Diagnostic Test Characteristics of Coefficient-Based Models for Sensitivity Cut Points, Using Symptom and Sign Model for Urine Sampling and Antibiotic Treatment and the Symptom, Sign, and Dipstick Model for Antibiotic Treatment
| Linear Predictor Cut Point, Shrunken Coefficients (≥)a | Linear Predictor Cut Point, Unshrunken Coefficients (≥)a | Sensitivity (95% CI) | Specificity (95% CI) | Positive Predictive Value (95% CI) | Negative Predictive Value (95% CI) | Children Clinical Rule Positive % (95% CI) |
|---|---|---|---|---|---|---|
| Symptom and sign model | Urine Sampled, Antibiotic Treatedb | |||||
| −0.504 | −0.195 | 20.0 (11.7–32.0) | 99.8 (99.5–99.9) | 66.7 (42.9–84.2) | 98.2 (97.7–98.7) | 0.7 (0.4–1.0) |
| −1.092 | −0.87 | 30.0 (19.8–2.7) | 99.5 (99.1–99.7) | 56.3 (39.0–72.1) | 98.4 (97.9–98.9) | 1.2 (0.8–1.6) |
| −1.813 | −1.698 | 40.0 (28.5–52.8) | 98.2 (97.6–98.6) | 32.9 (23.1–44.4) | 98.7 (98.1–99.0) | 2.7 (2.1–3.3) |
| −2.059 | −1.98 | 50.0 (37.6–62.4) | 97.5 (96.9–98.1) | 31.3 (22.8–41.2) | 98.9 (98.4–99.2) | 3.5 (2.9–4.3) |
| −2.372 | −2.34 | 60.0 (47.2–71.5) | 96.3 (95.5–97.0) | 26.7 (19.9–34.7) | 99.1 (98.6–99.4) | 4.9 (4.2–5.8) |
| −2.729 | −2.75 | 70.0 (57.3–80.2) | 94.6 (93.7–95.4) | 22.6 (17.1–29.1) | 99.3 (98.9–99.6) | 6.8 (5.9–7.8) |
| −3.396 | −3.515 | 80.0 (68.0–88.3) | 88.3 (87.0–89.4) | 13.3 (10.1–17.2) | 99.5 (99.1–99.7) | 13.2 (12.0–14.5) |
| −3.717 | −3.884 | 85.0 (73.6–92.0) | 83.9 (82.4–85.2) | 10.6 (8.1–13.6) | 99.6 (99.2–99.8) | 17.6 (16.2–19.1) |
| −4.567 | −4.86 | 93.3 (83.5–97.5) | 61.0 (59.1–62.8) | 5.1 (3.9–6.6) | 99.8 (99.4–99.9) | 40.2 (38.4–42.0) |
| −5.299 | −5.7 | 96.7 (87.6–99.2) | 37.8 (35.9–39.6) | 3.4 (2.6–4.3) | 99.8 (99.2–100.0) | 63.0 (61.2–64.8) |
| −6.138 | −6.664 | 100 | 15.7 (14.4–17.1) | 2.6 (2.0–3.3) | 100 | 84.6 (83.2–85.9) |
| Symptom, sign, and dipstick model | Antibiotic Treatedc | |||||
| 0.801 | 1.43 | 20.0 (11.7–32.0) | 99.9 (99.7–100.0) | 85.7 (57.3–96.4) | 98.2 (97.7–98.7) | 0.5 (0.3–0.9) |
| −0.122 | 0.321 | 40.0 (28.5–52.8) | 99.9 (99.7–100.0) | 88.9 (70.7–96.4) | 98.7 (98.2–99.0) | 1.0 (0.7–1.4) |
| −1.346 | −1.15 | 60.0 (47.2–71.5) | 99.3 (98.8–99.5) | 64.3 (51.0–75.7) | 99.1 (98.7–99.4) | 2.0 (1.6–2.6) |
| −3.114 | −3.275 | 80.0 (68.0–88.3) | 93.8 (92.9–94.7) | 22.5 (17.4–28.6) | 99.5 (99.2–99.7) | 7.8 (6.8–8.8) |
| −3.700 | −3.98 | 83.3 (71.7–90.8) | 88.3 (87.0–89.5) | 13.8 (10.6–17.7) | 99.6 (99.2–99.8) | 13.2 (12.0–14.6) |
| −4.746 | −5.237 | 96.7 (87.6–99.2) | 66.3 (64.5–68.1) | 6.0 (4.7–7.7) | 99.9 (99.6–100.0) | 35.0 (33.3–36.8) |
| −5.235 | −5.825 | 98.3 (89.1–99.8) | 53.1 (51.2–54.9) | 4.5 (3.5–5.7) | 99.9 (99.5–100.0) | 48.1 (46.2–49.9) |
| −5.955 | −6.69 | 100 | 29.5 (27.8–31.2) | 3.1 (2.4–3.9) | 100 | 71.2 (69.4–72.8) |
Note: Results based on models using multiple imputation to deal with missing values. For comparison, clinician diagnosis sensitivity was 46.6% and specificity was 94.7%.
↵a Derived from the coefficient-based models using multiple imputation where the coefficients are listed within the Supplemental Appendix, Table 5 (http://www.annfammed.org/content/14/4/325/suppl/DC1).
↵b Percentage of children who would be at or above this threshold assuming all children had a urine sample.
↵c Percentage of children who would be at or above this threshold assuming all children had a urine sample and dipstick test.