Childhood immunization during the COVID-19 pandemic: experiences in Haiti, Lesotho, Liberia and Malawi

Abstract Objective To examine changes in vaccination of children younger than 1 year during the coronavirus disease 2019 (COVID-19) pandemic (March 2020–August 2021) in Haiti, Lesotho, Liberia and Malawi. Methods We used data from health management information systems on vaccination of children aged 12 months or younger in districts supported by Partners In Health. We used data from January 2016 to February 2020 and a linear model with negative binomial distribution to estimate the expected immunization counts for March 2020–August 2021 with 95% prediction intervals, assuming no pandemic. We compared these expected levels with observed values and estimated the immunization deficits or excesses during the pandemic months. Findings Baseline vaccination counts varied substantially by country, with Lesotho having the lowest count and Haiti the highest. We observed declines in vaccination administration early in the COVID-19 pandemic in Haiti, Lesotho and Liberia. Continued declines largely corresponded to high rates of COVID-19 infection and discrete stock-outs. By August 2021, vaccination levels had returned to close to or above expected levels in Haiti, Liberia and Lesotho; in Malawi levels remained below expected. Conclusion Patterns of childhood immunization coverage varied by country over the course of the pandemic, with significantly lower than expected vaccination levels seen in one country during subsequent COVID-19 waves. Governments and health-care stakeholders should monitor vaccine coverage closely and consider interventions, such as community outreach, to avoid or combat the disruptions in childhood vaccination.


Introduction
The coronavirus disease 2019 (COVID-19) pandemic has severely affected health services globally, leading to concerns about disruptions to essential services such as immunizations. 1 Vaccinations, particularly those given in the first year of life, result in substantial reductions in mortality and are among the most cost-effective health interventions in low-and middleincome countries. [2][3][4] Before the pandemic, inequity in full vaccination coverage in low-and middle-income countries persisted due to inadequate health infrastructure, insufficient human resources and supply chain disruptions. People living in the poorest households and in remote areas are least likely to have optimal vaccination coverage and uptake. 5 These inequities were exacerbated in previous health emergencies, such as the H1N1 influenza and Ebola virus disease epidemics, 6,7 and vaccine interruptions have led to secondary disease outbreaks. 8,9 Early COVID-19 pandemic models predicted immunization interruptions and raised alarms about the possibility of increased mortality as a result. [9][10][11][12] For example, one scenario predicted that for every one excess COVID-19 death acquired during visits for routine vaccination, 84 deaths could be prevented by sustaining routine childhood immunization in Africa. 10 Governments have adopted policies such as curfews, travel bans and school closures to mitigate the spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). 13,14 However, these interventions can have unintended effects on health-service delivery. For example, of 105 countries included in the World Health Organization (WHO) pulse survey in mid-2020, 50% reported partial disruptions and 10% reported severe disruptions in facility-based immunizations. 15 Some early studies from both low-and middle-income countries and high-income countries observed reductions in vaccinations depending on geography and the corresponding pandemic prevention measures reported. [16][17][18][19][20][21][22] Of these studies, we identified just two studies showing rebounds in coverage alongside catch-up campaigns and the lifting of social distancing measures. 19,20 Important gaps in the literature exist, including limited sustained monitoring beyond initial prevention measures, no on-the-ground investigations in low-and middle-income countries outside Asia and little investigation within specific populations at risk. 23 Recognizing the potential disruption of childhood vaccinations due to COVID-19, we longitudinally monitored vaccination administration in four countries with geographic and population differences: Haiti, Lesotho, Liberia and Malawi. Within these countries, our teams support health-care delivery in specific districts, primarily sites that are rural and hard to Objective To examine changes in vaccination of children younger than 1 year during the coronavirus disease 2019 (COVID-19) pandemic (March 2020-August 2021) in Haiti, Lesotho, Liberia and Malawi. Methods We used data from health management information systems on vaccination of children aged 12 months or younger in districts supported by Partners In Health. We used data from January 2016 to February 2020 and a linear model with negative binomial distribution to estimate the expected immunization counts for March 2020-August 2021 with 95% prediction intervals, assuming no pandemic. We compared these expected levels with observed values and estimated the immunization deficits or excesses during the pandemic months. Findings Baseline vaccination counts varied substantially by country, with Lesotho having the lowest count and Haiti the highest. We observed declines in vaccination administration early in the COVID-19 pandemic in Haiti, Lesotho and Liberia. Continued declines largely corresponded to high rates of COVID-19 infection and discrete stock-outs. By August 2021, vaccination levels had returned to close to or above expected levels in Haiti, Liberia and Lesotho; in Malawi levels remained below expected. Conclusion Patterns of childhood immunization coverage varied by country over the course of the pandemic, with significantly lower than expected vaccination levels seen in one country during subsequent COVID-19 waves. Governments and health-care stakeholders should monitor vaccine coverage closely and consider interventions, such as community outreach, to avoid or combat the disruptions in childhood vaccination.
reach -the communities usually most at risk of disruption of services during an emergency. In this study, we examined how vaccinations of children younger than 1 year were affected during the COVID-19 pandemic and discuss potential ways immunization programmes can maintain essential services during acute health crises.

Study sites
Partners In Health is a global nongovernmental organization working through a so-called accompaniment model, employing close partnership with national and local governments to build strong and equitable health systems for the most vulnerable communities. 24 Partners In Health does not operate health facilities, but integrates into the public facility operations of health ministries. Therefore, support in the four countries included in this study: Haiti, Lesotho, Liberia and Malawiwhich served 10 districts, 41 facilities and about 6.4 million people, is adapted by country to fit the needs of the country's health system. Here we focus on districts supported by Partners In Health because (i) we use the health-service utilization assessment methods presented in this study to routinely monitor the pandemic there and therefore can rapidly obtain data and results; and (ii) as we are integrated into programming, we can assess the effect of the COVID-19 pandemic in the context of these locales. While Partners In Health operates in many countries, we focused on these four countries due to the availability of their data, their capacity to provide data and their prioritization of childhood immunization.
The study was approved by:

Analytical approach
As part of our support for the response to the COVID-19 pandemic at these facilities, we developed methods for ongoing monitoring of indicators of health-service utilization captured in existing routine health information systems. [25][26][27] Full details of our analytical approach are given elsewhere 25 and further specifications for this study are detailed in the supplementary material in the data repository. 28 In brief, we modelled monthly immunization counts for each facility with a negative binomial regression accounting for yearly trends and seasonality using historical data from January 2016 to February 2020, except for Haiti which started from January 2017. We did not include terms for autocorrelation, as no autocorrelation was detected in the residuals of these models. We used these models to extrapolate immunization counts to 2020 and 2021, providing estimates with 95% prediction intervals of what counts we would expect in the absence of the pandemic, aggregated across the included sites at the country level. We also computed the cumulative difference in number of vaccinations (observed -predicted vaccination counts) and per cent difference in vaccinations during the following periods of the COVID-19 pandemic: early (March-August 2020), middle (September 2020-February 2021), late (March-August 2021) and total (March 2020-August 2021). An observation was flagged as a statistically significant deviation from expected if it was less than zero and the 95% prediction interval did not contain zero.
We excluded a facility for a specific vaccine dose if: (i) that vaccine dose was missing data for more than 20% of baseline months; or (ii) that vaccine dose was missing data for any of the months of the evaluation period (March 2020-August 2021). For indicators missing baseline data, we fitted models assuming the data were missing completely at random. 25 Vaccine-dose data excluded due to missing data at the facility level were also excluded from the country-level summary reported in this study. Of the 301 facility-indicator combinations, eight (2.7%) were excluded. Data were checked for outliers and reviewed by site staff. All analyses and visualizations were done in R v4.0.4 (R Foundation, Vienna, Austria).

Immunization vaccine-dose combinations
We considered 14 vaccine-dose combinations administered to children younger than 1 year in the four countries (Table 1 available at: https:// www .who .int/ publications/ journals/ bulletin/ ). The combinations included bacillus Calmette-Guerin (BCG) vaccine, polio vaccine (oral polio vaccine or inactivated polio vaccine; doses 0-3), pentavalent vaccine (hepatitis B-Haemophilus influenzae type B-diphtheria-tetanuspertussis; doses 1-3), pneumococcal vaccine (doses 1-3), rotavirus vaccine (doses 1 and 2) and measles vaccine (dose 1). We report results for each vaccine dose by country; some combinations are not included for specific countries, as indicated in Table 2. We grouped vaccines into five classes based on age at administration: at birth, at 6 weeks, at 10 weeks, at 14 weeks and at 36 weeks. Table 3 reports the monthly volume of vaccinations administered during the baseline period (January 2016 to February 2020). The volume varied substantially by country due to catchment size, with Lesotho having the lowest number of monthly vaccinations, from 63.5 (interquartile range, IQR: 51.5 to 71.0) oral polio or inactivated polio vaccines (dose 0) administered a month to 100.0 (IQR: 79.0 to 106.0) pentavalent vaccine (dose 1) administered a month. Haiti had the highest number of vaccines administered, from 289.7 (IQR: 143.1 to 464.5) pneumococcal vaccines (dose 3) administered a month to 947.0 (IQR: 797.0 to 1114.5) BCG vaccines administered a month.

Vaccinations, March-August 2020
Overall, for all vaccines, March 2020 levels were close to expected but were followed by notable declines relative to expected levels throughout this early evaluation period, apart from in Malawi (Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Fig. 5). No country site had a statistically significant early decline in vaccines administered at birth (Fig. 1). However, Haiti, Lesotho and Liberia had a statistically significant decline in measles vaccinations early in this period with an upward trend at the end of the period (Fig. 5

Vaccinations, September 2020-February 2021
Early in the middle evaluation period (September-November 2020), most of the vaccinations administered early in childhood returned to within expected levels in all four countries ( Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Fig. 5). However, in the second half of this evaluation period, the expected vaccination numbers for dose 3 vaccines and the measles vaccine decreased in Lesotho, BCG vaccination decreased in Haiti, and BCG and doses 1-3 vaccination remained at expected or decreased levels in Malawi (Fig. 1,  Fig. 2, Fig. 3, Fig. 4 and Fig. 5). Liberia maintained vaccinations at predicted levels throughout the period except for sharp decreases in pentavalent vaccinations. For September 2020-February 2021, the median (range) cumulative percentage difference in the 14 vaccinedose combinations reported was 6.9% (-37.5% to 17.6%) greater than expected for Haiti, 11.5% (−15.7% to 17.4%) less for Lesotho, 27.8% (−12.5% to 56.5%) greater for Liberia and 9.0% (−18.0% to −3.9%) less for Malawi (Table 4); details in data repository. 28

Discussion
Pandemics can disrupt infrastructure and divert health resources. From four low-and middle-income countries in sub-Saharan Africa and the Caribbean, we found that vaccine service utilization was affected for children younger than 1 year, largely during the early months of the COVID-19 pandemic and during subsequent waves of infection. Early declines in vaccinations, with decreases of up to 75%, were also observed in the Netherlands, 19 Pakistan, 17 Singapore, 21 United Kingdom of Great Britain and Northern Ireland 18 and the United States of America (USA) 20 during full physical-distancing measures or restriction periods in early 2020. However, and importantly, the initial declines were not sustained in the facilities we studied, with rebounds to expected levels in all vaccines observed within 3 months of the initial pandemic period and in between waves of infection. These results are similar to the findings of two recently published studies on vaccination uptake. Across eight African countries from March to July 2020, initial declines were seen in immunization with the pentavalent 3 and BCG vaccines but the levels had returned to normal by July 2020. 32 In another study on health-service utilization in Kinshasa, Democratic Republic of the Congo until December 2020, no overall decrease in vaccine uptake was seen for children aged 12 months or younger. 22 We found that similar trends persisted in our study locations well into 2021. In our study, all countries except Malawi showed early declines in vaccination uptake, with Lesotho having the smallest disruptions and Liberia and Haiti having the largest near equal disruptions. The reasons for variations in vaccine coverage are likely complex and varied by country, and even between districts within a country. For example, initial restrictions during the COVID-19 pandemic differed between countries with Haiti, Liberia and Lesotho having very high initial COVID-19 stringency indexes in April 2020. 29 However, often the restrictions were not strictly adhered to in rural districts as seen across rural Africa, 33 which may account for the maintenance in immunization in Ma-    lawi. Furthermore, transient declines in BCG and polio vaccinations in Lesotho in October 2020 were due to stock-outs, demonstrating the pandemic's effect on supply chains. 12,34 Lastly, individuals were likely hesitant to visit health-care facilities, either because of the perceived risk of being infected by SARS-CoV-2 or problems with travel during restrictions. Reduced health-seeking behaviour for immunizations has been observed in past pandemics and outbreak emergen-cies 6,7 and studies early in the COVID-19 pandemic have suggested similar results. [17][18][19][20][21]32 The later declines in vaccination in 2021 corresponded with peaks in infection with higher cumulative cases and case fatality rates (Table 1). Malawi and Lesotho had large waves of infection in January-March 2021 and May-August 2021, which matched declines in several vaccinations. These declines were most pronounced in Malawi, probably be-cause of a lack of community outreach during infection waves ( Table 1). The declines in immunization are most likely due to renewed fear of visiting healthcare facilities coupled with increased restrictive measures. 33,35 In addition, stock-outs of vaccines in Haiti and Liberia contributed to discrete drops in administration of BCG and pentavalent vaccines (Table 1). 34 Generally, we found that the declines were least pronounced in vac-  cines administered at the time of birth (BCG vaccine and polio vaccine dose 0), except for transient stock-outs in Lesotho and Haiti and large infection waves in Lesotho and Malawi. The declines became more pronounced as the recommended age for a child to receive the vaccine increased. All sites had statistically significant declines in measles vaccinations administered at 9 months, especially early in the pandemic and with waves of infection, a concern that has been flagged for several low-and middle-income countries. 36 Even with restrictive measures and fear of infection, delivery in a health facility was seen as a vital health service during the pandemic, 37 which explains the small decline in the uptake of vaccines administered immediately after birth. Later vaccines often required families to travel from home to facilities; parental fear of exposing their children to infection during vaccination is a key challenge that has been observed in other studies. 16,17,33,38 Few studies thus far have reported trends in vaccinations past the early COVID-19 pandemic period in 2020. In the 10 districts in our study, uptake of childhood vaccines for infants aged 12 months or younger returned to or were maintained at expected levels within 6 months of the start of the COVID-19 pandemic and with periods of low CO-VID-19 infection rates. This finding is consistent with the one other study we identified that had extended monitoring, where the administration of first measles, mumps and rubella vaccine in the Netherlands returned to within 1-2% of baseline by September 2020. 19 The eight African countries study, which only followed immunizations until July 2020, observed a return to normal levels within 4 months. 32 Even with returns to expected levels in monthly administration through the pandemic, all countries in our study, except Liberia, had an overall cumulative decrease in median vaccination administration of between 2% and 6%. However, the only significant deviations were in both rotavirus doses in Liberia and all vaccinations in Malawi.
The maintenance of or return to close-to-normal monthly administration in Liberia, Haiti and Lesotho also likely has several reasons, including lifting of the strict restrictive measures, fewer COVID-19 cases, better public understanding of the pandemic and infection risks, and community outreach to improve vaccination. [18][19][20] In addition, the Partners In Health and district-level health ministry teams adopted several strategies to support vaccination administration during the pandemic. For example, our sites supported information campaigns through various outlets consistent with WHO recommendations for robust community health education 39 and clinical and preventive services for promotion of essential services. Other studies have shown that public health messages to encourage essential services such as vaccinations by national and international governing bodies 39 and community and/or individual vaccination awareness campaigns 16,19 have supported the uptake of vaccination services during the COVID-19 pandemic. In addition, providing extra staffing and space to separate children visiting clinics for immunizations was critical for maintaining vaccination uptake in the USA during the pandemic. 40 Furthermore, our teams provided logistical support for routine immunization campaigns, social or transport support and Liberia and Lesotho reported supporting logistics and incentives in special community catch-up immunization campaigns in their districts, which have been used elsewhere to maintain and increase vaccination coverage. 11,16,17,19,41 Despite this support, it is important to note that some teams reported persistent challenges in staffing, availability of safe spaces for vaccination, transport and cold chain equipment for separate or outreach vaccination; for example, in Malawi, especially during the second and third waves of infection, community outreach was suspended. Lastly, the COVID-19 vaccine campaign was introduced in all sites in March-May 2021. This factor may have exacerbated COVID-19 vaccine misinformation campaigns 42 and decreased utilization of routine immunization but we did not directly study this effect. Further work to understand the influence of the introduction of COVID-19 vaccination on routine im- munization is important. These issues must be addressed to maintain coverage during new waves of the COVID-19 pandemic and to reach and immunize the children missed during the early months of the pandemic.
Our study has several limitations. First, the data used from the District Health Information Software 2 are aggregated at the facility level. We cannot assess whether individual patients received all vaccines, if vaccinations were on time or delayed for a specific child, or if the catchment populations of the districts changed from the baseline period. Another challenge is the possibility that the COVID-19 pandemic affected the timeliness and completeness of facility-level reporting, particularly during restrictions; however, Partners In Health teams provided additional logistical and technical resources to ensure continued timely data collection in these 10 districts. Furthermore, the data are limited to the 10 districts within the four countries; while this may limit generalizability, these results represent a broad geography not yet included in reports on the effect of the COVID-19 pandemic.
Another important limitation is the exclusion of facilities with more than 20% missing baseline data. Country teams raised concerns that high levels of missing data at a facility could also suggest poor accuracy of the data reported, which in turn could lead to inaccurate models and predicted counts during the overall period. For this reason, in our ongoing monitoring, we do not report on sites with high levels of missing data. The countrylevel measures reported in our study are aggregated across the facility-level model predictions and observed values; 25 therefore, facilities excluded from the ongoing monitoring are excluded from these aggregate measures. While this may lead to an undercount in both the predicted and observed number of immunizations, we believe that the per cent deviations and significance of deviations are more accurate by excluding facilities with high levels of missing data. Our teams are currently exploring imputation methods so that facilities with high levels of missing data can still be included in these monitoring activities. 43 However, it is important to emphasize that we excluded very few facilities because of having more than 20% missing data, with a maximum number of four facilities, out of 41 facilities, excluded for a single vaccine dose indicator.
To ensure vaccine utilization rates are maintained despite continued waves of infection and potential renewed restrictions, we suggest governments and health-care stakeholders strengthen efforts for educating communities and parents on COVID-19 risks and the value of childhood vaccinations with targeted community outreach clinics. ■ Surveillance Working Group, which is composed of the following -Partners In