Two-by-two factorial randomised study within a trial (SWAT) to evaluate strategies for follow-up in a randomised prevention trial

Table 3 Collection of questionnaire data at 3, 6, 12 and 18 months according to short message service (SMS) notification allocation

	No SMS notification for questionnaires (n = 702)	SMS notification for questionnaires (n = 692)	Adjusted difference in % collection (95% CI)	Adjusted odds ratio (95% CI)
3 months
Collected	523 (75%)	535 (77%)	2.8% (−1.7 – 7.3%)	1.17 (0.91 – 1.49)
Not collected	179 (25%)	157 (23%)	2.8% (−1.7 – 7.3%)	1.17 (0.91 – 1.49)
6 months
Collected	528 (75%)	523 (76%)	0.4% (−4.1 – 4.9%)	1.02 (0.80 – 1.30)
Not collected	174 (25%)	169 (24%)	0.4% (−4.1 – 4.9%)	1.02 (0.80 – 1.30)
12 months
Collected	542 (77%)	516 (75%)	−2.6% (−7.1 – 1.8%)	0.87 (0.68 – 1.11)
Not collected	160 (23%)	176 (25%)	−2.6% (−7.1 – 1.8%)	0.87 (0.68 – 1.11)
18 months
Collected	506 (72%)	503 (73%)	0.6% (−4.1 – 5.3%)	1.03 (0.82 – 1.30)
Not collected	196 (28%)	189 (27%)	0.6% (−4.1 – 5.3%)	1.03 (0.82 – 1.30)

Data shown are n (%) using the number randomised to each group as the denominator
Estimates are adjusted for host trial allocation
5576 time points from 1394 participants included in model
Some evidence that SMS intervention interacts with time, becoming less effective at 6, 12 and 18 months, overall interaction p = 0.048; therefore, between-group estimates for the effect of prior SMS notification versus none are presented separately for each time point
If the interaction between the SMS intervention and time is ignored, the overall adjusted difference in data collection is 0.3% (95% CI −3.6 – 4.1%) and adjusted odds ratio is 1.02 (0.83 – 1.25)

ISSN: 1745-6215