Table 3.
Bivariate Associations between Air Pollution Exposures and Change in Global Cognitive Performance Mediated by Baseline Emotional Distress from the Women’s Health Initiative Memory Study (N = 6,118)
| Effect Estimates for 3-year average exposures before the baselinea | ||||
|---|---|---|---|---|
| PM2.5b | NO2c | |||
| Outcome | βd | (95% CI) | βd | (95% CI) |
| Estimatesa of Direct Effects | ||||
| Effect of exposure on annual linear changes in 3MS (βexposure on linear Δ3MS) | .029 | (−.061, .120) | −.037 | (−.155, .081) |
| Effect of exposure on annual quadratic changes in 3MS (βexposure on quadratic Δ3MS) | −.101 | (−.213, .010) | .038 | (−.090, .145) |
| Estimatesa of Indirect Effects and Associated Paths | ||||
| Effects of exposure on baseline emotional distress (βexposure on distress) | .016 | (−.020, .053) | .058 | (.018, .098) |
| Effects of baseline emotional distress on annual linear changes in 3MS (βdistress on linear Δ3MS) | .029 | (−.061, .120) | .032 | (−.059, .123) |
| Effects of baseline emotional distress on annual quadratic changes in 3MS (βdistress on quadratic Δ3MS) | −.040 | (−.139, .059) | −.042 | (−.142, .057) |
| Indirect effect of exposure on linear changes in 3MS (βexposure on distress * βdistress on linear Δ3MS) | .0005 | (−.002, .004)e | .002 | (−.004, .008)e |
| Indirect effect of exposure on quadratic changes in 3MS (βexposure on distress * βexposure on quadratic Δ3MS) | −.0006 | (−.005, .002)e | −.0013 | (−.001, .004)e |
| Total exposure effect on linear changes in 3MS | .030 | (−.062, .121) | −.035 | (−.155, .081) |
| Total exposure effect on quadratic changes in 3MS | −.101 | (−.214, .011) | .037 | (−.089, .144) |
Note: 3MS = Modified Mini Mental State Examination; Emotional distress = latent variable consisting of the 6-items of the Center for Epidemiological Studies Depression Scale and the 5-items of the Short Form Health Survey-36 emotional limitation subscale.
All effect estimates were derived from the bivariate multilevel parallel process structural equation model as depicted in panel B of Figure 2. The baseline emotional distress latent factor was scaled on a z-score metric.
PM2.5 estimates represent the average annual exposure (per interquartile range of 3.37 μg/m3) for the 3-years prior to the WHI baseline assessment.
NO2 estimates represent the average annual exposure (per interquartile range of 9.00 ppb) for the 3-years prior to the WHI baseline assessment.
All reported effect estimates on baseline emotional distress, annual linear and quadratic changes were adjusted for age at WHIMS baseline, race/ethnicity, geographic region of residence, education, household income, employment status, lifestyle factors (smoking, alcohol use, physical activities) and clinical characteristics (use of hormone treatment; hypercholesterolemia, hypertension, diabetes, and history of cardiovascular disease).
Effect estimates on annual linear and quadratic changes on the 3MS are adjusted for region of residence, race/ethnicity, employment status, age, alcohol use, diabetes, and hypertension.
Asymmetric 95% confidence interval for indirect effect estimated via Monte Carlo simulation.
Bolded estimates denote PM2.5 or NO2 effects that are statistically significant at p<.05.