The large population in the health plan allowed controls to be matched to PCR-positive children on many potential confounders, and matched controls were more similar to PCR-positive children than were PCR-negative controls on all measured potential confounders