Mortality prediction in ICU Using a Stacked Ensemble Model

Table 2

Examples of features in different types and the corresponding description of missing values.

Class	Name	Unit	Missing percentage (%)
Demographics	Age	Years	4.6
	BMI	kilograms/metres^2	3.7
	PILD	1	0
	⋯	⋯	⋯
Vital signs	d1_heartrate_max	Beats per minute	3
	d1_spo2_max	Percentage	0.4
	d1_temp_max	Degrees Celsius	23.7
	⋯	⋯	⋯
Laboratory features	h1_bun_min	g/L	11.5
	d1_albumin_min	g/L	91.4
	d1_glucose_max	mmol/L	57.4
	⋯	⋯	⋯
APACHE covariates	gcs_motor_apache	None	2.1
	heart_rate_apache	None	1
	⋯	⋯	⋯
APACHE comorbidity	Leukemia	None	0.8
	sodium_apache	mmol/L	0.8
	Aid	None	0.8
	diabetes_mellitus	None	0.8
	Immunosuppression	None	0.8
	⋯	⋯
APACHE prediction	AHDP	None	8.7
	AIDP	None	8.7

Abbreviation: PILD: pre_icu_los_days; AHDP: apache_4a_hospital_death_prob; AIDP: apache_4a_icu_death_prob.