Objective To identify the level and patterns of self-management in elderly patients with COPD, to analyze the characteristics of each distinct pattern, and to provide reference for targeted intervention. Methods A total of 627 elderly patients with COPD were investigated by using a general information questionnaire, the COPD Self-management Scale, the COPD Assessment Test, the FRAIL Scale, the Chinese Version of Housebound Scale, and the Social Support Rating Scale. Five dimensions of the COPD Self-management Scale were treated as clustering variables to yield the dendrogram, then the optimal k value was determined with the "Elbow method", at last K-means algorithm was performed to determine the final clustering. Results The total self-management score was 156.99 ± 25.13.Three different self-management groups were identified and designated as clusterⅠ: moderate symptom, dependent, negative and non-compliant profile; cluster Ⅱ: severe-symptom, restricted and blindly coping profile; and cluster Ⅲ: mild symptom, superior and self-confident profile. The three patterns showed significant differences in gender, body mass index, current residence, main caregivers, education level, per capita monthly income, medical payment type, the number of acute exacerbations in the past year, mMRC, smoking status, oxygen therapy, use of inhaled drugs, respiratory muscle training, comorbid disease, frailty, CAT result, housebound status and social support (P＜0.05,P＜0.01). Conclusion Self-management of elderly COPD patients is at a low to moderate level. The three patterns of self-management identified through cluster analysis show high heterogeneity.Interventions based on these specific patterns of self-management may improve self-management of patients.