Cohort data above are depicted at the patient level. The original Palliative Connect training cohort (‘Training data 2016’, above) consisted of a random 85/15 training/test split stratified by patient (eg, if a patient is selected for the training set, all their encounters are included in the training set), of all admissions in the 2016 calendar year for patients ≥18 years, excluding observation, obstetric, rehabilitation and hospice admissions (n=55 500 encounters corresponding to 40 000 unique patients). The test cohort for this evaluation project included all admissions from 2017 who met the same aforementioned inclusion criteria (n=58 464 encounters corresponding to 41 327 unique patients). ‘Died’ indicates that the patient died within 6 months of their last hospital encounter during the study period. Household income is the median household income for the zip code in which the patient lives. Education variables (less than ninth grade, high school graduate, some college, bachelor’s degree, graduate degree) correspond to the proportion of all residents ≥25 years old in the patient’s zip code for which this is the highest level of educational attainment achieved (eg, a bachelor’s degree but no further).