Sentinel versus population-based surveillance of pneumococcal conjugate vaccine effectiveness.

OBJECTIVE To compare sentinel and population-based surveillance of the effect of seven-valent pneumococcal conjugate vaccine (PCV7), introduced in 2000, on the hospitalization of children aged under 5 years with invasive pneumococcal disease (IPD) in the United States of America. METHODS Population surveillance data were used to identify children hospitalized between 1998 and 2006 with IPD caused by Streptococcus pneumoniae serotypes. The change from 1998 and 1999 (baseline) to 2006 in the number of hospitalized IPD cases recorded by sentinel surveillance systems involving single hospitals or groups of hospitals was compared with the change in the incidence of hospitalized IPD cases measured by population-based surveillance. FINDINGS The change in incidence in the eight surveillance areas varied from -37 to -82% for IPD caused by any serotype and from -96 to -100% for IPD caused by serotypes contained in PCV7. All individual sentinel hospitals with more than three cases annually at baseline reported a decrease in cases by 2006. In addition, over 95% of sentinel systems with an average of more than 30 cases annually at baseline recorded a change by 2006 in the number of cases caused by any serotype that fell within the 95% confidence interval for the change in the incidence of hospitalized cases in the corresponding population surveillance area. The change in cases caused by PCV7 serotypes was accurately measured by 93% and 100% of sentinel systems with ≤ 20 and > 20 cases annually at baseline, respectively. CONCLUSION Sentinel surveillance can accurately measure the effect of PCV7 on the number of children hospitalized with IPD, provided sufficient cases are detected at baseline. Serotyping increases accuracy.