Table 3

ORs with 95% CIs for three outcomes in three groups of frequency of visits to supermarkets, as derived from generalised estimating equation (GEE) models that included interaction between time segments and frequency of visits to supermarkets. Interactions test differential associations between time segments and outcomes, between exposure groups

	Frequency of visits to supermarkets
	Rarely	Sometimes	P for interaction*	Often	P for interaction*
Tobacco accessibility†
Time segments
2013	Ref	Ref		Ref
2014–2015	0.75 (0.50 to 1.13)	0.93 (0.69 to 1.24)	0.404	0.63 (0.43 to 0.94)	0.563
2016–2017	0.68 (0.44 to 1.03)	0.84 (0.62 to 1.13)	0.383	0.58 (0.37 to 0.87)	0.566
Smoking norm‡
Time segments
2013	Ref	Ref		Ref
2014–2015	0.84 (0.53 to 1.35)	0.83 (0.59 to 1.17)	0.975	0.74 (0.48 to 1.15)	0.684
2016–2017	0.87 (0.54 to 1.38)	1.13 (0.82 to 1.59)	0.329	0.90 (0.57 to 1.42)	0.910
Smoking attitude§
Time segments
2013	Ref	Ref		Ref
2014–2015	0.95 (0.53 to 1.72)	0.79 (0.53 to 1.18)	0.613	0.81 (0.48 to 1.39)	0.694
2016–2017	0.88 (0.49 to 1.59)	0.64 (0.42 to 0.96)	0.338	0.62 (0.36 to 1.08)	0.378

*Rarely was the reference category in the interaction analysis.
†Tobacco accessibility was defined as perceiving tobacco to be purchasable in shops by someone their age.
‡Smoking norm was defined as perceiving that friends think it is OK for people their age to smoke.
§Smoking attitude was defined as thinking it is OK for someone their age to smoke.