Background: The prevalence of CI and the utility of CI screening for community oncology practices are not well established. We used two tools to screen for CI in older patients (pts) enrolled onto a cluster randomized controlled trial and explored how CI screening influences conversations about cognition (CAC) between older pts, their caregivers, and oncologists. Methods: Pts aged ≥70 with advanced cancer were recruited (URCC 13070; PI: Mohile). CI screen, including Mini-Cog (MC) (normal/abnormal) and Blessed Orientation Memory Concentration Test (BOMC) (scored 0-28), were included in a Geriatric Assessment (GA). Practices were randomized to usual care (UC) vs GA intervention (GA summary provided to oncologists). Audio-recorded clinical encounters were transcribed by 2 blinded coders who coded CAC with a priori scheme as follows: cognition discussed (Y/N), type of concern, who initiated. MC and BOMC were compared and receiver operating characteristic (ROC) analyses identified BOMC score that best predicted abnormal MC. Results: Mean age was 77 (range 70-93); 2.2% screened positive by BOMC using standard score of ≥11 and 33.5% had abnormal MC. Pts with abnormal MC were more likely to have impaired activities of daily living (ADL) (34 vs 24%), Instrumental ADL (64 vs 52%), Timed Up and Go (47 vs 34%) and positive depression screen (28 vs 19%), (p < 0.05 for all). CAC occurred in 22% of encounters and were more common in the intervention arm (OR 4.64, 95%CI: 2.98- 7.21, p < 0.001). Differences in CACs were most notable for pts with abnormal MC (71% in intervention group vs 16% in UC, p < 0.001). Oncologists were more likely to be the initiator of CACs in the intervention arm (90% vs 57%, p < 0.001). The most common concerns were memory (54%) and comprehension (15%). A BOMC cutoff ≥4 was optimal for predicting abnormal MC (AUC = .73, sensitivity 59%, specificity 74%). Conclusions: In community oncology practices, 1/3 of older pts screen positive for CI and were more likely to have other GA impairments. Providing oncologists with results of a CI screen increases CACs. A lower BOMC cutoff predicted abnormal MC, consistent with studies in non-cancer pts where lower BOMC score predicted more mild CI.