We examine the targeting effects of stricter screening in the Dutch Disability Insurance (DI) program induced by a major nationwide reform. The drastic 2003 "Gatekeeper Protocol" raised DI application costs and revealed more information about individual true ability to work. Discontinuity-in-Time regressions on administrative data show substantial declines in DI application rates (a 40% decrease in one year), with the largest decline occurring in difficult-todiagnose impairments and less severe health disorders. This resulted in a more deserving pool of applicants. At the same time, those who stopped applying had worse health, worked less, and were more likely to be on UI and social assistance than workers who did not apply in the old system. There are no additional targeting gains at the point of the award decision, implying that changes in average health conditions of awardees were fully driven by self-screening and work resumption in the DI waiting period.