Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jul 29.
Published before final editing as: Soc Psychiatry Psychiatr Epidemiol. 2026 Mar 13:10.1007/s00127-026-03075-7. doi: 10.1007/s00127-026-03075-7

Midlife exposure to neighborhood greenness and later-life cognitive decline: The Multi-Ethnic Study of Atherosclerosis

Lilah M Besser a, Lun-Ching Chang b, Kelly R Evenson c, James E Galvin d, Susan R Heckbert e, Jana A Hirsch f, Peter James g, Marcia Pescador Jimenez h, Joel Kaufman i, Samuel N Lockhart j, Diana Mitsova k, Kari A Moore l, Bonnie C Sachs m, Timothy Hughes n,*, Ana V Diez Roux o,*
PMCID: PMC13409554  NIHMSID: NIHMS2190145  PMID: 41824016

Abstract

Purpose:

We investigated whether living in greener neighborhoods in midlife is associated with slower cognitive decline in later life.

Methods:

We used data on 2,881 participants from the population-based Multi-Ethnic Study of Atherosclerosis. Geocoded residential addresses (1980–2009) were used to derive midlife neighborhood greenness exposure defined as a 10-year mean of annual normalized difference vegetation index values (based on satellite imagery) during the midlife period (ages 45–54). Cognitive testing over ~10 years, when the participants were ≥55-year-olds, captured global cognition and processing speed. Multivariable linear mixed effects regression estimated associations between the 10-year midlife greenness measure and global cognition and processing speed z-scores in later life and whether greenness-cognition associations varied by age at first cognitive visit.

Results:

Greater midlife greenness was associated with slower annual decline in processing speed in the overall sample. We found no differences in associations by age at first cognitive visit.

Conclusion:

In an ethnoracially and geographically diverse US cohort, living in greener neighborhoods in midlife was associated with slower cognitive decline (i.e., processing speed) in later life.

Keywords: greenness, greenspace, midlife, cognition, cognitive decline, neighborhood

INTRODUCTION

Mild cognitive impairment affects ~23% of ≥65-year-olds, while dementia, which is cognitive decline severe enough to affect daily functioning and activities, affects ~11% [1]. Cognitive decline can be a marker for underlying neurodegenerative disease such as Alzheimer’s disease (AD), the leading cause of dementia among older adults. Over the next few decades, a rising number of individuals will be affected by cognitive decline and dementia due to the projected increase in population of older adults [2, 3].

In addition to older age and genetic risk factors (e.g., apolipoprotein E), individual-level modifiable risk factors for cognitive decline and dementia include physical inactivity, vascular disease, and diabetes, among other factors [4, 5]. The Lancet Commission’s 2024 report on dementia prevention, care, and intervention attributed 45% of dementias to modifiable risk factors (e.g., diabetes, hypertension, physical inactivity, obesity, social isolation)[4]. Increasingly, researchers have investigated community-level modifiable exposures that may influence cognitive function in older age [6–8]. Health-promoting neighborhood characteristics such as the presence of greenspaces (e.g., parks, tree canopy) have been associated with better brain health including cognitive functioning in later life [9–12]. For example, within our same study cohort, the Multi-Ethnic Study of Atherosclerosis (MESA), older adults living in neighborhoods with a greater percentage of park space were more likely to maintain/improve global cognitive function over time [13]. The aforementioned studies are part of broader body of research linking social and structural determinants of health (SDOH) to cognitive health and AD risk [14, 15], which recognize SDOH as upstream predictors of unhealthy behaviors (e.g., physical inactivity, etc.) that over time increase cognitive decline and dementia risk.

Neighborhood greenness is a modifiable environmental exposure associated with better cognitive function in older adults [16–19]. For instance, a UK-based Whitehall II cohort study of older adults found that an interquartile increment in greenness in the 1,000m surrounding the residence was associated with slower decline in global cognition over ~10 years [17]. Greenness is typically measured using the normalized difference vegetation index (NDVI), which is commonly derived from LANDSAT satellite imagery and quantifies the degree of healthy/dense vegetation in an area. NDVI and similar measures of neighborhood greenness have also been associated with the hypothesized mediators that link greater greenness to better cognitive function, including increased physical activity, reduced air pollution, greater social interaction, and better mental health [10, 20–24].

While several studies have found that neighborhood greenness is associated with cognitive function in older age [10, 16, 17, 19], most have measured greenness and/or cognitive function at one time point and no known studies have focused on greenness exposure among 40- and 50-year olds (the early midlife period) in relation to later life cognitive outcomes. This early midlife period is critical because this is when AD neuropathology can begin to develop [25]. Late-onset AD typically develops among ≥65-year-olds [26], and given that individuals with AD neuropathology may remain asymptomatic for decades prior to their dementia diagnosis [27], studies with greenness exposure starting at age 45 are needed to understand its potential for primary and secondary prevention of AD-associated cognitive decline. Also, in contrast to late-life measures, midlife measures of neighborhood greenness better match the hypothesized mechanisms linking greenness to cognitive decline (e.g., physical inactivity, cardiometabolic factors), as these mediating factors are expected to exert their influence on later-life cognition in a chronic way. Overall, midlife neighborhood greenness may promote healthy behaviors and reduce detrimental environmental exposures during the period when asymptomatic AD neuropathology can develop, ultimately preventing or delaying AD diagnosis and associated symptoms in later life. Thus, in this study, we used data from MESA to determine if greater neighborhood greenness in midlife is associated with slower decline in cognition over time in later life and whether associations varied by age group.

METHODS

Sample

Longitudinal data were obtained from MESA, a population-based cohort study that enrolled 45- to 84-year-olds from six US locations (i.e., Forsyth County, North Carolina; New York, New York; Baltimore, Maryland; St. Paul, Minnesota; Chicago, Illinois; Los Angeles, California) [28]. MESA was established to characterize subclinical cardiovascular disease and to investigate risk factors of disease progression. At the seven in-person exams conducted to date (spanning 2000–2024), participants provided their sociodemographics, comorbidities, and health history, and completed self-reported surveys, physical exams (e.g., height, weight, blood pressure), specimen collection (e.g., blood for genetic testing), and at some exams, cognitive testing, among other assessments. MESA received approval from the University of Washington Institutional Review Board and participants provided informed consent at each site. Additional details on the MESA study design are provided elsewhere [28]. In this study, we restricted to participants with Exam 5 cognitive test data (i.e., first cognitive visit) (n=4,716), with no ICD-9 diagnosis of dementia or medication use for cognitive decline as of Exam 5 (n=4,639) [29], valid test scores on the Cognitive Abilities Screening Instrument (CASI) (n=4,497), had residential address history data that covered the midlife period (ages 45–54 years) (n=3,112), and had no missing NDVI values due to cloud cover (final sample, n=2,881) (Figure 1).

Figure 1.

Figure 1.

Open in a new tab

Sample size flow diagram

a Cognitive data only available from Exam 5 on. b Determined from hospitalization records, death certificates, and self-reported medication use. c Excluded CASI scores deemed invalid by test administrator, more than 3 missing CASI components, or CASI score<20 (lacks face validity). d Only individuals who had addresses for their entire midlife period (from 45 to 54 years) are included. e Excluded if one or more annual NDVI values are missing for the midlife period (45 to 54 years)

Measures

Neighborhood Greenness.

Participants provided their residential address history back to 1980 using MESA’s residential history questionnaire. The addresses were geocoded to latitude-longitude coordinates, and we linked these to normalized difference vegetation index (NDVI) values from ~July 1 of each year (summer, greenest time of year) downloaded from Google Earth Engine. NDVI is calculated from LANDSAT satellite imagery by determining the difference between near infrared and red-light reflectance, quantifying the greenness of an area on a scale of −1 to 1 (values closer to 1=greener). NDVI values <1 indicate water, thus we set values <1 to 0 so that we did not down weight areas near water (i.e., NDVI range: 0 to 1). Based on the participants’ latitude-longitude coordinates, we derived annual NDVI measures for the 270m and 1230m circular buffers surrounding their residences (0.16 and 0.76 miles, respectively). We chose these values because they parallel the prior literature on 250m MODIS satellite data, and represent both the immediate area around the home, as well as the area reachable via walking distance around the home [30, 31]. NDVI values for 1980–1983 were set to the 1984 values given the lack of satellite imagery prior to 1984.

To determine midlife greenness exposure for each participant, we calculated the 10-year mean of the annual NDVI measures covering the entire midlife period. Sixty-seven percent of participants lived at one address during the midlife period (45–54 years), 25% lived at 2 addresses, 6% lived at 3 addresses, and 2% lived at ≥3 addresses. Annual NDVI values were based on the residential addresses for each year from ages 45 to 54 years, thus participants’ midlife mean NDVI values were weighted by the proportion of the 10-year period spent in each location. Individuals who lacked address history data and/or NDVI values (e.g., due to too much cloud cover) for any point of that midlife period were excluded from these calculations. The 1230m surrounding the residence was chosen as the primary buffer for our study because numerous studies have shown that neighborhood characteristics measured up to 1.5km (~1 mile) are associated with numerous health outcomes including cognitive function [13, 32].

Cognition.

MESA participants received cognitive testing at four time points spanning ~10 years, at Exam 5 (2010–2012), at Exam 6 (2016–2019), during the MESA MIND A ancillary study visit (years: 2019–2021), and at Exam 7 (2022–2023) (Figure 2). CASI is a global measure of cognitive function, assesses attention, concentration, orientation, short-term memory, long-term memory, language, visual construction, verbal fluency, abstraction, and judgment, with total scores ranging from 0 to 100 (higher: better) [33]. Digit Symbol Coding (DSC) scores ranged from 0–133 with higher scores indicating better processing speed [34, 35]. For both cognitive tests, z-scores were calculated by first subtracting from the participant’s score the sample’s mean score at Exam 5 and then dividing by the standard deviation at Exam 5. In this study, we focused on these measures of global cognition and processing speed because they were available across time and have been more consistently associated with neighborhood characteristics in prior MESA studies [36]. Participants were given the option to complete tests in English, Spanish, or Chinese [37]. Due to the COVID-19 pandemic, 56.1% (778 visits) of the MESA MIND A cognitive testing and 0.1% (1 visit) of Exam 7 cognitive testing were conducted by telephone or video call (all other cognitive tests done in person).

Figure 2.

Figure 2.

Open in a new tab

Timeline of greenness and cognition measurements

Covariates.

Individual-level sociodemographics included age at Exam 5, self-reported gender, self-reported ethnoracial group (Black, Chinese, Hispanic/Latino, non-Hispanic White), primary language (English, Spanish, Chinese), married/with partner at Exam 5 (yes versus no), education level at Exam 5 (4 categories), and income at Exam 5 (4 categories). Midlife neighborhood socioeconomic status (SES) was based on US Census (1980, 1990, 2000, 2010) measures of the percentage of residents in the Census tract living below the poverty line. We linearly interpolated intercensal year values, between 1980 and 1990, 1990 and 2000, and 2000 and 2010, to determine each participant’s midlife neighborhood SES value based on their address at age 45.

Comorbidities from Exam 5 included depressive symptoms (Center for Epidemiologic Studies Depression Scale (CES-D) score ≥16), diabetes (from fasting glucose level or medication use), hypertension diagnosis and/or medication use, cardiovascular disease (i.e., history of myocardial infarction, resuscitated cardiac arrest, congestive heart failure, coronary bypass, coronary heart disease), cerebrovascular disease (i.e., stroke or transient ischemic attack), and obesity (body mass index ≥30kg/m2).

Apolipoprotein E (APOE) isoforms were previously assayed using single nucleotide polymorphism (SNP) genotyping (rs7412 and rs429358). APOE4 carrier status (yes versus no) was defined as those with one or two ε4 alleles versus those with ε2 and/or ε3 alleles.

Analytic strategy

Descriptive statistics (e.g., means and standard deviation (SD)) described the sample’s sociodemographics, comorbidities, APOE4 carrier status, and cognitive test scores by tertile of neighborhood greenness (i.e., mean NDVI) in midlife (T1=0.00–0.24, T2=>0.24–0.44, T3=>0.44). Differences in these variables by tertile of greenness were tested using multinomial logistic regression (moderate (T2) and high (T3) greenness compared to low greenness, T1). Multivariable linear mixed effects regression models (with random intercepts and slopes for each person and robust standard errors [38]) were used to estimate the associations between neighborhood greenness in midlife and later-life measures of cognitive domain z-scores (CASI, DSC). Prior to running the regression models, we verified that there was a linear relationship between greenness and the cognitive outcomes via spaghetti plots (no indication of non-linear relationship) and residuals versus predicted plots (observed homoscedasticity/random scatter of residuals around horizontal zero line). Time (years from first cognitive visit at Exam 5) was included in all models and the greenness×time interaction term tested whether midlife greenness was associated with mean differences in the slope of cognitive decline. Time-invariant covariates were added to models in two steps: 1) controlling for covariates from the first cognitive visit, specifically age, gender, ethnoracial group, married/with partner, education, income, primary language, and site (as well as the interactions of these covariates with time, e.g., gender×time) (Model 1); and 2) additionally controlling for midlife neighborhood SES (as well as its interaction with time) (Model 2). In addition, we repeated the models described above to determine if associations differed by age group at first cognitive visit/Exam 5 (dichotomized at median: 55–65 years versus >65 years). Differences in associations by age group were tested by including interaction terms in the multivariable models (i.e.., greenness×age, greenness×time×age).

We conducted all analyses in SAS v9.4 and the linear mixed models were run using PROC MIXED. All regression findings are reported with respect to an IQR increment in midlife greenness, which provides an estimate of the difference in cognitive outcomes for the typical spread of values in greenness (the middle 50%).

Six sensitivity analyses were conducted to assess if findings differed significantly from our main analyses. First, we re-ran Models 1 and 2 described above but additionally controlled for APOE4 carrier status (and its interaction with time since initial cognitive visit, Exam 5), as it is known to increase risk for Alzheimer’s disease and thus can be associated with accelerated cognitive decline [39, 40]. Second, we ran the Models 1 and 2 additionally controlling for comorbidities measured at the first cognitive visit (i.e., hypertension, diabetes, obesity, cardiovascular and cerebrovascular disease, and depressive symptoms, and their interactions with time since initial cognitive visit), which could be confounders that we should adjust for but alternatively mediators that if adjusted for would obscure greenness-cognition associations. Third, we calculated propensity scores in logistic regression models based on participants’ age, gender, education, ethnoracial group, marital status, site, obesity, diabetes, current smoker, and hypertension status at Exam 1. The first logistic regression model estimated the probability of response (inclusion in our analytic sample based on our inclusion/exclusion criteria) based on participant characteristics. The second logistic regression model estimated the crude probability of response (without including the covariates in the first model). Then, a stabilized weight for each participant was constructed by dividing the probability from the second model by the probability in the first model. These weights were then applied to the linear mixed effects regression Models 1 and 2 from the main analyses to examine whether our main model findings were affected by selection bias from multiple sources (i.e., missing Exam 5, missing covariates, drop out/death prior to Exam 5) (using the ‘weight’ statement in SAS). Using inverse probability weighting, we weight each observation based on the participant’s probability of selection into the analysis, which provides estimates closer to what would have been observed if the entire original sample (n=6,814) was included in the analysis [41]. Fourth, we re-ran Models 1 and 2 using the midlife greenness measures based on 270m buffers instead of 1230m buffers to assess whether the definition of neighborhood boundary changed our findings. Fifth, we restricted Models 1 and 2 to those who completed the cognitive testing in person. Thus, we excluded cognitive test scores from the third and fourth round of testing for those completing it over the phone or video call due to the COVID-19 pandemic. Lastly, given that individual-level SES is expected to be a key confounder, we re-ran the main regression models allowing household income to change across the cognitive visits (i.e., accounting for time-varying income).

RESULTS

Our main analytic sample included 2,881 participants (Figure 1), with a range of 389 to 540 individuals representing each of the six MESA cities. Participants were 65.8±6.0 years old at their first cognitive test visit (Exam 5), 53% were women, 72% had at least some college education, and 28% were Black, 21% Hispanic/Latino, 10% Chinese, and 42% White (Table 1). Approximately 16% spoke a language other than English as their primary language (9% Spanish, 7% Chinese). Twenty-eight percent had ≥1 APOE4 allele (i.e., risk factor for AD). Depressive symptoms (i.e., CES-D) were present in 12% of participants, diabetes in 20%, hypertension in 51%, cardiovascular disease in 7%, cerebrovascular disease in 3%, and obesity in 39%. Mean CASI scores at baseline (Exam 5) were 89.2 (SD=7.5) and mean DSC scores were 53.8 (SD=17.3). Mean annual decline in CASI was −0.0059 units/year (95% CI=−0.0095, −0.0023) and mean annual decline for DSC was −0.0323 units/year (95% CI=−0.0354, −0.0293).

Table 1.

Participant characteristics by tertiles of midlife neighborhood greenness

Characteristica (at Exam 5) Total sample, n=2,881 Low Greenness, T1 Moderate Greenness, T2 High Greenness, T3
Age (years), mean (SD) 65.8 (6.0) 65.5 (6.0) 66.1 (6.1)* 65.8 (6.0)
Women, n (%) 1527 (53.0%) 541 (55.5%) 520 (52.9%) 458 (49.6%)*
Education†, n (%)
 <High school degree 330 (11.5%) 200 (20.5%) 96 (9.8%) 34 (3.7%)
 High school degree 489 (17.0%) 172 (17.7%) 156 (15.9%)**** 161 (17.5%)****
 Some college 864 (30.0%) 233 (23.9%) 313 (31.9%)**** 318 (34.5%)****
 Bachelor’s degree or higher 1193 (41.5%) 369 (37.9%) 416 (42.4%)**** 408 (44.3%)****
Household family income, n (%)
 <$30,000/year 789 (27.4%) 363 (37.3%) 255 (25.9%) 171 (18.5%)
 $30,000–49,999/year 639 (22.2%) 214 (22.0%) 231 (23.5%)*** 194 (21.0%)****
 $50,000–99,999/year 854 (29.6%) 219 (22.5%) 314 (31.9%)**** 321 (34.7%)****
 $100,000+ 599 (20.8%) 178 (18.3%) 183 (18.6%)** 238 (25.8%)****
Race/ethnicity group, n (%)
 White 1201 (41.7%) 282 (29.0%) 424 (43.1%) 495 (53.6%)
 Chinese 285 (9.9%) 127 (13.0%) 112 (11.4%)*** 46 (5.0%)****
 Black 804 (27.9%) 251 (25.8%) 242 (24.6%)*** 311 (33.7%)**
 Hispanic/Latino 591 (20.5%) 314 (32.2%) 205 (20.9%)**** 72 (7.8%)****
Married/living with partner, n (%) 1801 (63.9%) 545 (58.0%) 612 (63.6%)* 644 (70.4%)****
Primary language, n (%)
 English 2425 (84.2%) 671 (68.9%) 856 (87.1%) 898 (97.2%)
 Spanish 262 (9.1%) 199 (20.4%) 58 (5.9%)**** 5 (0.5%)****
 Chinese 194 (6.7%) 104 (10.7%) 69 (7.0%)**** 21 (2.3%)****
APOE4 carrier, n (%) 744 (27.5%) 258 (28.8%) 239 (26.0%) 247 (27.9%)
Depressive symptoms, mean (SD) 344 (12.4%) 143 (15.3%) 119 (12.6%) 82 (9.1%)****
Diabetes, n (%) 566 (19.7%) 201 (20.7%) 184 (18.8%) 181 (19.6%)
Hypertension, n (%) 1481 (51.4%) 498 (51.1%) 471 (47.9%) 512 (55.4%)
Cardiovascular disease, n (%)‡ 199 (6.9%) 55 (5.7%) 67 (6.8%) 65 (7.0%)
Cerebrovascular disease, n (%)§ 81 (2.8%) 56 (5.8%) 79 (8.1%)* 64 (6.9%)
Obesity (BMI≥30), n (%) 1114 (38.7%) 345 (35.5%) 392 (40.0%)* 377 (40.9%)*
CASI at Exam 5, mean (SD) 89.2 (7.5) 87.5 (8.2) 89.5 (7.3)**** 90.7 (6.5)****
DSC at Exam 5, mean (SD) 53.8 (17.3) 51.5 (18.7) 54.4 (17.2)*** 55.8 (15.4)****
Midlife neighborhood SES (% poverty at age 45), mean (SD) 15.3 (12.6) 22.3 (14.4) 15.1 (11.1)**** 8.0 (6.5)****
Open in a new tab

Abbreviations: APOE=apolipoprotein E; BMI=body mass index; CASI=Cognitive Abilities Screening Instrument; DSC=Digit Symbol Coding; SES=socioeconomic status;

a

Missing data: Married, n=63; Income, n=80; education, n=5; CES-D, n=105; obese, n=5; Diabetes, n=7; APOE, n=177; CASI, n=63; DS, n=240; cardiovascular or cerebrovascular disease, n=3;

†

Assessed at Exam 1;

‡

myocardial infarct, resuscitated cardiac arrest, congestive heart failure, coronary bypass, coronary heart disease;

§

transient ischemic attack or stroke; Compared to T1:

*

p<0.05

**

p<0.01

***

p<0.001

****

p<0.0001

Comparing participant characteristics by level of neighborhood greenness in midlife (i.e., versus low greenness, T1), fewer women lived in the highest tertile of greenness (T3), and education and income levels increased from the lowest to highest tertile of greenness (Table 1). A greater proportion of Hispanic/Latino and Chinese participants lived in low greenness neighborhoods in midlife (T1), whereas a greater proportion of Black participants lived in high greenness neighborhoods in midlife (T3). Individuals in high greenness neighborhoods in midlife (T3) less often reported later-life depressive symptoms. Later-life obesity was more frequent among individuals living in neighborhoods with moderate (T2) and high (T3) levels of neighborhood greenness in midlife. CASI and DSC scores were higher in those with moderate (T2) and high (T3) levels of midlife greenness. Finally, neighborhood SES (percentage in poverty) in midlife was lower among individuals in moderate (T2) and high (T3) greenness neighborhoods in midlife.

For the 1230m circular buffer surrounding the participant’s residence, mean greenness (NDVI) in midlife was 0.33 (SD=0.16, interquartile range (IQR)=0.276). For the 270m circular buffer surrounding the participant’s residence, mean greenness in midlife was 0.34 (SD=0.16, IQR=0.283). Mean midlife greenness values by site are presented in Supplemental Table 1 (ranged from 0.20 in New York to 0.50 in Forsyth County, North Carolina).

In the adjusted analyses controlling for demographics and site (Table 2, Model 1), living in a greener neighborhood in midlife was not associated with baseline CASI z-scores (mean difference in score per IQR increment=0.037; 95% CI=−0.034, 0.109) or baseline DSC z-scores (mean difference=−0.041, 95% CI=−0.114, 0.033). This remained consistent with further control of neighborhood SES in Model 2. In adjusted analyses controlling for demographics and site (Table 2, Model 1) living in a greener neighborhood in midlife was not associated with annual decline in CASI z-scores (mean difference in decline per IQR increment=0.005, 95% CI=−0.001, 0.011) but was associated with slower annual decline in DSC z-scores (mean difference in decline per IQR increment=0.012, 95% CI=0.006, 0.017). When additionally controlling for midlife neighborhood SES (Table 2, Model 2), the lack of an association remained between midlife greenness and annual decline in CASI, while the association remained between midlife greenness and slower annual decline in DSC (mean difference in decline per IQR increment=0.014, 95% CI=0.000, 0.020). Midlife greenness-cognition associations did not differ by age group at first cognitive visit (Supplemental Tables 2 and 3).

Table 2.

Association between neighborhood greenness at midlife and later-life cognition

Outcome Model 1a
 Model 2†
Estimate (95% CI) Estimate (95% CI)
CASI z-score Mean difference in baseline cognition per IQR increment of NDVI‡ 0.037 (−0.034, 0.109) 0.108 (−0.185, 0.401)
Mean difference in annual change in cognition per IQR increment of NDVI‡ 0.005 (−0.001, 0.011) 0.480 (−0.035, 0.017)
DSC z-score Mean difference in cognition at baseline per IQR increment of NDVI‡ −0.041 (−0.114, 0.033) −0.094 (0.000, −0.008)
Mean difference in annual change in cognition per IQR increment of NDVI‡ 0.012 (0.006, 0.017)**** 0.014 (0.000, 0.020)*
Open in a new tab

Abbreviations: CASI = Cognitive Abilities Screening Instrument; DS = Digit Symbol; NDVI=normalized difference vegetation index;

a

Model 1 controlled for age, ethnoracial group, gender, marital status, education, income, primary language, site (and interaction between these covariates and time from initial cognitive visit, Exam 5);

†

Model 2 controlled for Model 1 covariates plus neighborhood SES at age 45 (and interaction between neighborhood SES and time from initial cognitive visit, Exam 5);

‡

Mean of NDVI value for ~July 1 of each year during ages 45 to 54 (midlife) for the 1230m surrounding residence;

*

p<0.05,

**

p<0.01,

***

p<0.001,

****

p<0.0001

In sensitivity analyses, additionally controlling for APOE4 carriership and comorbidities resulted in minimal changes to the regression estimates in Model 2 (data not shown). After applying inverse probability weighting to account for possible selection bias, the estimates remained similar to those presented in Table 2 (mean annual change in DSC per IQR increment=0.016, 95% CI=0.009, 0.023) (Supplemental Table 4, Model 2). In the two sensitivity analyses that defined greenness based on the 270m circular buffer surrounding the residence and excluded telephone/video call visits, the estimates remained similar to Table 2 (mean annual change in DSC per IQR increment was 0.014 (95% CI=0.007, 0.020) for both analyses) (Supplemental Tables 5 and 6, Model 2). Lastly, models controlling for time-varying household income resulted in negligible changes from Table 2 findings, with no association observed between midlife greenness and CASI and a similar association observed between greater midlife greenness and slower annual decline in DSC (Model 2: mean annual change in DSC per IQR increment=0.015, 95% CI=0.008, 0.022) (data not shown).

DISCUSSION

In this study, we found that individuals who lived in greener neighborhoods in midlife had slower annual declines in processing speed in later life, with a mean annual decline that was 0.014 units/year slower for every IQR increment of midlife greenness. While initially this may seem like a small effect size, cognitive decline over time among middle- to older-age adults without dementia is known to be small to negligible (e.g., one study found <0.10 z-score unit change in processing speed from ages 55 to 60) [42, 43]. The observed association with midlife greenness is equivalent to slowing decline in processing speed by ~7.6 months per year (i.e., based on sample’s mean annual rate of decline of −0.022 units/year). These possible benefits to cognitive function over time may translate to delayed dementia onset in older age, which will need to be examined in future studies.

We found that the associations between midlife greenness and annual decline in global cognition and processing speed did not differ by age at first cognitive visit. In contrast, a study of participants in the Chinese Longitudinal Healthy Longevity Survey found that greater neighborhood greenness (mean NDVI in 500m surrounding residence) was associated with lower odds of cognitive impairment among the younger age group (65–79 years) but not the older age group (≥80 years) in their sample [44]. This may suggest that greenness has a larger impact on the cognitive function of younger older adults, possibly due to greater mobility and thus exposure to neighborhood greenspaces at younger than older ages. Additional investigations are needed to determine if greenness-cognition associations vary by age group in other cohorts.

A prior study investigated long-term greenness exposure that temporally preceded longitudinal change in cognition measured later in life. Using data on 16,962 women enrolled in the U.S. Nurses’ Health Study, the study found that an IQR increase in neighborhood greenness during late midlife (i.e., 61 to 69 years old) was associated with a 0.004-unit (95% CI=0.001, 0.006) slower decline in global cognition z-scores in later life (≥70-year-olds) [16]. The study differed from ours in that it focused on women nurses who mostly self-identified as White, and thus findings from that study may not extend or demonstrate the same effect sizes as for men or other ethnoracial or SES groups. Additionally, the study focused on greenness exposure among late midlife period (61–69 years), and thus did not capture exposure during the 40s and 50s when Alzheimer’s disease and other neurodegenerative pathologies may initially form in the brain. Even so, the findings from the prior study are very similar to ours, confirming that long-term exposure to greener neighborhoods may help slow cognitive decline in older age.

Despite few studies that are directly comparable to our study, several cross-sectional [19, 45–48] and longitudinal [17, 18] studies examined associations between neighborhood greenness, measured using NDVI, and cognition in middle to older age adults. In studies of ≥45-years-olds, greater neighborhood greenness was associated with better global cognition, psychomotor speed/attention, and verbal fluency in two cross-sectional studies, while no greenness-cognition associations were found in three other cross-sectional studies [19, 45–48]. Two studies of middle- to older-age adults found that greater neighborhood greenness at baseline was associated with slower longitudinal decline in global cognition, reasoning, and fluency, and a decrease in greenness was associated with greater odds of lower global cognitive function over time [17, 18]. Our study adds considerably to the literature by measuring long-term exposure to greenness in midlife in relation to longitudinal change in cognitive function in later life in an ethnoracially diverse cohort. The temporal and longitudinal nature of our study contributes to the extant evidence suggestion a possible causal association between greenness and cognitive function. Our study of exposure during midlife is significant because this is the period when pathological changes due to neurodegenerative disease including AD can take form. Also, residential greenness exposure likely exerts long-term/chronic effects on modifiable AD risk factors (e.g., physical activity, depression, diabetes), which are on the hypothesized causal pathways linking midlife greenness exposure to later life cognitive decline. Thus, the midlife period may be a critical period of exposure to health promoting neighborhood environments that reduces risk for later life cognitive decline and dementia onset.

Our study has several strengths. We employed longitudinal data on neighborhood greenness exposure that temporally preceded longitudinally measured cognitive test outcomes. MESA provides a geographically and ethnoracially diverse sample and data on important covariates including individual- and neighborhood-level SES, which helps reduce confounding bias. Our focus on midlife greenness exposure is key strength, as discussed further above. Nonetheless, limitations must be noted. Although the participants live in six US cities, findings may not be generalizable to other regions of the U.S. The longitudinal cognitive data included a global cognitive test and a test of speed of processing but did not cover other specific cognitive domains affected in aging and Alzheimer’s disease, limiting inference to specific aspects of cognition related to greenness. Participants self-reported their residential address history back to 1980, which may be subject to information bias (i.e., recall bias) and misclassification of the greenness exposure that may bias estimates (uncertain direction of bias). We demonstrated associations with both the 1230m and 270m buffer measures of midlife greenness; however, future studies will need to discern the most important residential buffer to consider with respect to the health benefits of greenness exposure. In addition, studies are needed to assess whether residential greenness exposure has a greater impact during critical periods over the life course (e.g., childhood, midlife, or late life). We excluded individuals who did not have residential history and/or were missing NDVI values covering their midlife period. To address possible selection bias due to missing data from multiple sources (missing Exam 5 visit, missing covariates, or drop out/death before Exam 5), we employed inverse probability weighting, which did not significantly change our regression estimates [49]. This study was not devised to investigate the causal mechanisms linking greenness exposure to cognitive decline, which will need to be explored in future studies. NDVI does not differentiate the specific greenspace/vegetation types (e.g., tree canopy, shrubs, park spaces) that have the greatest impact on cognitive decline, which will need to be determined in new studies to provide practicable evidence that can directly inform policies and urban plans. Cohort studies with data on time spent in greenspaces are needed to strengthen causal evidence for greenness-cognitive decline associations. While our measure of greenness exposure preceded the cognitive outcomes by at least 10 years, we cannot eliminate the possibility of some reverse causation in which individuals experiencing subtle cognitive declines choose to move/reside in greener neighborhoods (e.g., due to less cognitive confusion or stress caused by urban environments). Finally, in the regression models, we included marital status and income assessed at one time point, not accounting for changes over time (although sensitivity analyses revealed that accounting for time-varying income between Exams 5 and 7 resulted in negligible changes to Table 2 estimates). In addition, residual confounding by other time-varying exposures or by other modifiers of cognition decline associated with green space at midlife may confound our estimates.

Overall, our study suggests that midlife greenness exposure is associated with slower longitudinal decline in cognition in later life. Replication studies will be needed to ensure that the findings are generalizable to other geographic regions and robust to further confounder adjustment. In addition, new studies will be needed that explicate the causal mechanisms linking greenness and cognitive decline; explore effect modification by factors such as sex/gender, ethnoracial group, and SES; and examine whether there are critical time periods of greenness exposure to maximize brain health and slow cognitive decline in later life. Future studies will also be needed to explore the types of greenspaces most beneficial to cognitive health and whether midlife greenness exposure is associated with delayed dementia onset and/or lower incidence of dementia in older age. Finally, intervention studies and natural experiments (e.g., examining health effects of community greening initiatives) will be needed to strengthen evidence for causal associations between greenness and cognitive health, so that these research findings can be translated into urban planning and policies.

Supplementary Material

Supplemental Tables

Acknowledgments

The authors thank the other investigators, the staff, and the participants of the MESA study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org. This paper has been reviewed and approved by the MESA Publications and Presentations Committee. Dr. Pescador Jimenez is supported by NIH R01AG087199 and R00AG066949.

Sources of funding

This study was supported by National Institutes of Health R21AG075291, R01AG072634, R01HL071759, and R01 HL 127659. The MESA study was supported by contracts 75N92020D00001, HHSN268201500003I, N01-HC-95159, 75N92020D00005, N01-HC-95160, 75N92020D00002, N01-HC-95161, 75N92020D00003, N01-HC-95162, 75N92020D00006, N01-HC-95163, 75N92020D00004, N01-HC-95164, 75N92020D00007, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168 and N01-HC-95169 from the National Heart, Lung, and Blood Institute, and by grants UL1-TR-000040, UL1-TR-001079, and UL1-TR-001420 from the National Center for Advancing Translational Sciences (NCATS) and grants P30AG049638, R01AG054069, and R01AG058969 from the National Institute on Aging. All sources of funding listed above had no role in in study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

Footnotes

Consent Statement

Consent was not necessary as this was a secondary data analysis.

Ethics approval

MESA received approval from the University of Washington Institutional Review Board and participants provided informed consent at each site. The current study is a secondary data analysis and was deemed exempt by the University of Miami Institutional Review Board.

Competing interests

Dr. Lockhart is a full-time employee of Perceptive Inc. The authors declare that they have no conflicts of interest.

Data availability

Data can be requested from MESA at www.mesa-nhlbi.org.

References

  • [1].Rajan KB, Weuve J, Barnes LL, McAninch EA, Wilson RS, Evans DA. Population estimate of people with clinical Alzheimer’s disease and mild cognitive impairment in the United States (2020–2060). Alzheimers Dement. 2021;17:1966–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [2].2024 Alzheimer’s disease facts and figures. Alzheimers Dement. 2024;20:3708–821. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [3].Fang M, Hu J, Weiss J, Knopman DS, Albert M, Windham BG, et al. Lifetime risk and projected burden of dementia. Nat Med. 2025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [4].Livingston G, Huntley J, Liu KY, Costafreda SG, Selbaek G, Alladi S, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet. 2024;404:572–628. [DOI] [PubMed] [Google Scholar]
  • [5].Tilvis RS, Kahonen-Vare MH, Jolkkonen J, Valvanne J, Pitkala KH, Strandberg TE. Predictors of cognitive decline and mortality of aged people over a 10-year period. J Gerontol A Biol Sci Med Sci. 2004;59:268–74. [DOI] [PubMed] [Google Scholar]
  • [6].Besser LM, McDonald NC, Song Y, Kukull WA, Rodriguez DA. Neighborhood Environment and Cognition in Older Adults: A Systematic Review. Am J Prev Med. 2017;53(2):241–251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [7].Wu Y-T, Prina AM, Brayne C. The association between community environment and cognitive function: a systematic review. Social Psychiatry and Psychiatric Epidemiology. 2015;50:351–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [8].Peterson RL, George KM, Tran D, Malladi P, Gilsanz P, Kind AJH, et al. Operationalizing Social Environments in Cognitive Aging and Dementia Research: A Scoping Review. Int J Environ Res Public Health. 2021;18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [9].Kuhn S, Duzel S, Eibich P, Krekel C, Wustemann H, Kolbe J, et al. In search of features that constitute an “enriched environment” in humans: Associations between geographical properties and brain structure. Sci Rep. 2017;7:11920. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [10].Besser LM, Jimenez MP, Reimer CJ, Meyer OL, Mitsova D, George KM, et al. Diversity of Studies on Neighborhood Greenspace and Brain Health by Racialized/Ethnic Group and Geographic Region: A Rapid Review. Int J Environ Res Public Health. 2023;20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [11].Zagnoli F, Filippini T, Jimenez MP, Wise LA, Hatch EE, Vinceti M. Is Greenness Associated with Dementia? A Systematic Review and Dose-Response Meta-analysis. Curr Environ Health Rep. 2022;9:574–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [12].Da L, Song X, Jia Z, Myers NGL, Sun J, Wei J, et al. Objectively measured environmental features and their association with cognition and dementia: a systematic review and meta-analysis. Ageing Res Rev. 2024:102630. [DOI] [PubMed] [Google Scholar]
  • [13].Besser LM, Chang L, Evenson KR, Hirsch J, Michael Y, Galvin JE, et al. Associations between neighborhood park access and longitudinal change in cognition in older adults: The Multi-Ethnic Study of Atherosclerosis. J Alzheimers Dis. 2021; 82(1):221–233. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [14].Adkins-Jackson PB, George KM, Besser LM, Hyun J, Lamar M, Hill-Jarrett TG, et al. The structural and social determinants of Alzheimer’s disease related dementias. Alzheimers Dement. 2023;19:3171–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [15].Besser LM, Edwards K, Lobban NS, Tolea MI, Galvin JE. Social Determinants of Health, Risk and Resilience against Alzheimer’s Disease and Related Dementias: The Healthy Brain Initiative. J Alzheimers Dis Rep. 2024;8:637–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [16].Pescador Jimenez M, Wagner M, Laden F, Hart JE, Grodstein F, James P. Midlife Residential Greenness and Late-Life Cognitive Decline among Nurses’ Health Study Participants. Environ Health Perspect. 2024;132:77003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [17].de Keijzer C, Tonne C, Basagana X, Valentin A, Singh-Manoux A, Alonso J, et al. Residential Surrounding Greenness and Cognitive Decline: A 10-Year Follow-up of the Whitehall II Cohort. Environ Health Perspect. 2018;126:077003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [18].Zhu A, Wu C, Yan LL, Wu CD, Bai C, Shi X, et al. Association between residential greenness and cognitive function: analysis of the Chinese Longitudinal Healthy Longevity Survey. BMJ Nutr Prev Health. 2019;2:72–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [19].Jimenez MP, Elliott EG, DeVille NV, Laden F, Hart JE, Weuve J, et al. Residential Green Space and Cognitive Function in a Large Cohort of Middle-Aged Women. JAMA Netw Open. 2022;5:e229306. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [20].Yang BY, Hu LW, Jalaludin B, Knibbs LD, Markevych I, Heinrich J, et al. Association Between Residential Greenness, Cardiometabolic Disorders, and Cardiovascular Disease Among Adults in China. JAMA Netw Open. 2020;3:e2017507. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [21].Dalecka A, Ksinan A, Szabo D, Capkova N, Pikhart H, Bobak M. Neighborhood environment and cognitive functioning in middle-aged and older population: A mediating role of physical activity. Int J Hyg Environ Health. 2025;264:114521. [DOI] [PubMed] [Google Scholar]
  • [22].Jubaer A, Hossain R, Ahmed A, Hossain MS. Factors influencing spatiotemporal variability of NO(2) concentration in urban area: a GIS and remote sensing-based approach. Environ Monit Assess. 2025;197:167. [DOI] [PubMed] [Google Scholar]
  • [23].Motoc I, Ginos BNR, Goncalves Soares A, Elhakeem A, Voortman T, Kavousi M, et al. Examining associations of air pollution and green space with depressive symptoms in adults: A LongITools cross-cohort analysis. Environ Res. 2025;264:120321. [DOI] [PubMed] [Google Scholar]
  • [24].Villeneuve PJ, Gill GK, Cottagiri SA, Dales R, Rainham D, Ross NA, et al. Does urban greenness reduce loneliness and social isolation among Canadians? A cross-sectional study of middle-aged and older adults of the Canadian Longitudinal Study on Aging (CLSA). Can J Public Health. 2024;115:282–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [25].Bendlin BB, Carlsson CM, Gleason CE, Johnson SC, Sodhi A, Gallagher CL, et al. Midlife predictors of Alzheimer’s disease. Maturitas. 2010;65:131–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [26].Rabinovici GD. Late-onset Alzheimer Disease. Continuum (Minneap Minn). 2019;25:14–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [27].Jack CR Jr, Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, et al. Tracking pathophysiological processes in Alzheimer’s disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12:207–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [28].Bild DE, Bluemke DA, Burke GL, Detrano R, Diez Roux AV, Folsom AR, et al. Multi-Ethnic Study of Atherosclerosis: objectives and design. Am J Epidemiol. 2002;156:871–81. [DOI] [PubMed] [Google Scholar]
  • [29].Fujiyoshi A, Jacobs DR, Alonso A, Luchsinger JA, Rapp SR, Duprez DA. Validity of Death Certificate and Hospital Discharge ICD Codes for Dementia Diagnosis: The Multi-Ethnic Study of Atherosclerosis. Alzheimer Dis Assoc Disord. 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [30].James P, Berrigan D, Hart JE, Hipp JA, Hoehner CM, Kerr J, et al. Effects of buffer size and shape on associations between the built environment and energy balance. Health Place. 2014;27:162–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [31].James P, Hart JE, Banay RF, Laden F. Exposure to Greenness and Mortality in a Nationwide Prospective Cohort Study of Women. Environ Health Perspect. 2016;124:1344–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [32].Li J, Hirsch JA, Michael YL, Besser LM, Auchincloss AH, Hughes TM, et al. Spatial scale effects on associations between built environment and cognitive function: Multi-Ethnic Study of Atherosclerosis. Health Place. 2024;86:103181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [33].Teng EL, Hasegawa K, Homma A, Imai Y, Larson E, Graves A, et al. The Cognitive Abilities Screening Instrument (CASI): a practical test for cross-cultural epidemiological studies of dementia. Int Psychogeriatr. 1994;6:45–58; discussion 62. [DOI] [PubMed] [Google Scholar]
  • [34].Jaeger J Digit Symbol Substitution Test: The Case for Sensitivity Over Specificity in Neuropsychological Testing. J Clin Psychopharmacol. 2018;38:513–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [35].Wechsler D Wechsler Adult Intelligence Scale. Third edition ed. San Antonio, TX: The Psychological Corporation; 1997. [Google Scholar]
  • [36].Besser LM, Chang LC, Hirsch JA, Rodriguez DA, Renne J, Rapp SR, et al. Longitudinal Associations between the Neighborhood Built Environment and Cognition in US Older Adults: The Multi-Ethnic Study of Atherosclerosis. Int J Environ Res Public Health. 2021;18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [37].Fitzpatrick AL, Rapp SR, Luchsinger J, Hill-Briggs F, Alonso A, Gottesman R, et al. Sociodemographic Correlates of Cognition in the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Geriatr Psychiatry. 2015;23(7):684–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [38].Mansournia MA, Nazemipour M, Naimi AI, Collins GS, Campbell MJ. Reflection on modern methods: demystifying robust standard errors for epidemiologists. Int J Epidemiol. 2021;50:346–51. [DOI] [PubMed] [Google Scholar]
  • [39].Blair CK, Folsom AR, Knopman DS, Bray MS, Mosley TH, Boerwinkle E, et al. APOE genotype and cognitive decline in a middle-aged cohort. Neurology. 2005;64:268–76. [DOI] [PubMed] [Google Scholar]
  • [40].Lim YY, Williamson R, Laws SM, Villemagne VL, Bourgeat P, Fowler C, et al. Effect of APOE Genotype on Amyloid Deposition, Brain Volume, and Memory in Cognitively Normal Older Individuals. J Alzheimers Dis. 2017;58:1293–302. [DOI] [PubMed] [Google Scholar]
  • [41].Carry PM, Vanderlinden LA, Dong F, Buckner T, Litkowski E, Vigers T, et al. Inverse probability weighting is an effective method to address selection bias during the analysis of high dimensional data. Genet Epidemiol. 2021;45:593–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [42].Machulda MM, Pankratz VS, Christianson TJ, Ivnik RJ, Mielke MM, Roberts RO, et al. Practice effects and longitudinal cognitive change in normal aging vs. incident mild cognitive impairment and dementia in the Mayo Clinic Study of Aging. Clin Neuropsychol. 2013;27:1247–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [43].Salthouse TA. Trajectories of normal cognitive aging. Psychol Aging. 2019;34:17–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [44].Zhu A, Yan L, Shu C, Zeng Y, Ji JS. APOE epsilon4 Modifies Effect of Residential Greenness on Cognitive Function among Older Adults: A Longitudinal Analysis in China. Sci Rep. 2020;10:82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [45].Crous-Bou M, Gascon M, Gispert JD, Cirach M, Sanchez-Benavides G, Falcon C, et al. Impact of urban environmental exposures on cognitive performance and brain structure of healthy individuals at risk for Alzheimer’s dementia. Environ Int. 2020;138:105546. [DOI] [PubMed] [Google Scholar]
  • [46].Dzhambov AM, Bahchevanov KM, Chompalov KA, Atanassova PA. A feasibility study on the association between residential greenness and neurocognitive function in middle-aged Bulgarians. Arh Hig Rada Toksikol. 2019;70:173–85. [DOI] [PubMed] [Google Scholar]
  • [47].Wang D, Lau KK, Yu R, Wong SYS, Kwok TTY, Woo J. Neighbouring green space and mortality in community-dwelling elderly Hong Kong Chinese: a cohort study. BMJ Open. 2017;7:e015794. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [48].Yu R, Wang D, Leung J, Lau K, Kwok T, Woo J. Is Neighborhood Green Space Associated With Less Frailty? Evidence From the Mr. and Ms. Os (Hong Kong) Study. J Am Med Dir Assoc. 2018;19:528–34. [DOI] [PubMed] [Google Scholar]
  • [49].Seaman SR, White IR. Review of inverse probability weighting for dealing with missing data. Stat Methods Med Res. 2013;22:278–95. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Tables
NIHMS2190145-supplement-Supplemental_Tables.docx (22.3KB, docx)

Data Availability Statement

Data can be requested from MESA at www.mesa-nhlbi.org.

RESOURCES