Table 3.

Sensitivity, specificity, and observed 90 day covid-19 mortality risk at different centiles of predicted risk with QCOVID4 to predict covid-19 related death (n=461) in validation cohort of 145 397 people with a positive test result for SARS-CoV-2 infection

Top centile	Predicted 90 day risk threshold (%)	Sensitivity	Specificity	Observed risk (95% CI) at 90 days (%)
Top 5%	0.908	87.6	95.3	5.520 (5.018 to 6.070)
Top 10%	0.278	94.8	90.3	2.986 (2.722 to 3.276)
Top 15%	0.129	96.7	85.3	2.032 (1.853 to 2.228)
Top 20%	0.075	97.8	80.2	1.541 (1.406 to 1.689)
Top 25%	0.049	98.5	75.2	1.241 (1.132 to 1.360)
Top 30%	0.034	98.5	70.2	1.034 (0.943 to 1.133)
Top 35%	0.025	98.5	65.2	0.886 (0.809 to 0.972)
Top 40%	0.019	98.7	60.2	0.777 (0.709 to 0.852)
Top 45%	0.015	99.1	55.2	0.694 (0.633 to 0.761)
Top 50%	0.012	99.3	50.2	0.626 (0.571 to 0.686)
Top 55%	0.010	99.3	45.1	0.569 (0.519 to 0.624)
Top 60%	0.008	99.6	40.1	0.523 (0.477 to 0.573)
Top 65%	0.007	99.6	35.1	0.483 (0.440 to 0.529)
Top 70%	0.006	99.6	30.1	0.448 (0.409 to 0.491)
Top 75%	0.005	99.8	25.1	0.419 (0.382 to 0.459)
Top 80%	0.004	100.0	20.1	0.394 (0.359 to 0.431)
Top 85%	0.003	100.0	15.0	0.371 (0.338 to 0.406)
Top 90%	0.0025	100.0	10.0	0.350 (0.319 to 0.384)
Top 95%	0.0017	100.0	5.0	0.332 (0.303 to 0.363)

CI=confidence interval.