Comparison between machine learning methods for mortality prediction for sepsis patients with different social determinants

Table 1 Statistics of 5783 sepsis patients

Social Determinants	Category	n	% sepsis population	In-hospital mortality	% in-hospital mortality	Training	Testing
Race	Asian	179	3.10	26	14.53	129	50
	Black or African American	501	8.66	52	10.38	348	153
	Hispanic or Latino	188	3.25	18	9.57	132	56
	Other	714	12.35	165	23.11	527	187
	White	4201	72.64	575	13.69	2912	1289
Sex	Female	2562	44.30	384	14.99	1798	764
	Male	3221	55.70	452	14.03	2250	971
Marital status	Separated	398	6.88	52	13.07	287	111
	Significant other	2559	44.25	363	14.19	1788	771
	Single	1638	28.32	174	10.62	1157	481
	Unknown	332	5.74	102	30.72	248	84
	Widowed	856	14.80	145	16.94	568	288
Insurance type	Government	166	2.87	13	7.83	115	51
Insurance type	Medicaid	570	9.86	67	11.75	395	175
	Medicare	3358	58.07	560	16.68	2335	1023
	Private	1639	28.34	185	11.29	1168	471
	Self-pay	50	0.86	11	22.00	35	15
Language	English	5167	89.35	727	14.07	3631	1536
	Other	499	8.63	94	18.84	339	160
	Spanish	117	2.02	15	12.82	78	39

Categories of each social determinant are ranked alphabetically; n: number of sepsis patients in the category; % sepsis population: percentage of the number of sepsis patients in the category among the 5,783 sepsis patients; In-hospital mortality: number of patients in the category deceased in-hospital; % in-hospital mortality: percentage of patients in the category deceased in-hospital; Training: number of patients of the given category that were assigned to the training set during train-test split; Testing: number of patients of the given category that were assigned to the test set during train-test split

ISSN: 1472-6947