Figure 1.

Flowchart of data. We used a cohort study of 51,007 acute patient admissions at the emergency department with laboratory and demographical data. At data pre-processing we excluded 2166 records with more the 50% missing in data (4.3% of the total), removed outliers, imputed and scaled the data. In total 48,841 records were structured data, where 34,190 (70%) patient records were allocated to training data and 14,651 (30%) patient records were allocated to validation and test data.