Bulletin of the World Health Organization

Varicella and herpes zoster hospitalizations before and after implementation of one-dose varicella vaccination in Australia: an ecological study

Anita E Heywood a, Han Wang b, Kristine K Macartney b & Peter McIntyre b

a. School of Public Health and Community Medicine, University of New South Wales, Level 3, Samuels Building, Botany Road, Kensington, NSW 2052, Australia.
b. National Centre for Immunisation Research and Surveillance, The Children’s Hospital at Westmead, Sydney, Australia.

Correspondence to Anita E Heywood (email: a.heywood@unsw.edu.au).

(Submitted: 23 October 2013 – Revised version received: 20 February 2014 – Accepted: 05 March 2014 – Published online: 13 June 2014.)

Bulletin of the World Health Organization 2014;92:593-604. doi: http://dx.doi.org/10.2471/BLT.13.132142

Introduction

Prior to the introduction of varicella vaccination in Australia, primary infection with the varicella-zoster virus (VZV) was a common childhood disease, with the majority (88%) of the population experiencing infection by adolescence.1 Although the varicella vaccine has been available since 1995, few countries have recommended universal childhood vaccination, and even fewer have implemented publicly funded national varicella vaccination programmes.2 Most of the data on the impact of varicella vaccination come from studies conducted in the United States of America (USA), where, since 1996, vaccination has been recommended for children older than 12 months. However, vaccine uptake was slow, with one-dose coverage at 19–35 months of age not reaching more than 80% until 2002.3 During the USA one-dose era, significant declines in varicella ambulatory visits, hospitalizations and deaths were documented, including non-targeted age groups, consistent with a herd immunity effect.4,5 Nonetheless, continued disease transmission and outbreaks in highly vaccinated populations prompted the move from a one-dose to a two-dose schedule in both Germany6 and the USA.3

The World Health Organization has requested evidence for the impact of varicella and herpes zoster (HZ, a reactivation of latent VZV) vaccination programmes from countries with robust data to use for developing evidence-based recommendations.7 To date, evidence of programme impact on varicella zoster disease outside the USA is predominantly regional or limited to small population samples; no data are available for subpopulations with higher incidences. Additionally, increases in HZ are hypothesized to occur from a reduction in natural immunological boosting in previously infected individuals. However, data are limited and inconclusive.

Vaccines listed on Australia’s National Immunization Programme schedule are fully funded at a national level for eligible age groups, with programme delivery managed by each state and territory.8,9 The experience that followed the availability of varicella vaccines in Australia (Table 1) is unique in several aspects. Both licensed vaccines, Varivax® and Varilrix®, have been available since 2000, with Varilrix® used almost exclusively in the National Immunization Programme since November 2005. Furthermore the programme includes one-dose routine vaccination at 18 months of age and a single catch-up dose delivered via Australia’s school-based immunization programme at 12–13 years. Data on childhood vaccination coverage are available from the Australian Childhood Immunization Register and complete national hospitalization data are available by age for Aboriginal and Torres Strait Islanders and non-indigenous people. It is known that indigenous Australians have higher varicella hospitalization rates compared to non-indigenous Australians. These higher rates are possibly related to poorer access to primary care, particularly in remote areas, and/or higher rates of skin/soft tissue complications related to environmental living conditions.11

Using national data records on varicella immunization and hospitalization, we evaluated age-related trends in both varicella and HZ hospitalizations during periods of differing varicella vaccine coverage. We also aimed to assess the outcome of the immunization programme on varicella hospitalizations in Australia’s indigenous population.

Methods

Data sources

The Australian Childhood Immunization Register records vaccines given to all Medicare-enrolled children younger than seven years of age,12 which includes 99% of Australia’s annual births of approximately 300 000 children.13 Data on the proportion of eligible children who received the varicella vaccine were obtained quarterly, including indigenous status. Coverage data at 24 months of age were available for children born between October 2001 and September 2009. To determine timeliness of vaccination, we also assessed coverage at 60 months of age for children born between January 2003 and June 2008. State and territory summary data on the proportion of school enrolments vaccinated in the adolescent school-based catch-up programme in 2009 were provided by jurisdiction health departments. The school-based catch-up was conducted at 12 to 13 years of age. For these adolescents, parental reports of previous natural infection or vaccination are accepted as valid reasons for non-vaccination in the school-based vaccination programme. However, parental report data were not available for analysis.14

National, de-identified demographic and diagnostic data for individual hospitalizations (private and public hospitals) for varicella and HZ were obtained from the National Hospital Morbidity database for the period from July 1998 to June 2010.15 All episodes coded as varicella or its complications (codes B01-B01.9 in the International Statistical Classification of Diseases, 10th Revision, Australian Modification, ICD-10-AM) and HZ or its complications (codes B02-B02.9) in the principal or any diagnostic fields were obtained. Dual hospitalization coding for varicella and HZ were excluded (0.6% of all VZV-coded hospitalizations). Mid-year population estimates by age and indigenous status were obtained from the Australian Bureau of Statistics.13

We conducted a search in the Medline database for studies assessing the impact of National Immunization Programmes using the search terms “varicella” or “zoster” and “hospitalizations” and “vaccination” or “immunization”. The search was limited to English language articles published or available online from 1 January 1996 to 1 March 2013. The initial search identified 101 potential studies on varicella hospitalizations and 53 potential studies on HZ hospitalizations. After excluding studies that did not report average annual hospitalization rates for pre- and post-programme periods, eight varicella and three HZ studies were included in the review.

Data analysis

Annual crude and age-specific hospitalization rates for varicella and HZ in the population were calculated by Australian financial year of hospital discharge (1 July to 30 June). Average hospitalization rates for the periods of vaccine availability (Table 1) were calculated by date of hospital admission to reflect changes in vaccine availability. Reported rates refer to principal hospitalizations, unless otherwise stated. To account for temporal changes in population age structure, age-adjusted hospitalization rates were calculated, using direct standardization to the population of the publicly funded period when comparing vaccine availability periods and 2009/2010 financial year population when comparing annual changes.13 Analysis of hospitalization rates for Aboriginal and Torres Strait Islander peoples was restricted to four jurisdictions (Western Australia, South Australia, Queensland and the Northern Territory) due to known incomplete hospital records of indigenous status in other jurisdictions in earlier years.11 Indigenous Australians represent 2.4% of the total Australian population, with 60.1% residing in these four states and territories.13,16 Poisson regression was used to analyse yearly trends in crude, age-standardized and age-specific hospitalizations and to calculate average annual percentage change and accompanying P-values. Hospitalizations averted were calculated by applying pre-vaccination hospitalization rates to the funded immunization programme population. Ninety-five per cent confidence intervals (CIs) were calculated using the Poisson distribution for hospitalization counts and the log transformation method was used to obtain incidence rate ratios (IRR). Analysis was undertaken using SAS version 9.2 (SAS Institute, Cary, USA).

Results

Vaccination coverage

Before the immunization programme funding was introduced, the varicella vaccination coverage increased slowly (Fig. 1). According to the Australian Childhood Immunization Register, 9.7% of children aged 24 months had been vaccinated before the unfunded recommendations in late 2003. By the end of 2005, the second year of the unfunded recommendations, 20.9% vaccination coverage was achieved. After immunization programme funding, coverage at 24 months of age increased to 74.8% within the first year and exceeded 80% 2.5 years into the programme in 2008 (Fig. 1).

Fig. 1. Annual varicella hospitalization rates (principal diagnosis) for children aged 18–59 months and varicella vaccination coverage in children aged 24 and 60 months, Australia, July 1998 to June 2010
Fig. 1. Annual varicella hospitalization rates (principal diagnosis) for children aged 18–59 months and varicella vaccination coverage in children aged 24 and 60 months, Australia, July 1998 to June 2010
a Australian financial year from 1 July to 30 June.
Note: Coverage for 24 and 60 months are assessed for the same cohort of children, e.g. for the cohort born January–June 2002, 24-month coverage was assessed at January–June 2004 (12.1%) and 60-month coverage at January–June 2007 (21.0%).

Vaccine coverage at 60 months of age was consistently higher, reaching 90.0% by the end of 2012. Among the indigenous population, coverage was 1.2% and 5.4% lower at the age of 24 months in 2003 and 2005, respectively, compared with the non-indigenous population before the immunization programme funding. However, after the funding, vaccination increased rapidly to an estimated 81.9% of indigenous children aged 24 months by the end of 2010 and 91.6% of indigenous children aged 60 months by the end of 2011. School-based vaccination programmes had variable one-dose coverage across jurisdictions, ranging from 19% to 42% of enrolled students in 2009, which was similar to previous years.14

Varicella hospitalizations

Between 1 July 1998 and 30 June 2010, 16 261 varicella hospitalizations were recorded for all ages, including 10 632 (65.4%) principally coded as varicella. The annual crude principal hospitalization rates declined by an average of 21.4% (95% CI: 17.8–25.0) per year following the immunization programme funding. During the funded period, the age-standardized varicella hospitalization rate was 49.6% (95% CI: 47.3–51.9) lower than the pre-vaccine period and 40.3% (95% CI: 37.4–43.1) lower than in the recommended, unfunded period (Table 2). Significant average annual declines in hospitalization rates were observed for all age groups under 40 years (P < 0.001). The greatest reduction in rates was detected in children from 1 to 4 years old, with 72.5% (95% CI: 68.8–75.7) lower rates during the funded immunization programme period, compared with the pre-vaccine period and 58.6% (95% CI: 52.9–63.6) lower compared with the recommended, unfunded period. Notably, the age group of 1 to 4 years was the only one to show a statistically significant decline during the recommended, unfunded period (Table 2).

For the age group specifically targeted under the funded immunization programme, i.e. children aged 18–59 months (Fig. 1), the average hospitalization rate during the programme was a quarter of the average pre-vaccine rate (IRR 0.25, 95% CI: 0.22–0.29). For infants (under one year), not eligible for vaccination, hospitalization rates during the funded immunization programme were significantly lower than during the pre-vaccine and recommended, unfunded periods, 62.1% (95% CI: 54.7–68.3) and 53.3% (95% CI: 44.5–60.7), respectively. When applying pre-vaccine hospitalization rates to the funded period population, an estimated 686 varicella hospitalizations, including 369 in children younger than five years, were annually averted in Australia following the immunization programme funding.

Over the study period, there were 560 varicella hospitalizations, 333 (59.5%) of which were with a principal diagnosis, and where the patient was recorded as being Aboriginal and/or Torres Strait Islander. Prior to vaccine availability, indigenous Australians were hospitalized at a minimum of twice the rate of the non-indigenous population (IRR 2.6, 95% CI: 2.0–3.2). Following the funded immunization programme, overall hospitalization rates remained higher for the indigenous people (Fig. 2).

Fig. 2. Varicella hospitalization ratesa (principal diagnosis) and incidence rate ratios for indigenous and non-indigenous population by varicella vaccine availability and age group in four Australian jurisdictions,b July 1998 to June 2010
Fig. 2. Varicella hospitalization rates<sup>a</sup> (principal diagnosis) and incidence rate ratios for indigenous and non-indigenous population by varicella vaccine availability and age group in four Australian jurisdictions,<sup>b</sup> July 1998 to June 2010
a Average annual hospitalization rate per 100 000 population.
b Western Australia, South Australia, Northern Territory and Queensland.

For indigenous children aged 0–4 years, the hospitalization rates were also higher during the pre-vaccine period compared with non-indigenous children (IRR: 1.9; 95% CI: 1.4–2.7). However, the rate declined from 71.8 (95% CI: 51.1–98.2) per 100 000 population in the pre-vaccine period to 16.6 (95% CI: 11.1–24.1) per 100 000 population in the immunization programme period (IRR: 0.23; 95% CI: 0.14–0.38), reaching similar rates for both groups of children (IRR: 1.1; 95% CI: 0.7–1.6) (Fig. 2).

Herpes zoster hospitalizations

Between 1 July 1998 and 30 June 2010, 59 660 hospital episodes were coded as HZ, and 25 198 (42.2%) as the principal diagnosis. Crude principal HZ hospitalization rates increased by an average of 0.53% (95% CI: 0.18–0.89) per year. However, when rates were age-standardized, HZ hospitalization declined at an average of 0.57% (95% CI: 0.24–0.91%) per year (Fig. 3). Temporal changes for all HZ-coded hospitalizations were similar to principal HZ hospitalizations (Fig. 3). Compared with earlier time periods, age-standardized HZ rates were significantly lower during immunization programme funding and age-specific principal HZ hospitalization rates remained stable or lower (Table 3).

Fig. 3. Annual crude and age-standardized hospitalization rates for herpes zoster, Australia, 1998 to 2010
Fig. 3. Annual crude and age-standardized hospitalization rates for herpes zoster, Australia, 1998 to 2010
a Australian financial year from 1 July to 30 June.

Discussion

Australia is one of the few countries that has included varicella vaccination under its national immunization programme, distinctively funding one-dose routine childhood vaccination and an adolescent catch-up programme. Many European countries – despite a European consensus recommendation – have not introduced universal varicella vaccination.17 For example, the United Kingdom of Great Britain and Northern Ireland recently decided to fund an HZ immunization programme, but rejected a population-based varicella vaccination based on concerns about breakthrough varicella, a potential shift in varicella to older age groups and potential increases in HZ.18 However, we found no evidence for a shift in varicella hospitalization to older age groups in Australia after the implementation of universal varicella vaccination. Instead, we observed significant reductions in varicella hospitalizations for persons younger than 40 years. In line with our results, studies from other countries with universal varicella vaccination have shown marked declines in varicella hospitalizations in both targeted age groups and indirect effects to non-targeted age groups (Table 4). We also observed that the pre-existing twofold disparity in varicella hospitalization between indigenous and non-indigenous children was eliminated under the funded immunization programme. Demonstration of this broad beneficial outcome across population groups in Australia suggests that our experience is likely to be applicable to other countries with disparities in varicella disease.

Our data show evidence of herd immunity during the funded immunization programme. In Australia, similar to other countries, data on adult varicella vaccination coverage is not available. However, it is presumed to be low and would not account for the decline in hospitalizations observed in adults younger than 40 years of age. Our results show more than a 60% decline in hospitalizations for infants – not eligible for vaccination – which can only be attributed to herd immunity effects. This finding is consistent with declines in neonatal (68%) and congenital (79%) varicella from enhanced surveillance both in Australia26 and in the USA.22 Furthermore, concerns regarding increased risk of infection during pregnancy18 are not supported by our results, which show a decline in varicella hospitalizations in the 20–39 year age group which include women of child-bearing age. Furthermore, data from the USA one-dose programme show declines in hospitalisations across all age groups and no upward age shift in varicella hospitalizations.4,21,27,28 Overall, available data provide evidence that pregnant women and their infants benefitted from a universal childhood varicella vaccination programme due to herd immunity effects.

It was estimated that an Australian varicella immunization programme would directly avert 450 hospitalizations annually, saving up to 21 532 Australian dollars per hospitalization averted.29 Our results exceeded that prediction, with an average of 686 hospitalizations prevented annually in the first four and a half years of the funded programme. A re-evaluation of programme cost-effectiveness, including herd immunity effects, is therefore warranted. Funded catch-up vaccination is likely to be central to reducing the pool of susceptible adolescents and adults for whom varicella disease outcomes are more severe.30 However, few countries that have implemented a universal varicella vaccination programme have also implemented formal catch-up programmes. In our adolescent programme, the uptake matched our expectations, based on varicella seroprevalence data indicating more than 80% seropositivity in individuals aged 10–14 years before the vaccine was available.1,14 We also show incremental increases in coverage between 24 and 60 months of age, which indicate lack of timeliness and suggest uptake may be prompted by vaccination requirements at the time of school or childcare entry. The importance of vaccine availability for older children has been emphasized in progressively expansive catch-up recommendations in the USA.3 Similar to data from Ontario, Canada,23 we observed that a low universal vaccination coverage during the recommended but unfunded period had a minimal effect on the reported disease burden which was limited to the target age group. Accomplishing a rapid, high vaccination coverage and mechanisms for catch-up appear to be important components of a successful one-dose funded universal immunization programme.

Modelling studies have predicted higher HZ incidence among unvaccinated previously infected individuals (based on absence of immune boosting from exposure to circulating varicella zoster virus).30 Instead, we show that age-adjusted and age-specific HZ hospitalization rates did not increase over time, despite a high varicella vaccination coverage in Australia for almost five years. Our coverage approached the 90% coverage included in studies modelling the impact of varicella vaccination on the epidemiology of varicella and HZ. This is the scenario that is included in studies modelling the impact of varicella vaccination on the epidemiology of varicella and HZ.3032 Some epidemiological studies have reported temporal increases in the crude HZ hospitalization rates.19,33 However, only two other studies have assessed trends in age-standardized HZ rates before and after vaccine introduction, including hospitalizations and health care utilization, and showed no temporal increase in HZ (Table 5).35,36

Increasing age is the greatest risk factor for VZV reactivation, due to age-related decline in cellular immunity, high prevalence of chronic disease and use of immune-compromising medication.37 HZ hospitalization rates in persons older than 80 years are more than twice that of persons aged 70–79 years. Over the study period, the percentage of the Australian population over 80 years of age increased from 2.8% to 3.7%.13 Statistical adjustments for ageing populations are required in epidemiological studies to adequately determine temporal changes in HZ. A combined childhood varicella and older adult HZ vaccination programme is a potential comprehensive strategy for the prevention of VZV disease in the entire population, and vaccination against HZ has been recommended for Australians aged over 60 years.8 However, due to manufacturer supply issues, virtually no HZ vaccine has been available and there is currently no funding from the national immunization programme.

The strengths of our study include the use of comprehensive national population-based databases: 12 years of hospital admissions data, not limited by under-reporting, sampling or regional differences; and national vaccine coverage data where under-reporting for immunization programme-funded vaccines is minimal.12 Although several studies4,1921,2325,27,28 have demonstrated the early effect of a universal varicella immunization programme (Table 4), we included national data assessing both population-adjusted varicella and HZ over the pre- and post-programme periods. We used principal-coded hospitalizations, which potentially underestimates the total hospitalized disease burden, but reduces reporting of incidental hospitalizations and miscoding and is therefore likely to be a more accurate method than the use of all HZ-related hospitalizations. Although this is an ecological study, there is no evidence that other factors that could affect hospitalization rates have changed over time, such as better access to health care or changes in hospital admissions or coding practices. The high positive predictive value (95.7%) for varicella coding demonstrated in a hospitalized Australian paediatric population38 supports the robustness of using varicella-coded data. However, varicella hospitalization data in older adults may be limited by miscoding of HZ.35 The majority of HZ-related hospitalizations are likely to have been complicated by complex co-morbidities, particularly in the frail elderly and/or immunocompromised populations and may not reflect the principal cause for admission.39

Our study is limited to the inclusion of VZV infection requiring hospitalization. Varicella results in severe morbidity in only a minority of cases, for which hospitalization is a proxy measure. While the risk of complications is greater in adults and children with immunocompromising conditions, the highest absolute numbers of varicella hospitalizations are in otherwise healthy children.38 Approximately 2% of cases in children younger than two years require hospitalization.40 Hospitalization rates for HZ are two to four times the rate for varicella.41 VZV-related disease also significantly impacts health care utilization at the primary care level, but this was not assessed in our study. Assessing trends in non-hospitalized VZV disease will become more important as Australia’s varicella immunization programme matures, including monitoring outbreaks and breakthrough varicella.

Although one-dose programmes have been effective in preventing severe varicella disease, as further confirmed by our study, evidence suggests that a two-dose schedule is required to interrupt virus transmission. Ongoing school outbreaks and high rates of breakthrough varicella, although usually mild, have prompted some countries to implement a two-dose schedule.3,6 A submission to fund two-dose varicella vaccination under the Australian National Immunization Programme was rejected in 2008 due to uncertainty regarding the incremental cost-effectiveness of the second dose.42 However, emerging data on further declines during the two-dose programme in the USA43,44 and recent evidence indicating that breakthrough varicella was almost seven times less likely to occur in two- compared with one-dose vaccine recipients45 provide empirical evidence of the potential benefits of a two-dose schedule.

This study is a comprehensive analysis of national Australian population-based data comparing both varicella and HZ hospitalizations during periods of varicella vaccine availability, using robust national vaccine coverage data and the largest study reporting experience with Varilrix®. There are several differences in the approach to implementing varicella vaccination programmes internationally, including the age at vaccination, one or two-dose schedules, and inclusion of catch-up vaccination.

Australia’s experience with a one-dose funded varicella vaccination programme with rapidly attained high coverage is relevant to countries considering a universal programme. The beneficial outcome of the vaccination programme is expected to increase as the programme matures and re-examination of the cost-effectiveness of incorporating a second dose may be warranted over time.


Acknowledgements

The authors thank Brynley Hull for providing the coverage data from the Australian Childhood Immunization Register.

Funding:

The National Centre for Immunisation Research and Surveillance is supported by the Australian Government Department of Health and Ageing, the New South Wales Department of Health, and The Children’s Hospital, Westmead, NSW, Australia. No funding from external sources was received in relation to the writing of this article.

Competing interests:

AEH has received funding to conduct investigator-driven research from GlaxoSmithKline and Sanofi Pasteur. The other authors declare no conflicts of interest.

References

Share