Table 6

Test accuracy of the final risk score in the external validation sample

Discrimination characteristicsApplication of the score assuming 5.5% prevalence of undiagnosed COPD
Score thresholdSensitivity (%)Specificity (%)Correctly classified (%)LR+LR−PPV (%)NPV (%)Screening yield (%)NNSClinical assessments per case detected
0100035.015.55.501919
0.596.134.756.21.470.117.999.35.291913
190.651.965.51.880.189.999.04.982111
1.589.455.167.11.990.1910.498.94.922110
281.271.975.22.890.2614.498.54.47237
2.563.287.479.05.020.4222.697.63.48295
355.891.979.36.890.4828.697.33.07334
3.547.694.678.18.800.5533.996.92.62393
440.495.876.49.570.6235.896.52.22453
4.523.298.372.013.50.7844.095.61.28793
520.698.471.213.10.8144.395.51.13893
5.511.399.568.621.50.8955.895.10.621612
65.5399.766.718.90.9550.994.80.303292
6.53.2399.865.019.40.9752.594.70.185632
  • Correctly classified= proportion of participants with disease status correctly classified.

  • LR, likelihood ratio (ie, the ratio by which the pretest probability is altered by a positive or negative test result);

  • NPV, negative predictive value (proportion of all participants with a negative test who are disease free);

  • NNS, number-needed-to-screen (the number of patients or patient records the risk score would need to be applied to) to identify one patient with COPD; NPV, negative predictive value (proportion of all subjects with a negative test who are disease free); PPV, positive predictive value (proportion of all participants with a positive test who have disease); Screening yield, proportion of all patients subjected to the risk score who would be correctly identified as having COPD.

  • COPD, chronic obstructive pulmonary disease.