Abstract

The selection of initial antimicrobial treatment in a patient with community acquired pneumonia is an important clinical decision. Because this decision is usually made before the results of specific microbiological tests are available, we attempted to determine how well the presenting clinical features would allow prediction of microbial aetiology in 441 adults admitted to hospital with pneumonia. Five of 90 variable available on admission were selected for inclusion in a multivariate discriminant function analysis because of their strong association with one or more of the major aetiological subsets (Mycoplasma pneumoniae, Streptococcus pneumoniae, "other," and undetermined). These variables were age, number of days ill before admission, presence or absence of bloody sputum and of lobar infiltration on chest radiograph, and white blood cell count. The microbial aetiology was correctly predicted by this discriminant function analysis in only 42% of cases, which gives a quantitative estimate of the degree of difficulty encountered in determining the microbial aetiology at the time of admission for pneumonia. When a similar discriminant function analysis was applied to the third of patients in whom the microbial aetiology was never determined, most of these cases were predicted to be due to Streptococcus pneumoniae.