Table 5.

Additive predictive value of proposed variables compared with the refit GAP index

Model	C-statistic (95% CI)	Difference in C-statistic compared with GAP*		NRI (%) (95% CI)*	IDIevent (95% CI)*	IDInon-event (95% CI)*	IDI
		(95% CI)	P value
Original GAP index	0.683 (0.624 to 0.742)	−0.018 (−0.046 to 0.034)	0.208	−4.0 (−35.1 to 34.4)	−0.003 (−0.005 to 0.012)	−0.001 (−0.001 to 0.003)	−0.004 (−0.005–0.015)
Refit GAP index	0.701 (0.646 to 0.756)	–	–	–	–	–	–
GAP +6MWD††	0.738 (0.686 to 0.789)	0.037 (0.003 to 0.070)	0.031	46.6 (15.0 to 75.4)	0.034 (0.005 to 0.083)	0.007 (0.002 to 0.017)	0.041 (0.006–0.101)
GAP +EH	0.742 (0.691 to 0.793)	0.041 (0.004 to 0.078)	0.031	88.5 (51.8 to 108.3)	0.045 (0.013 to 0.084)	0.009 (0.003 to 0.018)	0.055 (0.016–0.100)
GAP +6 MWD††+EH	0.756 (0.706 to 0.806)	0.055 (0.011 to 0.098)	0.014	61.9 (26.9 to 102.9)	0.062 (0.026 to 0.119)	0.012 (0.005 to 0.025)	0.074 (0.032–0.143)
DO-GAP index‡‡	0.752 (0.701 to 0.802)	0.050 (0.004 to 0.097)	0.035	46.1 (19.6 to 66.2)	0.055 (0.022 to 0.101)	0.011 (0.004 to 0.022)	0.066 (0.026–0.122)

*Improvement comparisons made to the GAP index after refitting original GAP index to the derivation set.

†6MWD included as categorical variable.

‡Model based on point-score index (see study design and methods for further detail).

DO-GAP, distance-oxygen-gender-age-physiology; EH, exertional hypoxia; GAP, gender-age-physiology; IDI, incremental discrimination improvement; 6MWD, 6 min walk test distance; NRI, net reclassification improvement.