as low RoB.
Five studies [135,136,146,147] provided either low-or very low-quality evidence supporting the harmful influences of tobacco smoke exposure across various aggressive behavioural questionnaires. Very Low-quality evidence for a harmful association [132] between active or passive tobacco exposure and child self-reported anger and aggressive behaviour was found.
Three studies [116,141,149] examined the relationship between Nitrogen Dioxide (NO2) exposure and child self-reported conduct problems symptoms and reported inconsistent evidence for a harmful association (Very Low quality). No effect of Elemental Carbon Attributed to Traffic (ECAT) on parent-reported externalising behaviours (BASC-2) was found ( [145], Moderate quality).
In addition, there was inconsistent evidence for an association between exposure to particulate matter less than 2.5 microns (PM2.5), and child self-reported conduct problems [141,149]. No effect was found in a study that examined the influence of PM2.5 on parent reported conduct problem scores [107]. Another study by Loftus et al., 2020 [144] explored the influence of exposure to particulate matter less than 10 microns (PM10) on parent reported child aggressive behaviours but it did not show a significant association (Very Low quality). Ambient air lead exposure (PbA)) was associated with high parent-reported aggressive behaviour ( [148], Very Low quality).
The study by Muñoz-Reyes et al. [158] contrasted the frequency of observed aggressive behaviours during the warm season (summer/spring) with the frequency of such behaviours during the cold season (autumn/winter), reporting Very Low-quality evidence for harmful effect of warm seasonality. Low-quality evidence associated increased humidity with harmful increases in teacher-reported child aggressive behaviours [151,156]. Studies examining the effects of sunlight exposure on teacher reported [151,156] and child self-assessment [152] behavioural outcomes reported inconsistent or no evidence, respectively (very low quality). Low-and Moderate-quality evidence for the harmful influence of increased temperature on teacher and parent-reported child aggression symptoms was reported in three studies [151,156,159]. However, we found one Very Low-quality study that provided evidence for beneficial effect of temperature on children's self-reported anger [152]. Aggression during summer recess was lower compared with aggression during the school year ( [154]; Low quality). No association between hours of precipitation per day and children's self-reported anger was found ( [152], Very Low-quality evidence). Finally, a study carried out by Lochman et al. [153] examined longitudinal associations between tornado exposure and externalising symptoms, and reported a harmful associations of Moderate quality.

Spatial Density and Interior Design
Four observational and two longitudinal studies [34,[160][161][162][163]165] (n = 8568) from the USA and the UK examined spatial density and architectural design in relation to childhood aggression. RoB was judged as high in all studies except one rated as unclear [163] and one rated as low [34]. A study [165] reported a beneficial effect of increased playroom openness, but no effect of space per child or room group size on observed aggressive behaviours. Low-quality evidence assessing the association between high density (in comparison to low density) child playrooms and frequency of aggressive behaviours reported inconsistent results [160][161][162]. Moderate-quality evidence examining the effect of overcrowding in the home [34,163] reported inconsistent associations with parent-reported conduct problems but was associated with reduced teacher-reported externalising behaviours.
Three studies, two quasi-experimental and one longitudinal (n = 8257) conducted in Iran, the UK, and the USA examined the associations between interior design features and childhood aggression [34,179,181]. Low-quality evidence of association ( [181] unclear RoB) between red painted classroom walls and increased self-reported aggression was found.
In-patient psychiatric ward sensory room modifications were correlated with beneficial reductions in observer rated aggressive behaviour ( [179] unclear RoB Moderate quality). Additionally, presence of damp in the house was associated with elevated trajectories of conduct problems in children ( [34], Moderate quality).

Urbanicity and Rurality
Three longitudinal and three cross-sectional studies (n = 17,630) from the USA, the Netherlands, and Thailand explored the influence of urbanicity and rurality of residence on children's aggressive behavioural outcomes [34,164,168,170,172,176]. One study was rated as high RoB [168] with the remaining assessed as low or moderate RoB. One study [176] reported inconsistent associations between the location of the participants and scores across three self-reported aggression outcomes (Very Low quality). Moderate-quality inconsistent evidence [34,172] was reported for the effect of urban residence on child conduct problems and aggressive behaviour in parent-reported questionnaires, whilst evidence for a lack of association was found for teacher-completed aggression outcomes [172]. Another study [170] examined the effects of urban or rural settings on aggressive behaviours in schoolchildren attending schools from either setting. It reported no association of setting with parentreported behaviours, but a harmful effect of urban school location on teacher-assessed behaviours (both Very Low quality). Very Low-quality evidence reported no association between children recruited from rural or urban Head Start centres and teacher-reported anger ratings ( [168]: AML Behaviour Rating Scale). Neighbourhood urbanicity (mean number of addresses within a 1 km radius of participant's residence) was associated with increased teacher-reported child problem behaviours ( [166] Low quality).

Neurodiverse Children
Six studies (n = 79,249) from the USA, Pakistan, the UK, and Australia included neurodiverse participants exclusively [34,188,[190][191][192][193]. The studies are heterogenous utilising a variety of designs including longitudinal, cross-sectional, quasi-experimental, interventional, including two case studies. Two studies were judged to be of high RoB [190,192], one unclear [193] and the remaining three of low RoB.
Baird et al. [34] explored interaction effects between a sub-sample of children with intellectual disability (assessed via cognitive measures) from the Millennium Cohort Study (MCS) and various physical environmental exposures (neighbourhood greenspace: NDVI, access to a private garden, air pollution: NO 2 , urban or rural residence, household density, presence of damp. The authors reported no mediating influence of intellectual disability on the association between environmental exposures and children's conduct problem trajectories, except for household density (beneficial effect, moderate quality). Another study [188] reported no correlation between urban tree canopy coverage and frequency of aggressive behaviours in children with ASD but found an association between residing in lower urban tree canopy areas with increased parent-reported conduct problem severity. However, the evidence was deemed to be of low quality in both studies. The case study by Durand and Mapstone [190] examined the impact of fast and slow beat music on a child with intellectual disability, reporting reductions in observed frequency of aggressive behaviour during the fast beat condition and increases during the slow beat condition in comparison to a no-music baseline (Low quality). Additionally, a clinical trial of new age and classical music [191] provided Moderate-quality evidence for the beneficial effects of music on self-reported aggression in children with intellectual disabilities. A case study [192] assessing the impact of spatial proximity between an adolescent girl with intellectual disability and the therapist, provided a Very Low-quality evidence for a correlation between closer proximity and increased duration of observed aggressive behaviours. Finally, a study [193] examined the efficacy of modified sensory rooms in reducing distress in adolescent psychiatric inpatients, reporting additional benefits for individuals who had a history of aggression (Moderate quality).

Discussion
This is the first narrative review that updates previous literature across several environmental domains as well as including neurodiverse children, a previously under reported population in other reviews.

Greenspace
We found evidence that supports the therapeutic benefits of increased natural land and greenness surrounding child residences. Previous reviews have also shown associations between greenspace exposure and reductions in violent behaviours [194,195].
The greenspace evidence synthesised primarily supports the therapeutic influence of neighbourhood nature exposure on child aggressive behavioural outcomes. These effects, at least partially, were also present in NDDs populations. Whilst more epidemiological and experimental research paradigms are required to solidify the evidence for this therapeutic relationship and understand its underlying mechanistic pathways, we provide initial evidence for the role of nature in reducing aggression in neurotypical and diverse children. Initial attempts at establishing guidelines for integration and therapeutic adoption are beginning to be developed [196]. Studies examining socio-cultural barriers to children accessing urban greenspaces [197] are of crucial importance, but these findings need to be communicated to institutions and policy decision makers. We also recommend future experimental studies that aim to elucidate the underlying (neuro)mechanistic pathways via which nature exposure conveys these potential benefits. Advances in this regard would drastically redefine architectural and urban design for physical and mental health.

Noise Pollution
Children appear to consistently self-report higher aggressive and annoyance related behaviours related to environmental noise, whereas parent reported outcomes either show a lack of association or inconsistent associations both for harm and benefit. This may suggest that noise exposure operates on pathways involving subjective annoyance and irritation which may not translate into objective longer-term increases in aggression problems. Additionally, although noise annoyance may not play a direct role in the aetiology of those problems, noise exposure of individuals who experience frustration or irritable mood has been shown to increase its severity [25,26]. Noise pollution, therefore, may not operate as a causal mechanism of aggression, but exacerbate pre-existing manifestation, potentially via draining of attentional and cognitive resources, leading to increased self-regulatory difficulties [27].

Air Pollution
We found absent and inconsistent associations between ECAT, particulate matter less than PM2.5, particulate matter less than PM10 and NO 2 exposure and childhood aggression problems. Tobacco smoke exposure showed a harmful association with aggressive behavioural outcomes irrespective of who was the outcome assessor. We also found this harmful association for childhood exposure to ambient air lead exposure (PbA). The lack of association of PM2.5 and PM10 with these behaviours is potentially anomalous when considering research that has linked air pollution with increased risk of mental health disorders [198]. The harmful effects of tobacco smoke and ambient lead exposure may increase the risk of neuropsychiatric disorders and violent crime, possibly via neuroinflammation [19][20][21][22][23][24].
Whilst none of the retrieved articles examined the effects of air pollution on neurodiverse children, it was shown that families of these children disproportionately reside in areas of higher particulate concentration than those of typically developing children [49], as well as exhibiting elevated rates of aggressive behaviour [199,200].

Meteorological Effects
Summer seasonality, humidity, temperature, and previous tornado exposure were consistently correlated with increased childhood aggressive behaviours. We found little evidence for either harmful or beneficial effects of ambient temperature and seasonality. Previous studies suggest that humidity compounds the negative effects of heat on mental health [201], as well as being associated with increased emergency department visits for mental health problems [202]. Elevated temperature has also been associated with increased violent crime [203]; however, those associations warrant further examination.

Spatial Density
The negative impact of high spatial density on aggression in young people [204] and inpatients in psychiatric wards has been highlighted previously [205][206][207]. Notwithstanding the beneficial effects of increased playroom openness, inconsistent influences for other spatial characteristics prevent a firm explanation of findings. Theories have posited that high spatial density triggers perceptions of crowding and a subsequent physiological stress arousal response [30][31][32]. Further studies on possible mechanistic pathways between high spatial density and aggression in children could lead to therapeutic adaptations in clinical and residential spaces [208].

Urbanicity and Rurality of Residence
Due to the quality of retrieved evidence, we were unable to extricate any definitive conclusions for associations between urban or rural residence and childhood aggressive behaviours. This is potentially anomalous considering that children residing in rural areas are exposed to more greenspace which generally appears to have calming effects [209][210][211][212][213][214]. Rurality, however, is only one factor in a great number of confounders on childhood aggression. Furthermore, studies do not often use operationalised definitions of "rural" or "rurality" [215], potentially leading to heterogeneity in the underlying conceptual constructs being examined, limiting the replicability and specificity of results.

Interior Design and Housing Quality
Previous work has associated damp problems with increased toxic mould, contributing to poor air quality [216] and/or potential neuroinflammatory and/or neurotoxic responses [20,22]. Damp in a house may also be associated with other adversities such as low socio-economic status and household disruption [217] exemplified by previous research linking poor household conditions to psychological distress [218].
Whilst preliminary evidence from this review supported the positive impact of modified sensory rooms to de-escalate aggression, it is very limited in scope. One study [35] suggested that the increased reduction in distress related to sensory deficits may be attributable to a sense of autonomy children and young people may gain by using the room unaccompanied.

Music
We found preliminary evidence for the therapeutic potential of music in neurodiverse children which is similar with findings reported in adults [219]. Music listening has been associated with a broad range of psychological and physiological benefits [38][39][40]. Some [36,37] have stated that the therapeutic influences of music may operate mechanistically via enhancing emotional regulation, but such evidence is not yet available [41]. Music is a complex physical phenomenon, which requires additional targeted research to examine its effects on aggressive behaviours in typically developing and neurodiverse young people.

Strengths and Limitations
This review is comprehensive and has examined the evidence of a wide range of environmental exposures in relation to the display of aggressive behaviours in typical and neurodiverse children. To the best of our knowledge, this is the first review that comparatively examines available research on environmental determinants of aggression in these two groups. The review shows clearly the disproportionately sparse literature relating those children and the physical environment despite the fact that they are more likely to be affected by social adversity, poor housing, air pollutants, and poverty [48][49][50]220,221].
The incorporation of GRADE to assess the quality of evidence in this review may well be simultaneously both a strength and a limitation. Whilst it facilitated the examination of the certainty of included evidence, the adaptation of GRADE for use in a non-metaanalytic review including epidemiological studies, may, as highlighted previously [95], inadvertently result in downgrading of evidence irrespective of study quality. We also adopted a modified risk of bias protocol which may have impacted the RoB assessments of included studies. There may also have been potential conflicts of interest based on the source of funding which we did not consider in this review.
A final limitation of this evidence synthesis is the inclusion of studies that adopt a diverse range of heterogenous physical environmental exposures and metrics. As has been highlighted previously by experts in physical environmental epidemiological analysis on child socio-cognitive outcomes [116], further research is needed on improving environmental measures of aspects such as air pollution exposure, and access to and quality of children's greenspaces. Developing more holistic, accurate, and reliable measures of environmental exposures will facilitate novel research paradigms (computational, simulatory and experimental) that can elucidate the influences of these aspects, reciprocally informing direction for future research into (neurobiological)mechanistic pathways.

Conclusions
Physical environmental exposures sit at the intersection of social, biochemical, and (epi)genetic aetiological influences on the development and progression of a spectrum of physical and mental health outcomes. Further research can support stakeholders, ranging from city planners and environmental legislators to politicians and clinicians, in considering the role of the physical environment in the context of adverse impact on child (neuro)development. Whilst there is obvious need to further examine environmental and climate influences on mental health of all children, particular attention must be paid to neurodiverse children and their families. A recent report recommended that in order to pursue and achieve health parity for those children, we must "reduce poverty and improve living environments" [50]. Research focusing on that population will help to bridge the equity gap that has significant therapeutic and health implications for all citizens.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph20032549/s1, Table S1: search strategy.  Acknowledgments: We thank Nancy Kouroupa, Steven Naughton, Laura Paulauskaite, Peiyao Tang and Rachel Royston for assisting in co-screening studies; Bori Vegh for support with preparation of the manuscript; Louise Marston, for guidance on study heterogeneity and narrative synthesis reporting; Deborah Marletta for assistance with the electronic database protocol development.

Conflicts of Interest:
Multiple studies included in this review involved the authorship of Eirini Flouri. Flouri had no influence/involvement in the selection or appraisal of these studies. No other authors report any potential conflict of interest.