Results from an 18 country cross-sectional study examining experiences of nature for people with common mental health disorders

Exposure to natural environments is associated with a lower risk of common mental health disorders (CMDs), such as depression and anxiety, but we know little about nature-related motivations, practices and experiences of those already experiencing CMDs. We used data from an 18-country survey to explore these issues (n = 18,838), taking self-reported doctor-prescribed medication for depression and/or anxiety as an indicator of a CMD (n = 2698, 14%). Intrinsic motivation for visiting nature was high for all, though slightly lower for those with CMDs. Most individuals with a CMD reported visiting nature ≥ once a week. Although perceived social pressure to visit nature was associated with higher visit likelihood, it was also associated with lower intrinsic motivation, lower visit happiness and higher visit anxiety. Individuals with CMDs seem to be using nature for self-management, but ‘green prescription’ programmes need to be sensitive, and avoid undermining intrinsic motivation and nature-based experiences.

. Sample descriptive statistics for outcomes and moderator as a function of common mental health disorder (CMD) status. a Self-reported doctor-prescribed medication use. Data presented are the weighted means/SDs. Due to stratified sampling the raw and weighed data are almost identical. Missing data/'Unsure'/'Prefer not to answer': CMD status = 1; Intrinsic motivation = 187; Weekly visits = 5 PSP = 1039. Of the total sample 14,973 individuals reported a recent visit (79.5%), of those Missing data/'Unsure'/'Prefer not to answer': Happiness = 7; Anxiety = 8. Data for covariates presented in Supplementary Materials Table S4.

Results
Preliminary statistics. Table 1 presents descriptive statistics. There were n = 2,698 (14%) respondents who reported having at least one CMD, slightly below the 17% annual figure 25 . This is likely due to the fact we focused on medication use as a proxy for CMD, and current rather than annual rates. In terms of specific conditions the frequencies were: depression (only) n = 910 (4.8%); anxiety (only) n = 1013 (5.4%); both depression and anxiety n = 775 (4.1%). Supporting Hypothesis 1a, intrinsic motivation to spend time in nature, for the sample as a whole, was significantly above the mid-point of 4 (M = 5.80, SD = 1.36), t (18, 23; all ps < 0.001; all ds > 1.01. In terms of visit frequency, more than half of all participants in each of the four groups reported visiting nature ≥ once a week (differences between groups, i.e. Hypothesis 2, are tested below). Supporting Hypothesis 3a, overall sample happiness during the last visit was also significantly above the mid-point of 4 (M = 5.81, SD = 1.11), t (14,972  Potentially moderating role of perceived social pressure (PSP). The right hand columns in Table 2 add PSP and the interactions with the three CMD groups to the first set of models.     Fig. 1. As PSP increases for those without conditions, anxiety also increases (blue points). However, the relationship is even steeper for all CMD groups (red points), a synergistic effect suggesting that those with CMDs who feel pressured are also susceptible to feeling greater anxiety during visits.
Covariates. Due to space constraints results for covariates are presented and discussed in Supplementary Materials section 3.

Discussion
To our knowledge, this is the first multi-country analysis of nature-related motives, visits, and wellbeing experiences of people suffering from common mental health disorders (CMDs) such as anxiety and depression. We found that intrinsic motivation to spend time in nature was generally high, although lower among individuals Table 2. Intrinsic motivation to visit nature, likelihood of visiting nature and experiences on the most recent nature visit as a function of common mental health disorder (CMD) status and perceived social pressure to visit nature. *p < 0.05, **p < 0.01, ***p < 0.001. a Self-reported doctor-prescribed medication use. Different ns are due to missing data on predictor or outcome variables, with lower ns for the visit experiences due to only n = 14,973 people visiting a relevant location in the last four weeks. All analyses control for: sex, age, perceived financial strain, employment status, marital status, number of children in household, having a long-term limiting illness, smoking status, alcohol use, seasonal wave and country. Analyses for visit outcomes also controlled for number of companions, presence of dog, transport mode, travel time, visit duration and happiness or anxiety 'yesterday' (depending on outcome). Full models including all covariate data are available in Supplementary Materials. www.nature.com/scientificreports/ with CMDs, consistent with a general motivational deficit 18 . The majority of individuals with CMDs visited nature at least once a week, and contrary to experiential avoidance 19 , those with depression were just as likely, and those with anxiety more likely, to visit compared to those with no conditions. Consistent with research suggesting natural settings are 'calming' , 'stress relieving' 26 , and can help reduce negative, ruminative thoughts 26 , there was a tendency to report high levels of happiness, and low levels of anxiety, during recent blue space visits across the entire population; though experiences were slightly less positive among individuals with CMDs. Finally, consistent with self determination theory 21 perceived social pressure (PSP) to visit green/blue spaces was associated with lower intrinsic motivation among all groups, with the association for those with depression particularly pronounced. Higher PSP was also associated with lower visit happiness, and higher visit anxiety, especially for those with CMDs. However, PSP was associated with greater likelihood of visits, especially among those with anxiety. That intrinsic motivation to spend time in nature was high for people with CMDs and more than half were visiting nature ≥ once a week, suggests that many may be using nature for affect-regulation purposes 27 . Nevertheless, those with anxiety (in particular) reported lower visit happiness than those without conditions, and all three CMD groups reported higher anxiety. Although this may reflect lower intrinsic motivation (as predicted), one issue may be our focus on blue spaces, which offer potential threats (such as drowning) which may be particularly salient for people with anxiety. Although we think this unlikely, given most visits involved walking near blue spaces, further research focusing on various natural settings for people with different types of CMD is warranted. Another possibility may be that their experiences were genuinely less positive on average due to area-based factors. CMD rates are higher in low-income neighbourhoods 9 , and deprived areas tend to have lower quality natural environments 28 . Most blue space visits are local, and visits to poorer quality areas are associated with less positive experiential wellbeing outcomes 29 . Thus, individuals with CMDs may be reporting less positive experiences, in part, because they occur in poorer quality natural spaces, potentially exacerbating socio-economic related mental health inequalities 30 .
That higher PSP was associated with a greater likelihood of visiting nature, especially for those with anxiety, but lower visit-related happiness, and greater visit-related anxiety, suggests that although some (perceived) pressure may be effective at getting people out, it may undermine intrinsic pleasure from visiting nature 31 . However, due to the cross-sectional nature of the survey, we are unable to determine causality. It may be that this instead reflects less motivated people, who experience less positive wellbeing outcomes, going out to please others. More detailed longitudinal work is needed in the 'green prescription' field to unpack this issue.
Limitations. Although our sample was collected by an international polling company and was representative by age, gender and region within each country, we are not able to claim that the sample was fully representative in the respective countries. Moreover, the within country samples were also too small to test our hypotheses for each country separately. Further studies using larger samples which are able to be fully representative, such as Natural England's, Monitor of Engagement with the Natural Environment (MENE) survey (https ://www.gov. uk/gover nment /colle ction s/monit or-of-engag ement -with-the-natur al-envir onmen t-surve y-purpo se-and-resul ts) are needed across multiple countries to explore the generalisability of the current findings across different geographical and cultural contexts. We also recognise that depression/anxiety are not the only CMDs, and that CMDs are on a spectrum 32 , so the current findings are only meant as a first exploration and are by no means definitive. Further, all the data were self-reported, and we were unable to validate people's medication status or nature experiences. Although our prescription item is widely used 33 , it was also unable to account for length of use, dosage, access to other supporting services etc. or identify individuals who (a) meet criteria for a clinical diagnosis but are not currently receiving pharmaceutical treatment, (b) have particularly severe conditions, or (c) might be taking these drugs to help manage other conditions. Clearly, objective data on clinical diagnoses and treatment stage would be an important step for future research, especially as symptom severity and stage of treatment could be a critical factor in the success of green prescription uptake and adherence 34 . We are also aware that the explanatory power of our models was small (intrinsic motivation, visit likelihood) to moderate (visit experiences), and thus CMD status is only playing a very small role in these outcomes; there remain many other variables, beyond even our extensive set of covariates, which it is important to consider in improving our understanding of these outcomes. We also recognise that single items for measuring e.g. intrinsic and extrinsic motivation are not as robust as multi-item scales, but there is an inherent trade-off when collecting data from large-samples. Moreover, we recognise that we are assuming linearity in the response options for our outcome variables, when technically they could be considered ordinal scales. Nevertheless, it has long been recognised that findings are robust to this assumption for these kinds of dependent variables with results using linear and ordinal analyses producing essentially the same outcomes, with the linear approaches being far easier to interpret 35 . Finally, more in-depth qualitative work could enrich our understanding of whether and how people with CMDs are deliberately visiting nature for symptom self-management and how engagement with green/ blue prescription programmes might affect their motivations and experiences 34 .

Conclusions
Many individuals with CMDs are motivated to visit nature, and derive psychological benefits from such visits, though area level environmental inequalities may be undermining their potential for even better experiences. Nature based programmes such as 'green prescriptions' are becoming more prevalent. Our data suggest that perceived pressure to visit nature may increase visit frequency, but at the cost of undermining intrinsic motivation and the emotional benefits that might be achieved. Careful techniques to discuss accessing nature as a means of self-or supported-management (e.g. motivational interviewing 36 ), need to be integrated into these programmes if they are to offer clients the best support. The BIS was broadly concerned with people's experiences of the natural world and a range of health and wellbeing outcomes. There were seven main sections including: (1) Subjective well-being; (2) Visit frequency to a wide range of different natural environments; (3) Blue spaces in the local area and childhood experiences; (4) The most recent visit to a blue space; (5) Water quality; (6) Health and well-being generally; and (7) Demographics. Full methodological details are available on the Open Science Framework website 37 . For current purposes it is important to note that while some questions (e.g. visit frequency) asked about peoples' contact with a wide range of green (e.g. parks, woodlands, uplands) and blue (e.g. rivers, lakes, coasts) spaces, other questions, e.g. about the most recent visit, related only to blue spaces. Data will be made freely available on the internet in accordance with the European Union's Open Data strategy after a suitable embargo period.
Sampling. Samples of ~ 1000 respondents representative on sex, age, and geographical location were obtained from each country/region (total n = 18,838). To account for seasonal biases, sampling was undertaken in four waves between June 2017 and April 2018. All analyses used weights, created by YouGov, which accounted for selection, non-response, and population biases. Supplementary Tables S2 and S3.

Common mental health disorders (CMDs).
Our indicator of whether someone was currently experiencing a CMD was based on the following question: "During the past two weeks, have you used any medicines for any of the following conditions that were prescribed for you by a doctor? Please select all that apply", with 'yes'/'no' response options. Alongside physical health conditions, e.g. back/neck pain, were two mental health conditions: 'depression' and 'tension and anxiety' . The question was taken from the European Health Interview Survey (Eurostat, 2013), and we created three CMD groups: (a) those with depression, i.e. taking antidepressants; (b) those with anxiety, i.e. those taking anxiolytics; and (c) those with both depression and anxiety. We recognise that this is only a proxy measure and, among other limitations, may miss people with CMDs not currently on medication (see "Discussion"). Medication data was missing for one participant.
Intrinsic motivation. Adapting items from a measure used in the physical activity domain ("I find exercise fun" and "I enjoy my exercise sessions") 38 , a single-item asked participants the extent to which participants felt the following statement was true for them: "I find visiting green and blue space enjoyable or fun", with responses on a seven-point scale from 'not at all' (1) to 'very true' (7). We suggest that similarly to exercise, visiting natural spaces may be considered under the umbrella of activities that serve to increase health and wellbeing. As such it is proposed that adapting a measure of motivation from the physical activity domain was appropriate. 187 participants responded 'unsure' and were classed as 'missing' .
Visiting nature ≥ once a week. Participants were asked how often they had made recreational visits to a range of natural environments, including both green (vegetated) and blue (inland/coastal water) spaces, in the last four weeks (See Supplementary Table S1 for the full list) using four categorical response options ('not at all in the last four weeks' , 'once or twice in the last four weeks' , 'once a week' , 'several times a week'). The last four weeks was chosen as a recall period based on previous leisure visit surveys. 'Not at all' and 'once/twice' were collapsed together, and 'once a week' and 'several times' were collapsed together to provide a binary 'yes'/'no' indication of at least (≥) weekly nature visits, an important threshold in previous studies 39 . Nature visit frequency data was missing for five participants.
Blue space visit wellbeing. Consistent with the aims of the BlueHealth project, participants were asked to describe their most recent visit to any type of blue space within the last four weeks (see Supplementary Table S1 for taxonomy). 79.4% (n = 14,973), had made at least one such visit. Of these the majority involving walking, especially along footpaths and promenades. Other visits included sunbathing, swimming, playing with children, watersports, and ice skating (in winter months). Detailed analysis of visit type will be presented elsewhere. For current purposes the most important aspect was participant's recalled experiential wellbeing in terms of the degree to which they agreed with the following statements: "It made me feel happy" and "It made me feel anxious", with response options from 'strongly disagree' (1) to 'strongly agree' (7). These items were adapted from the OCED 40 measures of positive and negative experiential wellbeing respectively. Of the 14,973, seven had missing data for happiness and eight people had missing data for anxiety.

Scientific Reports
| (2020) 10:19408 | https://doi.org/10.1038/s41598-020-75825-9 www.nature.com/scientificreports/ Perceived social pressure. Perceived social pressure (PSP) was assessed using a single item adapted from the same motivation measure as for intrinsic motivation ("I feel under pressure from friends/family to exercise" 38 . Participants were asked the extent to which they felt the following statement was true for them: "I sometimes feel pressured by others (e.g. partner, friends) to visit green and blue spaces", with responses from 'not at all' (1) to 'very true' (7). 1,038 participants responded 'unsure' and were classed as 'missing' , alongside 1 person with missing data.
Sociodemographic and individual level controls. Regression analyses were adjusted for a number of sociodemographic indicators that have been found to be associated with nature visits and/or mental health in previous research. For full details and rationale see https ://bit.ly/BIS-Techn ical-Repor t (section 4.7). These included: sex (male = reference, female); age (18-29, 30-39, 40-49, 50-59, 60 + = reference); perceived financial strain (finding it very difficult on present income = reference, finding it difficult on present income, coping on present income, living comfortably on present income, don't know); employment status (employed, other = reference); marital status (married or in civil partnership, not married = reference); number of children in the household (none = reference, one, two or more); having a long-term limiting illness (yes, no = reference); smoking status (current smoker, previously smoked, never smoked = reference); and alcohol use (less than monthly = reference, up to once a week, and up to daily). Seasonal wave (spring = reference, summer, autumn, winter), and country (reference = UK) were included to control for seasonal and country level effects. Raw data included 'missing's is presented in Supplementary Table S4. Analyses for visit-related 'happiness' and 'anxiety' also controlled for visit-related factors that have been shown to be important for visit wellbeing in previous studies 29 including: number of companions; presence of dog (yes, no = reference); mode of transport to get to destination (private = reference, public, walking/cycling, other); travel time to destination (in minutes; 0-14 = reference, 15-29, 30-59, 60-119, 120 +); and visit duration (in minutes; 10-20 = reference, 30-50,60-80, 90-110, 120-170, 180 +). Finally, to reduce the possibility that visitrelated happiness and anxiety was merely due to general affective disposition, we also controlled for happiness and anxiety on the previous day as measures of general levels of experiential wellbeing 40 . Raw data included 'missing's is presented in Supplementary Table S4.

Analyses. Analyses were conducted in IBM SPSS Statistics 25. To test H1a and H3/b, one-sample t-tests
were conducted comparing the mean scores for the whole sample, and for sub-samples as a function of CMD group, to the scale-mid-points. Remaining hypotheses were tested using a series of linear and logistic (for the binary outcome of visit ≥ once week) regression models, with perceived social pressure (PSP) and interaction terms included to test the potential moderation effect of PSP (H4). Fully adjusted models are presented above with unadjusted and partially adjusted models presented in Supplementary Materials Table S5 to S8. Variance inflation factor (VIF) tests for all analyses suggested no evidence of multicollinearity between covariates in any of the analyses (Supplementary Materials Table S10). Country was added as a fixed effect to control for potential within-country clustering. This is similar to a random intercept model for country but with the advantage that inference on the relationship between country and outcomes is more straightforward. There were insufficient cases of medication use in each country to robustly explore country-specific medication effects (random slopes). Although we considered a matched-control design for the current analyses, i.e. only comparing those with CMDs to a reduced sample who were not on medication but similar in other respects, such designs tend to produce similar results to full sample analyses, but with reduced power 41 . Preliminary analysis suggested this was also the case here, so the full sample was retained. Loss of data due to missing/unsure/prefer not to answer responses (especially for the PSP measure) resulted in full sample (n = 18,838) models of n = 17,570 for intrinsic motivation and visits weekly (i.e. 93.3% included), and visit sub-sample (n = 14,973) analyses with n = 14,012 (93.6%) for visit happiness and n = 13,975 (93.3%) for visit anxiety. Of note, this reduced visit only sample was almost identical in terms of CMD status and demographic composition to the full sample (Supplementary  Table S9), reducing the risk of bias. To aid comparability across Models 1 (without PSP) and Models 2 (with PSP), both models only used the sample with PSP data.
The relationships between covariates and main study outcomes. Below we present a brief consideration of the significant relationships between the covariates included in the final, fully adjusted models for each outcome variable. These are presented in Tables S5 to S8  www.nature.com/scientificreports/ in winter (B = − 0.10, 95% CI = [− 0.05, − 0.14], p < 0.001) and spring (B = − 0.06, 95% CI = [− 0.12, − 0.01], p = 0.021), compared to Summer. It is logical that enjoyment would be higher in seasons that are typically associated with milder weather. Up to daily alcohol use was also positively associated with IM. This could be a result of the social interaction that may accompany regular alcohol intake and be associated with visits for recreational purposes and therefore increased enjoyment.  www.nature.com/scientificreports/ and may be due to greater responsibility for the safety of the children during the visit. As expected, anxiety yesterday (B = 0.15, 95% CI = [0.14, 0.15], p < 0.001) was related to visit anxiety, and drinking alcohol up to daily (B = 0.07, 95% CI = [0.01, 0.13], p = 0.017) was also related to increased anxiety compared to drinking less than monthly. Wave was also associated with anxiety; as perhaps may be expected, anxiety was higher during winter visits (B = 0.06, 95% CI = [0.00, 0.12], p = 0.050), compared to summer. It is logical that anxiety would be higher in seasons that are typically associated with more severe weather and activities that could be perceived as more risky such as ice skating etc. Compared to the UK, participants in the majority of countries (Canada, Esto , p < 0.001) was associated with reduced anxiety compared to using private transport, and as expected, travel times of 30 min or greater to reach the destination were also all associated with greater anxiety (all B's > 0.08, all p's < 0.05) compared to a travel time of less than 15 min. Finally, consistent with the happiness ratings, visit anxiety was lower for trips that lasted longer than 60 min, suggesting that longer periods of time spent in nature may be associated with increased wellbeing benefits (B's range from − 0.10 to − 0.31, all p's < .0.05).