TABLE 5

Diagnostic accuracy of the model in the external validation sample#

Cut-pointPatients at/above cut-pointSensitivitySpecificityCorrectly classifiedPositive likelihood ratioNegative likelihood ratioPositive predictive valueNegative predictive valueNND
≥2.588.297.4 (90.9–99.7)12.5 (10.5–14.7)18.61.11 (1.07–1.16)0.21 (0.05–0.82)7.9 (6.2–9.8)98.4 (94.5–99.8)13 (11–17)
≥5.053.780.5 (69.9–88.7)48.4 (45.3–51.5))50.71.56 (1.38–1.77)0.40 (0.25–0.64)10.7 (8.3–13.5)97.0 (95.1–98.3)10 (8–13)
≥7.533.968.8 (57.3–78.9)68.8 (65.8–71.6)68.82.21 (1.85–2.63)0.45 (0.32–0.63)14.4 (11.0–18.5)96.6 (95.1–97.8)7 (6–10)
≥10.024.854.5 (42.8–65.9)77.5 (74.8–80.1)75.92.43 (1.92–3.07)0.59 (0.46–0.75)15.7 (11.5–20.6)95.7 (94.1–97.0)7 (5–9)
≥12.519.846.8 (35.3–58.5)82.3 (79.8–84.6)79.82.64 (2.01–3.47)0.65 (0.52–0.80)16.8 (12.1–22.5)95.3 (93.7–96.6)6 (5–9)
≥15.015.041.6 (30.4–53.4)87.0 (84.7–89.0)83.83.19 (2.34–4.35)0.67 (0.56–0.81)19.6 (13.8–26.6)95.1 (93.5–96.4)6 (4–8)
≥17.59.933.8 (23.4–45.4)91.9 (90.0–93.5)87.74.14 (2.85–6.03)0.72 (0.61–0.85)24.1 (16.4–33.3)94.8 (93.2–96.1)5 (3–7)
≥20.08.328.6 (18.8–40.0)93.2 (91.5–94.7)88.64.23 (2.77–6.44)0.77 (0.66–0.88)24.4 (16.0–34.6)94.5 (92.9–95.8)5 (3–6)
≥22.55.920.8 (12.4–31.5)95.2 (93.7–96.5)89.94.36 (2.60–7.30)0.83 (0.74–0.93)25.0 (15.0–37.4)94.0 (92.4–95.4)4 (3–7)
≥25.03.714.3 (7.4–24.1)97.1 (95.9–98.1)91.24.96 (2.58–9.53)0.88 (0.81–0.97)27.5 (14.6–43.9)93.7 (92.0–95.1)4 (3–7)
≥30.02.614.3 (7.4–24.1)98.3 (97.3–99.0)92.38.45 (4.11–17.4)0.87 (0.80–0.96)39.3 (21.5–59.4)93.7 (92.1–95.1)3 (2–5)
≥35.01.910.4 (4.6–19.4)98.8 (97.9–99.4)92.58.71 (3.67–20.7)0.91 (0.84–0.98)40.0 (19.1–63.9)93.5 (91.9–94.9)3 (2–6)
≥40.01.26.5 (2.1–14.5)99.2 (98.4–99.7)92.68.17 (2.74–24.4)0.94 (0.89–1.00)38.5 (13.9–68.4)93.3 (91.6–94.7)3 (2–8)
≥45.00.82.6 (0.3–9.1)99.3 (98.6–99.7)92.43.73 (0.79–17.7)0.98 (0.95–1.02)22.2 (2.8–60.0)93.0 (91.3–94.5)5 (2–36)
≥50.00.62.6 (0.3–9.1)99.6 (99.0–99.9)92.76.53 (1.22–35.1)0.98 (0.94–1.01)33.3 (4.3–77.7)93.0 (91.3–94.5)3 (2–23)
  • Data are presented as % or n (95% CI). NND: number of diagnostic assessments needed per case detected. #: n=1083; : percentage of subjects with a predicted risk score at or above the cut-point.