Early prediction of in-hospital mortality utilizing multivariate predictive modelling of electronic medical records and socio-determinants of health of the first day of hospitalization

Table 3 Most influential categorical variables for support vector machine (SVM), logistic regression (LR) and Xgboost from 250 rounds of permutation importance, sorted by odds ratio

Selected variables in each model are highlighted in bold green. All variables show a p-value of < 0.0001 (chi-square test). Alanine-aminotransferase (ALAT), aspartate-aminotransferase (ASAT), red cell distribution width (RCW). *Modes of admission: “After consultation with doctor from the establishment* is when a patient is seen in a scheduled consultation and kept in hospital immediately afterwards, without going through the emergency room, therefore without immediate vital risk; whereas “Confirmed prescheduled appointment” is hospitalization programmed in advance, e.g., for non-urgent surgery, again without immediate vital risk

ISSN: 1472-6947