Table 5 Sensitivity, specificity, and number needed to screen for various cut-points and screening strategies

From: Development and validation of prediction algorithm to identify tuberculosis in two large California health systems

 

Sensitivity

Specificity

NNS

Internal validation, KPSC

Actual TB disease

 Actually screened in year before index date (5.2% of the population)

0.05

0.94

2745

 Model based screening of 5.2% of the population

0.41

0.94

320

 Actually screened during full study period (17.7% of the population)

0.15

0.76

3632

 Model based screening of 17.7% of the population

0.71

0.81

611

 Ideal EHR-based CDPH screening (29.7% of the population)

0.76

0.72

852

 Model based screening of 29.7% of the population

0.82

0.68

917

Actual and hypothetical TB disease with 95% SI

 Actually screened in year before index date (5.2% of the population)

0.06 (0.04, 0.08)

0.94

1806 (1287, 2577)

 Model based screening of 5.2% of the population

0.35 (0.32, 0.37)

0.94

300 (282, 315)

 Actually screened during full study period (17.7% of the population)

0.33 (0.27, 0.36)

0.76

1369 (1165, 1752)

 Model based screening of 17.7% of the population

0.64 (0.61–0.67)

0.81

548 (515, 576)

 Ideal EHR-based CDPH screening (29.7% of the population)

0.73 (0.69–0.75)

0.72

709 (662, 762)

 Model based screening of 29.7% of the population

0.76 (0.72–0.79)

0.68

788 (728, 839)

External validation, KPNC

Actual TB disease

 Actually screened in year before index date (5.2% of the population)

0.03

0.94

5580

 Model based screening of 5.2% of the population

0.47

0.94

393

 Actually screened during full study period (17.7% of the population)

0.14

0.76

5362

 Model based screening of 17.7% of the population

0.78

0.81

790

 Ideal EHR-based CDPH screening (29.7% of the population)

0.88

0.70

1079

 Model based screening of 29.7% of the population

0.82

0.73

1059

Actual and hypothetical TB disease with 95% SI

 Actually screened in year before index date (5.2% of the population)

0.04 (0.03, 0.05)

0.94

3222 (2727, 4026)

 Model based screening of 5.2% of the population

0.38 (0.37, 0.39)

0.94

361 (351, 369)

 Actually screened during full study period (17.7% of the population)

0.36 (0.34, 0.38)

0.76

1632 (1485, 1774)

 Model based screening of 17.7% of the population

0.70 (0.68–0.71)

0.81

662 (646, 679)

 Ideal EHR-based CDPH screening (29.7% of the population)

0.79 (0.77, 0.80)

0.73

832 (805, 855)

 Model based screening of 29.7% of the population

0.81 (0.80, 0.83)

0.70

877 (850, 901)

  1. Bold fonts indicate change in dataset and/or outcome definition for model evaluations.
  2. KPSC Kaiser Permanente Southern California, KPNC Kaiser Permanente Northern California, 95% SI 95 percent simulation interval, CDPH California Department of Public Health, TB tuberculosis, EHR electronic health record.
Back to article page