Table 2.
Classifier performance (recall, precision and F-score). SVM RBF with binary count feature extraction was the most effective combination to identify incident type and severity level
| Benchmark | Original | Independent | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Recall | Precision | F-score | Recall | Precision | F-score | Recall | Precision | F-score | |
| Incident type a | 78.3 | 78.3 | 78.3 | 73.9 | 73.9 | 73.9 | 68.5 | 68.5 | 68.5 |
| Falls | 96.2 | 83.3 | 89.3 | 95.6 | 96.6 | 96.1 | 91.3 | 86.5 | 88.8 |
| Medications | 76.9 | 76.9 | 76.9 | 80.9 | 91.7 | 85.9 | 81.1 | 78.6 | 79.8 |
| Pressure injury | 88.5 | 100.0 | 93.9 | 89.2 | 86.8 | 88.0 | 96.8 | 76.0 | 85.2 |
| Aggression | 92.3 | 88.9 | 90.6 | 81.6 | 76.9 | 79.2 | 81.5 | 62.2 | 70.6 |
| Documentation | 46.2 | 63.2 | 53.3 | 46.2 | 31.6 | 37.5 | 47.6 | 16.0 | 24.0 |
| Blood products | 80.8 | 95.5 | 87.5 | 100.0 | 62.5 | 76.9 | 83.1 | 43.0 | 56.6 |
| Patient identification | 84.6 | 61.1 | 71.0 | 71.4 | 25.0 | 37.0 | 23.3 | 44.4 | 30.5 |
| Infection | 92.3 | 88.9 | 90.6 | 83.3 | 38.5 | 52.6 | 40.9 | 13.2 | 20.0 |
| Clinical handover | 80.8 | 65.6 | 72.4 | 71.4 | 18.5 | 29.4 | 37.9 | 14.3 | 20.8 |
| Deteriorating patient | 92.3 | 85.7 | 88.9 | 100.0 | 25.0 | 40.0 | 21.4 | 17.6 | 19.4 |
| Others | 30.8 | 50.0 | 38.1 | 54.7 | 85.3 | 66.7 | 57.1 | 87.0 | 69.0 |
| SAC level a | 62.9 | 62.9 | 62.9 | 50.1 | 50.1 | 50.1 | 52.7 | 52.7 | 52.7 |
| SAC1 | 82.8 | 92.3 | 87.3 | 84.0 | 11.2 | 19.8 | 82.6 | 6.8 | 12.5 |
| SAC2 | 41.4 | 60.0 | 49.0 | 43.2 | 7.2 | 12.3 | 16.2 | 9.6 | 12.0 |
| SAC3 | 44.8 | 54.2 | 49.1 | 35.9 | 52.3 | 42.6 | 46.9 | 49.8 | 48.3 |
| SAC4 | 82.8 | 52.2 | 64.0 | 62.4 | 61.2 | 61.8 | 58.3 | 61.8 | 60.0 |
aMicro-averaging measures