OBJECTIVE: The ability to predict exacerbations in patients with COPD might permit more rational use of preventive interventions. Our objective was to develop risk indexes for exacerbations and hospitalizations due to exacerbations that might be applied to the individual patient. METHODS: Spirometry, demographics, and medical history were obtained at baseline in 1,829 patients with moderate-to-very severe COPD who entered a trial of inhaled tiotropium. Information about exacerbations and hospitalizations due to exacerbation was collected during the 6-month follow-up period. Analyses of first outcomes were modeled using univariable and multivariable Cox proportional hazards regressions. RESULTS: During follow-up, 551 patients had at least one exacerbation and 151 patients had at least one hospitalization due to exacerbation. In the multivariable model for exacerbation, older age, percentage of predicted FEV(1), duration of COPD, a productive cough, antibiotic or systemic corticosteroid use for COPD in the prior year, hospitalization for COPD in the prior year, and theophylline use at baseline predicted a higher risk. In the multivariable model for hospitalization, older age, percentage of predicted FEV(1), unscheduled clinic/emergency department visits for COPD in the prior year, any cardiovascular comorbidity, and prednisone use at baseline were associated with greater risk. Both the exacerbation and the hospitalization models provided moderately good discrimination, the validated concordance indexes being 0.66 and 0.73, respectively. Methods for calculating risk in individual patients are provided. CONCLUSIONS: Spirometry along with a few questions directed to the patient are strongly predictive of exacerbations and related hospitalizations over the ensuing 6 months.