Finally, this analysis considered only patients for whom both medical and pharmacy claims data were
available. Patients who had only medical or only phar-macy data included in the database were discounted, and
this might have affected the overall results.