Birgitta Versluijs

73 HRCT after hematopoietic cell transplantation TABLE 6. Logistic regression. Diagnostic variable Odds ratio a Regression coefficient b P Weight factor c Ground-glass pattern Airtrapping 1.04 (1.01-1.07) 1.06 (1.02-1.10) 0.036 0.053 0.01 0.005 1 (0.036/0.036) 1.5 (0.053/0.036) a Numbers in parentheses indicate 95% CIs. b Natural logarithm of odds ratio. c The weight factor was calculated by dividing the applicable regression coefficient by the regression coefficient for ground-glass pattern. TABLE 7. Test characteristics of the “allo-score” (ground-glass pattern + airtrapping) depending on different cutoff values. Cutoff Sensitivity Specificity PPV NPV Correct (%) > 5 1 0.38 0.46 1 60 > 10 0.89 0.56 0.52 0.90 67 > 15 0.89 0.65 0.57 0.92 73 > 20 0.67 0.79 0.63 0.82 75 > 25 0.67 0.91 0.80 0.84 83 > 30 0.50 0.88 0.69 0.77 75 > 35 0.17 0.94 0.60 0.68 67 > 40 0.17 0.97 0.75 0.69 69 > 45 0.11 0.97 0.67 0.67 67 > 50 0.06 0.97 0.50 0.66 65 > 55 0 1 NA 0.65 65 Note—Sensitivity is the probability that the test result is positive in a patient with alloim- mune-mediated lung syndromes (allo-LS). Specificity is the probability that the test result is negative in a patient without allo-LS. Positive predictive value (PPV) is the probability that the patient has allo-LS given a positive test result. Negative predictive value (NPV) is the probability that the patient does not have allo-LS given a negative test result. Correct (%) indicates the percentage of patients with the cor- rect diagnosis. NA indicates not applicable. 4