Background: Many studies reported a steep learning curve in adopting minimally invasive surgical technology.A regionalization policy that concentrates minimallyinvasive radical prostatectomy (MIRP) care over a limited number of surgeons may greatly improve outcome.However, the effect of regionalization on MIRP outcome remains unclear.Data and Method: This study uses State Inpatient Databases (SID) from 2003 to 2010 and hospital length of stay (LOS) following MIRP as thestudy outcome.We restrict the analysis to the States of New York, Maryland and Iowa to ensure the ability to tracksurgeons'MIRP volumes with accuracy.We estimate an orderdprobit model with MIRP volume spline interpolationto explain LOS of MIRP.The analysis is at the MIRP surgery level.Besides adjusting forpatient characteristics (age, race,comorbidity, and insurance coverage), the regression model further adopts surgeon and yearfixed effects to allow for surgeon learning heterogeneity and a generallearning trend, respectively.We extract surgeons' learning curves from the regression model.Next we use these curves and discrete optimization technique to identify the best within-state MIRP regionalization policy that minimizes LOS.Results: On average, prior to treating 13 patients a quarter, increased volume shortens LOS by about 6% from 1.76 days to 1.66 days.After treating 13 patients a quarter, higher volume does not improve care outcome.The observed average LOS is 1.69 days.In contrast, the average LOS under an optimal quarterly within-state regionalization allocation would be 1.001 days (95% confidence interval:[1, 1.013]).In total, best regionalization can reduce 4,260 fewer hospital days (out of 10,441 days)among the 6,172 patients receiving MIRPs in New York, Maryland, and Iowa between 2004 and 2010.Conclusion: Regionalization can significantly improve MIRP care outcome (LOS shortened by 40.8%).This warrants exploring policies that promote regionalization of MIRP in the current US care delivery system.