PURPOSE: Falls are a modifiable source of morbidity for older adults with cancer, yet are underassessed in oncology practice. In this secondary analysis of a nationwide cluster-randomized controlled trial, we examined characteristics associated with patient-oncologist conversations about falls, and whether oncologist knowledge of geriatric assessment (GA) resulted in more conversations. METHODS: Eligible patients (ClinicalTrials.gov identifier: NCT02107443 ) were age ≥ 70 years, had stage III/IV solid tumor or lymphoma, were being treated with noncurative treatment intent, and ≥ 1 GA domain impairment. Patients in both arms underwent GA. At practices randomly assigned to the intervention arm, oncologists were provided a GA summary with management recommendations. In both arms, patients had one clinical encounter audio-recorded, transcribed, and coded to categorize whether a conversation about falls occurred. Generalized linear mixed models adjusted for arm, practice site, and other important covariates were used to generate proportions and odds ratios (ORs) from the full sample. RESULTS: Of 541 patients (intervention N = 293 and usual care N = 248, mean age: 77 years, standard deviation: 5.3), 528 had evaluable audio recordings. More patients had conversations about falls in the intervention versus usual care arm (61.3% v 10.3%, P < .001). Controlling for the intervention and practice site, history of falls (OR, 2.1; 95% CI, 1.3 to 3.6; P = .005) and impaired physical performance (OR, 4.7; 95% CI, 1.7 to 12.8; P = .002) were significantly associated with patient-oncologist conversations about falls. CONCLUSION: GA intervention increased conversations about falls. History of falls and impaired physical performance were associated with patient-oncologist conversations about falls in community oncology practice.