Table 5.

Multi-valued mapping matches for the BP, HR and RWAT prediction tasks. For the BP task, there was a high match rate for both within the same patient records and across patients, suggesting an ill-conditioned problem. For the HR and RWAT tasks, the matches were much lower. Ground truth for RWAT is only available for the Aurora-BP dataset.

Task	MIMIC		Aurora-BP
	Intra-patient	Inter-patient	Intra-patient	Inter-patient
SBP	33.2%	15.0%	13.9%	16.2%
HR	0.75%	2.10%	0.02%	0.89%
RWAT	–	–	0.08%	4.78%