Table 8.

Multiple regression: effect of each factor on TCI total score.

Predictor (reference category)	Test of predictor‡			Fixed components				Random components
	δ	d.f.	P	Parameter	Estimate	95% CI	p	Parameter	σ	95% CI
Gender	4.34	1	0.036	constant	54.49	49.33, 59.65	<0.001	Practices	1.29	0.02, 2.57
				male	3.66	0.23, 7.08	0.04	Individuals	7.93	7.22, 8.64
Tenure	3.85	1	0.049	tenure/year	0.13	0.00, 0.26	0.05
Professional groups (GP)	7.57	4	0.109	nurse	-5.36	-9.31, -1.40	0.01
				admin	-3.42	-6.75, -0.08	0.04
				com nurse	-3.21	-7.56, 1.16	0.15
				Other	-1.60	-5.48, 2.28	0.41
Practice list size (≤ 3500)	5.20	2	0.074	3501–7000	-4.07	-7.43, -0.72	0.02
				>7000	-3.15	-6.84, 0.54	0.09
Number of GPs (1 or 2)	7.28	2	0.026	3–5	-2.53	-7.18, 2.13	0.29†
				6 or more	1.79	-3.24, 6.82	0.49†

‡ The relative importance of each predictor is assessed by comparing the change in -2 log-likelihood that occurs when a particular predictor is omitted from the final model with the percentage points of a chi-squared distribution with degrees of freedom given in the column titled d.f.

σ = standard deviation of TCI scores (either between practices or between individuals within practices as specified).

† These estimates suggest that the difference between the smallest practice and the medium size practices and the difference between the smallest practices and the largest practices are not statistically significant; however the difference between medium and large practices is 4.32 with 95% CI (1.37, 7.28).