Table 5 Performance of the algorithm in predicting readmission for Queensland compared to UK and US data
From: Using routine inpatient data to identify patients at risk of hospital readmission
| Â | Risk score threshold | ||
|---|---|---|---|
| Â | 50% | 70% | 80% |
Queensland data | Â | Â | Â |
Sensitivity (95% CI) | 44.7 (42.5–46.9) | 7.8 (6.7–9.1) | 1.1 (0.1–1.7) |
Specificity (95% CI) | 78.1 (76.4–79.6) | 97.5 (96.8–98.1) | 99.8 (99.6–99.9) |
False positive rate (95% CI) | 37.5 (35.0–40.0) | 28.2 (22.3–34.6) | 14.8 (4.2–33.7) |
LR+ (95% CI) | 2.04 (1.86–2.23) | 3.11 (2.33–4.15) | 7.02 (2.43–20.28) |
LR- (95% CI) | 0.71 (0.68–0.74) | 0.95 (0.93–0.96) | 0.99 (0.99–1.00) |
UK study[11] | Â | Â | Â |
Sensitivity | 54.3 | 17.8 | 8.1 |
Specificity | 72.2 | 95.0 | 98.6 |
False positive rate | 34.7 | 22.6 | 15.7 |
LR+ (95% CI) | 2.0 | 3.6 | 5.8 |
LR- (95% CI) | 0.6 | 0.9 | 0.9 |
US study[14] | Â | Â | Â |
Sensitivity | 57.9 | Not reported | Not reported |
Specificity | 70.7 | Â | Â |
False positive rate | 34.1 | Â | Â |
LR+ (95% CI) | 2.0 | Â | Â |
LR- (95% CI) | 0.6 | Â | Â |