Table 3.

Predictive performance of physical activity and sedentary behaviour composite scores 1 and 2 for mortality risk in the Whitehall II accelerometer sub-study (N cases/N total = 410/3991, mean [standard deviation] follow-up = 8.1 [1.3] years).

	HR (95% CI) PA and SB composite scores	Royston's R2 (95% CI)	AIC (95% CI)	ΔAIC (95% CI)	ΔAIC (95% CI)	Sensitivity % (95% CI)a	Specificity % (95% CI)a	C-index (95% CI)	ΔC-index (95% CI)	ΔC-index (95% CI)
Model 1 b	NA	0.366 (0.328, 0.424)	6374.6 (5801.3, 6989.1)	Ref.	NA	69.9 (56.4, 87.9)	66.4 (47.1, 82.3)	0.751 (0.731, 0.778)	Ref.	NA
Model 2 c	PA and SB score 1: 1.10 (1.06, 1.14), p < 0.001	0.382 (0.344, 0.439)	6355.3 (5794.5, 6960.7)	−19.3 (−45.1, −5.6)e	Ref.	70.3 (53.2, 85.2)	67.1 (51.4, 83.7)	0.758 (0.738, 0.784)	0.007 (0.002, 0.014)e	Ref.
Model 3 d	PA and SB score 1: 1.10 (1.05, 1.15), p < 0.001	0.381 (0.345, 0.439)	6357.2 (5798.4, 6963.5)	−17.4 (−44.5, −4.3)e	2.0 (−7.5, 1.9)	67.2 (50.5, 86.1)	70.4 (52.9, 86.1)	0.758 (0.738, 0.784)	0.007 (0.003, 0.014)e	−0.0001 (−0.0003, 0.0041)
	PA and SB score 2: 0.99 (0.89, 1.10), p = 0.84

Abbreviations: CI: confidence interval; AIC: Akaike information criterion; C-index: Harrell's C-index.

^aYouden index cutoff points for the calculation of the sensitivity and specificity: 0.363 (95% CI: [0.301, 0.475]) for model 1, 0.374 (95% CI: [0.318, 0.489]) for model 2, and 0.381 (95% CI: [0.354, 0.424]) for Model 3.

^bModel 1: Model includes socio-demographic, behavioural, and health-related factors.

^cModel 2: Model 1 additionally adjusted for the physical activity and sedentary behaviour composite score 1.

^dModel 3: Model 2 additionally adjusted for the physical activity and sedentary behaviour composite score 2.

^eConfidence intervals (calculated based on 1000 bootstrap samples) not containing zero indicate significant differences in predictive performance between Model 1 and Model 2, and between Model 2 and Model 3.