Objective: To assess the effects of an active pharmacogenetic screening policy for antiepileptic drug (AED) therapy on everyday clinical practice and clinical outcomes. Methods: We extracted data covering all public hospitals and clinics in Hong Kong for patients who were newly commenced on carbamazepine or other AEDs, or were tested for HLA-B∗15:02 3 years before policy implementation (prepolicy: September 16, 2005 to September 15, 2008) and 3 years after (postpolicy: September 16, 2008 to September 15, 2011). We compared AED prescriptions and the incidence of SJS/TEN between the 2 periods and analyzed adherence to the policy. Results: A total of 111,242 patients were included and 4,149 were tested for HLA-B∗15:02. As a proportion of all new AED prescriptions, carbamazepine declined from 16.2% (10,077/62,056) in the pre-policy period to 2.6% (1,910/74,606) in the post-policy period (p < 0.001) while other AEDs increased. Among patients started on their first-ever AEDs, incidence of Stevens-Johnson syndrome (SJS)/toxic epidermal necrolysis (TEN) induced by carbamazepine reduced from 0.24% (20/8,284) to 0% (0/1,076; p = 0.027), but SJS/TEN induced by phenytoin increased (0.15% [18/11,839] vs 0.26% [33/12,618], p 5 0.058), and the overall incidence of AED-induced SJS/TEN remained unchanged (0.09% [42/45,832] vs 0.07% [39/55,326], p = 0.238). Test-prescription practice was adherent to the policy in only 26.4% (1,302/4,929) of relevant patients. Conclusion: The screening policy was associated with prevention of carbamazepine-induced SJS/TEN without reducing the overall burden of AED-induced SJS/TEN, likely because of clinicians preferring AEDs that do not require genetic screening but may also induce SJS/TEN.