Table 1.
Summary of EHR datasets by phenotype. The filter positive set was defined as the set of patients passing the clinically established filter criterion as defined in the Method Section. Labeled set was defined as the small set of patients with manually curated gold-standard labels, which was randomly sampled from filter positive set. The prevalence is the proportion of subjects with positive phenotype status among those for whom labels are provided.
| EHR Platform | Phenotype | Sample Size of Filter Positive Patients | No. of Labeled Samples (%) | Prevalence (%) |
|---|---|---|---|---|
| MGB | Asthma | 7289 | 183 (2.5 %) | 47.5 (%) |
| Breast Cancer | 2002 | 94 (4.7 %) | 77.6 | |
| COPD | 3021 | 153 (5.1 %) | 43.1 | |
| Depression | 10189 | 252 (2.5 %) | 54.8 | |
| Epilepsy | 2225 | 117 (5.3 %) | 47.9 | |
| Hypertension | 19853 | 390 (2.0 %) | 79.0 | |
| SCZ | 456 | 108 (23.7 %) | 17.6 | |
| T1DM | 2111 | 128 (6.1 %) | 16.4 | |
| RA | 987 | 153 (15.5%) | 36.6 | |
| CAD | 3793 | 186 (4.9 %) | 37.1 | |
| CD | 519 | 136 (26.2 %) | 53.7 | |
| UC | 476 | 126 (26.5 %) | 49.2 | |
| T2DM | 3460 | 280 (8.1 %) | 35.7 | |
| MS | 136 | 101 (74.3 %) | 52.5 | |
| Stroke | 2052 | 128 ( 6.2 %) | 36.7 | |
|
| ||||
| BWH | Pseudogout | 12035 | 365 (3.0 %) | 21.9 |
|
| ||||
| BCH | ARDS | 2201 | 44 (1.9 %) | 40.9 |