Effects of neck-exercise and health promotion on headache outcomes in office workers: secondary analysis of the NEXpro stepped wedge cluster randomised controlled trial

Table 2 Binominal models for headache occurrence

Outcome Variable: Headache occurrence last four weeks
Binominal model: Headache: Yes (1) or No (0)	Model 1 ~ Intervention	Model 2 ~ Intervention + Period	Model 3 ~ Intervention x Period	Model 4 ~ Period + Exposure time (Intervention nested in exposure time)
Independent variable	Odds ratio (95%CI)	Odds ratio (95%CI)	Odds ratio (95%CI)	Odds ratio (95%CI)
Constant	6.06 (3.15 to 11.67)	6.36 (2.94 to 13.77)	6.66 (3.01 to 14.74)	6.48 (2.96 to 14.17)
Intervention	0.46 (0.25 to 0.84)	0.98 (0.40 to 2.42)		-
Period 1		1.10 (0.48 to 2.50)	0.98 (0.40 to 2.35)	0.99 (0.43 to 2.28)
Period 2		0.76 (0.33 to 1.75)	0.94 (0.38 to 2.35)	0.88 (0.36 to 2.16)
Period 3		0.63 (0.23 to 1.73)	0.41 (0.12 to 1.39)	0.35 (0.12 to 1.05)
Period 4		0.32 (0.10 to 1.07)	0.28 (0.12 to 0.65)	0.24 (0.06 to 1.04)
Intervention x Period 1			1.38 (0.32 to 5.89)
Intervention x Period 2			0.28 (0.0 to 1.14)
Intervention x Period 3			1.67 (0.43 to 6.50)
Exposure: 4 months				1.29 (0.49 to 3.37)
Exposure: 8 months				0.54 (0.15 to 1.87)
Exposure: 12 months				4.30 (0.75 to 24.71)
Exposure: 16 months				2.46 (0.34 to 18.00)
Observations	520	520	520	520
Log likelihood	−267.9	−265.1	−262.9	−261.1
Akaike Information Criterion (AIC)	543.8	546.2	545.7	544.3
Bayesian Information Criterion (BIC)	560.8	580.2	588.3	591.1

Values are odds ratios with 95 % Confidence intervals in brackets, all back transformed from logit scale. Figures in bold indicate the best fitting Model
Model one: the overall effect of the intervention had an odds ratio of 0.46 (95% confidence interval: 0.25 to 0.84)
Model two: By adding period to the model, no significant additional effect for any period was found. A trend for slightly larger effects, irrespective of the intervention, within later periods was found (See also Fig. 2)
Model three: No significant interaction for intervention and period was found. At period four when all participants had received the intervention, the intervention effect amounted to an odds ratio of 0.28 (95% CI: 0.12 to 0.65)
Statistical fit between models one, two and three was tested by ANOVA and demonstrated no better fit for Model two (p= 0.34) or three (p=0.11) compared to Model one
Model four: No significant effect for period and the intervention, nested in exposure time, was demonstrated. As was shown for Models two and three, a trend for a larger effect in later periods was found. AIC and BIC demonstrated a lower statistical fit, compared to other models and especially Model one

ISSN: 1129-2377