A machine learning model to predict the risk of 30-day readmissions in patients with heart failure: a retrospective analysis of electronic medical records data

Table 1 Target periods and value expressions of each data type

Data Type	Data Source	Target periods	Value expression
Demographics	EDW	Patient-level, no applicable target period	Binary (0 or 1); Continuous/discrete value
Admissions	EDW	Admission date to discharge date	Continuous/discrete value
Diagnoses	EDW	2 years pre-discharge date to discharge date	Binary of occurrence in the target period
Labs	EDW	Admission date to 1 week after admission date; 1 week before discharge date to discharge date	Number of occurrences in the target period ^a
Medications	EDW/RPDR	Admission date to discharge date	Number of occurrences in the target period
Procedures	EDW	Admission date to discharge date	Number of occurrences in the target period
Notes	RPDR	Admission date to discharge date ^b	Binary of occurrence in the target period ^c

^a abnormal occurrences, ^b except Social History = penultimate to admission date, ^c except Allergies = number of occurrences in the target period

ISSN: 1472-6947