Table 3.

Subgroup analysis for residential greenness and cognitive function

Covariates	Serial cross-sectional analysis (n=38 327)		Longitudinal analysis (n=19 726)
	Linear regression	Logistic regression	Linear mixed effects regression	Mixed effects logistic regression
	Coefficient for change of MMSE score (95% CI)	OR of cognitive impairment (95% CI)	Coefficient for change of MMSE score (95% CI)	OR of cognitive impairment (95% CI)
Stratified by gender
Males (n=15 747)	0.20 (0.11 to 0.30)	0.93 (0.90 to 0.96)	0.083 (−0.0042 to 0.17)	0.98 (0.95 to 1.02)
Females (n=22 580)	0.24 (0.15 to 0.33)	0.94 (0.92 to 0.97)	0.060 (−0.032 to 0.15)	0.99 (0.96 to 1.02)
P value for interaction	0.31	0.003	0.051	0.044
Stratified by residence
Urban area (n=9076)	0.36 (0.24 to 0.49)	0.92 (0.89 to 0.96)	0.050 (−0.025 to 0.12)	0.99 (0.94 to 1.04)
Rural area (n=29 251)	0.16 (0.081 to 0.24)	0.95 (0.93 to 0.97)	0.11 (−0.023 to 0.24)	0.99 (0.96 to 1.02)
P value for interaction	0.007	0.168	0.333	0.902
Stratified by physical activity
Exercise (n=10 923)	0.084 (−0.013 to 0.18)	0.97 (0.94 to 1.01)	0.027 (−0.069 to 0.12)	1.02 (0.98 to 1.06)
Do not exercise (n=27 404)	0.24 (0.16 to 0.33)	0.93 (0.90 to 0.95)	0.069 (−0.015 to 0.15)	0.97 (0.95 to 1.00)
P value for interaction	0.025	<0.001	0.136	0.009

All the regression models on 0.1-unit NDVI and cognitive function were adjusted for age, gender, ethnicity, marital status, urban/rural residence, education, occupation, financial support, social and leisure activity, smoking status, alcohol consumption and physical activity at baseline. The longitudinal analysis was additionally adjusted for the time to reflect the number of years for each follow-up survey since entering the cohort.

MMSE, Mini-Mental State Examination; NDVI, Normalised Difference Vegetation Index.