Variable | Prevalence % | Positive serology | S_{e} | S_{p} | PPV | NPV |

Discrete | 20 | >-0.35 | 49 | 95 | 71 | 88 |

25 | >-0.44 | 55 | 93 | 72 | 86 | |

35^{¶} | >-0.53 | 61 | 89 | 75 | 81 | |

Dichotomous | 20 | ND | ||||

25 | >-0.76 | 60 | 85 | 57 | 86 | |

35^{¶} | >-0.78 | 66 | 80 | 64 | 81 |

S

_{e}: sensitivity (probability of a positive test outcome in a hypersensitivity pneumonitis (HP) individual); S_{p}: specificity (probability of a negative test outcome in a non-HP individual); PPV: positive predictive value (probability of having HP when serological score is above the cut-off value); NPV: negative predictive value (probability of having HP when serological score is equal to or below the cut-off value); ND: nondiscriminative.^{#}: S = ((1–P)/P)((C_{FP}–C_{TN})/(C_{FN}–C_{TP})), where S is the slope of the receiver-operating characteristic curve at the optimal operating point, in which “optimal” is in terms of minimising costs (C). “Costs” can be identified here as patient morbidity, where C_{FP}, C_{FN}, C_{TN}and C_{TP}represent the costs of false-positive, false-negative, true-negative and true-positive results, respectively. Here, it was assumed that C_{FP}= C_{FN}and C_{TP}= C_{TN}. Therefore, here S = ((1–P)/P), where P denotes the prevalence in the target population.^{¶}: prevalence in study population. To use this table for a given patient from, for example, a department of respiratory disease where the prevalence of HP is nearly 35%, with a serological score of -0.43 (>-0.53) with discrete variables and 0.18 (>-0.78) with dichotomous variables, both serological scores indicate a diagnosis in favour of HP. To calculate scores, see table 3⇑.