Table 2.
Comparison of the scores achieved by GPT-4 and Japanese medical residents across various clinical fields (N=137).
| Fields | Questions, n (%) | Examinees, % (95% CI) | GPT-4, % (95% CI) | Differences | P value |
| General practice | 19 (13.9) | 71.8 (61.0-82.6) | 63.2 (41.5-84.8) | –8.6 | .40 |
| Internal medicine | 48 (35.0) | 55.2 (49.4-60.9) | 81.3 (70.2-92.3) | 26.1 | <.001a |
| Surgery | 9 (6.6) | 57.6 (41.3-74.0) | 77.8 (50.6-105) | 20.2 | .22 |
| Pediatrics | 12 (8.8) | 55.1 (39.6-70.5) | 66.7 (40.0-93.3) | 11.6 | .42 |
| Obstetrics and gynecology | 15 (10.9) | 49.1 (38.8-59.4) | 80.0 (59.6-100) | 30.9 | .02a |
| Emergency | 19 (13.8) | 48.1 (37.7-58.5) | 57.9 (35.7-80.1) | 9.8 | .39 |
| Psychiatry | 15 (10.9) | 53.8 (40.4-67.2) | 46.7 (21.4-71.9) | –7.1 | .58 |
aStatistically significant.