Table 2.

 Predictors of in hospital mortality

	Odds ratio (95% CI)	P value
Multivariable non-linear mixed model*
Intercept	0.001 (0.000 to 0.001)	<0.001
Log EuroSCORE†	8.822 (7.701 to 10.107)	<0.001
CABG	1
CABG + other	1.271 (0.938 to 1.721)	<0.122
CABG + valve(s)	1.331 (1.147 to 1.545)	<0.001
CABG + valve(s) + other	2.189 (1.422 to 3.371)	<0.001
Other	1.897 (1.464 to 2.457)	<0.001
Valve(s) only	0.788 (0.675 to 0.919)	<0.003
Valve(s) + other	1.369 (0.962 to 1.949)	0.081
Carstairs score‡	1.029 (1.011 to 1.048)	<0.002
Multivariable non-linear mixed model plus predictors of social deprivation*§
Intercept	0.005 (0.002 to 0.010)	<0.001
Log EuroSCORE†	8.532 (7.429 to 9.797)	<0.001
Diabetes	1.167 (1.010 to 1.347)	<0.036
CABG	1
CABG + other	1.284 (0.947 to 1.741)	<0.108
CABG + valve(s)	1.300 (1.118 to 1.511)	<0.001
CABG + valve(s) + other	2.232 (1.447 to 3.444)	<0.001
Other	1.740 (1.323 to 2.289)	<0.001
Valve(s) only	0.743 (0.633 to 0.872)	<0.001
Valve(s) + other	1.343 (0.942 to 1.916)	0.103
Carstairs score‡	1.025 (1.006 to 1.043)	0.009
BMI	0.924 (0.900 to 0.950)	<0.001
BMI 1¶	1.000 (1.000 to 1.000)	<0.707
BMI 2¶	1.003 (1.001 to 1.006)	<0.010
BMI 3¶	0.994 (0.987 to 1.000)	<0.050

CABG=coronary artery bypass grafting (reference procedure); BMI=body mass index.

*Includes consultant surgeons as random effects (P=0.004).

†Odds ratio represents change in 1 natural log transformed EuroSCORE point.

‡Odds ratio represents change in 1 untransformed Carstairs score point.

§Smoking behaviour, diabetes, and BMI (all related to deprivation) added as additional candidate explanatory variables, although smoking behaviour did not improve model fit and so was excluded in stepwise selection process.

¶Higher order terms from fitting restricted cubic spline with 5 knots to describe effects of BMI as non-linear function. Knots for restricted cubic spline for BMI placed at 21.1, 24.9, 27.2, 29.8, and 35.38.