Table 4.
Operating Characteristics of Alternate Tuberculosis Medication Adherence Measuresa as Compared With Nonadherence by Urine Testing
| Sample | No. of Participants in Sample | Sensitivity (95% CI), % | Specificity (95% CI), % | Positive Predictive Value (95% CI), % | Negative Predictive Value (95% CI), % | Area Under the Receiver Operating Characteristic Curve |
|---|---|---|---|---|---|---|
| 99DOTS patient-reported doses alone | 608 | 61 (57–65) | 70 (58–81) | 94 (91–96) | 18 (14–23) | 0.65 |
| 99DOTS patient- and provider-reported doses | 608 | 89 (86–91) | 33 (22–45) | 91 (89–94) | 27 (18–38) | 0.61 |
| Pill estimate | 650 | 89 (86–91) | 21 (12–32) | 89 (86–92) | 20 (12–30) | 0.55 |
| 4-day recall | 650 | 97 (95–98) | 23 (14–34) | 90 (88–93) | 50 (33–67) | 0.60 |
| Last missed dose question | 650 | 78 (74–81) | 52 (40–63) | 92 (90–95) | 24 (18–31) | 0.65 |
| Last missed dose question and 4-day recall | 650 | 77 (73–81) | 55 (43–66) | 93 (90–95) | 24 (18–31) | 0.65 |
| Last missed dose question and pill estimate | 650 | 73 (69–76) | 56 (44–67) | 92 (90–95) | 22 (16–28) | 0.64 |
| Last missed dose question and 4-day recall and pill estimate | 650 | 72 (68–76) | 58 (47–70) | 93 (90–95) | 22 (16–28) | 0.65 |
For this analysis, alternate adherence measures were classified as binary variables, as described in Table 1.