key: cord-0301307-ebnxgm1u
authors: Maes, M. J. A.; Pirani, M.; Booth, E. R.; Shen, C.; Milligan, B.; Jones, K. E.; Toledano, M. B.
title: Benefits of natural habitat particularly woodland on children's cognition and mental health
date: 2021-01-13
journal: nan
DOI: 10.1101/2021.01.12.21249675
sha: 450e17ff8fbfcfbcf247292ce21af5193f477cc1
doc_id: 301307
cord_uid: ebnxgm1u

Life in urban areas is associated with adverse human health effects, including risks of developing cognitive problems and mental health issues. Many epidemiological studies have established associations between urban nature, cognitive development and mental health, but why specifically we receive these health benefits remains unclear, especially in children. Here, we used longitudinal data in a cohort of 3,568 children aged 9 to 15 years at 31 schools across London to develop a model and examine the associations between natural habitat type, and children's cognitive development and mental health. We show that, after adjusting for other environmental, demographic and socioeconomic variables, higher daily exposure rates to natural habitat and particularly woodland were associated with enhanced cognitive development and mental health from late childhood to early adolescence. Our results suggest that optimising ecosystem services linked to cognitive development and mental health benefits should prioritise the type of natural habitat for sustainable urban planning decisions.

studies have established associations between urban nature, cognitive development 26 and mental health, but why specifically we receive these health benefits remains 27 unclear, especially in children. Here, we used longitudinal data in a cohort of 3,568 28 children aged 9 to 15 years at 31 schools across London to develop a model and 29 examine the associations between natural habitat type, and children's cognitive 30 development and mental health. We show that, after adjusting for other 31 environmental, demographic and socioeconomic variables, higher daily exposure 32 rates to natural habitat and particularly woodland were associated with enhanced 33 cognitive development and mental health from late childhood to early adolescence. 34

Our results suggest that optimising ecosystem services linked to cognitive 35 development and mental health benefits should prioritise the type of natural habitat 36 for sustainable urban planning decisions. 37 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint

Over 55% of the human population is now living in cities, with an estimated 2.5 billion 39 more people to be added to the urban population by 2050 1 . Although urban 40 populations often have better socioeconomic livelihoods, living in urban areas is 41 associated with a number of adverse human health effects 2 . In particular, 42 urbanisation is associated with risks of developing cognitive problems and mental 43 health issues 3,4 , and is linked with various demographic and socioeconomic 44 factors 5,6 . The COVID-19 pandemic has further exacerbated mental health problems 45 such as stress, anxiety and depressive symptoms, amongst others 7,8 . In London, for 46 example, an estimated 1 in 4 individuals will experience a diagnosable mental health 47 condition in any given year, costing £26 billion annually through poorer education, 48 employment and quality of life, affecting London's economy, population and 49 infrastructure 9 . The renewed focus on cognition and mental health due to the 50 negative effects of the COVID-19 pandemic highlight the importance to understand 51 the mechanisms and dynamic interactions attributed to a higher risk of cognitive 52 problems and mental health issues in urban areas, which until now remain unclear. 53

Emerging evidence suggests that surrounding environments, particularly exposure to 54 natural areas plays an important role in cognitive development and mental health 10-55 12 , and are part of a range of ecosystem services (ES) characterised to impact 56 human health and well-being 13 . Relative risk estimates of the association between 57 natural areas and cognitive development and mental health have been comparable 58 in magnitude to family history and parental age, and higher than the degree of 59 urbanisation 11 . However, these associations lack a mechanistic understanding 14 . 60 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint 7 type of urban natural habitat by fitting our longitudinal models (Supplementary 131 Methods 1). We found that children's cognitive development did improve with higher 132 DER to natural space. A difference of one in natural space DER corresponded to an 133 expected positive difference of 0.03 (95% credible interval [CI]: 0.01, 0.06) points in 134 the child's cognitive development using the EF score when controlling for other 135 demographic, environmental and socio-economic fixed effects ( Fig. 2a and  136 Supplementary Figure 1a) . We also provide the results for our mental health 137 outcomes with natural space DER (Fig. 2b,c and Supplementary Figure 1b,c) , where 138 we found no improvement of mental health with higher DER to natural space, 139 meaning the 95% CI included the null effect for both models. Our results for the tier 2 140 models, where both green and blue space was used for the natural habitat 141 characterisation were almost identical to our tier 1 models for natural space DER. 142 This is probably due to a high collinearity between our DER for natural space and 143 green space (Supplementary Table 1 ) as children's DER to blue space was low. This 144 also meant that our models did not find an improvement of children's cognitive 145 development and mental health with DER of blue space ( To further assess the role of different types of urban natural habitat to children's 148 cognitive development and mental health, we characterised green space into two 149 distinctive natural habitat types, i.e. grassland and woodland. We found that 150 children's cognitive development and mental health did improve with higher DER to 151 woodland. When all other risk factors were held constant, there was a beneficial 152 contribution to cognitive development by 0.42 (95% CI: 0.21, 0.57) points using the 153 EF score and a reduction in the risk of emotional and behavioural problems by -0.17 154 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint 8 (95% CI: -0.32, -0.03) points using the SDQ total difficulties score ( Fig. 2 and  155 Supplementary Figure 3) . We found no improvement of overall well-being with higher 156 DER to woodland ( Fig. 2c and Supplementary Figure 3c) . When comparing those 157 children exposed to the highest level of woodland (~38%) to those exposed to the 158 lowest level of woodland (0%) in our study, we estimated a percent change in 159 cognitive development of 6.83% (95% CI: 3.41, 9.11) using the EF score, and a 160 percent change in the risk of emotional and behavioural problems of -16.36% (95% 161 CI: -27.49, -3.50) using the SDQ total difficulties score. We found no improvement of 162 children's cognitive development and mental health with a higher DER to grassland 163 with the exception of our outcome for overall well-being using the HRQoL score ( The role of other risk factors for cognition and mental health. We fitted our 166 longitudinal models with a number of other risk factors to account for demographic, 167 environmental and socio-economic factors that are known to influence children's 168 cognitive development and mental health 5,6 . We found that our outcomes for 169 children's cognitive development and mental health were influenced by the child's 170 age, ethnic background, gender, parental occupation and type of school 171 (Supplementary Table 2 ,3,4). When compared to independent schools for example, 172 state schools were predicted to result in a negative contribution to children's 173 cognitive development and mental health by a percent change decrease of -5.10% 174 (95% CI: -6.05, -4.30) using the EF score, a 10% (95% CI: 5, 15) increase in the risk 175 of emotional and behavioural problems using the SDQ total difficulties score, and an 176 increase in odds of exhibiting low overall well-being by 57% using the HRQoL score 177 (95% CI: 19, 104). In addition to this, we found that air pollution influenced children's 178 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint cognitive development in most of our models (Supplementary Table 2 ). When 179 removing demographic, environmental, and socio-economic factors from our models, 180 we showed that modelled environmental variables were, in general, tenfold smaller 181 than the contribution of our demographic and socio-economic variables 182 (Supplementary Table 5 ). This stepwise exclusion of fixed effects from our models 183 highlights the relative importance of our demographic and socio-economic variables 184 to children's cognitive development and mental health. 185

To test the robustness of our findings, we did a series of sensitivity analyses to 186 assess which models perform best for evaluating the association between children's 187 cognitive development and mental health, and types of urban natural habitat. This previously hypothesised by other studies 25 . When using a different weighting for our 197 DER, we found that our models showed consistent patterns when we modelled with 198 a DER based on a daytime or full day weighting (Supplementary Table 2 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint

To our knowledge, this is the largest epidemiological study to report on the impact of 201 natural habitat type exposure on cognitive development and mental health in children 202 from late childhood to early adolescence. Our results showed a strong association 203 between woodland exposure, children's cognitive development and mental health. 204

We also found that exposure to natural space or green space was associated with a 205 beneficial contribution to cognitive development, while there was a weaker 206 association for our mental health outcomes. Finally, we did not find a consistent 207 association of blue space or grassland exposure with cognitive development and 208 mental health. 209

Overall, we observed that exposure to woodland was associated with a beneficial 210 contribution to cognitive development and a lower risk of emotional and behavioural 211 difficulties from late childhood to early adolescence. This is in line with previous 212

reports of positive impacts from woodland on physical and mental health 19,23,30 , with 213 the exception of a study performed in central Scotland 31 . Forest bathing, for 214 example, is a relaxation therapy that has been associated with physiological 215 benefits, supporting the human immune function, reducing heart rate variability and 216 salivary cortisol, and psychological benefits such as reduced feelings of hostility and 217 depression 19,30 . However, the hypothetical mechanisms why we experience these 218 psychological benefits from woodland remain unknown. Higher audio-visual 219 exposure through vegetation and animal abundance has been documented to 220 improve mental health, of which both features are expected in higher abundance in 221 woodland 17,32 . Even though our results show that urban woodland is associated with 222 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint children's cognitive development and mental health, the mechanistic pathway to 223 explain this association in urban areas remains unknown. 224

Our results also showed that exposure to natural space or green space was 225 associated with a beneficial contribution to children's cognitive development, which 226 was consistent with previous studies 12,33 . For the mental health outcomes, our 227 findings for weaker associations with exposure to natural space or green space is 228 consistent with the variability in these relationships found in previous studies 10,11,21,34 . 229

One reason explaining this variability may be that most studies, including this study, 230

do not account for the quality of green space, which has been proposed as more 231

important than the quantity of green space 35 . Nevertheless, systematic reviews 232 suggest that nature positively influences mental health; even though, evidence is 233 often limited to cross-sectional studies, and inadequate particularly for children 28 . 234

We did not find a consistent association between blue space exposure, cognitive 235 development and mental health. However, we cannot dismiss that blue space may is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. follow-up assessment (Fig. 1a) . This subset excluded 8 schools due to low sampling 328 size (< 15 children per school). Included children were on average 12 and 14.2 years 329 old during the baseline and follow-up assessment respectively, and 57.9% of them 330 were female. The children (n = 3,568) were part of 31 schools across London, of 331 which 12 were independent schools and 19 were state schools. Of the 31 332 participating schools, 3 were located outside the Greater London Authority (GLA) 333 administrative area (Fig. 1a) . During the assessments, information was gathered on is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. higher EF values indicated better cognitive performance (Fig. 1b) . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint

We assessed children's mental health from the self-reported SDQ and the 364 KIDSCREEN-10 Questionnaire taken by each child 49 . The SDQ total difficulties 365 score assesses the emotion and behaviour of children and was calculated by 366 summing the scores for the four difficulties subscales on emotional problems, 367 conduct, hyperactivity and peer problems. Each subscale comprised of five items 368 that can be scored 0, 1 or 2 and each subscale score can therefore range from 0 to 369

10. An SDQ total difficulties score was treated as count data where a higher value 370 represented more behavioural difficulties (Fig. 1c) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint Quantification of natural habitat composition. Our exposure assessment of 388 natural habitat was based on a three-tier stepwise characterisation: (1) natural 389 space, (2) green and blue space, and (3) grassland and woodland. We used different 390 data sources to quantify the natural habitats surrounding the residential and school 391 area of each child. Firstly, we generated a NDVI spatial layer of our study area using 392 data from the Sentinel-2 satellite at 10 m spatial resolution 52 . NDVI is a unit-less 393 index of relative overall vegetation density and quality based on differential surface is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint

We calculated each child's proportionate DER to each natural habitat 411 characterisation in buffer areas of 50 m, 100 m, 250 m and 500 m around the 412 residential and school area: 413

where DER is the daily exposure rate, RER is the residential exposure rate and SER 415 is the school exposure rate. We assumed each child spent the weekend in their 416 residential area, while we weighted weekdays by the daytime (12 hours) children 417

were assumed to spend at home (4 hours) and at school (8 hours). We selected 418 different buffer areas to assess the consistency of our results in a comparable 419 manner with previous studies 11,12,25 . Based on the above formula, we calculated 420 natural space DER by combining our NVDI and water maps. Then, we calculated 421 green and blue space DER by using our NDVI and water maps separately. Finally, 422

we calculated grassland and woodland DER by combining our NDVI and height 423 strata map. The different spatial resolutions of our NDVI and height strata map 424 resulted in classification errors where pixels were misclassified as grassland or 425 woodland when in fact it was part of the built environment. To correct for this, we 426 excluded buildings from these maps using the buildings feature from OS Open Map 427 (Supplementary Figure 4d) 55 . It was difficult to use blue space DER of the 3,568 428 participants because 2,383 children (66.8%) had, for example, no blue space within 429 250 m. We therefore reclassified blue space into tertiles (three levels: level 1 -no 430 blue space, level 2 -blue space with a DER below the mean, and level 3 -blue 431 space with a DER above the mean). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint 21 woodland) to 0.99 (between natural space and green space) (Supplementary Table  457 1). The high Pearson's correlation coefficient was not considered a problem because We assumed a Gaussian, Poisson and Binomial distribution for the EF score, SDQ 464 total difficulties score and HRQoL score, respectively. We included a random effect 465 term for child identifier to allow for between-child variance, while we used a random 466 effect term for tests at the time of visit (two levels: baseline or follow-up) for each 467 child to introduce correlation among the repeated measurements. School was not 468 added as an additional random effect in our multilevel model because it did not 469 improve the model fit, and three different cross-validation techniques were used for 470 model comparison and selection (Supplementary Table 8 ,9,10). Fully adjusted 471 models included natural habitat DERs, age, area-level deprivation, ethnicity, gender, 472 parental occupation and school type, and models with EF score were additionally 473 adjusted for air pollution. Additionally, we did a stratified analysis to investigate 474 potential changes in point estimates and avoid potential bias from over adjustment 475 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. Figure 5) . 482

We performed the following sensitivity analyses to determine the best models for 483 evaluating the association with natural habitat DER by fitting additional Bayesian 484 mixed-effect models for (i) the association with different buffer areas (Supplementary 485 Figure 1,2,3) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. The full model outputs that support the findings of this study are available in the 507 Supplementary Information. 508

The source code to compute NDVI from satellite data using Google Earth Engine is 510 available at earthengine.google.com. The code for processing raw LiDAR data, 511 creating our environmental exposure variables and modelling our data is available at 512 github.com/MikaelMaes/HumanExposure.git. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint

World Urbanization Prospects: The

City planning and population health: a global challenge

Urban 519 mental health

Impact of urbanization on 521 cognitive disorders

Gender differences in mental health

Greener living environment healthier people? Exploring 600 green space, physical activity and health in the Doetinchem Cohort Study

Exploring the relationship between Public Open Space attributes and mental 604 health in Perth

Residential exposure 606 to visible blue space (but not green space) associated with lower psychological 607 distress in a capital city

A new approach to quantify and 609 map carbon stored, sequestered and emissions avoided by urban forests

What can global positioning systems tell us about the contribution of different 613 types of urban greenspace to children's physical activity? Health Place 18

System of Environmental-Economic Accounting -Ecosystem Accounting: 616 Draft for the Global Consultation on the complete document

Socio-Economic Determinants of Crime Rates

Modelling Local Area Police-Recorded Crime

Cohort Profile: The Study of Cognition

Mobile Phones (SCAMP)

The National Statistics Socio-economic 622 Classification: Origins, Development and Use

Deprivation and health in Scotland. Health Bull. 625 (Raleigh)

Office of National Statistics

Assessment of neuropsychological function 629 through use of the Cambridge Neuropsychological Testing Automated Battery: 630 Performance in 4-to 12-year-old children

Memory Task Predicts Arithmetical Performance 2 Years Later

Trail Making Test A and B: Normative data stratified by age 636 and education

Theory and methodology in executive function research

The Strengths and Difficulties

Questionnaire: a pilot study on the validity of the self-report version

The KIDSCREEN Group Europe. The Kidscreen questionnaires--quality of life 644 questionnaires for children and adolescents

Quality of Life Based on the KIDSCREEN-27: Child Self-Report, Parent 647 Ratings and Child-Parent Agreement in a Swedish Random Population 648

Sentinel-2 User Handbook

Normalized difference vegetation index (NDVI) as a marker 651 of surrounding greenness in epidemiological studies: The case of Barcelona 652 city

Google Earth Engine: Planetary-scale geospatial analysis for 654 everyone

Characterizing forest ecological structure using pulse 657 types and heights of airborne laser scanning

The association between greenness and traffic-related air 660 pollution at schools

Association between Traffic-Related Air Pollution in Schools 662 and Cognitive Development in Primary School Children: A Prospective Cohort 663

Approximate Bayesian inference for latent 665

Gaussian models by using integrated nested Laplace approximations

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted January 13, 2021. ; https://doi.org/10.1101/2021.01.12.21249675 doi: medRxiv preprint