Background: Oncologists commonly use performance status scores such as ECOG and Karnofsky when making treatment decisions. However, there is limited data on the utility of these measures in selecting therapy and predicting treatment-related complications in the elderly. In July 2016, medical oncologists at our institution made it standard practice to use the Cancer and Aging Research Group toxicity prediction score (CARG) to identify elderly patients who are “fit” for conventional therapy versus those who are “frail” and may experience increased complications. Methods: We retrospectively reviewed referrals to our outpatient medical oncology clinic for patients 75 years of age and older. Referrals leading to systemic therapy (i.e. chemotherapy, targeted therapy, and immunotherapy) or changes to therapy based on CARG were included in our study. We evaluated CARG utilization during the 6 months before and the 6 months after its institution. We also assessed how changes to therapy due to CARG impacted admissions and emergency department (ED) visits. Results: Thirty-eight patients with mean age of 81.2 years, 52.6% male, met inclusion criteria. Their diagnoses included GI (36.8%), lung (21.1%), hematologic (18.4%), breast (10.5%), GU (2.6%), and other (10.5%) malignancies. During 2016, CARG was assessed in 14 patients, 9 of whom received systemic therapy. CARG was assessed in 12.5% and 41.2% of patients receiving systemic therapy in the 6-month periods before and after its incorporation, respectively. There were fewer mean admissions (0.7 versus 2.25), days of admission (4.3 versus 8), and ED visits (1.1 versus 2.5) when therapy was decreased in intensity or withheld based on CARG. However, the study lacked power to establish statistical significance. Conclusions: A CARG assessment occurred in less than 50% of elderly cancer patients who received systemic therapy. Therapy modifications based on CARG may lead to fewer hospitalizations and ED visits, but more consistent use of the score and further investigation are necessary.