In our nationally representative sample of U.S community-dwelling older adults, we evaluated whether greenness was associated with three short-term mental ill health indictors. We found that increased greenness was significantly associated with lower self-perceived stress during the past week, upon controlling for confounders. We also found partial mediation of the greenness-stress association by physical activity and history of respiratory diseases. Associations were modified by several individual-specific characteristics, with the significantly inverse association among participants who were relatively young, White, well-educated, more socially active, and those had higher BMI or lived in the Northeast. Although greenness was not significantly associated with anxiety and depression across all participants, our results show that greenness was associated with larger reductions in anxiety and depression scores in Whites, individuals who had higher SES, and were physically active, as compared to their counterparts. A directed acyclic graph (DAG) was shown in Fig. 2 to illustrate the role of confounders, mediators, and modifiers in the association between greenness and perceived stress.
Our findings of inverse association of greenness with perceived stress overall, and with anxiety/depression for certain subgroups are generally consistent with previous longitudinal studies of greenness and mental distress, a global measure of mental health [9,10,11,12,13] and cross-sectional studies on self-reported stress [7, 40]. For instance, three studies of the British Household Panel Survey found significantly lower mental distress associated with living in greener urban areas in England [10,11,12], while a Swedish study found no link between access to greenness and mental distress [13], perhaps due to its use of a relatively crude exposure measure (e.g., serene, lush) that may not capture subtle differences in greenness as was possible using the NDVI. Our observed protective effect of greenness on mental health is also in line with a British study that showed that pregnant women living in the greener quintiles were 18–23% less likely to report depressive symptoms than those living in the least green quintile [9]. While supportive, the relevance of findings from these European studies to our study may be limited, given differences in housing and land use [41].
Comparison of our findings to those from U.S studies, however, are limited, as the only U.S studies to date are cross-sectional studies conducted in cities or other localized geographic regions [18, 19, 21], with mixed results. Our results for perceived stress are most consistent with that from Fan et al. [18], who showed a significant mitigation of perceived stress (log scale, β = − 0.044) associated with neighborhood greenness in a probability sample of Chicago residents, and Beyer et al. [21], who reported significant beneficial effect of greenness (i.e., tree canopy & NDVI combined) on perceived stress (β = − 0.735) among Wisconsin residents. Our study, however, differs from a recent Seattle, Washington study of 4338 monozygotic twins that found no association between NDVI and perceived stress upon adjusting for income and physical activity [19]. This discrepancy may result from our substantially different study populations, as the twin population consisted of predominately White middle-aged women, whereas our study population consisted of older, more ethnically diverse men and women living across the U.S Notably, we found region of the country to modify our greenness-stress association, with this association being significant and strongest in the Northeastern U.S, but not in other regions including the West, where the majority of the twins in the Seattle study resided. Strong beneficial effects of greenness on stress for participants in the Northeast may be attributed to the Northeast region’s high population and housing density and large proportion of urban land [42, 43]. As a result, small changes in the level greenness may have greater effects on mental health. Further research should aim to explore possible regional differences in more detail.
Similarly, more research is needed to support our finding of effect modification by SES, where we showed the significant inverse association between greenness and perceived stress only for individuals with higher education. This finding is counter to prevailing theory that the beneficial effect of greenness is stronger among individuals with lower education or SES due to the fact that these individuals spend more time near their homes and thus use and interact more frequently with their immediate surrounding environment [9, 15]. However, it is possible that the key modifying factor is not SES but rather time spent near home, where children, housewives, and elderly (e.g., community-dwelling older adults assessed in our study) may spend more time near their homes as compared to the other populations [e.g., pregnant women [9]] examined in previous studies [15]. Our finding on the modifying effect of BMI lends support to our argument, in which obese individuals tended to stay around their residence than those with lower BMI, thus interact more with their surrounding greenness.
Notably, our findings that show greenness to be associated with lower stress levels through increased physical activity are consistent with theories that relate green vegetation in the natural environment with more positive emotional states, physiological activity levels, and behavior and cognitive functioning [5, 44]. Several mechanisms through which greenness influences mental health have been postulated, including the enhancement of health and well-being [15], promotion of physical activity [9, 45], provision of healthy environment and better air quality [46], and facilitation of social contact [8]. While consistent with prevailing hypotheses and some recent literature [7, 9, 40, 47, 48], our findings are in contrast with other studies that found little or no mediation by physical activity [18, 40, 49]. Possible explanations for the conflicting findings include difference in geographic area coverage, where more localized geographic area (e.g., Chicago) may result in lower variation in exposure in some studies [18, 40], and in the examination of recreational walking in the English study as compared to rigorous physical activity (e.g., sports, exercise classes or heavy housework) in the current study [49]. Further, although we did not find significant positive mediation by social support, we did find socially active individuals to have lower perceived stress associated with greenness as compared to their socially inactive counterparts, which is consistent with prevailing hypotheses [50]. It is possible that the insignificance of social support as a mediator may result from our measure of social support, which was relatively crude as compared to the Social Support List, a modified 19-items questionnaire with good construct validity and high reliability, that was used in previous study that reported partial mediation of the relation between green space and health by social contact [26]. Alternatively, our findings may reflect differing impacts of social support on greenness and mental health, as results from a Chicago study provided yet a different relation of social support, greenness, and mental health, finding social support to counteract the beneficial effect of neighborhood greenness on perceived stress among Black and Hispanic adults. The authors suggested that the high levels of park accessibility in the Chicago neighborhoods may obscure their study findings [18].
In addition to physical activity, we observed history of respiratory diseases to partially mediate the association of decreasing greenness and increased stress, although the indirect effect was small. Contrary to the common belief that greenness should have a beneficial effect on health, we found greenness to be positively associated with history of respiratory disease. Our finding is supported by previous studies that showed that greenness was associated with increased respiratory diseases (e.g., asthma), possibly reflecting the mediating influence of pollen and other aeroallergens on this association [51, 52]. Our positive associations between greenness and respiratory disease may further reflect the fact that over 90% of our cohort resided in MSAs, with most not moving during our study period, given results from several studies showing that associations between greenness and respiratory conditions tend to be positive for populations living in urban areas [53,54,55]. Fuertes et al. [55], for example, found that green space measured by NDVI was inversely associated with allergic rhinitis and eye and nose symptoms among children living in rural areas, but was positive associated among children living in urban areas. Alternatively, it may also be possible that individuals with emphysema or COPD may tend to move away from urbanized (polluted) environments to greener (less-polluted) areas; however, this hypothesis has not yet been examined.
As with perceived stress, we found greenness to be associated with decreased anxiety and depressive symptoms in specific segments of our population, consistent with existing hypotheses that anxiety and depression are influenced by the amount of neighborhood greenness [8, 9, 19, 24]. We found greater greenness to be associated with lower depression and anxiety scores only in men, which is similar to a British longitudinal study reporting associations between improved mental distress and greenness among men, with only moderate associations among older women [11]. We observed greenness to be significantly associated with reduced anxiety/depression symptoms among White but not Black and Hispanic participants, and among individuals with higher SES or living in higher SES neighborhoods. Furthermore, we found the effect modification by physical activity, in which greater greenness was associated with improved anxiety and depression symptoms only among individuals who were physically active. These findings are similar with previous evidence showing that the association between greenness and depressive symptoms was significant for individuals who were active [9], and also with collective evidence from literature review that the natural environment increases the positive effects of physical activity on well-being [45].
Our study has several limitations. First, CESD and HADS-A are not clinical diagnostic instruments, nor are they designed to assess chronic mental disorder. However, these questionnaires are widely used screening tools for current depressive and anxiety symptom severity in the somatic, psychiatric and general population settings [56]. Likewise, high PSS has been shown to correlate with high serum cortisol, a stress biomarker [57]. Second, exposure misclassification remains a concern in environmental epidemiology. While NDVI is a valid measure of greenness [58], it cannot distinguish between different types of green area such as tree canopy or parks, which may influence mental health through different mechanisms [59]. Third, we could not rule out the likelihood of self-selection bias in that participants with higher SES being more likely to select living in greener neighborhoods and to have lower mental health scores. Lastly, findings from the current study may not be generalizable to younger age groups.
Despite these limitations, our study is the first to examine the association of neighborhood greenness and mental health in a nationally representative sample of older, community-dwelling Americans. The longitudinal nature of our study design is a major strength, as previous U.S studies were cross-sectional design. We evaluated two affective and one psychological measures of mental health to provide a comprehensive picture of effect of greenness on mental health. We were also one of the first to explore how the greenness-mental health associations might be modified or mediated by a host of individual and/or neighborhood characteristics, and we are not aware of previous published results in older adults. In addition, our study was well-powered to detect meaningful associations and adjusted for confounding from individual- and census-level SES, air pollution and weather.