Table 3.

Odds ratios (95% CI) from multilevel logistic regression for the relationship between neighbourhood SES trajectory and birth outcomes.

A)	Full sample
Neighbourhood SES trajectory	Preterm birth		SGA
	Model 1	Model 2	Model 1	Model 2
Stable High	REF	REF	REF	REF
Stable Medium	1·06 (1·02-1·10)	1·04 (0·99-1·09)	1·07 (1·04-1·10)	1·03 (0·99-1·07)
Stable Low	1·20 (1·15-1·25)	1·12 (1·07-1·17)	1·37 (1·32-1·42)	1·19 (1·15-1·23)
Improving to High	0·99 (0·91-1·07)	0·98 (0·91-1·06)	0·98 (0·92-1·04)	0·97 (0·92-1·03)
Improving to Medium	1·13 (1·05-1·22)	1·09 (1·02-1·18)	1·11 (1·05-1·18)	1·04 (0·98-1·10)
Declining to Medium	1·04 (0·97-1·11)	1·03 (0·97-1·10)	1·05 (1·00-1·10)	1·03 (0·98-1·08)
Declining to Low	1·17 (1·10-1·25)	1·12 (1·05-1·20)	1·23 (1·17-1·29)	1·12 (1·06-1·18)

B)	Subsample of women who remained in the same address throughout entire exposure period
	Preterm birth		SGA
	Model 1	Model 2	Model 1	Model 2
Stable High	REF	REF	REF	REF
Stable Medium	1·07 (1·00-1·14)	1·05 (0·99;1·13)	1·05 (1·00;1·10)	1·02 (0·97;1·06)
Stable Low	1·26 (1·18-1·34)	1·17 (1·09;1·25)	1·34 (1·27;1·41)	1·17 (1·11;1·23)
Improving to High	1·03 (0·91-1·06)	1·02 (0·90;1·14)	0·98 (0·90;1·08)	0·98 (0·90;1·07)
Improving to Medium	1·17 (1·05-1·31)	1·12 (1·01;1·26)	1·07 (0·98;1·16)	1·00 (0·92;1·09)
Declining to Medium	1·01 (0·92-1·12)	1·00 (0·91;1·11)	1·06 (0·98;1·14)	1·04 (0·96;1·12)
Declining to Low	1·15 (1·04-1·28)	1·09 (1·01;1·18)	1·26 (1·14;1·32)	1·12 (1·04;1·20)

Model 1: Including only time-point dummy variables and time-period × neighbourhood SES trajectory interactions.

Model 2: Including Model 1 terms and adjusting for individual-level characteristics: maternal age, parity, migration background and household income.

Stable High trajectory (most advantaged) as reference category (REF).

First time-period (2003-2006) used as reference.

Number of preterm births and SGA births in each SES category are displayed in Table 2.

Part B corresponds to estimates from sensitivity analysis 3b.