Josephine van Dongen

Chapter 2 36 In the absence of therapeutic interventions for AGE, prevention is the approach to decrease the burden of disease in this vulnerable patient population. In the Netherlands a targeted rotavirus vaccine program would be cost-effective as suggested by two previous studies 5,14 . However due to changing epidemiology 33,34 the effectiveness of a selected vaccine strategy might be modified. Other European countries have implemented rotavirus vaccination in their national immunization program, thereby creating herd immunity (indirect protection) 35,36. Our study results suggest that these infants with medical risk conditions would benefit from AGE prevention. As strengths, this study covers the full burden of disease by combining incidence rate, severity, healthcare attendance and family impact of AGE in infants with medical risk conditions. Fur thermore, this vulnerable patient population is generally not studied. Therefore this study provides unique information on community disease burden of AGE. By comparing par ticipants and non-par ticipants, it appeared that the study population is representable for the group of infants with medical risk conditions in the Netherlands. This study has several limitations. First, data on pathogen and MVS scale were missing for about half of the AGE episodes, because parents failed to take a stool sample and complete the diaries. To reduce bias, we used multiple imputation in calculating pathogen specific IRs and in the analysis on risk factors for severe AGE. While this method accounts non-random missing data and thereby adjusts for bias by complete case analysis, it also has limitations by assuming (too much) variance in outcome variables based on the imputation procedure 37,38. However, with multiple imputation the sample size is maintained and it yields unbiased standard errors 37,38 . Propor tion of missing data is not a guide for using multiple imputation, as with correct specified imputation procedures estimates obtained will be less biased than by complete case analysis 39 . Fur thermore we checked for differential missingness and showed the results of complete case and imputed data analyses together.The high rate of missing data in community studies on AGE could be overcome by using modern interactive technologies to monitor par ticipants 28 . In the RotaFam study, symptom data was in 97% complete and stool samples were obtained from 87% of AGE episodes, compared to 45% in our study. Secondly, in our study 33% of par ticipants were lost to follow up before 18 months of age. Possibly, because parents taking care of a child with special medical needs were overburdened and thereby limited in their capacity to adhere to study procedures, illustrated by the median days of hospitalization (35 days, range 3-439). Still, the mean follow up was 14 months. In addition, the survival analysis is used to consider loss to follow up and observation time 40 .