300 Background: A clinical pathways program requires regular updates to pathways content, real-time decision support, and data collection and analysis of decisions made. A goal of Dana-Farber Pathways (DFP) is to analyze data to reduce unwarranted variation and inform granular, warranted variation. Methods: Each of our 31 medical oncology pathways is created and updated by DF experts (clinicians and scientists). DFP is implemented through a web-based portal that provides real-time decision support. DFP captures on- and off- pathway treatment decisions and reasons for off-pathway decisions; this is shared with each disease group monthly. For each interval review meeting we assess on-pathway rates and decisions for each node in the pathway. Low on-pathway rates prompt discussion about possible action, including provider education or pathway modification. Results: In 2017, 7,460 total treatment decisions were collected through DFP; 78% were on-pathway. We have clinical reasoning for 90% of the off-pathway decisions. Off-pathway analysis has been used in several important ways: 1. Catalyzing change: we detected early adoption of immunotherapy in small cell lung cancer and discussed the relevant data. The pathway was updated to adopt this class of treatment. 2. Understanding unexpected events: we identified a recent etoposide shortage and discussed alternate recommendations. They were added to the pathway in case of future shortages. 3. Provider education: we detected a consistently low on-pathway rate in one location, mostly driven by a specific provider. This provided a mechanism to discuss practice patterns and provide targeted education. Conclusions: Off-pathway analysis provides insight into user variation, fosters and supports peer coaching, and supports the creation of dynamic, up-to-date pathways.[Table: see text]