Background: Older adults with cancer are at increased risk of treatment-related toxicity and mortality. A comprehensive geriatric assessment (GA) may uncover aging associated vulnerability and identify those at greatest risk of adverse outcomes. We studied the association between a novel frailty index and treatment-related morbidity and mortality among older adults with GI malignancies. Methods: Older adults (≥60y) referred for initial consultation at the UAB GI oncology clinic between 9/2017 to 12/2020 were enrolled in a prospective Cancer and Aging Resilience Evaluation (CARE) registry. All participants underwent a patient-reported GA at baseline as previously described (Williams et al J Geriat Oncol 2020). Using this information, we constructed a 44-item frailty index using a deficit accumulation approach. Vital status was acquired by linking with death records and chart review. In a subgroup of patients continuing care at UAB, we collected information on toxicity for the first 6 months of treatment via chart review using CTCAE v5.0. We used Kaplan Meier Methods and log-rank test to compare survival distributions, and a multivariate cox regression to adjust for potential confounders. We compared the toxicity rates across frailty subgroups using risk ratio (RR) calculated from general linear models. Results: Of 765 consecutive older adults referred to GI oncology clinic, 590 (77%) had available data to measure frailty index. Median age at enrollment was 68y; with 59% males and 72% White. Common cancer types included colorectal (30%) and pancreatic cancer (26%); mostly with advanced stage disease (stage III 28%; IV 46%). Overall, 168 (28%) were characterized as pre-frail and 230 (39.3%) as frail. As compared to non-frail, those who were frail were more likely to be Black (33% vs 20%; p < 0.01) and have pancreatic cancer (33.6% vs 21.8%; p < 0.01). Over a median follow up of 22 months, 212 (36%) patients had died. The 2y overall survival among non-frail, pre-frail and frail patients was 71%, 63% and 51%, respectively (log rank p value < 0.001). In a multivariate cox regression, as compared to non-frail patients, frailty was associated with worse OS (HR 1.75; 95% CI 1.13-2.70; p = 0.01) after adjusting for age, sex, race, cancer stage, cancer type, line of therapy and performance status. In a subset of 168 patients with available data, baseline frailty was associated with increased risk of ≥grade 3 non-hematologic toxicity (RR 2.23; 95% CI 1.27-3.92; p < 0.01) but not ≥grade 3 hematologic toxicity (RR 1.03; 95% CI 0.67-1.58; p = 0.90) as compared to non-frail patients Conclusions: The CARE-Frailty Index is a novel frailty index built on the principle of deficit accumulation using a patient-reported GA, and appears to be a robust predictor of survival and may predict treatment related toxicity among older adults with GI malignancies.