Table 2.

Distributions of medical events across various demographic groups. The data in this table correspond to MIMIC-III data that are unaltered (i.e. no missingness has been simulated). There are 3950 patients in the training set and 1692 patients in the test set. The mean number of medical events for each patient is 315, with a minimum of 7 and maximum of 3463. Across all patients, there are a total of 994,320 medical concepts. Mean and median number of medical events per patient record are evenly distributed across age groups. There are more older patients than younger patients in the training data. We also see that mean and median number of medical events per patient record are evenly distributed across sex. Proportions of male-to-female patients are relatively even in the training data.

Age	Mean Events	Median Events	Standard Deviation	Proportion of Patients
17–54	336	252	310	0.254
55–70+	315	261	221	0.716
Sex
Male	323	261	236	0.541
Female	317	258	258	0.428
Insurance Type
Government	323	262	251	0.706
Private or Self	313	250	239	0.263
Race
Black	355	280	325	0.119
White	317	259	228	0.722