Hepatitis B vaccination timing: results from demographic health surveys in 47 countries

Abstract Objective To examine the impact of hepatitis B vaccination schedules and types of vaccines on hepatitis B vaccination timing. Methods We used data for 211 643 children from demographic and health surveys in 47 low- and middle-income countries (median study year 2012). Data were from vaccination cards and maternal interviews. We grouped countries according to the vaccination schedule and type of vaccine used (monovalent or combination). For each country, we calculated hepatitis B vaccination coverage and timely receipt of vaccine doses. We used multivariable logistic regression models to study the effect of vaccination schedules and types on vaccination delay. Findings Substantial delays in vaccination were observed even in countries with fairly high coverage of all doses. Median delay was 1.0 week (interquartile range, IQR: 0.3 to 3.6) for the first dose (n = 108 626 children) and 3.7 weeks (IQR: 1.4 to 9.3) for the third dose (n = 101  542). We observed a tendency of lower odds of delays in vaccination schedules starting at 6 and at 9 weeks of age. For the first vaccine dose, we recorded lower odds of delays for combination vaccines than for monovalent vaccines (adjusted odds ratio, aOR: 0.76, 95% confidence interval, CI: 0.71 to 0.81). Conclusion Wide variations in hepatitis B vaccination coverage and adherence to vaccination schedules across countries underscore the continued need to strengthen national immunization systems. Timely initiation of the vaccination process might lead to timely receipt of successive doses and improved overall coverage. We suggest incorporating vaccination timing as a performance indicator of vaccination programmes to complement coverage metrics.


Introduction
Chronic hepatitis B virus (HBV) infection continues to make a substantial contribution to the global burden of disease. 1,2 The risk of developing chronic HBV is inversely related to the age at acquisition of infection. 3,4 Immunization is the most effective measure to prevent the transmission of HBV. 5,6 In 2014, the World Health Organization (WHO) reaffirmed the need for hepatitis B vaccines to become an integral part of national immunization schedules. 7 WHO recommends a birth dose within 24 hours of birth to prevent perinatal and early horizontal HBV transmission. 8 The birth dose should be followed by 2 or 3 doses of monovalent or multivalent hepatitis B vaccines. 8 Vaccination coverage estimates from WHO and the United Nations Children's Fund (UNICEF) capture the proportion of vaccinated children in specific age groups. However, these estimates provide little insight into the extent to which vaccinations are administered on time and they tend to understate the susceptibility to HBV infection in a population. [9][10][11] In practice, vaccinations are more likely to be received late than early. 12,13 When hepatitis B vaccination is delayed, children fail to receive adequate protection when they are most vulnerable. Moreover, by increasing the period of susceptibility to infection, 8 late vaccinations raise the risk of HBV infection 14 and hence the risk of chronicity. Furthermore, a delay in one dose may lead to delays in further doses, 15 thereby extending the at-risk period. This has important implications in countries that are highly endemic for HBV infection. In this situation, catch-up vaccination of older children has relatively little impact because they might already be infected by the time they present for vaccination. 8 There are multiple options for incorporating hepatitis B vaccines into national immunization programmes and the choice of vaccination schedule depends primarily on programmatic considerations. 8 From a policy perspective, data from a large number of countries are necessary to evaluate the impact of existing hepatitis B vaccination schedules and vaccine types on hepatitis B vaccination timing. Thus far, analyses of hepatitis B vaccinations have been limited in scope [16][17][18] and have not tackled this aspect. The demographic and health surveys (DHS) provide data on childhood vaccinations based on vaccination cards and maternal interviews. Data compiled through DHS are nationally representative and are considered to be the best available data on vaccination coverage. 19 We estimated vaccination coverage and timing, and examined the impact of hepatitis B vaccination schedules and vaccine types on vaccination timing in countries for which DHS data were publicly available.

Study design
Full details of DHS methods have been reported elsewhere. 20,21 DHS data on hepatitis B vaccination were available for 54 countries. For every country, we used the most recent survey available until the end of 2015. Seven surveys were excluded due to incomplete data or non-standard recording of dates. We therefore included 47 countries with survey years ranging from 2005 to 2014. We grouped countries based on their vaccination schedule and type of vaccine (monovalent or combination) in use (Table 1, available at http://www.who.int/bulletin/volumes/95/3/16.178822). In Objective To examine the impact of hepatitis B vaccination schedules and types of vaccines on hepatitis B vaccination timing. Methods We used data for 211 643 children from demographic and health surveys in 47 low-and middle-income countries (median study year 2012). Data were from vaccination cards and maternal interviews. We grouped countries according to the vaccination schedule and type of vaccine used (monovalent or combination). For each country, we calculated hepatitis B vaccination coverage and timely receipt of vaccine doses. We used multivariable logistic regression models to study the effect of vaccination schedules and types on vaccination delay. Findings Substantial delays in vaccination were observed even in countries with fairly high coverage of all doses. Median delay was 1.0 week (interquartile range, IQR: 0.3 to 3.6) for the first dose (n = 108 626 children) and 3.7 weeks (IQR: 1.4 to 9.3) for the third dose (n = 101 542). We observed a tendency of lower odds of delays in vaccination schedules starting at 6 and at 9 weeks of age. For the first vaccine dose, we recorded lower odds of delays for combination vaccines than for monovalent vaccines (adjusted odds ratio, aOR: 0.76, 95% confidence interval, CI: 0.71 to 0.81). Conclusion Wide variations in hepatitis B vaccination coverage and adherence to vaccination schedules across countries underscore the continued need to strengthen national immunization systems. Timely initiation of the vaccination process might lead to timely receipt of successive doses and improved overall coverage. We suggest incorporating vaccination timing as a performance indicator of vaccination programmes to complement coverage metrics.
countries that had altered their schedules before the DHS survey we limited our analyses to the more established vaccination schedule.
We identified and analysed individual vaccine doses according to the respective country's national immunization schedule. To assess vaccination coverage, we used only documented vaccinations (with or without specific dates marked) for each vaccine dose. Vaccination coverage was categorized as complete if the child was recorded as fully immunized with three or four doses of the vaccine according to the country's national immunization schedule. Vaccination coverage was categorized as incomplete if any of the recommended doses were recorded as 0 (not given), including when data on other doses was missing. 8 We excluded children younger than 12 months to avoid the drawback of censored observations. The denominator for coverage was the DHS sample of surviving children born in the past 5 years before the survey (or sometimes 3 years, depending on the DHS interval). To address potential bias from maternal recall, 24,25 we estimated crude vaccination coverage and completeness (from vaccination card plus maternal recall).
To assess vaccination timing, we compared each child's recorded vaccination dates with those recommended in the country's national immunization schedule. Age at vaccination was determined by subtracting the child's date of birth from valid vaccination dates. Vaccinations were categorized as timely if administered within 4 weeks of the recommended age, or delayed if administered more than 4 weeks after the recommended age. We calculated the percentage of children receiving delayed or timely vaccinations. The denominator for calculating timing included children vaccinated early, i.e. before the recommended age. National immunization schedules often do not specify when to give the birth-dose vaccine. 26 We therefore defined a timely birth dose as received within 7 days after delivery, based on the evidence on effective prevention of perinatal hepatitis B transmission. 27 We also computed estimates based on the WHO recommendation of giving hepatitis B vaccine within 24 hours of birth. 8

Statistical analysis
We performed all analyses with the survey functions of Stata statistical software, version 14 (Stata Corp., College Station, United States of America), using a significance level of ≤ 0.05.
We took account of the complex DHS survey design and used sample weights provided in the available data sets. Using Spearman rank correlations, we analysed the relationship between vaccination timing and coverage of the third dose of vaccine across countries.
We then used binary multivariable logistic regression models to calculate adjusted odds ratios (aOR) and 95% confidence intervals (CI) to investigate the impact of vaccination schedule and vaccine type on hepatitis B vaccination timing. Vaccinations were dichotomized as delayed or timely. We constructed pooled models for two outcomes: delayed first dose and delayed third dose. The main independent variables were the recommended week of the vaccination schedule and vaccine type (monovalent or combination). We categorized reported vaccination schedules as follows: starting at birth i.e. ≤ 1 week of age (reference category), 4, 6, 9 and 13 weeks, respectively. We incorporated covariates chosen for their possible or demonstrated associations with vaccination measures. 16,28 In an additional pooled model, we assessed the impact of the timing of the first dose on the timing of the third dose. The dependent variable was timing of the third dose and the main independent variable was timing of the first dose.

Results
Data were analysed for 211 643 children aged 12-60 months who had valid records of date of birth and date of mother's interview. The median survey year was 2012 (interquartile range, IQR: 2010 to 2013). Reported vaccination dates were almost all complete and valid. Overall, vaccination cards were available for 123 679 (weighted count) of the children aged 12-60 months.
At the time of the surveys, 24 countries used the three-dose standard schedule for hepatitis B vaccine (doses at 6, 10 and 14 weeks), four countries vaccinated at 9, 17 and 26 weeks and the remaining countries used other threedose schedules, some of which included an extra dose at birth, i.e. four doses in total (Table 1). Thirteen countries reported a vaccine dose at birth; eight included a birth dose in their threedose schedule and five used a four-dose schedule. Combination vaccine, mostly a pentavalent vaccine, was used in 29 countries, while monovalent vaccine was used in 18 countries. Fig. 1 shows the pooled distribution of ages at vaccination for 108 626 (first dose) and 101 542 (third dose) children aged 12-60 months at the time of the mother's interview, using data from vaccination cards only. Both the first and third doses had peak numbers of children vaccinated around the recom-  mended target ages, followed by tails to the right, indicating delays in vaccination. The different peaks in the distributions of first and third doses reflect the diverse immunization schedules and recommended target ages for these doses across the 47 countries.
Coverage of the birth dose ranged from 26% to 99% of children across the 13 countries using this dose. The percentage of children receiving birth-dose vaccinations on time ranged from 23% to 94% across countries (Fig. 2). The proportion of timely vaccinations was lower when we defined the birth dose as administered within 24 hours rather than within 7 days of birth.

Vaccination coverage
Coverage for all doses, and for complete coverage varied greatly, even across countries following the same vaccination schedule and vaccine type ( Table 2, available at http://www.who. int/bulletin/volumes/95/3/16.178822). For example, complete coverage for countries using the 6-, 10-, and 14-week schedule ranged from 13% in Mali to 93% in Swaziland. Overall, we recorded a drop in coverage in particular of the third dose compared to the first dose, irrespective of the vaccination schedule and vaccine type in use. This was particularly prominent in some countries, such as Azerbaijan (where coverage dropped from 69% to 48%) and Côte d'Ivoire (from 74% to 58%).

Vaccination delays
We observed a substantial variation in delays in receipt of the first and third doses across countries having the same vaccination schedule and vaccine type ( Table 3). We noted a drop in timely vaccinations between the first and third doses, irrespective of the vaccination schedule and vaccine type in use.
For the 47 countries overall, the median of the median delays for the first vaccine dose was 1.0 week, and the 75th percentile was 3.6 weeks, i.e. in 25% of the countries the median delay was more than 3.6 weeks. For the third dose, the delays were more than twice as long ( Table 4). The country-specific distribution of ages at vaccination had long tails, and delays at the 90th percentile were at least twice as long as the 75th percentile ( Table 5, available at http://www.who. int/bulletin/volumes/95/3/16.178822). Overall, WHO African Region countries tended to have lower vaccination coverage and poorer timing compared with countries in the Americas and Europe. Delays were recorded even in countries with high coverage, such as Bangladesh and Burkina Faso. We found a weak positive correlation (Spearman rho = 0.28; P = 0.05) between vaccination timing and coverage. Fig. 3 shows the timing and the corresponding coverage of the third vaccine dose for each of the 47 countries, using data from vaccination cards.   Table 2 and Table 5

Discussion
Our analysis of survey data from 47 lowand middle-income countries, inhabited by around 1.2 billion people, 29 showed a wide variation in hepatitis B vaccination coverage and timing across countries.
The results highlight differences in vaccination implementation, and in adherence to national immunization schedules. This may reflect differences in barriers to immunization, in inequities in health-care delivery and access, as upper-middle-income countries tended to have better coverage and timing than lower-middle and low-income countries. Most countries had fairly high coverage (> 80%), in particular for the first dose, and delivered vaccines on time.
Although this finding is encouraging, in most countries coverage decreased and delays increased with subsequent doses, irrespective of a country's specific vaccination schedule. Crucially, vaccination coverage was low (< 50%) and vaccinations were delayed in populous countries that are highly endemic for HBV infection, such as Nigeria. Despite WHO recommendations on hepatitis B vaccination within 24 hours, 8 only 13 countries in our analysis reported using a birth dose, with wide variations in its coverage and timing. Due to existing sociocultural, financial, infrastructural and logistic constraints on vaccine delivery, many countries do not require the birth dose to be strictly administered within 24 hours of birth. 26,30 A major challenge, particularly in highly endemic, resource-poor countries with a high proportion of home deliveries, is ensuring the timely administration of the birth dose to every child irrespective of where he or she is born. 30,31 Most countries where the HBV epidemic is concentrated have adopted the three-dose combination vaccine delivered at 6, 10 and 14 weeks. 30 Our analysis gave some indication that vac-cination delays were lower with vaccination schedules starting at 6 or 9 weeks of age compared with those starting at or before 1 week of age, and with combination vaccines as compared with monovalent vaccines. This might be attributable to increased compliance by vaccine recipients due to the reduced number of injections and fewer visits required to health-care facilities. 32 That said, administering combination vaccinations at 6 or 9 weeks of age, while cost-effective and simple, cannot prevent vertical and early horizontal transmission. 30 It has been suggested that, due to the predominantly horizontal routes of HBV transmission in Africa, the benefit of implementing a birth dose would not justify the necessary financial, human resource and infrastructure investments. 33 This is based on the premise that perinatal transmission is not a major factor in HBV transmission due to the lower prevalence of hepatitis B e-antigen (HBeAg) positivity in pregnant women in Africa. However, studies suggest that up to 38% of pregnant African women with chronic HBV are positive for HBeAg and hence at high risk of transmitting infection to their infants. [34][35][36] Data on the epidemiology of HBV, particularly transmission routes, 30 and on the benefits of birthdose vaccination are scarce in Africa. 37 Nevertheless, in our view, the benefits We found lower compliance with national schedules for the second and third vaccine doses and a weak correlation of timing with coverage. This implies that even in countries with relatively high coverage, children who achieve complete vaccination may spend a considerable period of time with no or incomplete protection. This is particularly concerning in countries with a high burden of infection. 3 Our analysis also indicates that the third dose of vaccine is more likely to be delayed among those who received a delayed first dose. This suggests that prioritizing timely first vaccinations could result in the timely receipt of successive doses 38 and avert delays that would require catch-up regimens. Given the existent challenges in providing hepatitis B vaccination in resourcepoor settings, catch-up regimens might decrease the likelihood of the timely completion of the hepatitis B vaccina-   Table 2 and Table 3. Dates of vaccination were based on observations with available vaccination dates recorded on vaccination cards. We included children vaccinated before the recommended age (early vaccinations) in the denominator when calculating delayed and timely vaccination rates. Estimates of early vaccinations are not shown in the figure. The following countries reported > 10% children vaccinated before the recommended age for the third dose: Azerbaijan (50%), Plurinational State of Bolivia (12%), Colombia (12%), Kyrgyzstan (60%), Nigeria (12%) and Tajikistan (56%). tion series. 38,39 This underscores the need to incorporate the monitoring of vaccination timing, in addition to coverage, into vaccination programmes.
Interrupting transmission routes for HBV warrants comprehensive strategies to prevent mother-to-child transmission and to deliver adequate and timely immunoprophylaxis in newborns 40 and infants. 41,42 In remote, resource-constrained settings, integrating vaccine administration with assisted home deliveries and employing out-ofcold-chain strategies might be possible solutions to improve timely vaccination coverage. [43][44][45] Furthermore, mathematical models, calibrated to country-specific HBV epidemiology might be useful to quantify the burden of infection attributable to delayed vaccinations. In this context, models could be developed to assess the infections and deaths averted by prioritizing timely vaccinations that use alternative vaccination schedules and diverse outreach strategies.

Limitations
The main limitation of this analysis is related to the available data from DHS. The survey years varied substantially across countries, and therefore caution is warranted when interpreting international comparisons. 20 Most surveys were fairly recently conducted -the median survey year was 2012-and provide useful insights into the quality (timing) and quantity (coverage) of current hepatitis B vaccination programmes. However, some of the older surveys, notably in the Republic of Moldova and Swaziland, may not reflect the current situation.
The distribution of ages at vaccination are only crude indicators of the timing issue, since each country's contribution was determined by the size of its survey sample, which varied among countries and did not reflect actual population sizes.
Our coverage estimates vary to some extent from available estimates 46 due to some aspects of our method: the use of DHS survey data, the age groups included and the reliance on documented vaccinations. Multisurvey prospective data were unavailable for most countries. We could not therefore assess temporal changes in vaccination measures and the effects of changes in vaccination schedules or vaccine types on the studied outcomes. Furthermore, some vaccination schedules included in the analysis were used only by a small number of countries, which impeded any conclusions about the effects of specific schedules. We restricted our analysis to established vaccination schedules.
This might lead to underestimates or overestimates depending on the uptake of newer vaccines and schedules by countries. Data on vaccination service providers were not available which might have provided valuable insights into the issue of hepatitis B vaccination timing.
We excluded undocumented vaccinations from the analysis and therefore coverage and delays may be underestimates, since undocumented vaccinations including lost or misplaced vaccination cards were not captured. 19 Vaccination information was based only on maternal recall in approximately 30% of the observations, with higher figures in some countries (such as the Democratic Republic of the Congo and Nigeria). However, no noteworthy differences in coverage were detected for most countries when we included maternal reports (data are available from the corresponding author).
A disadvantage of cross-sectional studies is the potential for survivor bias. Our analysis did not include deceased children since the included surveys did not record vaccination data for this subgroup. We might have overestimated vaccination measures slightly since it is unlikely that deceased children would have better vaccination parameters than surviving children. 47 The cross-sectional nature of the data also precluded our drawing causal inferences. Additionally, it is likely that there was residual confounding that was not adjusted for in our models. To enable more in-depth analyses, future surveys need to incorporate sufficiently detailed questions on barriers to immunization, e.g. vaccine availability in the health system, and on parental and provider vaccination practices. Lastly, the surveyed countries were not randomly sampled. Hence the external validity of the results for other low-and middle-income countries might be limited, particularly for those using different vaccination schedules than those in the current analysis. The available data were primarily from coun-tries in the WHO African, European and Americas Regions, with limited data from the Eastern Mediterranean, South-East Asian and Western Pacific Regions.

Fecha de vacunación de la hepatitis B: resultados de encuestas sobre demografía y salud en 47 países
Objetivo Examinar el impacto de los calendarios de vacunación de la hepatitis B y los tipos de vacunas en los plazos de vacunación de la hepatitis B. Métodos Se utilizaron datos de 211 643 niños de encuestas demográficas y de salud en 47 países con ingresos bajos y medios (año promedio de estudio 2012). La información provenía de las tarjetas de vacunación y de entrevistas a las madres. Se agruparon los países según el calendario de vacunación y el tipo de vacuna utilizada (monovalente o combinada). Para cada país, se calculó una cobertura de vacunación contra la hepatitis B y la recepción oportuna de las dosis de la vacuna. Se utilizaron modelos de regresión logística multivariable para estudiar el efecto de los calendarios de vacunación y los tipos en el retraso en la administración de vacunas.
Resultados Se observaron grandes retrasos en la vacunación, incluso en países con una cobertura bastante alta de todas las dosis. El retraso medio era de 1,0 semanas (rango intercuartílico, ICR: 0,3 a 3,6) para la primera dosis (n = 108 626 niños) y de 3,7 semanas (ICR: 1,4 a 9,3) para la tercera dosis (n = 101 542). Se observó una tendencia de menores probabilidades de retraso en los calendarios de vacunación que empezaban a las 6 y 9 semanas de edad. Para la primera dosis de la vacuna, se registraron menos probabilidades de retraso para las vacunas combinadas que para las monovalentes (coeficiente de posibilidades ajustado: 0,76, intervalo de confianza (IC) del 95%: 0,71 a 0,81). Conclusión Las grandes diferencias en la cobertura de vacunación contra la hepatitis B y la adherencia a los calendarios de vacunación entre países destacan la continua necesidad de mejorar los sistemas nacionales de inmunización. La iniciación oportuna del proceso de vacunación puede dar lugar a la recepción oportuna de dosis sucesivas y a la mejora de la cobertura general. Sugerimos la incorporación de la fecha de vacunación como un indicador de rendimiento de los programas de vacunación para completar el cálculo de la cobertura.  Notes: We examined data quality for all children covered by the surveys. Vaccination dates were counted as invalid if day, month or year were missing, or if the date was implausible, e.g. before the date of birth of the child or after the date of mother's interview or with erroneous dates (e.g. as year 9998). We only considered vaccination cards as available if seen by the interviewer. Excluded surveys: Ethiopia (non-standard date recording), Indonesia (date of birth not available), Morocco (only first dose reported), Nepal (non-standard date recording), Nicaragua (key missing variables, e.g. wealth index), Philippines (date of birth not available),and Turkey (date of birth not available). Countries that altered their national immunization schedules within 5 years of the survey were: Armenia (pentavalent introduced in 2009), Gabon (pentavalent introduced in 2010), Kyrgyzstan (pentavalent introduced in 2009) and Tajikistan (pentavalent introduced in 2008-09). Hence, we adopted the previous immunization schedule for these nations in our analysis. For Cambodia and Colombia, and the United Republic of Tanzania, data on multiple vaccine types (monovalent and combination) were reported. We based our estimates on monovalent vaccination in Colombia, pentavalent in Cambodia and tetravalent in the United Republic of Tanzania. The decision was based on schedules (vaccines) reported in the relevant annual UNICEF/WHO immunization reports and the available data sets.   (. . .continued)