Estimating global changes in routine childhood vaccination coverage during the COVID-19 pandemic, 2020–2021

Objectives The COVID-19 pandemic has disrupted the distribution of routine immunizations globally. Multi-country studies assessing a wide spectrum of vaccines and their coverage rates are needed to determine global performance in achieving vaccination goals. Methods Global vaccine coverage data for 16 antigens were obtained from WHO/UNICEF Estimates of National Immunization Coverage. Tobit regression was performed for all country-antigen pairs for which data were continuously available between 2015–2020 or 2015–2021 to predict vaccine coverage in 2020/2021. Vaccines for which multi-dose data were available were assessed to determine whether vaccine coverage for subsequent doses were lower than that of first doses. Results Vaccine coverage was significantly lower-than-predicted for 13/16 antigens in 2020 and all assessed antigens in 2021. Lower-than-predicted vaccine coverage was typically observed in South America, Africa, Eastern Europe, and Southeast Asia. There was a statistically significant coverage drop for subsequent doses of the diphtheria-tetanus-pertussis, pneumococcus, and rotavirus vaccines compared to first doses in 2020 and 2021. Conclusion The COVID-19 pandemic exerted larger disruptions to routine vaccination services in 2021 than in 2020. Global efforts will be needed to recoup vaccine coverage losses sustained during the pandemic and broaden vaccine access in areas where coverage was previously inadequate.


Introduction
The COVID-19 pandemic era has been characterized by global disruptions in the provision and distribution of essential health services. A pulse survey conducted by the World Health Organization (WHO) in summer 2020 found that nearly all 105 participating countries were facing COVID-19-related difficulties across a range of services spanning noncommunicable disease, mental health, maternal and child/adolescent health, and nutrition; notably, rou-tine immunization was the most frequently reported COVID-19 service casualty [1]. Fear of COVID-19 contraction dissuaded many families from presenting to clinical spaces for vaccine administration, and pandemic-related staff shortages, supply-chain disruptions, and lockdowns further hampered routine immunization and ongoing campaigns [2,3]. Preliminary estimates from 2020 suggest that initial health service disruptions considerably impacted routine immunization in at least 68 countries, affecting approximately 80 million children under the age of one [4]. Of 68 Gavi-supported vaccine introductions slated for 2020, 39 were delayed by COVID-19 [5]. A third pulse survey conducted in winter 2021 found that despite evidence of some service recovery, more than half of the 129 participating countries reported continued disruptions in routine immunization, particularly in the wake of the Delta and Omicron variants [6]. In many countries, COVID-19 vaccination campaigns overshadowed routine immunization activities [6].
Systematic reviews of COVID-19 era vaccine disruption research found global reductions in third-dose diphtheria-tetanus-pertussis (DTP) coverage, often considered a proxy of overall vaccine coverage, during the first four months of 2020 [7]. In some cases, decreases in pediatric vaccination coverage of up to 80 % were detected [3], reversing years' worth of progress. Initial quantitative assessments found that third-dose DTP vaccine and first-dose measles-containing vaccine (MCV) coverage were 7.7 % and 7.9 % lower than expected in 2020 compared to previous years, respectively [8]. An analysis of BCG immunization coverage in 29 countries during the early phase of the pandemic found relative reductions in vaccine coverage ranging from 3 to 96 % among seven study nations [9]. A multi-country analysis of the African continent in 2020 found that 13 of 15 countries demonstrated evidence of declines in the monthly number of vaccine doses provided across several antigens, with a greater than 10 % reduction noted in six countries [10]; similar decreases in routine vaccination followed by measles resurgences were previously observed in Guinea [11] and Liberia [12] during Ebola outbreaks between 2013 and 2015.
Only 11 countries/territories were found to have met the 2011-2020 Global Vaccine Action Plan (GVAP) target of 90 % coverage across the assessed vaccines in 2019 even before COVID-19related disruptions [13]. The pandemic has nudged these targets out of reach, widened systemic inequalities in vaccine administration [2,14], and revived concerns for comebacks in pediatric illnesses [15], including nearly-eradicated poliomyelitis [16,17]. The 2021-2030 Immunization Agenda (IA2030) is more ambitious than GVAP, striving for a reduction in the number of no-dose children by 50 % and the introduction of 500 new vaccines in low-and middle-income countries by 2030, in addition to 90 % coverage targets across vaccines [18]. However, evaluations of global vaccine coverage during 2021, the inaugural year of IA2030, remain scarce.
In this study, we use global vaccination data from WHO/UNICEF Estimates of National Immunization Coverage (WUENIC) to assess global vaccine coverage across 16 antigens, including Bacillus Calmette-Guérin (BCG), Diphtheria-Tetanus-Pertussis (DTP, doses 1 and 3), Hepatitis B virus (HBV, birth dose and dose 3), Haemophilus influenzae type B (HIB, dose 3), Human Papillomavirus (HPV), inactivated and oral poliomyelitis virus (IPV/OPV), measles (doses 1 and 2), pneumococcus (doses 1 and 2), rubella (dose 1), and rotavirus (dose 1 and final dose). Specifically, we use historical, annual vaccine coverage data from 2015 to 2019 to estimate COVID-19related disruptions in vaccine administration globally during 2020 and 2021. The data presented may guide future efforts to restart immunization campaigns, implement catch-up vaccination programs, and ration limited vaccine supplies based on countrylevel need.

Data
In this study, we use the estimates of vaccine coverage for each country from global vaccination data from WUENIC [19]. The details of the dataset are described elsewhere [20]. Briefly, the estimates are based on quantitative data from two sources: (1) country reported coverage data reported by national authorities based on aggregated administrative reports from health service providers on the number of vaccinations administered during a given period and reported target population data; (2) survey coverage based on estimated coverage from population-based household surveys, such as the Expanded Program on Immunization cluster survey, the UNICEF Multiple Indicators Cluster Survey, and the Demographic and Health Survey, among children aged 12-23 months or 24-35 months following a review of survey methods and results. Survey results are considered for the appropriate birth cohort based on the period of data collection, and statistics are tabulated for within-target schedule vaccine administration. All data is available on the WUENIC website. For the purposes of the current study, official statistics (as opposed to administrative statistics) were analyzed for the following antigens (vaccine code): BCG vaccine (BCG), DTP-containing vaccines dose 1 (DTPCV1) and 3 (DTPCV3), HBV birth dose (HEPB_BD) and dose 3 (HEPB3), HIB dose 3, HPV (females only), IPV dose 1 (IPV1), measlescontaining vaccine dose 1 (MCV1) and 2 (MCV2), pneumococcal conjugate vaccine dose 1 (PCV1) and 2 (PCV2), third-dose poliocontaining vaccine (POL3, includes both OPV and IPV), rubellacontaining vaccine dose 1 (RCV1), and rotavirus vaccine dose 1 (ROTA1) and final dose (ROTAC).

Statistical analysis
The present analysis considered only those country-antigen pairs for which annual, continuous data between 2015-2020 or 2015-2021 were available. As shown in Fig. 1, the final number of country-antigen pairs included in the analyses was 1517 for 2020 and 1376 for 2021. Because vaccine coverage cannot exceed 100 %, we used a Tobit regression model truncated at 100 % with year as the independent variable and vaccine coverage as the dependent variable. The regression parameters were estimated based on five years of historical data (2015-2019) for each country-antigen pair; the estimated parameters were used to predict vaccine coverage in 2020 and 2021. For a small subset of country-antigen pairs, model convergence was not achieved because the variance in coverage over time was too small or all values between 2015 and 2019 were identical; for these cases, we used a (Gauss-Markov type) linear regression model or constant term without truncation at 100 % ( Fig. 1). Because values above 100 % were censored in the Tobit regression model, all predicted values above 100 % were manually set to 100 %. We then calculated the difference between the observed and predicted coverage for each country-antigen pair. These values were illustrated using global maps stratified by antigen and year (2020 vs 2021) using R version 4.1.1 (rnaturalearth package). For ease of map visibility, only maps for DTPCV3, HEPB3, MCV1, and POL3 were presented in the main text; the additional 12 maps for the remaining antigens are presented in the Supplementary Material. For each antigen, we then calculated the average difference between observed and predicted vaccine coverage for all countries included in the analysis, stratified by year (2020 vs 2021). The average observed coverage for each antigen across all countries for which data was available between 2015 and 2019 was subjected to a Tobit regression, and predictions for 2020 and 2021 were calculated. The difference between observed and predicted values was calculated with corresponding 95 % prediction intervals.
Vaccine antigens for which data for two doses were available were separately assessed to test whether subsequent doses of the same vaccine were affected differently during the COVID-19 pandemic. We constructed violin plots showing the difference between observed and predicted vaccine coverage for the following antigens: DTP (DTPCV1 vs DTPCV3), HBV (HEPB_BD vs HEPB3), MCV (MCV1 vs MCV2), PCV (PCV1 vs PCV3), and ROTA (ROTA1 vs ROTAC). Data were stratified by year (2020 vs 2021). Gaussian kernels with the bandwidth proposed by Silverman's 'rule of thumb' were used for the density estimation in the violin plots [21]. Only countries included in both the first and subsequent dose were included in the analysis. We used two-sided t-tests to test for significant differences between the change in vaccine coverage between first and subsequent doses of the same antigen in any given year. P-values of less than 0.05 were regarded as statistically significant.

Sensitivity analysis
To check the sensitivity of our results, we repeated our analyses using 2019 vaccine coverage data as the 2020 and 2021 predicted vaccine coverage for each country-antigen pair. We limited country-antigen pairs to those included in the main analysis for ease of comparison with the main analysis. Because vaccine coverage on average generally increases over time [13], the sensitivity analysis is likely to underestimate changes between predicted and observed values of vaccine coverage.
Additionally, to assess our model's predictive accuracy, we used 2014-2018 and 2013-2017 data to predict 2019 coverage values and then compared them to actual 2019 values (i.e., a one-year look-ahead and two-year look-ahead, respectively). Due to issues with data and convergence, the number of countries and antigens included in these results may not necessarily match those in the original model. Results can be found in the Supplemental Data and are briefly discussed in the Supplementary Material.

Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or the writing of this report.

Patient and public involvement
Patients and the public were not involved in the design, conduct, or reporting of this study utilizing publicly available data.

Change in vaccination coverage during 2020 and 2021
The observed and predicted vaccine coverage for the 16 assessed antigens are shown in Fig. 2 (DTPCV3, HEPB3, MCV1, and POL3), eFigure 1 (BCG, DTPCV1, HEPB_BD, HIB3, HPV_FEM, IPV1, MCV2, PCV1, PCV2, RCV1, ROTA1, and ROTAC), and Table 1, stratified by year. In 2020, vaccine coverage point estimates were 1.11 (BCG) to 7.91 % (IPV1) lower than predicted across all antigens. In 2021, vaccine coverage point estimates were 1.52 (BCG) to 12.76 % (HPV_FEM) lower than predicted across all antigens. Statistically significant decreases in vaccine coverage were observed for 13 of 16 antigens in 2020 and all assessed antigens in 2021. Though findings varied by antigen, lower-than-predicted vaccine coverage was typically observed in South America, Africa, Eastern Europe, and Southeast Asia in both 2020 and 2021. Data for each country-antigen pair, stratified by year, are available in the Supplemental Data.
The corresponding results of the sensitivity analysis are shown in eFigure 2 and eTable 1. In 2020, vaccine coverage was À0.05 (ROTA1) to 3.09 % (RCV1) lower than predicted across all antigens. In 2021, vaccine coverage was 1.65 (BCG) to 8.76 % (HPV_FEM) lower than predicted across all antigens. Statistically significant decreases in vaccine coverage were observed for 9 of 16 antigens in 2020 and 11 of 16 antigens in 2021. The geographical distribution of vaccine coverage differences reflected that of the main analysis.

Differences in first vs. subsequent vaccine coverage
The difference between observed and predicted vaccine coverage for selected antigens for which data for two doses were available are shown in Fig. 3; the results of significance testing are displayed in Table 2. For all five vaccine types in both 2020 and 2021, the difference between observed and predicted vaccine coverage was always higher for subsequent doses compared to first doses. However, these differences were only found to be significantly different among four of the five vaccine types (DTPCV, HBV vaccine, PCV, and rotavirus vaccine) for 2020 and three of the five (DTPCV, PCV, and rotavirus vaccine) in 2021. The results of the sensitivity analyses, are shown in eFigure 3 and eTable 2.

Discussion
We present one of the first assessments of global vaccine coverage across multiple antigens during the first two years of the COVID-19 pandemic. We found widespread evidence of lowerthan-expected vaccine coverage globally, which worsened in 2021 compared to the previous year. Further analysis also revealed that multi-dose vaccines showed larger deviations from predicted coverage for subsequent as opposed to first doses. These findings comprise valuable public health information that should be used to guide policy efforts to recoup pandemic-related vaccine coverage regression and expand access to routine immunization in areas where coverage was previously inadequate.
The most commonly cited reason for decreasing rates of routine immunization is fear of exposure to COVID-19 in clinical spaces, resulting in the voluntary avoidance of pediatrician offices and vaccine clinics [3,16,22]. In combination with efforts to maintain social distancing or comply with lockdowns [3,16], routine immunizations were often forgotten or willingly put off, especially during periods of high COVID-19 transmission. A WHO pulse survey found that 33 % of disruptions to routine scheduled primary care services was due to decreased care-seeking [6]. In some countries, renewed debates about the safety and efficacy of the rapidly developed COVID-19 vaccine led to ''spillover hesitancy," in which COVID-19 vaccine concerns spread into the realm of routine vaccinations, encouraging a subset of families to leave their children unimmunized against diseases for which safe vaccines have been available for decades [23]. As the pandemic continues, targeted public health messaging regarding the importance of routine immunizations is of paramount importance [  should focus on educating parents regarding not only the benefits of routine vaccinations, but also the measures that many clinics have implemented to minimize the risk of COVID-19 exposures, including environmental sanitation, improved ventilation, and separate clinic spaces for unwell patients [2,9,24,25]. In fact, previous research has shown that the benefits of receiving routine pediatric vaccinations outweigh the risks associated with possible COVID-19 exposure [26].
The COVID-19 pandemic led to considerable difficulties accessing healthcare, both from the provider and patient sides. Routine vaccination centers closed [3], or in some cases were repurposed for COVID-19 vaccination such that routine immunization was  moved to the backburner [1]. Even when spaces to administer vaccines were available, provider shortages due to overburdened systems and staff redeployments limited patient outreach [3,16]. School closures and remote learning further contributed to decreasing vaccination rates as many immunization programs in LMICs are operated through school-based channels [16,22]. Restrictions in outreach activities have also contributed to reductions in vaccine coverage, as oftentimes immunization provision is more dependent on outreach that it is on facility-based services, especially in remote areas of LMICs. Movement and travel restrictions, combined with difficulties accessing transportation to and from healthcare institutions, during the pandemic created additional barriers to vaccine administration [2,3,16]. As our grasp on the COVID-19 pandemic continues to improve, the reopening of public health centers and routine vaccination facilities and the restarting of school-based vaccination campaigns will comprise the first step in recouping vaccine losses from the past two years [3]. However, the widening of systemic inequities in healthcare access during the pandemic will require more proactive approaches to vaccine administration, including drive-through clinics and door-to-door vaccination programs [3,9,25]. National governments should make an effort to engage community-based and nongovernmental organizations to strengthen primary healthcare such that routine services can be maintained. In addition to changes in vaccine demand, border closures, travel restrictions, and shipment delays led to significant supply chain difficulties that hindered routine immunization program [3,5,27]. As the world continues to return to normal, international vaccine  organizations including WHO, UNICEF, and Gavi, the Vaccine Alliance should be leveraged to help close the routine immunization gap that COVID-19 has exacerbated [2,10]. Previously, countries also reported a lack of physical equipment, such as syringes, in addition to vaccine stockouts [27]; typical equipment necessary for mass immunization programs should also be prioritized when distributing vaccine doses. Of note, the distribution of the COVID-19 vaccine itself has been subject to substantial inequities and further entrenched the divide between high-, middle-, and low-income countries [28]. Lessons learned from COVID-19 Vaccines Global Access (COVAX), which was established specifically to minimize these inequities, should be applied to future routine immunization programs [8]. Similarly, existing COVAX distribution mechanisms should be repurposed and diversified such that they can also be utilized for routine immunization purposes including reaching zero-dose children [29,30].
Our findings also highlighted that there was a significant dropout rate for many vaccines with multiple doses. Parents may have prioritized vaccines for their younger children [3]; by the same token, many births continued to take place in health settings, allowing for easy access to newborn vaccines (e.g., HBV) to be administered just after birth [2]. For many LMICs, minimizing dropout rates has been historically challenging because routine outreach may be limited in remote areas (e.g., once per year). However, even disruptions among newborn vaccine doses were observed, possibly because a combination of fear and overstretched medical institutions encouraged some mothers to labor at home, thus precluding the swift administration of the birth dose of the HBV or BCG vaccines [3,9]. Catch-up campaigns targeting children who missed new vaccine doses or boosters will be a crucial public health priority in the aftermath of COVID-19 [25]. However, in order to ensure that catch-up campaigns do not collaterally disrupt other routine health services, governments should also prioritize strengthening routine immunization services and building health system resilience in preparation for future pandemics. Simple measures such as immunization reminders [24,31] and opportunistic vaccination [31] should also be encouraged by local governments and pediatric medicine societies in order to reach as many children as possible. As the COVID-19 vaccine begins receiving approval for children under the age of five and many parents subsequently take their children to clinics, pediatricians can suggest bundling missed doses of other vaccines into the same visit.
Though vaccination rates have generally fallen during the COVID-19 pandemic, non-pharmaceutical interventions such as masking, social distancing, school and daycare closures, workfrom-home, and lockdowns have all contributed to suppressed transmission of pediatric infectious disease, especially respiratory infections [32] but also enteric pathogens [33]. However, as restrictions are lifted and the susceptible population of children grows, the risk of outbreaks of vaccine-preventable illnesses will continue to increase. For example, the increase in the number of poliomyelitis cases in endemic countries during the early phases of the pandemic has been attributed to suspended vaccine campaigns [16]. For the first time since 2014, a case of wild-type poliovirus spread internationally from the Afghanistan/Pakistan epidemiological block to Malawi; similarly, vaccine-derived poliovirus type 2 was detected in multiple new countries [17]. The recent detection of related strains of vaccine-derived poliovirus in UK sewage water suggests that high-income countries will not be spared from these new trends [34]. As much of the world makes efforts to return to normal, heightened surveillance of pediatric infectious disease should be encouraged to catch outbreaks early and focus catchup vaccination efforts. The global community must work in tandem to prevent outbreaks at the local level to prevent further stress on already overburdened healthcare systems and crossborder spillover events.

Limitations
This study has several limitations. First, the WUENIC data uses a combination of government reporting, survey data, and modeling to calculate yearly estimates of vaccine coverage. When based on reporting from local governments, data are subject to reporting bias. In cases where reporting and data are lacking, estimates may reflect modeling alone, which may overestimate vaccine coverage during the pandemic given that modeling would use prepandemic values as a baseline. The coverage data for 18 countries in the 2021 cohort (9 % of member states) were based on extrapolations of prior data [19]. Furthermore, the data may not include those who received vaccination outside the target schedule established by a given country or during catch-up campaigns. Second, to keep our model as simple as possible, we limited our independent variable to year and did not consider other factors such as stockouts, political turmoil, natural disasters, or economic regressions. Future country-level studies may benefit from more nuanced analyses of vaccine coverage. Third, because vaccine coverage was limited for many countries and antigens, we limited our baseline modeling to 2015-2019; thus, prediction accuracy may be limited by the use of only recent data. Further studies using longer-term data with higher granularity (e.g., monthly, weekly, or daily) may be warranted. Finally, our analyses focus on national-level coverage among countries for which data were available. To further inform public health efforts, additional studies that examine subnational trends and estimate coverage in countries where data are scarce will be needed.

Conclusions
The COVID-19 pandemic has affected all facets of healthcare, including routine immunizations. Though the past several decades have been characterized by widespread gains in vaccine coverage, the first two years of the pandemic were met with regressions in vaccine uptake in pediatric populations globally. If the world hopes to meet IA2030 targets, significant efforts will be needed to reverse the losses seen during the early pandemic years and further accelerate routine immunization programs.

Contributors
Conception/design of the work: CG, AE, KSL, and SN; acquisition of data: CG, KSL; data curation: AE; analysis of data: CG, AE; interpretation of findings: all authors; drafting of the work: CG, KSL; substantially revised the work: all authors.

Data sharing
Data are openly available on the WHO website (https://immunizationdata.who.int/listing.html?topic=coverage).

Funding
This work was supported by a grant PRESTO (JPMJPR22R8) from the Japan Science and Technology Agency. The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the paper. The authors had full access to all the data in the study and had final responsibility to submit for publication.

Ethics statement
Ethical approval was not required for this secondary analysis of publicly available data.

Data availability
The data is openly available online.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.