Background: COPD exacerbations are correlated with higher hospitalization rate  and mortality. Our objective was to find risk factors for exacerbations in primary care cohort and to develop statistical model to predict exacerbation.

Methods: In questionnaire-based cohort, COPD patients enrolled in the cohort. Data was collected for 2 years. COPD  patients were seen by their GPs at least twice a year. Data was split into training (75%) and validation (25%) datasets. We developed negative binomial regression model using training dataset to predict exacerbation rate within 1 year. Based on Akaike?s information criterion, an exacerbation prediction model was developed, overall performance was externally validated in validation dataset. Then we created prediction nomogram.

Results: 229 COPD patients (65%male, 67yrs) were analyzed. While 77% of patients had no exacerbation during the follow-up there were 73 recurrent exacerbations in total. The average number of exacerbations per subject was 0.32 during a median of 1.5yrs. The best subset in training dataset found that lower forced expiratory volume, high scores on MRC dyspnoea scale, exacerbation history, and not being on combination therapy at baseline were associated with higher rate of exacerbation.  When validated, the area-under-curve (AUC) was 0.75 for one or more exacerbations as well as 0.75 for 2 or more exacerbations. Calibration was accurate. 

Conclusion: Previous exacerbations and severe symptoms appear to be predictors for exacerbation in  COPD patients. Nomograms built from these models can assist clinicians and patients in shared decision-making process of their care.