Background: Several oncology guidelines recommend using oral drugs vs. IV to minimize COVID-19 risk for patients with cancer. We examined the association between prescribing patterns of oral capecitabine vs. IV 5FU for GI cancers and social distancing, measured by the change in population mobility patterns in response to shelter-in place policies, during the pandemic. Methods: Using claims data for commercially insured members, we included patients 18 years of age or older with colorectal, gastroesophageal, or pancreatic cancer, who had continuous health plan coverage for at least 2 months before and 1 month after initiating chemotherapy with capecitabine or 5-FU from January 2017 to August 2020. We analyzed unadjusted trends in proportion of chemotherapy that was oral during pandemic (March 1^st to August 31^st, 2020) compared to previous years. Then, we conducted difference-in-differences analysis using COVID-19 Community Mobility Reports, by Google, and utilizing different levels of changes in mobility trends across states over time. In our main model, we used a 20% decrease in retail and recreation visits as our threshold and compared the prescribing rates in states below and above the threshold as well as before and after the pandemic began. We also used different thresholds and categories of places to check the sensitivity of our findings. Models are adjusted for age, gender, month of year, urban status, comorbidities, and state of residence at chemotherapy start date. Results: A total of 17,414 nationally distributed patients (69% colorectal, 13% gastroesophageal, 18% pancreatic) were included (mean age, 58.8 years; 41% female). During the pandemic, 1,875 patients (65% colorectal, 15% gastroesophageal, 20% pancreatic) were identified. The proportion of oral regimens did not change significantly for colorectal and gastroesophageal patients and decreased by 7.4 percentage points (pp) (p < 0.01) for pancreatic patients. In regression modelling with mobility data, oral prescribing rates for colorectal patients increased by 3.1 pp (p < 0.01), largely driven by increases for female patients (9.2 pp, p = 0.02). We observed a decrease in oral prescribing rates among pancreatic patients (-1.20 pp, p = 0.04) and did not observe a significant change for gastroesophageal patients. Our results are not sensitive to different social distancing specifications. Conclusions: We observed differential impact of the pandemic on oral prescribing rates by GI cancer type and gender. Oral prescribing increased among colorectal cancer patients driven mostly by higher oral prescribing in females. For pancreatic and gastroesophageal patients, oral prescriptions either remained unchanged or decreased. This observation may reflect a variable impact of the pandemic on women as compared to men and might involve heightened caregiving responsibilities for women.