Background. The wide use of antibiotics for treatment of exacerbations of chronic obstructive pulmonary disease (COPD) lacks evidence. The efficacy is debatable, and bacterial involvement in exacerbation is difficult to verify. The aim of this prospective study was to identify factors that can help to estimate the probability that a microorganism is involved in exacerbation of COPD and, therefore, predict the success of antibiotic treatment. Methods. Clinical data and sputum samples were obtained from 116 patients during exacerbation of COPD. Bacterial infection was defined by the abundant presence of ≥ 1 potential pathological microorganism in relation to the normal flora in sputum. Results. Of 116 exacerbations, 22 (19%) had bacterial involvement. The combination of a negative result of a sputum Gram stain, a relevant nonclinical decrease in lung function (compared with baseline measurements), and occurrence of <2 exacerbations in the previous year were 100% predictive of a nonbacterial origin of the exacerbation. The presence of all 3 of these clinical characteristics yielded a positive predictive value of 67% for a bacterial exacerbation. Conclusions. Patients presenting with an exacerbation who have a negative result of a sputum Gram stain, do not have a clinically relevant decrease in lung function, and who have not experienced <2 exacerbations of COPD in the previous year do not require antibiotic treatment. A treatment protocol taking into account these variables might lead to a 5%-24% reduction in unnecessary treatment with antibiotics, depending on actual prescription rates.