Table 3.
Triage efficiency for the external validation set
| External validation set | Tianyou Hospital dataset | Xianning Central Hospital dataset | The Second Xiangya Hospital dataset | ||
|---|---|---|---|---|---|
| True positive scans, n | 698 | 511 | 129 | 58 | |
| Median draft report time (IQR), min* | 16·21 (11·67–25·71) | 14·50 (10·75–21·11) | 24·75 (15·67–43·72) | 25·73 (20·10–38·84) | |
| Median report approval time (IQR), min* | 23·06 (15·67–39·20) | 19·23 (14·33–27·33) | 47·37 (27·12–96·35) | 198·77 (45·85–675·83) | |
| Median AI triage time (IQR), min* | 0·55 (0·43–0·63) | 0·58 (0·45–0·64) | 0·50 (0·42–0·58) | 0·48 (0·44–0·53) | |
| Median reduction in triage time under scan-to-second-reader triage workflow (IQR), min | 15·73 (11·05–25·25) | 14·03 (10·13–20·55) | 24·31 (15·13–43·40) | 25·24 (19·65–38·48) | |
| p value† | <0·0001 | <0·0001 | <0·0001 | <0·0001 | |
| t | 257·42 | 243·09 | 114·59 | 69·32 | |
| Median reduction in triage time under scan-to-fever-clinician triage workflow (IQR), min | 22·62 (15·12–38·63) | 18·77 (13·88–26·73) | 47·03 (26·53–95·83) | 198·28 (45·31–675·26) | |
| p value† | <0·0001 | <0·0001 | <0·0001 | <0·0001 | |
| t | 188·08 | 253·61 | 87·37 | 50·78 | |
*
Raw data did not follow a normal distribution.
†
Calculated by comparing the results with zero.