Learning to predict in-hospital mortality risk in the intensive care unit with attention-based temporal convolution network

Table 3 The baseline of patients in training and testing dataset

Variables	Training (n = 15,331)	Testing (n = 2763)	P
Age	67.3(54.0–78.8)	67.7(53.9–79.2)	0.527
Sex (F/M)	6861/8470	1229/1534	0.791
ICU admission			0.014
CCU	2071	380
CSRU	2768	572*
MICU	5919	1037
SICU	2654	455
TSICU	1919	319
survival/Death	12,910/2421	2389/374*	0.003
ICU length of stay (hours)	88.8 (63.7–149.9)	86.9 (62.5–147.0)	0.180

Mean (SD) presented for normally distributed continuous variables, while median (IQR) was given to those with non-normally distributed continuous variable. Unless otherwise state n is as indicated in the column headings. The portion of admission in different ICU was statistically compared with the training dataset (*P < 0.05). F female, M male, CCU Coronary Care Unit, CSRU Cardiac Surgery Recovery Unit, MICU Medical ICU, SICU Surgical ICU, TSICU Trauma Surgical intense care unit

ISSN: 1471-2253