Machine learning approaches to the human metabolome in sepsis identify metabolic links with survival

Intensive Care Medicine Experimental

Table 1 Baseline patient demographics

	Overall (n = 60)	Survived (n = 35)	Died (n = 25)	p-value*
Age, years	58 (47,62)	53 (46, 63)	62 (48, 67)	0.33
Women	27 (45%)	17 (49%)	10 (40%)	0.51
Race/ethnicity				0.41
White	53 (88%)	29 (83%)	22 (88%)
Black	3 (5%)	4 (11%)	1 (4%)
Hispanic	3 (5%)	2 (6%)	1 (4%)
Asian	1 (2%)	0 (0%)	1 (4%)
Comorbidities
Diabetes	12 (20%)	8 (23%)	4 (16%)	0.44
Malignancy	23 (48%)	10 (29%)	19 (76%)	< 0.001
CKD**	16 (27%)	6 (18%)	10 (40%)	0.06
Liver disease	5 (8%)	3 (9%)	2 (8%)	0.91
COPD	7 (12%)	4 (11%)	3 (12%)	0.95
Illness severity
ARDS	30 (50%)	12 (34%)	18 (72%)	< 0.001
Respiratory failure	51 (85%)	27 (77%)	24 (96%)	0.03
SOFA score	9 (6, 12)	8 (5, 10)	11 (7, 15)	0.004
APACHE II score	30 (23,37)	29 (22, 33)	35 (26, 39)	0.02

Data are presented as number (proportion) or median (interquartile range). Percentages may not sum to 100 due to rounding. APACHE Acute Physiology and Chronic Health Evaluation, ARDS acute respiratory distress syndrome, CKD chronic kidney disease, COPD chronic obstructive pulmonary disease. *Comparing survivors and non-survivors (Chi-squared likelihood test for categorical variables and Wilcoxon rank sum test for continuous variables). **Median (IQR) estimated glomerular filtration rate at admission = 47.6 (25.1, 88.3) mL/min/1.73 m² in n = 60 patients in the cohort