Exposure to Greenness and Mortality in a Nationwide Prospective Cohort Study of Women

The researchers tested whether the amount of vegetation (greenness) around a woman's home was associated with her risk of dying over an 8-year period. They compared women living in the greenest areas (top 20% of greenness) to women living in the least green areas (bottom 20% of greenness). The primary outcome was all-cause nonaccidental mortality (death from any cause except accidents, suicide, or homicide). Secondary outcomes included deaths from specific causes: cardiovascular disease, respiratory disease, cancer, and other causes. They also tested whether the relationship was stronger for certain distances from the home (250 metres vs. 1,250 metres) and whether the effect could be explained by factors like physical activity, air pollution exposure, social engagement, and depression.

The study included 108,630 women from the Nurses' Health Study, a long-running prospective cohort of registered nurses in the United States. At the start of the follow-up period in 2000, participants were aged approximately 44–69 years (the cohort was established in 1976 with nurses aged 30–55, so by 2000 most were in their 50s and 60s). All were female, predominantly white (over 90%), and living across all 50 U.S. states. They were generally healthier and of higher socioeconomic status than the general U.S. population. The analysis excluded women who had died before 2000, those with missing address data, and those who lived in areas where satellite greenness data could not be reliably calculated (e.g., very dense urban cores with no vegetation). During the 8-year follow-up, 8,604 deaths occurred among the 108,630 women.

Residential greenness was measured using the Normalized Difference Vegetation Index (NDVI), a satellite-based measure that quantifies vegetation density on a scale from -1 to +1. Higher values indicate more green vegetation. NDVI was calculated from satellite imagery at a 30-metre resolution (roughly the size of a house and yard). For each participant, the researchers calculated the average NDVI within two buffer zones around their home address: a 250-metre radius (about a 3-minute walk) and a 1,250-metre radius (about a 15-minute walk). Greenness was measured four times per year (seasonally) and then averaged over the entire follow-up period, accounting for any changes in residence. Participants were divided into five groups (quintiles) based on their cumulative average greenness.

Mortality was determined through the National Death Index, state vital statistics records, and reports from family members or the postal service. Cause of death was coded using the International Classification of Diseases (ICD) system. Physical activity was measured using a validated questionnaire asking about hours per week of moderate-to-vigorous activity (e.g., brisk walking, jogging, cycling). Air pollution (particulate matter smaller than 2.5 micrometres, or PM2.5) was estimated using a national spatiotemporal model. Social engagement was assessed by asking how often participants attended religious services, club meetings, or group activities. Depression was measured using the Mental Health Index from the Medical Outcomes Study Short Form-36 (SF-36), a validated 5-item scale.

**Study design:** This is a prospective observational cohort study. The researchers followed a large group of women forward in time (2000–2008) and compared mortality rates between those with different levels of residential greenness at the start and during follow-up.

**Key design features:** The study used a time-varying exposure approach, meaning that greenness was recalculated each time a participant moved to a new address. This is important because people do move, and their exposure to greenness changes. The researchers also used cumulative average greenness (averaging all seasonal measurements over the entire follow-up), which reduces the influence of a single bad year or season. They controlled for a wide range of potential confounders measured at baseline: age, race/ethnicity, smoking status (never, past, current), body mass index (BMI), individual-level socioeconomic status (education, husband's education, occupation), and area-level socioeconomic status (median household income, median home value, and poverty rate at the census tract level). They also adjusted for region of the country and urbanicity (whether the address was in a city, suburb, or rural area).

**Statistical approach:** They used Cox proportional hazards models, which are standard for time-to-event data (time until death). Results are reported as hazard ratios (HR) with 95% confidence intervals (CI). A hazard ratio of 0.88 means an 12% lower rate of death. They also conducted a mediation analysis to test whether physical activity, air pollution, social engagement, and depression could statistically explain part of the greenness–mortality link. Mediation analysis tests whether a third variable (e.g., physical activity) lies on the causal pathway between the exposure (greenness) and the outcome (mortality).

**What this design can prove:** A prospective cohort study can establish that greenness precedes mortality (temporal order), which is a necessary condition for causality. The large sample size and long follow-up provide statistical power to detect modest effects. The extensive adjustment for confounders reduces the chance that the observed association is due to other factors (e.g., wealthier people live in greener areas and also have better health).

**What this design cannot prove:** This is not a randomised controlled trial. The researchers did not assign women to live in green or non-green areas. Therefore, the study cannot prove that increasing greenness causes lower mortality. There may be unmeasured confounders—factors that differ between women in green vs. non-green areas that also affect mortality risk. For example, women who choose to live in greener areas may also be more health-conscious in other ways (diet, healthcare utilisation, stress management) that were not fully measured. The mediation analysis is suggestive but cannot prove causation; it relies on assumptions about the direction of relationships and the absence of unmeasured confounders of the mediator–outcome relationship.

**Major methodological weaknesses:** The study population is not representative of the general U.S. population (all female, mostly white, all nurses, relatively high socioeconomic status). Greenness was measured from satellite imagery, which captures overall vegetation but not the type of green space (e.g., a manicured lawn vs. a forest vs. a park with benches) or its quality (e.g., safety, accessibility, maintenance). The study did not measure actual use of green spaces—only proximity. The mediation analysis used variables measured at a single time point (2000), but physical activity, air pollution, social engagement, and depression can change over time. Finally, the study did not account for the possibility that people who become ill might move to less green areas (reverse causation), although the time-varying exposure approach partially addresses this.

**Primary outcome (all-cause nonaccidental mortality):** Women in the highest quintile of cumulative average greenness (within 250 metres of home) had a 12% lower rate of death compared to women in the lowest quintile (HR = 0.88; 95% CI: 0.82, 0.94; p < 0.001). This means that over the 8-year follow-up, the death rate was 12% lower in the greenest areas.

**Dose-response relationship:** There was a graded relationship across quintiles. Compared to the lowest quintile, the hazard ratios for quintiles 2 through 5 were: 0.95 (95% CI: 0.89, 1.01), 0.93 (95% CI: 0.87, 0.99), 0.90 (95% CI: 0.84, 0.96), and 0.88 (95% CI: 0.82, 0.94). This suggests that more greenness is associated with progressively lower mortality risk.

**Distance from home:** The association was slightly weaker for the 1,250-metre buffer. For the highest vs. lowest quintile at 1,250 metres, the HR was 0.91 (95% CI: 0.85, 0.97). This suggests that greenness very close to home (within a 3-minute walk) may matter more than greenness further away.

**Cause-specific mortality:**

- **Respiratory mortality:** Strongest association. Highest vs. lowest quintile of greenness (250 m): HR = 0.66 (95% CI: 0.48, 0.91). This is a 34% lower rate of death from respiratory diseases (e.g., COPD, pneumonia).

- **Cancer mortality:** HR = 0.87 (95% CI: 0.78, 0.97). A 13% lower rate of cancer death.

- **Cardiovascular mortality:** HR = 0.92 (95% CI: 0.79, 1.07). This was not statistically significant (the confidence interval includes 1.00).

- **Other causes:** HR = 0.88 (95% CI: 0.76, 1.02). Not statistically significant.

**Mediation analysis:** The researchers estimated that the following factors could statistically explain part of the greenness–mortality association:

- Physical activity: mediated approximately 10–15% of the effect

- PM2.5 air pollution: mediated approximately 10–15% of the effect

- Social engagement: mediated approximately 5–10% of the effect

- Depression: mediated approximately 5–10% of the effect

- Together, these four factors accounted for about 30–40% of the total association, meaning that more than half of the effect remains unexplained.

**Subgroup analyses:** The association between greenness and mortality was consistent across strata of age, smoking status, BMI, and urbanicity. It was slightly stronger among women living in the Northeast and Midwest compared to the South and West.

A 12% reduction in mortality rate over 8 years is a moderate effect. To put it in context: if the annual death rate in the least green areas was about 1% per year (roughly 10 deaths per 1,000 women per year), then in the greenest areas it would be about 0.88% per year (8.8 deaths per 1,000 women per year). Over 8 years, that translates to roughly 9 fewer deaths per 1,000 women in the greenest vs. least green areas. This effect size is comparable to the mortality benefit seen with moderate physical activity (e.g., walking 30 minutes per day, 5 days per week) or with reducing PM2.5 air pollution by about 5 micrograms per cubic metre. For respiratory mortality specifically, a 34% reduction is a large effect—comparable to the benefit of quitting smoking for some respiratory conditions. However, the confidence interval for this finding is wide (0.48 to 0.91), meaning the true effect could be as small as a 9% reduction or as large as a 52% reduction.

**Observational design:** Cannot prove causation. Unmeasured confounders (e.g., health consciousness, diet quality, healthcare access, noise pollution, crime rates) could explain the findings.

**Population limited to women:** All participants were female, mostly white, and all were nurses. Results may not generalise to men, other racial/ethnic groups, or non-healthcare workers.

**Greenness measurement:** NDVI measures total vegetation from satellite imagery but does not distinguish between types of green space (e.g., a golf course vs. a community garden vs. a forest) or their quality (e.g., accessibility, safety, maintenance). A high NDVI could come from agricultural fields that are not accessible for recreation.

**No measure of actual use:** The study assumed that living near greenness means exposure to greenness. It did not measure how often women actually visited green spaces, how long they spent there, or what they did there.

**Mediation analysis limitations:** The mediators (physical activity, PM2.5, social engagement, depression) were measured at a single time point (2000) but could change over the 8-year follow-up. The analysis assumes no unmeasured confounders of the mediator–outcome relationship, which is unlikely to be true.

**Residual confounding by socioeconomic status:** Although the researchers adjusted for individual- and area-level SES, wealthier people tend to live in greener areas and also have better health outcomes. It is possible that the adjustment did not fully capture all aspects of socioeconomic advantage.

**No data on noise or heat:** The study did not measure noise pollution or extreme heat, both of which are environmental exposures that greenness might mitigate. If greenness reduces noise or heat, those effects were not tested.

**Short follow-up for some outcomes:** Eight years is moderate but may be too short to capture effects on slowly developing diseases like cancer. The cancer mortality finding (13% reduction) could be influenced by pre-existing disease at baseline.

**The authors acknowledge** that the study cannot rule out residual confounding, that the NDVI measure is imperfect, and that the mediation analysis is exploratory.

For someone running their own n=1 experiment:

**What to test:**

Increase your exposure to green vegetation within a 250-metre radius of your home. This could mean:

- Moving to a neighbourhood with more trees, parks, or gardens

- Adding potted plants, a balcony garden, or a small lawn if you have outdoor space

- Spending more time in nearby parks or green spaces (even if your home itself is not green)

- If you live in an apartment, consider a window box, indoor plants, or visiting a community garden

The "dose" that showed benefit in this study was living in the top 20% of greenness nationally. A practical target: aim for an NDVI value above 0.4–0.5 (you can check this using Google Earth Engine or the National Land Cover Database). More practically, aim to have visible trees, grass, or gardens within a 3-minute walk of your front door.

**Minimum meaningful duration:**

The study followed women for 8 years, but the greenness measure was cumulative and time-varying. For a self-experiment, a minimum of 3–6 months is reasonable to see changes in intermediate outcomes (mood, physical activity, stress). For mortality effects, you cannot test that directly. Instead, test the proposed mediators.

For testing effects on mental health or physical activity, 4–8 weeks of consistent exposure (e.g., daily walks in a park) should be sufficient to detect changes.

**What to measure (specific metrics):**

**Primary outcome (for a self-experiment):** All-cause mortality is not testable. Instead, measure the proposed mediators:

- **Physical activity:** Steps per day (using a pedometer or smartphone), minutes of moderate-to-vigorous activity per week (self-report or wearable)

- **Air pollution exposure:** If you have access to a portable PM2.5 monitor (e.g., PurpleAir, AirVisual), measure indoor and outdoor levels. Alternatively, check local air quality monitoring data

- **Social engagement:** Number of social interactions per week, time spent with friends/family, attendance at community events

- **Depression/mood:** Patient Health Questionnaire-9 (PHQ-9, 0–27 scale, lower = better), or the Positive and Negative Affect Schedule (PANAS). Measure weekly

- **Stress:** Perceived Stress Scale (PSS-10, 0–40 scale, lower = better). Measure weekly

- **Sleep:** Pittsburgh Sleep Quality Index (PSQI, 0–21 scale, lower = better) or a sleep tracker (e.g., Oura Ring, Fitbit)

**Secondary outcomes:** Blood pressure (measured at the same time each day), heart rate variability (HRV, measured with a chest strap or smartwatch),