Background: Geriatric Assessment (GA) can help oncologists determine fitness of their older patients (pt) for anti-cancer therapy. Our objective was to compare routine provider assessment (PA) and GA of older adults with gastroesophageal cancer (GEC). Methods: Patients ≥65 with any stage of GEC completed a GA. The pt’s provider completed a PA and abnormalities detected by both assessments were centrally reviewed and compared. GA and PA assessed functional status, nutrition, comorbidities, psychological distress, cognition, social support, and chemotherapy toxicity risk. Validated assessment tools were used for the GA. Data collected 3 months post-enrollment included hospitalization, ≥ grade 3 toxicities, and treatment delays. We compared the proportions detected/not detected abnormalities by PA vs GA using McNemar’s test for paired data, and we measured agreement using Kappa statistics (1=full, 0=no, -1=disagreement). Results: 82 pts were enrolled, majority male (74%), median age 73 (65-91), stage III/IV (82%) disease on first line therapy (79%). Cancer sites included gastric (32%), esophageal (43%), GEJ (26%). Pts’ demographic and clinical characteristics did not predict for the identification of GA abnormalities. GA detected 196 abnormalities and PA detected 86. Majority of pts (84%) had ≥ 1 unidentified abnormality by PA. Providers identified more pts with clinically significant comorbidities as compared to GA. However, the GA identified more abnormalities compared to PA in all other evaluated domains (Table). Low agreement was found between PA and GA in any of the domains. Pts scoring high on the CARG chemotherapy toxicity prediction tool (p=0.004) and pts with evidence of psychological distress (p=0.049) had more ≥ grade 3 toxicity. Pts with abnormalities in functional status, nutrition, psychological distress, and a high CARG score were more likely to be hospitalized during therapy. Conclusions: Clinical characteristics of older pts with GEC are not predictive of GA abnormalities. While providers were better at identifying comorbidities, the GA resulted in a comprehensive evaluation of all domains and identified significant abnormalities that affect treatment outcomes in this population.