Assessing the impact of a national clinical guideline for the management of chronic pain on opioid prescribing rates: a controlled interrupted time series analysis

Table 4 Results of the single interrupted time series analyses of weak and strong opioid prescribing rates in Scotland

	Weak¹			Strong²
	Estimate (95% confidence interval)	Standard error	P value	Estimate (95% confidence interval)	Standard error	P value
Intercept, β₀	189.29 (185.48, 193.09)	1.90	<0.01	35.17 (34.15, 36.19)	0.51	<0.01
Pre-intervention trend, β₁	1.36 (1.18, 1.54)	0.09	<0.01	0.82 (0.77, 0.87)	0.02	<0.01
Change in level, β₂	−0.34 (−6.15, 5.47)	2.90	0.91	−0.75 (−2.30, 0.80)	0.77	0.34
Change in trend, β₃	−2.27 (−2.61, −1.93)	0.17	<0.01	−0.55 (−0.64, −0.46)	0.05	<0.01
Post-intervention trend, β₁₊₃	−0.91 (−1.17, −0.64)	0.13	<0.01	0.27 (0.18, 0.36)	0.04	<0.01
Relative change, %^a	−21.68 (−24.73, −18.60)	n/a	n/a	−17.49 (−20.11, −14.78)	n/a	n/a
Durbin-Watson statistic	1.63	n/a	n/a	1.57	n/a	n/a

β_0–3 are coefficients from the single-group interrupted time series analysis model; the intercept represents the outcome at the start of the study period; the relative change is calculated compared to the predicted value at the same time point had the pre-intervention trend continued
¹Weak opioids are the following and their compounds: codeine, dihydrocodeine, meptazinol and tramadol
²Strong opioids are the following and their compounds: buprenorphine, diamorphine, dipipanone, fentanyl, hydromorphone, methadone, morphine, oxycodone, papaveretum, pentazocine, pethidine and tapentadol
^aCalculated at quarter 2 2020. 95% confidence interval calculated using model bootstrapping

ISSN: 1748-5908