Josephine van Dongen

Chapter 2 34 activities, like grocery shopping, cleaning or leisure activities ( Table 2 ). Risk factors for (severe) AGE 312 par ticipants (49%) had at least one AGE episode during follow up and multiple episodes (varying from two-seven) were repor ted in 88 infants (14%), differences in their characteristics are shown in eTable 4. The most impor tant differences between the groups were more frequent daycare attendance and higher family education among infants with at least one AGE episode. Infants with severe AGE did not differ from those with non-severe AGE. Mean age for first severe AGE episode was 8.3 months (95% CI 7.2-9.5) and 14.2 months (95% CI 13.5-14.8) for non-severe AGE. The multivariate Cox regression analysis resulted in a final model with only seasonal months and daycare attendance statistically significantly associated with severity ( Table 3 ). Table 3. Risk factors associated with time to first severe AGE episode Risk factor Univariate HR Multivariate HR* 95% CI Adjusted HR after MI* 95% CI Seasonal months (Oct- April) 27.01 27.12 (10.60-69.39) 27.24 (12.96-57.26) Sibling 0.77 0.42 (0.17-1.02) 0.76 (0.41-1.41) Breastfed 2.55 1.28 (0.46-3.56) 2.25 (1.18-4.30) Born in sea- sonal months 0.92 0.62 (0.28-1.42) 0.82 (0.43-1.48) Daycare 5.72 4.05 (1.35-12.17) 1.24 (0.43-3.55) Abbreviations: HR = hazard ratio, CI = confidence interval, MI = multiple imputation, RV = rotavirus, Jan = January. * Adjusted for seasonal months, siblings, breastfed, born in seasonal months and daycare attendance. Because of potential non-random missing information of risk factors, we compared complete cases versus those with at least one missing covariate, and found that complete cases were more frequently older at AGE episode or when censored.They less often had an older sibling in their household, attended daycare from six months onwards and had an episode in season ( eTable 5 ).The estimates after multiple imputation account for the missing not at random.