Table 2 Parameter estimates from the full and the most parsimonious (final) segmented regression models predicting the mean monthly number of proton pump inhibitor units used per 10 000 patient-days

	Estimate coefficient	SD	p Value
Full model*
1st segment (before the 1st intervention)
Intercept α₁	885.88	44.65	<0.0001
Baseline trend β₁	4.55	2.60	0.0846
2nd segment (during the 1st intervention)
Level change after June 2003 α₂	−358.51	80.91	<0.0001
Trend change β₂	−28.88	9.06	0.0022
3rd segment (between the two interventions)
Level change after July 2004 α₃	122.25	83.05	0.1458
Trend change β₃	38.38	9.97	0.0003
4th segment (after the 2nd intervention)
Level change after February 2006 α₄	−65.76	88.70	0.4611
Trend change β₄	−8.16	10.95	0.4588
Final model†
1st segment (before the 1st intervention)
Intercept α₁	954.16	22.06	<0.0001
2nd segment (during the 1st intervention)
Level change after June 2003 α₂	−294.78	73.30	0.0001
Trend change β₂	−24.33	8.81	0.0074
3rd segment (between the two interventions)
Level change after July 2004 α₃	171.19	76.09	0.0277
Trend change β₃	32.45	9.12	0.0007

*The full model contains all the explanatory variables used for testing the effects of interventions.
†The final model (ie, the most parsimonious model) retains all the significant terms of the full model. This model contains the significant changes associated with interventions.