Validation of a model to identify young patients for lymph node biopsy

A model to identify 9- to 25-year-old patients with malignant or granulomatous lymphadenopathy demonstrated a sensitivity/specificity of 0.95/0.96 in the original sample. This study tests the performance of the model in three new samples (n = 34, 87, 46) of young patients with peripheral lymphadenopathy. Patient charts were reviewed for the biopsy diagnosis and for the three variables in the model (lymph node size, recent ear-nose-throat symptoms, and chest roentgenogram). Prevalence rates of malignancy or granuloma in the three samples were 0.71, 0.37, and 0.37. The model's sensitivity/specificity in the three samples were 0.96/1.00, 0.88/0.80, and 0.82/0.83. Receiver operating characteristic curve areas for the model ranged from 0.87 to 0.98, indicating excellent discrimination. Likelihood ratios were calculated for all scores to allow adjustment of the model's predictions for population differences in the prevalence of malignancy or granuloma. We conclude that the model performs well and can help physicians to select those young patients with peripheral lymphadenopathy who should undergo biopsy.