Purpose: To determine clinical characteristics that predict long-term response to infliximab therapy in patients with Crohn's disease (CD), and to develop a model that reliably predicts long-term response to infliximab in patients with CD. Methods: In this single center retrospective cohort study, patients receiving maintenance infliximab therapy were selected, and clinical variables were collected. The clinical response at 2 years was determined and univariate analysis was done to determine predictive factors of this outcome. A classification tree with a recursive partitioning model to predict a 2-year clinical response to infliximab was created. The cohort of patients was randomly partitioned into two groups; 70% of patients (development cohort) were used to develop the classification tree, and the remaining 30% of patients (validation cohort) were used to validate the tree. Results: One hundred ninety four patients were treated with infliximab for IBD from 1998–2005. Seventy patients were not included in the study because of the following reasons: patient did not have CD (38 patients), patient was lost to follow up within 1 month of starting infliximab (2 patients), patient received episodic infliximab treatment (21 patients), and inadequate clinical data on patient (9 patients). Therefore, 124 patients were included in our study. One hundred and one patients were either treated with infliximab for at least 2 years, or stopped treatment within 2 years. Fifty-seven patients (56%) were classified as having a clinical response to infliximab at 2 years, 49 patients (49%) were classified as having a complete response to infliximab at 2 years, and 44 patients (44%) stopped infliximab within 2 years of initiating therapy. A higher proportion of patients no longer on infliximab at 2 years were on prednisone 1 month after starting treatment, and had more operations prior to starting treatment compared to the patients still responding to infliximab after 2 years. The classification tree developed had a sensitivity and specificity of 83.3% and 86.2%, respectively in the development cohort. It had a sensitivity and specificity of 71.4% and 46.7%, respectively in the validation cohort. Conclusions: The long-term response to infliximab in patients with CD is approximately 50%. The derived clinical predictive rule using a classification tree based on recursive partitioning may play a role in the future to guide physicians in selecting patients most likely to benefit from infliximab.