Published in:
Open Access 01-12-2011 | Research article
Hospital-based, prospective, multicentre surveillance to determine the incidence of intussusception in children aged below 15 years in Germany
Published in: BMC Gastroenterology | Issue 1/2011
Login to get accessAbstract
Background
A new vaccine against Rotavirus (RV) gastroenteritis was introduced in Germany in 2006. In 1997 the first RV vaccine was withdrawn due to an increased incidence in intussusception (IS). Thus, an accurate estimation of the incidence of IS is important for post-licensure surveillance.
Methods
IS-Data were obtained from the 'Erhebungseinheit für seltene pädiatrische Erkrankungen Deutschland' (ESPED, German surveillance unit for rare pediatric diseases) collaborations' central register where all cases of intussusception in Germany for the years 2006 and 2007 are collected (n = 1200). In order to obtain an unbiased estimate of the incidence, it is necessary to determine the population under risk out of which these cases originated, and the proportion of real cases not reported to the registry (underreporting). In order to assess underreporting, a random sample of 31 hospitals was re-assessed by an outside reviewer. The estimation of incidence was done using a single Maximum-Likelihood (ML) estimator based on data from both the registry and the sample.
Results
The uncorrected observed incidence was calculated to be 26.6/100,000 child-years for children below 1 year old, 23.8 for those below 2 years old, and 5.2 for those below 15 years old. The review revealed a mean reporting quota of about 41% and the ML approach yielded an incidence of 51.5/100,000 child-years (95%CI [41.7;61.1]) for children below 2 years of age.
Conclusions
While substantial under-reporting led to very conservative estimates of the IS incidence, the approach described here allows an accurate estimation of IS incidence including corresponding confidence bands. Therefore, ML estimation is a straightforward instrument to derive stable, unbiased estimates in epidemiological studies with incomplete data.