Computer program supporting the diagnostic accuracy of cellular BALF analysis: a new release

Recently, we developed a validated computer program based on polychotomous logistic regression analysis using bronchoalveolar lavage fluid (BALF) resu lts to distinguish between the three most common interstitial lung diseases (ILD): sarcoidosis, idiopathic pulmonary fibrosis (IPF) and extrinsic alle rgic alveolitis (EAA) or drug-induced pneumonitis. One of the limitations o f this program was that it was not useful in discriminating between infecti ous disorders and noninfectious disorders. Therefore, we added BALF samples obtained from patients with a confirmed bacterial pulmonary infection base d on culture results greater than or equal to 10(4) cfuml(-1) (group I: n=3 1) to the study population mentioned above (group II: n=272). Notably, just one variable, i.e. the percentage of polymorpho nuclear neutrophils, allow ed us to distinguish between infectious and non-infectious disorders. The a greement of predicted with the actual diagnostic group membership was 99-67 % (groups I and II). Additionally, 91.2% of the cases with ILD were correct ly classified. In conclusion, this updated Windows version 2000 of the vali dated computer program provides a very reliable prediction of the correct d iagnosis for an arbitrary patient with suspected pneumonia or with ILD give n information obtained from BALF analysis results, and is thought to improv e the diagnostic power of BALF analysis. (C) 2001 Harcourt Publishers Ltd.