5; group 2, 0.5 to <0.6; group 3, 0.6 to <0.7; and group 4, ≥0.7),1 migrant status (migrant: migration from outside the Epi-DSS area between 2000 and 2006),
and month of birth, and compared coverage across strata using chi-square tests. For children with vaccine cards, we obtained coverage at specific time points and median and inter-quartile ranges for age at vaccination. We constructed inverse Kaplan–Meier survival curves for immunization with one, two and three NVP-BKM120 molecular weight doses of pentavalent vaccine and compared time-to-immunization across strata using log-rank tests. We built multivariable Cox proportional hazards models to investigate the effects of travel time to vaccine clinics, sex, ethnic group, maternal education, migration and season (rainy:
April–June and October–November) on time-to-immunization with any dose of pentavalent vaccine, this website with each child contributing survival time from 14 days of age for dose one and from the date of the previous dose for doses two and three. Children with missing dates of vaccination were excluded from individual analyses as appropriate. We used a spatial bootstrap method with 100 repetitions to account for the intra-subject correlation induced by repeat observations from individual children and the inter-subject correlation engendered by spatial clustering of immunization events. In each repetition, we randomly selected 40 sublocations (with replacement) and estimated the proportional hazards model on all data from the selected sublocations. Variables without statistically significant effects (at the 0.05 level) based on Wald tests were dropped from the multivariable models. Complementary
log–log graphs and Wald tests for time-varying covariates were used to assess the validity of the proportional-hazards assumption. All analyses were conducted in Stata 9.2 (StataCorp, College Station, TX). We randomly selected 2504 eligible subjects from the population register. Of these, 1804 were enrolled on the first home visit and an additional 271 (of 509), 82 (of 180) and 12 (of 28) were enrolled on a second, third and fourth visit, for an overall enrollment rate of 86.6% (2169/2504). Reasons for non-enrollment included refusal to participate (23, 6.9%), loss to follow-up after three others or more unsuccessful visits (77, 23%), out-migration to an unknown location (48, 14.3%), out-migration outside the Epi-DSS area (136, 40.6%), database error (e.g. mapping error, age error: 47, 14%), and fieldwork error (4, 1.2%). Enrollment attained 95.4% when out-migrants and database errors were excluded. Monthly enrollment ranged from 79% to 93.7%, with 155–303 subjects visited each month (83 in December 2007). Survey respondents for the 2169 enrolled children included 1859 mothers, 131 fathers and 179 other relatives. Vaccine cards were available for 1870 subjects (86.2%).