TABLE 1.

Accuracy of predicted FVIII levels in the simulation experiment

Sampling strategy		Standard DCM		DCM with initialization
	n	Train	Test	Train	Test
t = 0.5, 4, 12, 24, 36, 48	120	99.0 ± 0.21	99.1 ± 0.25	99.6 ± 0.12	99.4 ± 0.16
	60	93.3 ± 13.0	93.0 ± 12.5	98.9 ± 0.42	97.9 ± 0.18
	20	89.5 ± 1.09	84.4 ± 1.79	92.8 ± 1.76	88.7 ± 3.27
t = 4, 24, 48	120	65.2 ± 8.68	65.3 ± 8.86	97.8 ± 0.33	97.8 ± 0.41
	60	60.7 ± 0.61	59.5 ± 0.62	96.0 ± 0.85	94.8 ± 0.97
	20	58.2 ± 0.99	59.1 ± 0.71	96.3 ± 1.18	90.1 ± 2.00
t = 8, 30	120	75.9 ± 0.65	76.1 ± 1.08	90.8 ± 6.63	90.3 ± 6.19
	60	72.4 ± 1.33	73.6 ± 1.19	81.4 ± 3.29	83.0 ± 3.08
	20	66.8 ± 1.78	61.2 ± 1.41	77.7 ± 4.82	76.5 ± 2.19
t = 24	120	28.6 ± 3.69	28.9 ± 5.31	76.2 ± 2.74	76.0 ± 2.41
	60	29.2 ± 1.21	29.4 ± 1.02	66.8 ± 2.23	65.2 ± 2.14
	20	29.6 ± 2.68	32.2 ± 1.92	73.7 ± 1.83	72.9 ± 1.80

Note: Train and test accuracy is represented as the percentage of predictions within 0.05 IU ml⁻¹ of true simulated FVIII levels ≥0.15 and within 0.02 IU ml⁻¹ of levels <0.15. Time points are in hours. $n$ is the number of patients in the train set. Test set size is the remainder of 500 – $n$. Values are represented as the mean ± one SD of five replicates.

Abbreviations: DCM, deep compartment model; FVIII, factor VIII.