Table 3.
Sensitivity and specificity of different cut-offs for the LR model in comparison to the baseline model for predicting retention in care as defined by 12-month gap and NHAS-outcome measures
| Outcome measure for retention in care | Model cutoff | Sensitivity (%, 95% CI) |
Specificity (%, 95% CI) |
|---|---|---|---|
| Twelve-month gap measure | LR model (predicted probability × 100) | ||
| ≥ 80 | 93 (93, 94) | 19 (16, 21) | |
| ≥ 86 | 84 (84, 84) | 36 (34, 38) | |
| ≥ 89 | 73 (73, 74) | 51 (48, 53) | |
| ≥ 91 | 63 (62, 63) | 64 (62, 67) | |
| ≥ 92 | 56 (55, 56) | 71 (69, 74) | |
| ≥ 94 | 38 (38, 39) | 85 (83, 87) | |
| ≥ 95 | 29 (28, 29) | 90 (89, 92) | |
| Baseline model (predicted probability × 100) | |||
| ≥ 87 | 90 (89, 90) | 21 (19, 23) | |
| ≥ 88.8 | 80 (80, 80) | 33 (31, 35) | |
| ≥ 90.3 | 73 (73, 73) | 42 (40, 45) | |
| ≥ 91.5 | 64 (64, 65) | 52 (50, 55) | |
| ≥ 92.2 | 51 (51, 51) | 60 (59, 63) | |
| ≥ 92.5 | 45 (45, 46) | 65 (63, 68) | |
| ≥ 93.1 | 33 (33, 34) | 77 (75, 79) | |
| NHAS-outcome measure | LR model (predicted probability × 100) | ||
| ≥ 82 | 80 (80, 80) | 32 (32, 33) | |
| ≥ 84 | 70 (69, 70) | 44 (43, 45) | |
| ≥ 86 | 60 (60, 60) | 53 (52, 54) | |
| ≥ 87 | 56 (56, 57) | 57 (56, 57) | |
| ≥ 89 | 41 (41, 42) | 70 (69, 71) | |
| ≥ 91 | 28 (28,29) | 80 (79, 81) | |
| Baseline model (predicted probability × 100) | |||
| ≥ 84.2 | 80 (80, 81) | 31 (30, 32) | |
| ≥ 86.3 | 69 (68, 69) | 44 (43,45) | |
| ≥ 87.5 | 60 (60, 60) | 53 (52, 54) | |
| ≥ 88.5 | 49 (49, 50) | 62 (61, 61) | |
| ≥ 89.4 | 40 (39, 40) | 70 (69, 71) | |
| ≥ 90 | 34 (34, 34) | 74 (73, 75) |
Model cut-offs are represented as predicted probabilities × 100
LR logistic regression, CI confidence interval, NHAS National HIV/AIDS Strategy