Evaluation of a targeted, theory-informed implementation intervention designed to increase uptake of emergency management recommendations regarding adult patients with mild traumatic brain injury: results of the NET cluster randomised trial

Table 5 Estimated effects of the intervention on clinical practice outcomes

	NET control¹		NET intervention²
	No. of patients	No. of (%)	No. of patients	No. of (%)	Adj. ORs^**§§	95%CI	p value	Adj. ARD %^{^}	95%CI
Outcomes measuring the implementation of single clinical recommendations
Appropriate post-traumatic amnesia screening (PTA)^*	1050	12 (1.1)	893	117 (13)	20.1	(6.8, 59.3)	< 0.001	14	(8, 19)
PTA screening-tool	1050	15 (1.4)	893	152 (17)	19.7	(6.6, 58.1)	< 0.001	17	(11, 23)
Memory-clinical assessment	1050	272 (26)	893	303 (34)	1.6	(1.2, 2.1)	0.001	9.5	(4.0, 15.1)
CT scan-clinical criteria (CT)^§	494	337 (68)	491	352 (72)	1.2	(0.8, 1.6)	0.375	3.2	(− 3.7, 10.0)
CT scan (all)	1050	458 (44)	893	446 (50)	1.2	(0.9, 1.6)	0.142	4.5	(− 1.5, 10.5)
Provision of written patient information (INFO)	944	175 (19)	785	160 (20)	1.2	(0.8, 1.8)	0.302	3.1	(− 3.0, 9.3)
Outcomes measuring the implementation of composite recommendations
Safe discharge based on PTA and INFO	944	2 (0.2)	785	45 (6)	27.6	(6.9, 110.5)	< 0.001	5.8	(2.7, 8.9)
Safe discharge based on PTA, CT, and INFO^§§§	413	0 (0)	402	14 (3.5)	1.8	(1.1, 3.0)	0.022	3.5	(1.0, 6.0)

¹Number of clusters = 17
²Number of clusters = 14
ORs = odds ratios
^*Primary outcome
^§Criteria that justify a scan are age 65 or older; GCS < 15, amnesia, suspected skull fracture, vomiting and coagulopathy. Only the subset of patients who have these symptoms noted in the medical records are included in the analysis
^**Adjusted odds ratios estimated from marginal logistic regression models using generalised estimating equations with an exchangeable correlation structure (unless otherwise noted) and robust variance estimation to allow for clustering of responses within EDs
^§§All models (unless otherwise noted) adjusted for the minimisation factors and pre-specified confounders (see ‘Effectiveness analyses’ section)
^§§§For this outcome, because there were no safe discharges in the control group, a cluster-level analysis was undertaken resulting in a ratio of geometric mean proportions. Details available in Additional file 2
^{^}ARD calculated from marginal probabilities [75]. Confidence intervals for the metric were obtained by a pairwise comparison of margins after fitting a GEE model using Stata [43] allowing for clustering of observations within EDs

ISSN: 1748-5908