Applying artificial neural network for early detection of sepsis with intentionally preserved highly missing real-world data for simulating clinical situation

Table 1 Clinical time series data: vital signs (rows 1–7), demographics (rows 8–13), and outcome (row 14)

	Measurement	Description
1	HR	Heart rate (beats per minute)
2	O₂Sat	Pulse oximetry (%)
3	Temp	Temperature (°C)
4	SBP	Systolic BP (mm Hg)
5	MAP	Mean arterial pressure (mm Hg)
6	DBP	Diastolic BP (mm Hg)
7	Resp	Respiration rate (breaths per minute)
8	Age	Age (years)
9	Gender	Female (0) or male (1)
10	Unit1	Administrative identifier for ICU unit (MICU); false (0) or true (1)
11	Unit2	Administrative identifier for ICU unit (SICU); false (0) or true (1)
12	HospAdmTime	Time between hospital and ICU admission (hours since ICU admission)
13	ICULOS	ICU length of stay (hours since ICU admission)
14	SepsisLabel^*	For septic patients, SepsisLabel is 1 if t ≥ t_sepsis and 0 if t < t_sepsis
14	SepsisLabel^*	For non-septic patients, SepsisLabel is 0

*We changed the definition of SepsisLabel to correspond to our experiment. The original definition is as follows: For septic patients, SepsisLabel is 1 if t ≥ t_sepsis − 6 and 0 if t < t_sepsis − 6. For nonsepsis patients, SepsisLabel is 0

ISSN: 1472-6947