Background: IO agents are being adopted rapidly into clinical care; however, variation in speed and breadth of adoption across oncology practices remains unknown. Our objective was to evaluate adoption patterns in the treatment of aNSCLC and identify practice characteristics associated with adoption trajectories. Methods: 43,697 patients diagnosed with aNSCLC from Jan’11-Dec’17 were obtained from the Flatiron Health electronic health record-derived database, a national sample of academic and community practices. We estimated the proportion treated each month with IO (nivolumab, pembrolizumab or atezolizumab) versus other therapies from time of first IO approval (Mar‘15) through Dec’17 in 123 practices that treated aNSCLC patients during this time. We used k-means clustering to identify patterns of IO adoption. Multivariable logistic regression models were used to adjust for differences in case-mix and evaluate association of practice size, location, and Quality Oncology Practice Initiative (QOPI) certification program with IO adoption. Results: We identified 4 distinct groups of practices based on trajectories that differed in speed and extent of IO adoption (Table). 17% of practices adopted IO rapidly and extensively; 28% were slower and more limited in their adoption; 24% initially had limited IO use, but adoption accelerated rapidly after 18 months; 32% initially adopted rapidly, but slowed markedly after 1 year. In multivariable analyses, we found no significant association between a practice’s size, location, or QOPI certification and IO adoption trajectory. Conclusions: There is significant variability in adoption of IO therapy by oncology practices. Further research is needed to characterize drivers of this variation at the physician level and its impact on patient outcomes. Understanding variability in the diffusion of new innovations could guide development of targeted educational interventions to optimize use of new effective therapies.