To evaluate the effect of the 21-gene recurrence score (RS) assay in breast cancer-specific mortality (BCSM) and decision-making for chemotherapy in older (aged 65 years) breast cancer. Methods: We retrospectively included older patients with T1-2N0 and estrogen receptor-positive breast cancer in the Surveillance, Epidemiology, and End Results database. Cox regression model and competing-risks model were used for data analysis. Results: This study included 8524 patients, 1987 (23.3%) had low RS, 5059 (59.4%) had intermediate RS, and 1478 (17.3%) had high RS. Chemotherapy was administrated in 2.0%, 8.6%, and 51.2% for low, intermediate, and high RS cohorts, respectively (P < 0.001). A total of 597 deaths were recorded, including one-quarter of breast cancer-related deaths and three-quarters as competing causes of death. The 5-year BCSM was 5.4%, 4.7%, and 9.1% for low, intermediate, and high RS cohorts, respectively (P < 0.001), using the Cox regression model, and was 0.8%, 0.9%, and 5.2% for low, intermediate, and high RS cohorts using the competing-risks regression, respectively (P < 0.001). RS was independently correlated with BCSM in both prognostic models. The stratified analysis demonstrated that chemotherapy was not correlated with a lower risk of BCSM in intermediate and high RS cohorts in both prognostic models. Sensitivity analyses replicated similar findings after stratification by the year of diagnosis and patients' age. Conclusions: The competing-risks model is useful in dealing with multiple end events for older breast cancer patients. 21-gene RS was independently associated with BCSM. However, chemotherapy did not significantly decrease the risk of BCSM in intermediate and high RS cohorts.