Table 6.

Applications of Bayesian additive regression trees (BART) to data from male donors in the INTERVAL trial assuming the target is to maximize the utility score. The trade-off parameter b in the utility function varies from 1 to 5 at an increment of 1.

		Trade-off Parameter
Criteria	Assignment Rule	b = 1	b = 2	b = 3	b = 4	b = 5
BART ITRAssignment Percentages	12 weeks	0.0	3.6	10.7	17.5	26.7
	10 weeks	3.2	9.4	18.8	25.5	28.4
	8 weeks	96.8	87.1	70.5	57.0	44.9
ITR Effects on UtilityPosterior Mean [95% Credible Interval]	All 12 weeks	−1.037[−1.118,−0.956]	−0.827 [−0.915,−0.739]	−0.618 [−0.712,−0.524]	−0.408 [−0.511,−0.305]	−0.199 [−0.317,−0.082]
	All 10 weeks	−0.025 [−0.106, 0.055]	0.027 [−0.060, 0.114]	0.079 [−0.017, 0.174]	0.131 [0.027, 0.232]	0.183 [0.068, 0.297]
	All 8 weeks	1.054 [0.973, 1.134]	0.794 [0.708, 0.883]	0.535 [0.437, 0.628]	0.275 [0.169, 0.379]	0.016 [−0.101, 0.135]
	BART ITR	1.064 [0.983, 1.144]	0.876 [0.793, 0.959]	0.750 [0.643, 0.843]	0.671 [0.553, 0.780]	0.732 [0.548, 0.852]
	Optimized ITR	1.082 [1.000, 1.159]	0.898 [0.814, 0.981]	0.802 [0.701, 0.893]	0.770 [0.671, 0.870]	0.823 [0.710, 0.928]

Note: Assignment proportions of the BART ITR in percentage and the posterior mean [95% equal tail credible interval] of the ITR effect on the utility outcome for five donor assignment rules are reported. Assignment rules include the following: recommend all male donors to (i) donate every eight weeks, (ii) donate every 10 weeks, (iii) donate every 12 weeks, (iv) donate according to the BART ITR, and (v) donate according to the optimized ITR (non-achievable in practice). A larger ITR effect on utility is more desirable.