Table 3.

Performance of deep learning model (area under receiver operating characteristic curve score) across different races/ethnicities and demographics in the UK Biobank test dataset

Demographic segment		AUROC‡	Group size	N where PCE ≥7.5%
Sex assigned at birth	Female	0.88	4615	217
	Male	0.89	3991	1824
Race/ethnicity†	Non-Hispanic White	0.87	8023	1945
	South Asian	0.83	206	44
	East Asian
	Black/African American	0.89	166	24
	Other/multiracial	0.93	202	25
Age bracket	Age ≤60	0.72	5112	417
	Age >60	0.84	3494	1624

AUROC = area under the receiver operating characteristic curve; PCE = pooled cohort equation.

^†Participants self-identified as belonging to a particular race/ethnicity group (Supplemental Appendix 2).

^‡Sex subgroups comparison (female to male), Delong test statistic: -0.6114, P value: .54. Age subgroups comparison, Delong test: -14.3876, P value: .000. Race/ethnicity subgroup comparisons: White vs Asian, Delong test statistic: 1.513, P value: 0.13; White vs Black, Delong test statistic: -0.814, P value: .416; Asian vs Black, Delong test statistic: -1.681, P value: .09.