PREDICTION OF EXISTENCE OF PNEUMONIA IN A DULTS WITH FEVER

BACKGROUND: This study was performed to define what clinical and analy tical variables were significantly associated with the presence of pne umonia in febrile patients. A predictive model which may rationalize r adiologic explorations is presented. METHODS: A prospective study was carried out in two Emergency Departments in Madrid, Spain. One hundred forty-one patients presenting fever of more than 48 hours in evolutio n with symptomatology of the lower airway or without focal data were i ncluded. Following anamnesis, physical exploration and elemental hemog ram the physician established a tentative diagnosis. Subsequently, on observation of thorax X-ray film definitive diagnosis was made. RESULT S: Thirty-eight percent of the patients had pneumonia. The variables w hich were significantly associated with the presence of pneumonia were : history of smoking (odds ratio [OR] 2.06), alcohol abuse (OR 3.18), chronic obstructive bronchopneumopathy (OR 4.35), respiratory symptoms (OR 5.73), cough (OR 5.37), expectoration (OR 2.08), dyspnea (OR 7.03 ), pleural pain (OR 13.38), pathologic auscultation (OR 7.46), leukocy tosis (OR 6.34) and neutrophilia (OR 8.10). Clinical prior to radiolog ic diagnosis showed a sensitivity of 45.3% and specificity of 93.2%. P neumonia was demonstrated in 10% of the cases with respiratory symptom atology with normal auscultation and without neutrophilia. A logistic regression model is proposed for the diagnosis of pneumonia. CONCLUSIO NS: Clinical judgement prior to observation of radiography has a low s ensitivity in the diagnosis of pneumonia. The use of logistic regressi on techniques allows considerable improvement in clinical judgement.