https://orcid.org/0000-0001-5473-8930
Abstract
Declining childhood immunization represents a serious public health problem globally and in New Zealand. To guide efforts to increase immunization coverage, this study monitors nationwide change in immunization coverage since the introduction of the National Immunisation Register (NIR) in 2005 and spatiotemporal patterns of immunization coverage from 2006 to 2017.
The study population consisted of 4 482 499 individual immunization records that were obtained from the NIR (2005–2017). Data on yearly and average immunization coverage in census area units (CAUs) in New Zealand were calculated by milestone age (6/8/12/18/24/60/144 months). Data for 2005 were excluded due to missing records in the introductory period of the NIR. We analyzed spatial and spatiotemporal patterns using Gi* and SaTScan methods.
Immunization coverage improved since the introduction of the NIR in 2005, reaching a peak in 2014 and 2015 with a slight decrease in 2016 and 2017. Well and insufficiently immunized areas were identified with spatial autocorrelation analyses highlighting several hot- and cold-spots. Comparison of CAUs with neighbouring CAUs allowed for the identification of places where immunization coverage was significantly higher or lower than expected, over both time and space.
We provide the first spatiotemporal analysis of childhood immunization in New Zealand that utilizes a large sample of over 4.4 million individual immunization records. Our spatial analyses enable policymakers to understand the development of childhood immunization coverage and make more effective prevention strategies in New Zealand.
Introduction
Universal vaccination programmes that increase immunization coverage have reduced the burden of infectious diseases in both developing and developed countries1 and the benefits of childhood vaccinations are now scientifically unquestioned.2 However, while immunization coverage has improved in many parts of the world, recent outbreaks serve to remind us just how these gains can be lost.3 Over the course of 2018, four of the six World Health Organization regions had substantial measles outbreaks, and measles was once again endemic in every region of the world.3 Declining immunization coverage was said to be related to supply shortages and growing vaccine hesitancy.4 In another example, the United Kingdom has also seen a decrease in coverage of the measles-mumps-rubella vaccine to 91.2%, the fourth annual decline in a row and to its lowest level since 2011–12.5 Collectively, this evidence highlights the significant and urgent need for surveillance of immunization coverage in developed countries such as New Zealand.
Research on immunization often focuses on attitudes towards immunization,6 surveillance and outbreaks events7,8 or uses survey-based methods.9 Several pioneering global epidemiological studies have painted a picture of immunization coverage with a broad brush; however, these studies often focus on developing countries and rarely depict detailed patterns that show within country or region variability which are important in developed countries such as New Zealand to inform research and policy.2,10,11 Indeed, several highly powered studies acknowledge that immunization coverage is likely to show variability within the larger geographical areas that were analyzed.10,12,13 Data at fine geographical scales and analyses that examine change over time are required to fully understand patterns of local immunization coverage. Understanding where and why immunization coverage may be declining over time and space is important as these areas can be at risk of losing their herd immunity.
Emerging evidence utilizing geospatial analyses have shown clusters of low immunization coverage.14–17 In one of the only spatiotemporal analyses investigating vaccine hesitancy, the largest increases were shown to have occurred in a relatively small proportion of regions throughout the state of California.11 Despite this emerging evidence, few studies to the authors knowledge have examined such spatiotemporal complexities using nationwide registry data.11 We extend evidence by using National Immunisation Register (NIR) data from 2006 to 2017 at a fine geographical scale [i.e. census area unit (CAU)] in a sample of over 4.4 million childhood immunization records from across New Zealand. This study investigates how immunization coverage has developed in New Zealand since the introduction of the NIR in 2005. It then examines nationwide spatial and spatiotemporal trends and patterns of immunization coverage.
Methods
Data
The New Zealand Ministry of Health provided the NIR for the years 2005–2017 at the meshblock geographical level (with identifiers of higher administrative units), including information about sex, prioritized ethnicity, milestone age in months, milestone year and milestone month, and census year. The full data consisted of 4 362 674 records representing an aggregation of 4 482 499 individual immunization events. From those, only events of fully immunized children were extracted (3 341 715 records of 3 438 236 events). Data from 2005 were excluded from the analysis due to numerous missing records. All analyses were performed at the level of CAUs geographies, based on the information about residency meshblock of the immunized children. CAUs are census administrative areas that, in urban areas, contain 3000–5000 people (2004 CAUs in 2013). The birth data in CAUs were provided by Statistics New Zealand and included the number of live-born children in CAUs between 2005 and 2018. These counts were rounded to the base of three to ensure the confidentiality of individuals before providing the official data.
Calculation of rates and trends of immunization coverage development over time
Counts of immunization events and counts of live-born children were aggregated in CAUs. The immunization rates for each milestone age (6/8/12/18/24/60/144 months) and year (2006–17) were calculated as a percentage of fully immunized children and the total amount of children of applicable milestone age for each area. The month of immunization and month of birth were considered to calculate the milestone age-specific immunization since.
Spatial and spatiotemporal trends and patterns in immunization coverage
The analysis of spatial and spatiotemporal autocorrelation enables the identification of areas that perform better or worse than their local neighbourhoods. This study investigated spatial and spatiotemporal trends and patterns of milestone age-specific immunization coverage around New Zealand between 2006 and 2017. We used local Gi*18 to explore local spatial autocorrelation in the data. It is used as an indicator of local clustering that measures a concentration of spatial data19 allowing for subsequent visualization of the location of identified clusters.18 The results distinguish between statistically significant clusters (P values < 0.05) of high and low immunization rates. We used the second-order queen’s contiguity scheme of the neighbourhood to obtain spatial weights for individual CAUs.20 GeoDA software package21 was used for the processing of the data and computing of the measures, while QGIS 3.222 was used for geovisualization of the clusters.
Spatiotemporal scan statistics were then used to identify clusters of high and low rate areas simultaneously over space and time. This was processed using the open-source software SaTScan 9.3.23 QGIS 3.2 was used for geovisualization of clusters. Input data consisted of: (1) immunization events aggregated into CAUs and grouped by year of the event; (2) demographic structure of CAUs represented by aggregated counts of births in CAUs; and (3) coordinates of CAU centroids. The retrospective space-time analysis of high and low rate clusters was based on input data applying the Discrete Poisson model.24 SaTScan was set to find clusters within a dynamic circular window including up to 10% of the population, while the maximum time span of the cluster was set to 90% of the study period. These values were selected due to focusing on rather small areas appearing long-term in time while trying to eliminate areas with an early deployment of NIR. We adjusted for the time by non-parametric stratified randomization to ensure the comparability of rates within various time periods.25 Resulting indirectly standardized rates (expressed as the relative risk) for each identified geographic cluster were estimated, and only significant clusters (P values ≤ 0.001) remained in the results. The clusters consisting of only one CAU, as well as clusters appearing for a period of the first reliable year (2006 for 6/8/12-months milestone age, 2007 for 18/24-months milestone age and 2010 for 60-months milestone age), were not considered due to the early introduction period of the NIR.
Results
Mapping of immunization coverage and exploratory analysis using spatial autocorrelation
Animated maps depicting the immunization coverage for each milestone age can be found in the supplementary online materials (Supplementary figure 1). In the years following the introductory period of the NIR (2005–06), the overall immunization coverage grew, more in the South Island than the North Island, until reaching a peak in the years 2014 and 2015. A slight decrease in immunization rates is visible since then. Immunization rates are generally lower for milestone ages of 6, 8 and 18 months than for ‘complete years’ milestone ages of 12, 24 and 60 months. However, the immunization rates at eight-month milestone age recorded the most rapid development in terms of improvement levelling the 12-month milestone age immunization uptake during recent years.
Figure 1 shows clusters of CAUs with an average immunization coverage significantly different from their neighbourhood CAUs. Clusters of cold-spots of low immunization coverage (purple shade) can be found within the major population centres of Auckland, Christchurch and Wellington regardless of the milestone age that is assessed. Cold-spots are the most notable for the immunization coverage at six-months milestone age with large cold-spots identifiable in Northland and through the central North Island. This observation was also found for 18-month milestone age. The persisting cold-spots are not age specific, except six-months milestone age located in the north-west coast region. Most hot-spots (light green shade) are found in rural areas of the central South Island. Both hot- and cold-spots are generally smaller for 8/12/24/60-months milestone age immunization coverage than for 6- and 18-months milestone age.
Hot- and cold-spots of average immunisation coverage by milestone age
Spatiotemporal patterns
The following set of geovisualizations (figure 2) represent spatiotemporal clusters of CAUs. Resulting significant clusters (P values ≤ 0.001) indicate higher (positive) or lower (negative) levels of immunization coverage when compared with neighbouring CAUs. The geovisualizations within figures 2 locate the cluster on the map, while Supplementary tables 1–6 in the supplementary material provide information on the start and end years of individual clusters, number of CAUs included, number of expected immunization events and observed immunization events, and the type of the cluster. The colours of the column match the colours in the maps.
Geovisualisation of spatiotemporal clusters of immunisation coverage in New Zealand
Spatiotemporal clusters of CAUs with low immunization coverage were generally identified in densely populated areas of big cities such as Auckland, Christchurch and Wellington as well as in the central North Island. A high number of clusters identified by individual analyses of milestone age-specific immunization coverage indicates considerable spatial variation in the case of immunization at 6, 8 and 18-months when compared with immunization coverage at 12, 24 and 60-months (figure 2, Supplementary figure 2 shows animated full resolution maps). Clusters of low immunization coverage were not found elsewhere in the South Island other than in Christchurch urban area regardless of the milestone age. All maps and tables showing and describing spatiotemporal clusters are located in the supplementary online material.
Main urban centres of Auckland, Christchurch and Wellington
Many of the spatiotemporal clusters, of both low and high immunized areas, appeared in Auckland and Counties Manukau. However, most of these clusters are historical since they had lasted for a certain time and ended earlier than 2017 (Supplementary tables 1–6). This means local differences of immunization coverage in the area slowly disappeared. The split between the eastern coast (higher immunization rates) and western coast (lower immunization rates) is evident around Auckland in the case of 6, 8 and 18-months immunization coverage. Only two historical clusters of low immunization coverage were identified at 12 and 24-months milestone age. However, three persisting bigger clusters emerged on the west coast when looking at 60-months milestone age.
No clusters were detected in Wellington for immunization coverage at 12-months and 60-months milestone age. However, a temporary cluster (2006/2007–15) of low immunization covering in the central city appeared for the remaining milestone ages of 6/8/18 months. Cluster 14 was classified as persisting in case of immunization coverage at 24-month milestone age (figure 2, Supplementary table 5). A similar situation was found in Christchurch with differences being a pair of clusters instead of a single cluster. All clusters in Christchurch were temporary. The clusters of high immunization rates were regularly detected in semi-rural areas south of Christchurch.
Other areas
A large persisting cluster of low immunization coverage at six-month milestone age was located in Northland. However, no clusters were identified for other milestone ages. The central North Island seems to be the region where clusters of low immunization coverage were detected most often, they were also the largest ones. Although most of the clusters (regardless of the milestone age) lasted only for a limited time, the Bay of Plenty area seems to underperform in long-term immunization coverage at 6, 18 and 60-months milestone age. The situation is analogous in the case of Lakes District Health Board in the central North Island where a persisting cluster of low immunization coverage was identified for 6-, 24- and 60-months milestone age (figure 2).
Clusters were often identified in urban areas of Hamilton and Tauranga. Mostly clusters of high immunization coverage persisting over time were detected in Tauranga for milestone ages of 18-, 24- and 60-months. The situation in Hamilton (and its vicinity) seems to be more complex. Both types of clusters assessed as persisting were located there. However, clusters of CAUs with low immunization coverage generally appeared in the central urban area, while high immunization clusters were located in rather suburban areas around the outskirts.
Discussion
We provide the first nationwide spatiotemporal investigation of childhood immunization in a large sample of over four million individual immunization records from 2005 to 2017 across New Zealand at fine geographical scale. We respond to calls to monitor change in immunization over time26 and extend knowledge by showing spatiotemporal changes in immunization coverage. This study shows that since the introduction of the NIR in New Zealand, immunization coverage has improved significantly reaching a peak in the years 2014 and 2015. However, a slight decrease in immunization coverage is visible from 2015 to 2017. This study also demonstrates several areas of spatial autocorrelation. For example, several cold-spots where there are clusters of CAUs with lower immunization coverage than in their neighbouring CAUs were identified in the major population centres of Auckland, Christchurch and Wellington regardless of the milestone age. Finally, this study contributes significantly to evidence by highlighting how spatial clusters of high and low immunization change over time. We provide specific evidence for policy by highlighting the start and the end of individual clusters, number of CAUs included and the number of expected immunization events and observed immunization events.
Emerging evidence suggests that areas of low immunization coverage cluster geographically.17,27,28 For instance, US evidence from California showed that of 50 233 children in the study population evaluated (2010–12), 10 144 (21%) were identified as being within a cluster of low immunization.17 Despite this, previous evidence is often restricted by smaller sample sizes,29 coarse geographic scales,26 limited geographical extents,30 limited temporal scope31 and often by not simultaneously investigating both spatial and temporal complexity culminating in spatiotemporal trends. Our nationwide and spatiotemporal findings over a decade provide important evidence that help identify clusters of immunization over time. Findings also help policymakers understand the development of childhood immunization in New Zealand and help them to focus future efforts to increase vaccination coverage in specific geographical areas. This is important as previous evidence shows that low immunization is associated with elevated risk of various outbreaks including measles and pertussis.32–34
Child wellbeing and better population health outcomes, supported by a strong and equitable public health system, are key priorities for the New Zealand Ministry of Health in 2019–20.35 In order to evaluate progress on these priorities, empirical evidence detailing each stage of childhood is needed. Thus, by considering temporal and spatial patterns at different stages of children’s lives, we are able to provide a comprehensive picture of child wellbeing and inform policy directed towards government priorities. Our findings assist both the Ministry of Health and regional district health boards in directing resources to either geographic areas or specific populations to improve health outcomes for children overall.
Our retrospective analyses help to better understand national and regional evaluation of the efficacy of immunization campaigns and outreach activities. Findings also enable improved area-specific understandings of policy and resource allocation. However, recent evidence suggests existing structural, economic, cultural and other factors are possible barriers to health care access36 and immunization.37 Hence, findings should be paired with information on the socioeconomic and demographic structure of clusters to identify not only places where it may be most appropriate to intervene, but also an appropriate intervention strategy (e.g. cultural). Locations of well-immunized child populations (hot-spots) indicate areas of good health outcomes for children. Consequently, such areas put less pressure on the health system due to healthier populations that are resilient to outbreaks of preventable diseases.
Our findings should be interpreted in light of study limitations. While NIR data were obtained from 2005–17, the first years (2005 and 2006) were often not usable due missing data. This reflects a real-world scenario where the introduction of a reporting system such as the NIR needs time to be implemented. While data were harmonized to reflect the change of census geographies, the milestone ages of immunization schedules are not always round years. This means the number of children born in CAUs needed to be recalculated in order to comply with reporting of immunization events in CAUs. For instance, children born in July or later are usually fully immunized during the following calendar year in the milestone age of six months. In addition to this, data used for the estimation of immunization rates did not account for people’s mobility and therefore did not consider if a family moved to another CAU. When interpreting the results of the cluster analysis, it is necessary to realize that both methods, spatial and spatiotemporal scanning, provide estimates of local hot-spots and cold-spots. However, this does not have to be the case when compared to universal values such as a national target or national average.
Moreover, this report analysed records of fully immunized children only. It also used rates computed for CAUs based on the NIR and live births provided by Statistics New Zealand. The estimates of immunization rates in less populated CAUs may therefore be less precise due to the confidentiality of data. For instance, in the case of the birth registry, random rounding to the base of three protected the privacy in the data. Finally, the selection of suitable parameters is crucial for the spatiotemporal scan statistic. Combinations of other settings including spatial, temporal and analytical parameters were tested to evaluate the sensitivity of the method and results did not change considerably under any such parameter changes. Only larger and more populated clusters were identified as the result of aggregation. That means some of the local variations were hidden although the locations of primary clusters and their characteristics tend to be similar.
Conclusion
Our study used NIR data on over 4.4 million childhood immunization events to examine nationwide spatiotemporal trends of immunization coverage from 2006 to 2017. We identified areas of low immunization coverage clustering across New Zealand. Although some of the identified clusters were only temporary, there are regions in New Zealand where low immunization coverage is persistent which may indicate possible structural problems or inequities. Findings help policymakers to understand the development of childhood immunization coverage in New Zealand and will inform effective future prevention strategies. In addition, an analysis of partial immunizations and vaccination hesitancy (both area-based and individual-based) could provide further insight.
Supplementary data
Supplementary data are available at EURPUB online.
Acknowledgement
This research was carried out as part of the GeoHealth Laboratory work programme at the University of Canterbury, funded by the New Zealand Ministry of Health. Access to the data used in this study was provided by the New Zealand Ministry of Health under conditions designed to keep individual information confidential and secure in accordance with requirements of the Health Information Privacy Code 1994 and the Privacy Act 1993.
Funding
This research was carried out as part of the GeoHealth Laboratory work programme at the University of Canterbury, funded by the New Zealand Ministry of Health.
Conflicts of interest
None declared.
National Immunisation Register was used to identify well and insufficiently immunized areas in New Zealand over time.
Some of the identified clusters were only temporary; however, there are regions in New Zealand where low immunization coverage is persistent which may indicate possible structural problems or inequities.
Findings help policymakers to understand the development of childhood immunization coverage in New Zealand and will inform effective future prevention strategies.
References
World Health Organization (WHO). Measles cases spike globally due to gaps in vaccination coverage.
QGIS Development Team. QGIS Geographic Information System. Open Source Geospatial Foundation Project.
Ministry of Health | Manatū Haoura. Work programme 2019/20.
Comments