Table 2.

Stepwise Cox proportional hazard regression for the individual contribution of cognitive activity, social memberships and social participation

Model	Variable	HR (s.e.)	95% CI	p-Value
Model 1	Low number of cognitive activities	[Reference]
	High number of cognitive activities	0.47 (0.08)	0.33–0.67	<0.0011
Model 2	Low number of cognitive activities	[Reference]
	High number of cognitive activities	0.49 (0.09)	0.35–0.70	<0.0011
Model 3	Low number of cognitive activities	[Reference]
	High number of cognitive activities	0.54 (0.10)	0.37–0.78	0.0011
Model 4	Low number of cognitive activities	[Reference]
	High number of cognitive activities	0.58 (0.11)	0.40–0.84	0.0041
Model 1	Low number of social memberships	[Reference]
	High number of social memberships	0.56 (0.07)	0.44–0.72	<0.0011
Model 2	Low number of social memberships	[Reference]
	High number of social memberships	0.57 (0.07)	0.45–0.73	<0.0011
Model 3	Low number of social memberships	[Reference]
	High number of social memberships	0.61 (0.08)	0.47–0.78	<0.0011
Model 4	Low number of social memberships	[Reference]
	High number of social memberships	0.65 (0.09)	0.50–0.84	0.0011
Model 1	Low frequency of social participation	[Reference]
	High frequency of social participation	0.59 (0.08)	0.45–0.78	<0.0011
Model 2	Low frequency of social participation	[Reference]
	High frequency of social participation	0.61 (0.09)	0.46–0.81	0.0011
Model 3	Low frequency of social participation	[Reference]
	High frequency of social participation	0.65 (0.09)	0.49–0.86	0.0031
Model 4	Low frequency of social participation	[Reference]
	High frequency of social participation	0.71 (0.10)	0.54–0.95	0.0211

HR, hazard ratio; s.e., standard error; LIBRA_adj, LIfestyle for BRAin health score without the weight for cognitive activity.

Model 1 = age and sex; Model 2 = model 1 + education; Model 3 = model 2 + wealth; Model 4 = model 3 + LIBRA_adj.

¹Statistically significant at p≤0.05.